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>
89 #include <linux/magic.h>
91 #include <asm/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/sockios.h>
104 #include <linux/atalk.h>
106 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
107 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
108 unsigned long nr_segs, loff_t pos);
109 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
113 static int sock_close(struct inode *inode, struct file *file);
114 static unsigned int sock_poll(struct file *file,
115 struct poll_table_struct *wait);
116 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
118 static long compat_sock_ioctl(struct file *file,
119 unsigned int cmd, unsigned long arg);
121 static int sock_fasync(int fd, struct file *filp, int on);
122 static ssize_t sock_sendpage(struct file *file, struct page *page,
123 int offset, size_t size, loff_t *ppos, int more);
124 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
125 struct pipe_inode_info *pipe, size_t len,
129 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
130 * in the operation structures but are done directly via the socketcall() multiplexor.
133 static const struct file_operations socket_file_ops = {
134 .owner = THIS_MODULE,
136 .aio_read = sock_aio_read,
137 .aio_write = sock_aio_write,
139 .unlocked_ioctl = sock_ioctl,
141 .compat_ioctl = compat_sock_ioctl,
144 .open = sock_no_open, /* special open code to disallow open via /proc */
145 .release = sock_close,
146 .fasync = sock_fasync,
147 .sendpage = sock_sendpage,
148 .splice_write = generic_splice_sendpage,
149 .splice_read = sock_splice_read,
153 * The protocol list. Each protocol is registered in here.
156 static DEFINE_SPINLOCK(net_family_lock);
157 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
160 * Statistics counters of the socket lists
163 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
171 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
172 16 for IP, 16 for IPX,
175 must be at least one bigger than
176 the AF_UNIX size (see net/unix/af_unix.c
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
191 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
193 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
197 if (copy_from_user(kaddr, uaddr, ulen))
199 return audit_sockaddr(ulen, kaddr);
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
219 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
225 err = get_user(len, ulen);
230 if (len < 0 || len > sizeof(struct sockaddr_storage))
233 if (audit_sockaddr(klen, kaddr))
235 if (copy_to_user(uaddr, kaddr, len))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen, ulen);
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
247 static struct inode *sock_alloc_inode(struct super_block *sb)
249 struct socket_alloc *ei;
251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254 init_waitqueue_head(&ei->socket.wait);
256 ei->socket.fasync_list = NULL;
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 kmem_cache_free(sock_inode_cachep,
269 container_of(inode, struct socket_alloc, vfs_inode));
272 static void init_once(void *foo)
274 struct socket_alloc *ei = (struct socket_alloc *)foo;
276 inode_init_once(&ei->vfs_inode);
279 static int init_inodecache(void)
281 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
282 sizeof(struct socket_alloc),
284 (SLAB_HWCACHE_ALIGN |
285 SLAB_RECLAIM_ACCOUNT |
288 if (sock_inode_cachep == NULL)
293 static const struct super_operations sockfs_ops = {
294 .alloc_inode = sock_alloc_inode,
295 .destroy_inode =sock_destroy_inode,
296 .statfs = simple_statfs,
299 static int sockfs_get_sb(struct file_system_type *fs_type,
300 int flags, const char *dev_name, void *data,
301 struct vfsmount *mnt)
303 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
307 static struct vfsmount *sock_mnt __read_mostly;
309 static struct file_system_type sock_fs_type = {
311 .get_sb = sockfs_get_sb,
312 .kill_sb = kill_anon_super,
315 static int sockfs_delete_dentry(struct dentry *dentry)
318 * At creation time, we pretended this dentry was hashed
319 * (by clearing DCACHE_UNHASHED bit in d_flags)
320 * At delete time, we restore the truth : not hashed.
321 * (so that dput() can proceed correctly)
323 dentry->d_flags |= DCACHE_UNHASHED;
328 * sockfs_dname() is called from d_path().
330 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
332 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
333 dentry->d_inode->i_ino);
336 static const struct dentry_operations sockfs_dentry_operations = {
337 .d_delete = sockfs_delete_dentry,
338 .d_dname = sockfs_dname,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
360 struct qstr name = { .name = "" };
361 struct dentry *dentry;
365 fd = get_unused_fd_flags(flags);
366 if (unlikely(fd < 0))
369 file = get_empty_filp();
371 if (unlikely(!file)) {
376 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
377 if (unlikely(!dentry)) {
383 dentry->d_op = &sockfs_dentry_operations;
385 * We dont want to push this dentry into global dentry hash table.
