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>
99 #include <trace/socket.h>
101 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
102 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
103 unsigned long nr_segs, loff_t pos);
104 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
105 unsigned long nr_segs, loff_t pos);
106 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
108 static int sock_close(struct inode *inode, struct file *file);
109 static unsigned int sock_poll(struct file *file,
110 struct poll_table_struct *wait);
111 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
113 static long compat_sock_ioctl(struct file *file,
114 unsigned int cmd, unsigned long arg);
116 static int sock_fasync(int fd, struct file *filp, int on);
117 static ssize_t sock_sendpage(struct file *file, struct page *page,
118 int offset, size_t size, loff_t *ppos, int more);
119 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
120 struct pipe_inode_info *pipe, size_t len,
124 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
125 * in the operation structures but are done directly via the socketcall() multiplexor.
128 static const struct file_operations socket_file_ops = {
129 .owner = THIS_MODULE,
131 .aio_read = sock_aio_read,
132 .aio_write = sock_aio_write,
134 .unlocked_ioctl = sock_ioctl,
136 .compat_ioctl = compat_sock_ioctl,
139 .open = sock_no_open, /* special open code to disallow open via /proc */
140 .release = sock_close,
141 .fasync = sock_fasync,
142 .sendpage = sock_sendpage,
143 .splice_write = generic_splice_sendpage,
144 .splice_read = sock_splice_read,
148 * The protocol list. Each protocol is registered in here.
151 static DEFINE_SPINLOCK(net_family_lock);
152 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
155 * Statistics counters of the socket lists
158 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
162 * Move socket addresses back and forth across the kernel/user
163 * divide and look after the messy bits.
166 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
167 16 for IP, 16 for IPX,
170 must be at least one bigger than
171 the AF_UNIX size (see net/unix/af_unix.c
176 * move_addr_to_kernel - copy a socket address into kernel space
177 * @uaddr: Address in user space
178 * @kaddr: Address in kernel space
179 * @ulen: Length in user space
181 * The address is copied into kernel space. If the provided address is
182 * too long an error code of -EINVAL is returned. If the copy gives
183 * invalid addresses -EFAULT is returned. On a success 0 is returned.
186 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
188 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
192 if (copy_from_user(kaddr, uaddr, ulen))
194 return audit_sockaddr(ulen, kaddr);
198 * move_addr_to_user - copy an address to user space
199 * @kaddr: kernel space address
200 * @klen: length of address in kernel
201 * @uaddr: user space address
202 * @ulen: pointer to user length field
204 * The value pointed to by ulen on entry is the buffer length available.
205 * This is overwritten with the buffer space used. -EINVAL is returned
206 * if an overlong buffer is specified or a negative buffer size. -EFAULT
207 * is returned if either the buffer or the length field are not
209 * After copying the data up to the limit the user specifies, the true
210 * length of the data is written over the length limit the user
211 * specified. Zero is returned for a success.
214 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
220 err = get_user(len, ulen);
225 if (len < 0 || len > sizeof(struct sockaddr_storage))
228 if (audit_sockaddr(klen, kaddr))
230 if (copy_to_user(uaddr, kaddr, len))
234 * "fromlen shall refer to the value before truncation.."
237 return __put_user(klen, ulen);
240 #define SOCKFS_MAGIC 0x534F434B
242 static struct kmem_cache *sock_inode_cachep __read_mostly;
244 static struct inode *sock_alloc_inode(struct super_block *sb)
246 struct socket_alloc *ei;
248 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251 init_waitqueue_head(&ei->socket.wait);
253 ei->socket.fasync_list = NULL;
254 ei->socket.state = SS_UNCONNECTED;
255 ei->socket.flags = 0;
256 ei->socket.ops = NULL;
257 ei->socket.sk = NULL;
258 ei->socket.file = NULL;
260 return &ei->vfs_inode;
263 static void sock_destroy_inode(struct inode *inode)
265 kmem_cache_free(sock_inode_cachep,
266 container_of(inode, struct socket_alloc, vfs_inode));
269 static void init_once(void *foo)
271 struct socket_alloc *ei = (struct socket_alloc *)foo;
273 inode_init_once(&ei->vfs_inode);
276 static int init_inodecache(void)
278 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
279 sizeof(struct socket_alloc),
281 (SLAB_HWCACHE_ALIGN |
282 SLAB_RECLAIM_ACCOUNT |
285 if (sock_inode_cachep == NULL)
290 static struct super_operations sockfs_ops = {
291 .alloc_inode = sock_alloc_inode,
292 .destroy_inode =sock_destroy_inode,
293 .statfs = simple_statfs,
296 static int sockfs_get_sb(struct file_system_type *fs_type,
297 int flags, const char *dev_name, void *data,
298 struct vfsmount *mnt)
300 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
304 static struct vfsmount *sock_mnt __read_mostly;
306 static struct file_system_type sock_fs_type = {
308 .get_sb = sockfs_get_sb,
309 .kill_sb = kill_anon_super,
312 static int sockfs_delete_dentry(struct dentry *dentry)
315 * At creation time, we pretended this dentry was hashed
316 * (by clearing DCACHE_UNHASHED bit in d_flags)
317 * At delete time, we restore the truth : not hashed.
318 * (so that dput() can proceed correctly)
320 dentry->d_flags |= DCACHE_UNHASHED;
325 * sockfs_dname() is called from d_path().
327 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
329 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
330 dentry->d_inode->i_ino);
333 static struct dentry_operations sockfs_dentry_operations = {
334 .d_delete = sockfs_delete_dentry,
335 .d_dname = sockfs_dname,
339 * Obtains the first available file descriptor and sets it up for use.
341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
355 static int sock_alloc_fd(struct file **filep, int flags)
359 fd = get_unused_fd_flags(flags);
360 if (likely(fd >= 0)) {
361 struct file *file = get_empty_filp();
364 if (unlikely(!file)) {
373 static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
375 struct dentry *dentry;
376 struct qstr name = { .name = "" };
378 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
379 if (unlikely(!dentry))
382 dentry->d_op = &sockfs_dentry_operations;
384 * We dont want to push this dentry into global dentry hash table.
385 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
386 * This permits a working /proc/$pid/fd/XXX on sockets
388 dentry->d_flags &= ~DCACHE_UNHASHED;
389 d_instantiate(dentry, SOCK_INODE(sock));
392 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
394 SOCK_INODE(sock)->i_fop = &socket_file_ops;
395 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
397 file->private_data = sock;
402 int sock_map_fd(struct socket *sock, int flags)
404 struct file *newfile;
405 int fd = sock_alloc_fd(&newfile, flags);
407 if (likely(fd >= 0)) {
408 int err = sock_attach_fd(sock, newfile, flags);
410 if (unlikely(err < 0)) {
415 fd_install(fd, newfile);
420 static struct socket *sock_from_file(struct file *file, int *err)
422 if (file->f_op == &socket_file_ops)
423 return file->private_data; /* set in sock_map_fd */
430 * sockfd_lookup - Go from a file number to its socket slot
432 * @err: pointer to an error code return
434 * The file handle passed in is locked and the socket it is bound
435 * too is returned. If an error occurs the err pointer is overwritten
436 * with a negative errno code and NULL is returned. The function checks
437 * for both invalid handles and passing a handle which is not a socket.
439 * On a success the socket object pointer is returned.
442 struct socket *sockfd_lookup(int fd, int *err)
453 sock = sock_from_file(file, err);
459 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
465 file = fget_light(fd, fput_needed);
467 sock = sock_from_file(file, err);
470 fput_light(file, *fput_needed);
476 * sock_alloc - allocate a socket
478 * Allocate a new inode and socket object. The two are bound together
479 * and initialised. The socket is then returned. If we are out of inodes
483 static struct socket *sock_alloc(void)
488 inode = new_inode(sock_mnt->mnt_sb);
492 sock = SOCKET_I(inode);
494 inode->i_mode = S_IFSOCK | S_IRWXUGO;
495 inode->i_uid = current->fsuid;
496 inode->i_gid = current->fsgid;
498 get_cpu_var(sockets_in_use)++;
499 put_cpu_var(sockets_in_use);
504 * In theory you can't get an open on this inode, but /proc provides
505 * a back door. Remember to keep it shut otherwise you'll let the
506 * creepy crawlies in.
509 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
514 const struct file_operations bad_sock_fops = {
515 .owner = THIS_MODULE,
516 .open = sock_no_open,
520 * sock_release - close a socket
521 * @sock: socket to close
523 * The socket is released from the protocol stack if it has a release
524 * callback, and the inode is then released if the socket is bound to
525 * an inode not a file.
528 void sock_release(struct socket *sock)
531 struct module *owner = sock->ops->owner;
533 sock->ops->release(sock);
538 if (sock->fasync_list)
539 printk(KERN_ERR "sock_release: fasync list not empty!\n");
541 get_cpu_var(sockets_in_use)--;
542 put_cpu_var(sockets_in_use);
544 iput(SOCK_INODE(sock));
550 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
551 struct msghdr *msg, size_t size)
553 struct sock_iocb *si = kiocb_to_siocb(iocb);
561 err = security_socket_sendmsg(sock, msg, size);
565 return sock->ops->sendmsg(iocb, sock, msg, size);
568 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
571 struct sock_iocb siocb;
574 init_sync_kiocb(&iocb, NULL);
575 iocb.private = &siocb;
576 ret = __sock_sendmsg(&iocb, sock, msg, size);
577 if (-EIOCBQUEUED == ret)
578 ret = wait_on_sync_kiocb(&iocb);
579 trace_socket_sendmsg(sock, msg, size, ret);
583 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
584 struct kvec *vec, size_t num, size_t size)
586 mm_segment_t oldfs = get_fs();
591 * the following is safe, since for compiler definitions of kvec and
592 * iovec are identical, yielding the same in-core layout and alignment
594 msg->msg_iov = (struct iovec *)vec;
595 msg->msg_iovlen = num;
596 result = sock_sendmsg(sock, msg, size);
602 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
604 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
607 ktime_t kt = skb->tstamp;
609 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
611 /* Race occurred between timestamp enabling and packet
612 receiving. Fill in the current time for now. */
614 kt = ktime_get_real();
616 tv = ktime_to_timeval(kt);
617 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
620 /* Race occurred between timestamp enabling and packet
621 receiving. Fill in the current time for now. */
623 kt = ktime_get_real();
625 ts = ktime_to_timespec(kt);
626 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
630 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
632 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
633 struct msghdr *msg, size_t size, int flags)
636 struct sock_iocb *si = kiocb_to_siocb(iocb);
644 err = security_socket_recvmsg(sock, msg, size, flags);
648 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
651 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