386 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
387 * This permits a working /proc/$pid/fd/XXX on sockets
389 dentry->d_flags &= ~DCACHE_UNHASHED;
390 d_instantiate(dentry, SOCK_INODE(sock));
393 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
395 SOCK_INODE(sock)->i_fop = &socket_file_ops;
396 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
398 file->private_data = sock;
404 int sock_map_fd(struct socket *sock, int flags)
406 struct file *newfile;
407 int fd = sock_alloc_file(sock, &newfile, flags);
410 fd_install(fd, newfile);
415 static struct socket *sock_from_file(struct file *file, int *err)
417 if (file->f_op == &socket_file_ops)
418 return file->private_data; /* set in sock_map_fd */
425 * sockfd_lookup - Go from a file number to its socket slot
427 * @err: pointer to an error code return
429 * The file handle passed in is locked and the socket it is bound
430 * too is returned. If an error occurs the err pointer is overwritten
431 * with a negative errno code and NULL is returned. The function checks
432 * for both invalid handles and passing a handle which is not a socket.
434 * On a success the socket object pointer is returned.
437 struct socket *sockfd_lookup(int fd, int *err)
448 sock = sock_from_file(file, err);
454 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
460 file = fget_light(fd, fput_needed);
462 sock = sock_from_file(file, err);
465 fput_light(file, *fput_needed);
471 * sock_alloc - allocate a socket
473 * Allocate a new inode and socket object. The two are bound together
474 * and initialised. The socket is then returned. If we are out of inodes
478 static struct socket *sock_alloc(void)
483 inode = new_inode(sock_mnt->mnt_sb);
487 sock = SOCKET_I(inode);
489 kmemcheck_annotate_bitfield(sock, type);
490 inode->i_mode = S_IFSOCK | S_IRWXUGO;
491 inode->i_uid = current_fsuid();
492 inode->i_gid = current_fsgid();
494 percpu_add(sockets_in_use, 1);
499 * In theory you can't get an open on this inode, but /proc provides
500 * a back door. Remember to keep it shut otherwise you'll let the
501 * creepy crawlies in.
504 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
509 const struct file_operations bad_sock_fops = {
510 .owner = THIS_MODULE,
511 .open = sock_no_open,
515 * sock_release - close a socket
516 * @sock: socket to close
518 * The socket is released from the protocol stack if it has a release
519 * callback, and the inode is then released if the socket is bound to
520 * an inode not a file.
523 void sock_release(struct socket *sock)
526 struct module *owner = sock->ops->owner;
528 sock->ops->release(sock);
533 if (sock->fasync_list)
534 printk(KERN_ERR "sock_release: fasync list not empty!\n");
536 percpu_sub(sockets_in_use, 1);
538 iput(SOCK_INODE(sock));
544 int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
545 union skb_shared_tx *shtx)
548 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
550 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
554 EXPORT_SYMBOL(sock_tx_timestamp);
556 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
557 struct msghdr *msg, size_t size)
559 struct sock_iocb *si = kiocb_to_siocb(iocb);
567 err = security_socket_sendmsg(sock, msg, size);
571 return sock->ops->sendmsg(iocb, sock, msg, size);
574 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
577 struct sock_iocb siocb;
580 init_sync_kiocb(&iocb, NULL);
581 iocb.private = &siocb;
582 ret = __sock_sendmsg(&iocb, sock, msg, size);
583 if (-EIOCBQUEUED == ret)
584 ret = wait_on_sync_kiocb(&iocb);
588 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
589 struct kvec *vec, size_t num, size_t size)
591 mm_segment_t oldfs = get_fs();
596 * the following is safe, since for compiler definitions of kvec and
597 * iovec are identical, yielding the same in-core layout and alignment
599 msg->msg_iov = (struct iovec *)vec;
600 msg->msg_iovlen = num;
601 result = sock_sendmsg(sock, msg, size);
606 static int ktime2ts(ktime_t kt, struct timespec *ts)
609 *ts = ktime_to_timespec(kt);
617 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
619 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
622 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
623 struct timespec ts[3];
625 struct skb_shared_hwtstamps *shhwtstamps =
628 /* Race occurred between timestamp enabling and packet
629 receiving. Fill in the current time for now. */
630 if (need_software_tstamp && skb->tstamp.tv64 == 0)
631 __net_timestamp(skb);
633 if (need_software_tstamp) {
634 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
636 skb_get_timestamp(skb, &tv);
637 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
641 skb_get_timestampns(skb, &ts);
642 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
648 memset(ts, 0, sizeof(ts));
649 if (skb->tstamp.tv64 &&
650 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