652 size_t size, int flags)
655 struct sock_iocb siocb;
658 init_sync_kiocb(&iocb, NULL);
660 iocb.private = &siocb;
661 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
662 if (-EIOCBQUEUED == ret)
663 ret = wait_on_sync_kiocb(&iocb);
664 trace_socket_recvmsg(sock, msg, size, flags, ret);
668 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
669 struct kvec *vec, size_t num, size_t size, int flags)
671 mm_segment_t oldfs = get_fs();
676 * the following is safe, since for compiler definitions of kvec and
677 * iovec are identical, yielding the same in-core layout and alignment
679 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
680 result = sock_recvmsg(sock, msg, size, flags);
685 static void sock_aio_dtor(struct kiocb *iocb)
687 kfree(iocb->private);
690 static ssize_t sock_sendpage(struct file *file, struct page *page,
691 int offset, size_t size, loff_t *ppos, int more)
696 sock = file->private_data;
698 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
702 return sock->ops->sendpage(sock, page, offset, size, flags);
705 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
706 struct pipe_inode_info *pipe, size_t len,
709 struct socket *sock = file->private_data;
711 if (unlikely(!sock->ops->splice_read))
714 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
717 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
718 struct sock_iocb *siocb)
720 if (!is_sync_kiocb(iocb)) {
721 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
724 iocb->ki_dtor = sock_aio_dtor;
728 iocb->private = siocb;
732 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
733 struct file *file, const struct iovec *iov,
734 unsigned long nr_segs)
736 struct socket *sock = file->private_data;
740 for (i = 0; i < nr_segs; i++)
741 size += iov[i].iov_len;
743 msg->msg_name = NULL;
744 msg->msg_namelen = 0;
745 msg->msg_control = NULL;
746 msg->msg_controllen = 0;
747 msg->msg_iov = (struct iovec *)iov;
748 msg->msg_iovlen = nr_segs;
749 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
751 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
754 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
755 unsigned long nr_segs, loff_t pos)
757 struct sock_iocb siocb, *x;
762 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
766 x = alloc_sock_iocb(iocb, &siocb);
769 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
772 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
773 struct file *file, const struct iovec *iov,
774 unsigned long nr_segs)
776 struct socket *sock = file->private_data;
780 for (i = 0; i < nr_segs; i++)
781 size += iov[i].iov_len;
783 msg->msg_name = NULL;
784 msg->msg_namelen = 0;
785 msg->msg_control = NULL;
786 msg->msg_controllen = 0;
787 msg->msg_iov = (struct iovec *)iov;
788 msg->msg_iovlen = nr_segs;
789 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
790 if (sock->type == SOCK_SEQPACKET)
791 msg->msg_flags |= MSG_EOR;
793 return __sock_sendmsg(iocb, sock, msg, size);
796 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
797 unsigned long nr_segs, loff_t pos)
799 struct sock_iocb siocb, *x;
804 x = alloc_sock_iocb(iocb, &siocb);
808 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
812 * Atomic setting of ioctl hooks to avoid race
813 * with module unload.
816 static DEFINE_MUTEX(br_ioctl_mutex);
817 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
819 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
821 mutex_lock(&br_ioctl_mutex);
822 br_ioctl_hook = hook;
823 mutex_unlock(&br_ioctl_mutex);
826 EXPORT_SYMBOL(brioctl_set);
828 static DEFINE_MUTEX(vlan_ioctl_mutex);
829 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
831 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
833 mutex_lock(&vlan_ioctl_mutex);
834 vlan_ioctl_hook = hook;
835 mutex_unlock(&vlan_ioctl_mutex);
838 EXPORT_SYMBOL(vlan_ioctl_set);
840 static DEFINE_MUTEX(dlci_ioctl_mutex);
841 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
843 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
845 mutex_lock(&dlci_ioctl_mutex);
846 dlci_ioctl_hook = hook;
847 mutex_unlock(&dlci_ioctl_mutex);
850 EXPORT_SYMBOL(dlci_ioctl_set);
853 * With an ioctl, arg may well be a user mode pointer, but we don't know
854 * what to do with it - that's up to the protocol still.
857 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
861 void __user *argp = (void __user *)arg;
865 sock = file->private_data;
868 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
869 err = dev_ioctl(net, cmd, argp);
871 #ifdef CONFIG_WIRELESS_EXT
872 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
873 err = dev_ioctl(net, cmd, argp);
875 #endif /* CONFIG_WIRELESS_EXT */
880 if (get_user(pid, (int __user *)argp))
882 err = f_setown(sock->file, pid, 1);
886 err = put_user(f_getown(sock->file),
895 request_module("bridge");
897 mutex_lock(&br_ioctl_mutex);
899 err = br_ioctl_hook(net, cmd, argp);
900 mutex_unlock(&br_ioctl_mutex);
905 if (!vlan_ioctl_hook)
906 request_module("8021q");
908 mutex_lock(&vlan_ioctl_mutex);
910 err = vlan_ioctl_hook(net, argp);
911 mutex_unlock(&vlan_ioctl_mutex);
916 if (!dlci_ioctl_hook)
917 request_module("dlci");
919 mutex_lock(&dlci_ioctl_mutex);
921 err = dlci_ioctl_hook(cmd, argp);
922 mutex_unlock(&dlci_ioctl_mutex);
925 err = sock->ops->ioctl(sock, cmd, arg);
928 * If this ioctl is unknown try to hand it down
931 if (err == -ENOIOCTLCMD)
932 err = dev_ioctl(net, cmd, argp);
938 int sock_create_lite(int family, int type, int protocol, struct socket **res)
941 struct socket *sock = NULL;
943 err = security_socket_create(family, type, protocol, 1);
954 err = security_socket_post_create(sock, family, type, protocol, 1);
967 /* No kernel lock held - perfect */
968 static unsigned int sock_poll(struct file *file, poll_table *wait)
973 * We can't return errors to poll, so it's either yes or no.