651 skb_get_timestampns(skb, ts + 0);
655 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
656 ktime2ts(shhwtstamps->syststamp, ts + 1))
658 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
659 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
663 put_cmsg(msg, SOL_SOCKET,
664 SCM_TIMESTAMPING, sizeof(ts), &ts);
667 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
669 inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
671 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
672 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
673 sizeof(__u32), &skb->dropcount);
676 void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
679 sock_recv_timestamp(msg, sk, skb);
680 sock_recv_drops(msg, sk, skb);
682 EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops);
684 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
685 struct msghdr *msg, size_t size, int flags)
687 struct sock_iocb *si = kiocb_to_siocb(iocb);
695 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
698 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
699 struct msghdr *msg, size_t size, int flags)
701 int err = security_socket_recvmsg(sock, msg, size, flags);
703 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
706 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
707 size_t size, int flags)
710 struct sock_iocb siocb;
713 init_sync_kiocb(&iocb, NULL);
714 iocb.private = &siocb;
715 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
716 if (-EIOCBQUEUED == ret)
717 ret = wait_on_sync_kiocb(&iocb);
721 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
722 size_t size, int flags)
725 struct sock_iocb siocb;
728 init_sync_kiocb(&iocb, NULL);
729 iocb.private = &siocb;
730 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
731 if (-EIOCBQUEUED == ret)
732 ret = wait_on_sync_kiocb(&iocb);
736 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
737 struct kvec *vec, size_t num, size_t size, int flags)
739 mm_segment_t oldfs = get_fs();
744 * the following is safe, since for compiler definitions of kvec and
745 * iovec are identical, yielding the same in-core layout and alignment
747 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
748 result = sock_recvmsg(sock, msg, size, flags);
753 static void sock_aio_dtor(struct kiocb *iocb)
755 kfree(iocb->private);
758 static ssize_t sock_sendpage(struct file *file, struct page *page,
759 int offset, size_t size, loff_t *ppos, int more)
764 sock = file->private_data;
766 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
770 return kernel_sendpage(sock, page, offset, size, flags);
773 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
774 struct pipe_inode_info *pipe, size_t len,
777 struct socket *sock = file->private_data;
779 if (unlikely(!sock->ops->splice_read))
782 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
785 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
786 struct sock_iocb *siocb)
788 if (!is_sync_kiocb(iocb)) {
789 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
792 iocb->ki_dtor = sock_aio_dtor;
796 iocb->private = siocb;
800 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
801 struct file *file, const struct iovec *iov,
802 unsigned long nr_segs)
804 struct socket *sock = file->private_data;
808 for (i = 0; i < nr_segs; i++)
809 size += iov[i].iov_len;
811 msg->msg_name = NULL;
812 msg->msg_namelen = 0;
813 msg->msg_control = NULL;
814 msg->msg_controllen = 0;
815 msg->msg_iov = (struct iovec *)iov;
816 msg->msg_iovlen = nr_segs;
817 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
819 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
822 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
823 unsigned long nr_segs, loff_t pos)
825 struct sock_iocb siocb, *x;
830 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
834 x = alloc_sock_iocb(iocb, &siocb);
837 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
840 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
841 struct file *file, const struct iovec *iov,
842 unsigned long nr_segs)
844 struct socket *sock = file->private_data;
848 for (i = 0; i < nr_segs; i++)
849 size += iov[i].iov_len;
851 msg->msg_name = NULL;
852 msg->msg_namelen = 0;
853 msg->msg_control = NULL;
854 msg->msg_controllen = 0;
855 msg->msg_iov = (struct iovec *)iov;
856 msg->msg_iovlen = nr_segs;
857 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
858 if (sock->type == SOCK_SEQPACKET)
859 msg->msg_flags |= MSG_EOR;
861 return __sock_sendmsg(iocb, sock, msg, size);
864 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
865 unsigned long nr_segs, loff_t pos)
867 struct sock_iocb siocb, *x;
872 x = alloc_sock_iocb(iocb, &siocb);
876 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
880 * Atomic setting of ioctl hooks to avoid race
881 * with module unload.