975 sock = file->private_data;
976 return sock->ops->poll(file, sock, wait);
979 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
981 struct socket *sock = file->private_data;
983 return sock->ops->mmap(file, sock, vma);
986 static int sock_close(struct inode *inode, struct file *filp)
989 * It was possible the inode is NULL we were
990 * closing an unfinished socket.
994 printk(KERN_DEBUG "sock_close: NULL inode\n");
997 sock_fasync(-1, filp, 0);
998 sock_release(SOCKET_I(inode));
1003 * Update the socket async list
1005 * Fasync_list locking strategy.
1007 * 1. fasync_list is modified only under process context socket lock
1008 * i.e. under semaphore.
1009 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1010 * or under socket lock.
1011 * 3. fasync_list can be used from softirq context, so that
1012 * modification under socket lock have to be enhanced with
1013 * write_lock_bh(&sk->sk_callback_lock).
1017 static int sock_fasync(int fd, struct file *filp, int on)
1019 struct fasync_struct *fa, *fna = NULL, **prev;
1020 struct socket *sock;
1024 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1029 sock = filp->private_data;
1039 prev = &(sock->fasync_list);
1041 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1042 if (fa->fa_file == filp)
1047 write_lock_bh(&sk->sk_callback_lock);
1049 write_unlock_bh(&sk->sk_callback_lock);
1054 fna->fa_file = filp;
1056 fna->magic = FASYNC_MAGIC;
1057 fna->fa_next = sock->fasync_list;
1058 write_lock_bh(&sk->sk_callback_lock);
1059 sock->fasync_list = fna;
1060 write_unlock_bh(&sk->sk_callback_lock);
1063 write_lock_bh(&sk->sk_callback_lock);
1064 *prev = fa->fa_next;
1065 write_unlock_bh(&sk->sk_callback_lock);
1071 release_sock(sock->sk);
1075 /* This function may be called only under socket lock or callback_lock */
1077 int sock_wake_async(struct socket *sock, int how, int band)
1079 if (!sock || !sock->fasync_list)
1082 case SOCK_WAKE_WAITD:
1083 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1086 case SOCK_WAKE_SPACE:
1087 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1092 __kill_fasync(sock->fasync_list, SIGIO, band);
1095 __kill_fasync(sock->fasync_list, SIGURG, band);
1100 static int __sock_create(struct net *net, int family, int type, int protocol,
1101 struct socket **res, int kern)
1104 struct socket *sock;
1105 const struct net_proto_family *pf;
1108 * Check protocol is in range
1110 if (family < 0 || family >= NPROTO)
1111 return -EAFNOSUPPORT;
1112 if (type < 0 || type >= SOCK_MAX)
1117 This uglymoron is moved from INET layer to here to avoid
1118 deadlock in module load.
1120 if (family == PF_INET && type == SOCK_PACKET) {
1124 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1130 err = security_socket_create(family, type, protocol, kern);
1135 * Allocate the socket and allow the family to set things up. if
1136 * the protocol is 0, the family is instructed to select an appropriate
1139 sock = sock_alloc();
1141 if (net_ratelimit())
1142 printk(KERN_WARNING "socket: no more sockets\n");
1143 return -ENFILE; /* Not exactly a match, but its the
1144 closest posix thing */
1149 #if defined(CONFIG_KMOD)
1150 /* Attempt to load a protocol module if the find failed.
1152 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1153 * requested real, full-featured networking support upon configuration.
1154 * Otherwise module support will break!
1156 if (net_families[family] == NULL)
1157 request_module("net-pf-%d", family);
1161 pf = rcu_dereference(net_families[family]);
1162 err = -EAFNOSUPPORT;
1167 * We will call the ->create function, that possibly is in a loadable
1168 * module, so we have to bump that loadable module refcnt first.
1170 if (!try_module_get(pf->owner))
1173 /* Now protected by module ref count */
1176 err = pf->create(net, sock, protocol);
1178 goto out_module_put;
1181 * Now to bump the refcnt of the [loadable] module that owns this
1182 * socket at sock_release time we decrement its refcnt.
1184 if (!try_module_get(sock->ops->owner))
1185 goto out_module_busy;
1188 * Now that we're done with the ->create function, the [loadable]
1189 * module can have its refcnt decremented
1191 module_put(pf->owner);
1192 err = security_socket_post_create(sock, family, type, protocol, kern);
1194 goto out_sock_release;
1200 err = -EAFNOSUPPORT;
1203 module_put(pf->owner);
1210 goto out_sock_release;
1213 int sock_create(int family, int type, int protocol, struct socket **res)
1215 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1218 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1220 return __sock_create(&init_net, family, type, protocol, res, 1);
1223 asmlinkage long sys_socket(int family, int type, int protocol)
1226 struct socket *sock;
1229 /* Check the SOCK_* constants for consistency. */
1230 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1231 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1232 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1233 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1235 flags = type & ~SOCK_TYPE_MASK;
1236 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1238 type &= SOCK_TYPE_MASK;
1240 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1241 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1243 retval = sock_create(family, type, protocol, &sock);
1247 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1251 trace_socket_create(sock, retval);
1253 /* It may be already another descriptor 8) Not kernel problem. */
1262 * Create a pair of connected sockets.
1265 asmlinkage long sys_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 err = audit_fd_pair(fd1, fd2);
1330 fd_install(fd1, newfile1);
1331 fd_install(fd2, newfile2);
1332 /* fd1 and fd2 may be already another descriptors.
1333 * Not kernel problem.
1336 err = put_user(fd1, &usockvec[0]);
1338 err = put_user(fd2, &usockvec[1]);
1347 sock_release(sock2);
1349 sock_release(sock1);
1355 sock_release(sock1);
1358 sock_release(sock2);
1366 * Bind a name to a socket. Nothing much to do here since it's
1367 * the protocol's responsibility to handle the local address.
1369 * We move the socket address to kernel space before we call
1370 * the protocol layer (having also checked the address is ok).
1373 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1375 struct socket *sock;
1376 struct sockaddr_storage address;
1377 int err, fput_needed;
1379 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1381 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1383 err = security_socket_bind(sock,
1384 (struct sockaddr *)&address,
1387 err = sock->ops->bind(sock,
1391 fput_light(sock->file, fput_needed);
1397 * Perform a listen. Basically, we allow the protocol to do anything
1398 * necessary for a listen, and if that works, we mark the socket as
1399 * ready for listening.
1402 asmlinkage long sys_listen(int fd, int backlog)
1404 struct socket *sock;
1405 int err, fput_needed;
1408 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1410 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1411 if ((unsigned)backlog > somaxconn)
1412 backlog = somaxconn;
1414 err = security_socket_listen(sock, backlog);
1416 err = sock->ops->listen(sock, backlog);
1418 fput_light(sock->file, fput_needed);
1424 * For accept, we attempt to create a new socket, set up the link
1425 * with the client, wake up the client, then return the new
1426 * connected fd. We collect the address of the connector in kernel
1427 * space and move it to user at the very end. This is unclean because
1428 * we open the socket then return an error.
1430 * 1003.1g adds the ability to recvmsg() to query connection pending
1431 * status to recvmsg. We need to add that support in a way thats
1432 * clean when we restucture accept also.
1435 long do_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1436 int __user *upeer_addrlen, int flags)
1438 struct socket *sock, *newsock;
1439 struct file *newfile;
1440 int err, len, newfd, fput_needed;
1441 struct sockaddr_storage address;
1443 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1446 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1447 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1449 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1454 if (!(newsock = sock_alloc()))
1457 newsock->type = sock->type;
1458 newsock->ops = sock->ops;
1461 * We don't need try_module_get here, as the listening socket (sock)
1462 * has the protocol module (sock->ops->owner) held.
1464 __module_get(newsock->ops->owner);
1466 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
1467 if (unlikely(newfd < 0)) {
1469 sock_release(newsock);
1473 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
1477 err = security_socket_accept(sock, newsock);
1481 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1485 if (upeer_sockaddr) {
1486 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1488 err = -ECONNABORTED;
1491 err = move_addr_to_user((struct sockaddr *)&address,
1492 len, upeer_sockaddr, upeer_addrlen);
1497 /* File flags are not inherited via accept() unlike another OSes. */
1499 fd_install(newfd, newfile);
1502 security_socket_post_accept(sock, newsock);
1505 fput_light(sock->file, fput_needed);
1509 sock_release(newsock);
1511 put_unused_fd(newfd);
1515 put_unused_fd(newfd);
1519 #ifdef HAVE_SET_RESTORE_SIGMASK
1520 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1521 int __user *upeer_addrlen,
1522 const sigset_t __user *sigmask,
1523 size_t sigsetsize, int flags)
1525 sigset_t ksigmask, sigsaved;
1529 /* XXX: Don't preclude handling different sized sigset_t's. */
1530 if (sigsetsize != sizeof(sigset_t))
1532 if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
1535 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1536 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1539 ret = do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1541 if (ret < 0 && signal_pending(current)) {
1543 * Don't restore the signal mask yet. Let do_signal() deliver
1544 * the signal on the way back to userspace, before the signal
1548 memcpy(¤t->saved_sigmask, &sigsaved,
1550 set_restore_sigmask();
1553 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1558 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1559 int __user *upeer_addrlen,
1560 const sigset_t __user *sigmask,
1561 size_t sigsetsize, int flags)
1563 /* The platform does not support restoring the signal mask in the
1564 * return path. So we do not allow using paccept() with a signal
1569 return do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1573 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1574 int __user *upeer_addrlen)
1576 return do_accept(fd, upeer_sockaddr, upeer_addrlen, 0);
1580 * Attempt to connect to a socket with the server address. The address
1581 * is in user space so we verify it is OK and move it to kernel space.
1583 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1586 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1587 * other SEQPACKET protocols that take time to connect() as it doesn't
1588 * include the -EINPROGRESS status for such sockets.
1591 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1594 struct socket *sock;
1595 struct sockaddr_storage address;
1596 int err, fput_needed;
1598 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1601 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1606 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1610 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1611 sock->file->f_flags);
1613 fput_light(sock->file, fput_needed);
1619 * Get the local address ('name') of a socket object. Move the obtained
1620 * name to user space.
1623 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1624 int __user *usockaddr_len)
1626 struct socket *sock;
1627 struct sockaddr_storage address;
1628 int len, err, fput_needed;
1630 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1634 err = security_socket_getsockname(sock);
1638 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1641 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1644 fput_light(sock->file, fput_needed);
1650 * Get the remote address ('name') of a socket object. Move the obtained
1651 * name to user space.
1654 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1655 int __user *usockaddr_len)
1657 struct socket *sock;
1658 struct sockaddr_storage address;
1659 int len, err, fput_needed;
1661 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1663 err = security_socket_getpeername(sock);
1665 fput_light(sock->file, fput_needed);
1670 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1673 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1675 fput_light(sock->file, fput_needed);
1681 * Send a datagram to a given address. We move the address into kernel
1682 * space and check the user space data area is readable before invoking
1686 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1687 unsigned flags, struct sockaddr __user *addr,
1690 struct socket *sock;
1691 struct sockaddr_storage address;
1697 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1701 iov.iov_base = buff;
1703 msg.msg_name = NULL;
1706 msg.msg_control = NULL;
1707 msg.msg_controllen = 0;
1708 msg.msg_namelen = 0;
1710 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1713 msg.msg_name = (struct sockaddr *)&address;
1714 msg.msg_namelen = addr_len;
1716 if (sock->file->f_flags & O_NONBLOCK)
1717 flags |= MSG_DONTWAIT;
1718 msg.msg_flags = flags;
1719 err = sock_sendmsg(sock, &msg, len);
1722 fput_light(sock->file, fput_needed);
1728 * Send a datagram down a socket.
1731 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1733 return sys_sendto(fd, buff, len, flags, NULL, 0);
1737 * Receive a frame from the socket and optionally record the address of the
1738 * sender. We verify the buffers are writable and if needed move the
1739 * sender address from kernel to user space.
1742 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1743 unsigned flags, struct sockaddr __user *addr,
1744 int __user *addr_len)
1746 struct socket *sock;
1749 struct sockaddr_storage address;
1753 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1757 msg.msg_control = NULL;
1758 msg.msg_controllen = 0;
1762 iov.iov_base = ubuf;
1763 msg.msg_name = (struct sockaddr *)&address;
1764 msg.msg_namelen = sizeof(address);
1765 if (sock->file->f_flags & O_NONBLOCK)
1766 flags |= MSG_DONTWAIT;
1767 err = sock_recvmsg(sock, &msg, size, flags);
1769 if (err >= 0 && addr != NULL) {
1770 err2 = move_addr_to_user((struct sockaddr *)&address,
1771 msg.msg_namelen, addr, addr_len);
1776 fput_light(sock->file, fput_needed);
1782 * Receive a datagram from a socket.
1785 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1788 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1792 * Set a socket option. Because we don't know the option lengths we have
1793 * to pass the user mode parameter for the protocols to sort out.
1796 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1797 char __user *optval, int optlen)
1799 int err, fput_needed;
1800 struct socket *sock;
1805 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1807 err = security_socket_setsockopt(sock, level, optname);
1811 if (level == SOL_SOCKET)
1813 sock_setsockopt(sock, level, optname, optval,
1817 sock->ops->setsockopt(sock, level, optname, optval,
1820 fput_light(sock->file, fput_needed);
1826 * Get a socket option. Because we don't know the option lengths we have
1827 * to pass a user mode parameter for the protocols to sort out.
1830 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1831 char __user *optval, int __user *optlen)
1833 int err, fput_needed;
1834 struct socket *sock;
1836 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1838 err = security_socket_getsockopt(sock, level, optname);
1842 if (level == SOL_SOCKET)
1844 sock_getsockopt(sock, level, optname, optval,
1848 sock->ops->getsockopt(sock, level, optname, optval,
1851 fput_light(sock->file, fput_needed);
1857 * Shutdown a socket.
1860 asmlinkage long sys_shutdown(int fd, int how)
1862 int err, fput_needed;
1863 struct socket *sock;
1865 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1867 err = security_socket_shutdown(sock, how);
1869 err = sock->ops->shutdown(sock, how);
1870 fput_light(sock->file, fput_needed);
1875 /* A couple of helpful macros for getting the address of the 32/64 bit
1876 * fields which are the same type (int / unsigned) on our platforms.
1878 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1879 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1880 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1883 * BSD sendmsg interface
1886 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1888 struct compat_msghdr __user *msg_compat =
1889 (struct compat_msghdr __user *)msg;
1890 struct socket *sock;
1891 struct sockaddr_storage address;
1892 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1893 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1894 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1895 /* 20 is size of ipv6_pktinfo */
1896 unsigned char *ctl_buf = ctl;
1897 struct msghdr msg_sys;
1898 int err, ctl_len, iov_size, total_len;
1902 if (MSG_CMSG_COMPAT & flags) {
1903 if (get_compat_msghdr(&msg_sys, msg_compat))
1906 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1909 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1913 /* do not move before msg_sys is valid */
1915 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1918 /* Check whether to allocate the iovec area */
1920 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1921 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1922 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1927 /* This will also move the address data into kernel space */
1928 if (MSG_CMSG_COMPAT & flags) {
1929 err = verify_compat_iovec(&msg_sys, iov,
1930 (struct sockaddr *)&address,
1933 err = verify_iovec(&msg_sys, iov,
1934 (struct sockaddr *)&address,
1942 if (msg_sys.msg_controllen > INT_MAX)
1944 ctl_len = msg_sys.msg_controllen;
1945 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1947 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1951 ctl_buf = msg_sys.msg_control;
1952 ctl_len = msg_sys.msg_controllen;
1953 } else if (ctl_len) {
1954 if (ctl_len > sizeof(ctl)) {
1955 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1956 if (ctl_buf == NULL)
1961 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1962 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1963 * checking falls down on this.
1965 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1968 msg_sys.msg_control = ctl_buf;
1970 msg_sys.msg_flags = flags;
1972 if (sock->file->f_flags & O_NONBLOCK)
1973 msg_sys.msg_flags |= MSG_DONTWAIT;
1974 err = sock_sendmsg(sock, &msg_sys, total_len);
1978 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1980 if (iov != iovstack)
1981 sock_kfree_s(sock->sk, iov, iov_size);
1983 fput_light(sock->file, fput_needed);
1989 * BSD recvmsg interface
1992 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1995 struct compat_msghdr __user *msg_compat =
1996 (struct compat_msghdr __user *)msg;
1997 struct socket *sock;
1998 struct iovec iovstack[UIO_FASTIOV];
1999 struct iovec *iov = iovstack;
2000 struct msghdr msg_sys;
2001 unsigned long cmsg_ptr;
2002 int err, iov_size, total_len, len;
2005 /* kernel mode address */
2006 struct sockaddr_storage addr;
2008 /* user mode address pointers */
2009 struct sockaddr __user *uaddr;
2010 int __user *uaddr_len;
2012 if (MSG_CMSG_COMPAT & flags) {
2013 if (get_compat_msghdr(&msg_sys, msg_compat))
2016 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
2019 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2024 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2027 /* Check whether to allocate the iovec area */
2029 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2030 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2031 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2037 * Save the user-mode address (verify_iovec will change the
2038 * kernel msghdr to use the kernel address space)
2041 uaddr = (__force void __user *)msg_sys.msg_name;
2042 uaddr_len = COMPAT_NAMELEN(msg);
2043 if (MSG_CMSG_COMPAT & flags) {
2044 err = verify_compat_iovec(&msg_sys, iov,
2045 (struct sockaddr *)&addr,
2048 err = verify_iovec(&msg_sys, iov,
2049 (struct sockaddr *)&addr,
2055 cmsg_ptr = (unsigned long)msg_sys.msg_control;
2056 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2058 if (sock->file->f_flags & O_NONBLOCK)
2059 flags |= MSG_DONTWAIT;
2060 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2065 if (uaddr != NULL) {
2066 err = move_addr_to_user((struct sockaddr *)&addr,
2067 msg_sys.msg_namelen, uaddr,
2072 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2076 if (MSG_CMSG_COMPAT & flags)
2077 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2078 &msg_compat->msg_controllen);
2080 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2081 &msg->msg_controllen);
2087 if (iov != iovstack)
2088 sock_kfree_s(sock->sk, iov, iov_size);
2090 fput_light(sock->file, fput_needed);
2095 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2097 /* Argument list sizes for sys_socketcall */
2098 #define AL(x) ((x) * sizeof(unsigned long))
2099 static const unsigned char nargs[19]={
2100 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2101 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2102 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2109 * System call vectors.
2111 * Argument checking cleaned up. Saved 20% in size.
2112 * This function doesn't need to set the kernel lock because
2113 * it is set by the callees.
2116 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2119 unsigned long a0, a1;
2122 if (call < 1 || call > SYS_PACCEPT)
2125 /* copy_from_user should be SMP safe. */
2126 if (copy_from_user(a, args, nargs[call]))
2129 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2136 trace_socket_call(call, a0);
2140 err = sys_socket(a0, a1, a[2]);
2143 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2146 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2149 err = sys_listen(a0, a1);
2153 do_accept(a0, (struct sockaddr __user *)a1,
2154 (int __user *)a[2], 0);
2156 case SYS_GETSOCKNAME:
2158 sys_getsockname(a0, (struct sockaddr __user *)a1,
2159 (int __user *)a[2]);
2161 case SYS_GETPEERNAME:
2163 sys_getpeername(a0, (struct sockaddr __user *)a1,
2164 (int __user *)a[2]);
2166 case SYS_SOCKETPAIR:
2167 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2170 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2173 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2174 (struct sockaddr __user *)a[4], a[5]);
2177 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2180 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2181 (struct sockaddr __user *)a[4],
2182 (int __user *)a[5]);
2185 err = sys_shutdown(a0, a1);
2187 case SYS_SETSOCKOPT:
2188 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2190 case SYS_GETSOCKOPT:
2192 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2193 (int __user *)a[4]);
2196 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2199 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2203 sys_paccept(a0, (struct sockaddr __user *)a1,
2205 (const sigset_t __user *) a[3],
2215 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2218 * sock_register - add a socket protocol handler
2219 * @ops: description of protocol
2221 * This function is called by a protocol handler that wants to
2222 * advertise its address family, and have it linked into the
2223 * socket interface. The value ops->family coresponds to the
2224 * socket system call protocol family.
2226 int sock_register(const struct net_proto_family *ops)
2230 if (ops->family >= NPROTO) {
2231 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2236 spin_lock(&net_family_lock);
2237 if (net_families[ops->family])
2240 net_families[ops->family] = ops;
2243 spin_unlock(&net_family_lock);
2245 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2250 * sock_unregister - remove a protocol handler
2251 * @family: protocol family to remove
2253 * This function is called by a protocol handler that wants to
2254 * remove its address family, and have it unlinked from the
2255 * new socket creation.
2257 * If protocol handler is a module, then it can use module reference
2258 * counts to protect against new references. If protocol handler is not
2259 * a module then it needs to provide its own protection in
2260 * the ops->create routine.
2262 void sock_unregister(int family)
2264 BUG_ON(family < 0 || family >= NPROTO);
2266 spin_lock(&net_family_lock);
2267 net_families[family] = NULL;
2268 spin_unlock(&net_family_lock);
2272 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2275 static int __init sock_init(void)
2278 * Initialize sock SLAB cache.
2284 * Initialize skbuff SLAB cache
2289 * Initialize the protocols module.
2293 register_filesystem(&sock_fs_type);
2294 sock_mnt = kern_mount(&sock_fs_type);
2296 /* The real protocol initialization is performed in later initcalls.
2299 #ifdef CONFIG_NETFILTER
2306 core_initcall(sock_init); /* early initcall */
2308 #ifdef CONFIG_PROC_FS
2309 void socket_seq_show(struct seq_file *seq)
2314 for_each_possible_cpu(cpu)
2315 counter += per_cpu(sockets_in_use, cpu);
2317 /* It can be negative, by the way. 8) */
2321 seq_printf(seq, "sockets: used %d\n", counter);
2323 #endif /* CONFIG_PROC_FS */
2325 #ifdef CONFIG_COMPAT
2326 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2329 struct socket *sock = file->private_data;
2330 int ret = -ENOIOCTLCMD;
2337 if (sock->ops->compat_ioctl)
2338 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2340 if (ret == -ENOIOCTLCMD &&
2341 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2342 ret = compat_wext_handle_ioctl(net, cmd, arg);
2348 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2350 return sock->ops->bind(sock, addr, addrlen);
2353 int kernel_listen(struct socket *sock, int backlog)
2355 return sock->ops->listen(sock, backlog);
2358 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2360 struct sock *sk = sock->sk;
2363 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2368 err = sock->ops->accept(sock, *newsock, flags);
2370 sock_release(*newsock);
2375 (*newsock)->ops = sock->ops;
2381 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2384 return sock->ops->connect(sock, addr, addrlen, flags);
2387 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2390 return sock->ops->getname(sock, addr, addrlen, 0);
2393 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2396 return sock->ops->getname(sock, addr, addrlen, 1);
2399 int kernel_getsockopt(struct socket *sock, int level, int optname,
2400 char *optval, int *optlen)
2402 mm_segment_t oldfs = get_fs();
2406 if (level == SOL_SOCKET)
2407 err = sock_getsockopt(sock, level, optname, optval, optlen);
2409 err = sock->ops->getsockopt(sock, level, optname, optval,
2415 int kernel_setsockopt(struct socket *sock, int level, int optname,
2416 char *optval, int optlen)
2418 mm_segment_t oldfs = get_fs();
2422 if (level == SOL_SOCKET)
2423 err = sock_setsockopt(sock, level, optname, optval, optlen);
2425 err = sock->ops->setsockopt(sock, level, optname, optval,
2431 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2432 size_t size, int flags)
2434 if (sock->ops->sendpage)
2435 return sock->ops->sendpage(sock, page, offset, size, flags);
2437 return sock_no_sendpage(sock, page, offset, size, flags);
2440 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2442 mm_segment_t oldfs = get_fs();
2446 err = sock->ops->ioctl(sock, cmd, arg);
2452 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2454 return sock->ops->shutdown(sock, how);
2457 EXPORT_SYMBOL(sock_create);
2458 EXPORT_SYMBOL(sock_create_kern);
2459 EXPORT_SYMBOL(sock_create_lite);
2460 EXPORT_SYMBOL(sock_map_fd);
2461 EXPORT_SYMBOL(sock_recvmsg);
2462 EXPORT_SYMBOL(sock_register);
2463 EXPORT_SYMBOL(sock_release);
2464 EXPORT_SYMBOL(sock_sendmsg);
2465 EXPORT_SYMBOL(sock_unregister);
2466 EXPORT_SYMBOL(sock_wake_async);
2467 EXPORT_SYMBOL(sockfd_lookup);
2468 EXPORT_SYMBOL(kernel_sendmsg);
2469 EXPORT_SYMBOL(kernel_recvmsg);
2470 EXPORT_SYMBOL(kernel_bind);
2471 EXPORT_SYMBOL(kernel_listen);
2472 EXPORT_SYMBOL(kernel_accept);
2473 EXPORT_SYMBOL(kernel_connect);
2474 EXPORT_SYMBOL(kernel_getsockname);
2475 EXPORT_SYMBOL(kernel_getpeername);
2476 EXPORT_SYMBOL(kernel_getsockopt);
2477 EXPORT_SYMBOL(kernel_setsockopt);
2478 EXPORT_SYMBOL(kernel_sendpage);
2479 EXPORT_SYMBOL(kernel_sock_ioctl);
2480 EXPORT_SYMBOL(kernel_sock_shutdown);