884 static DEFINE_MUTEX(br_ioctl_mutex);
885 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
887 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
889 mutex_lock(&br_ioctl_mutex);
890 br_ioctl_hook = hook;
891 mutex_unlock(&br_ioctl_mutex);
894 EXPORT_SYMBOL(brioctl_set);
896 static DEFINE_MUTEX(vlan_ioctl_mutex);
897 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
899 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
901 mutex_lock(&vlan_ioctl_mutex);
902 vlan_ioctl_hook = hook;
903 mutex_unlock(&vlan_ioctl_mutex);
906 EXPORT_SYMBOL(vlan_ioctl_set);
908 static DEFINE_MUTEX(dlci_ioctl_mutex);
909 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
911 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
913 mutex_lock(&dlci_ioctl_mutex);
914 dlci_ioctl_hook = hook;
915 mutex_unlock(&dlci_ioctl_mutex);
918 EXPORT_SYMBOL(dlci_ioctl_set);
920 static long sock_do_ioctl(struct net *net, struct socket *sock,
921 unsigned int cmd, unsigned long arg)
924 void __user *argp = (void __user *)arg;
926 err = sock->ops->ioctl(sock, cmd, arg);
929 * If this ioctl is unknown try to hand it down
932 if (err == -ENOIOCTLCMD)
933 err = dev_ioctl(net, cmd, argp);
939 * With an ioctl, arg may well be a user mode pointer, but we don't know
940 * what to do with it - that's up to the protocol still.
943 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
947 void __user *argp = (void __user *)arg;
951 sock = file->private_data;
954 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
955 err = dev_ioctl(net, cmd, argp);
957 #ifdef CONFIG_WEXT_CORE
958 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
959 err = dev_ioctl(net, cmd, argp);
966 if (get_user(pid, (int __user *)argp))
968 err = f_setown(sock->file, pid, 1);
972 err = put_user(f_getown(sock->file),
981 request_module("bridge");
983 mutex_lock(&br_ioctl_mutex);
985 err = br_ioctl_hook(net, cmd, argp);
986 mutex_unlock(&br_ioctl_mutex);
991 if (!vlan_ioctl_hook)
992 request_module("8021q");
994 mutex_lock(&vlan_ioctl_mutex);
996 err = vlan_ioctl_hook(net, argp);
997 mutex_unlock(&vlan_ioctl_mutex);
1002 if (!dlci_ioctl_hook)
1003 request_module("dlci");
1005 mutex_lock(&dlci_ioctl_mutex);
1006 if (dlci_ioctl_hook)
1007 err = dlci_ioctl_hook(cmd, argp);
1008 mutex_unlock(&dlci_ioctl_mutex);
1011 err = sock_do_ioctl(net, sock, cmd, arg);
1017 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1020 struct socket *sock = NULL;
1022 err = security_socket_create(family, type, protocol, 1);
1026 sock = sock_alloc();
1033 err = security_socket_post_create(sock, family, type, protocol, 1);
1046 /* No kernel lock held - perfect */
1047 static unsigned int sock_poll(struct file *file, poll_table *wait)
1049 struct socket *sock;
1052 * We can't return errors to poll, so it's either yes or no.
1054 sock = file->private_data;
1055 return sock->ops->poll(file, sock, wait);
1058 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1060 struct socket *sock = file->private_data;
1062 return sock->ops->mmap(file, sock, vma);
1065 static int sock_close(struct inode *inode, struct file *filp)
1068 * It was possible the inode is NULL we were
1069 * closing an unfinished socket.
1073 printk(KERN_DEBUG "sock_close: NULL inode\n");
1076 sock_release(SOCKET_I(inode));
1081 * Update the socket async list
1083 * Fasync_list locking strategy.
1085 * 1. fasync_list is modified only under process context socket lock
1086 * i.e. under semaphore.
1087 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1088 * or under socket lock.
1089 * 3. fasync_list can be used from softirq context, so that
1090 * modification under socket lock have to be enhanced with
1091 * write_lock_bh(&sk->sk_callback_lock).
1095 static int sock_fasync(int fd, struct file *filp, int on)
1097 struct fasync_struct *fa, *fna = NULL, **prev;
1098 struct socket *sock;
1102 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1107 sock = filp->private_data;
1117 spin_lock(&filp->f_lock);
1119 filp->f_flags |= FASYNC;
1121 filp->f_flags &= ~FASYNC;
1122 spin_unlock(&filp->f_lock);
1124 prev = &(sock->fasync_list);
1126 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1127 if (fa->fa_file == filp)
1132 write_lock_bh(&sk->sk_callback_lock);
1134 write_unlock_bh(&sk->sk_callback_lock);
1139 fna->fa_file = filp;
1141 fna->magic = FASYNC_MAGIC;
1142 fna->fa_next = sock->fasync_list;
1143 write_lock_bh(&sk->sk_callback_lock);
1144 sock->fasync_list = fna;
1145 sock_set_flag(sk, SOCK_FASYNC);
1146 write_unlock_bh(&sk->sk_callback_lock);
1149 write_lock_bh(&sk->sk_callback_lock);
1150 *prev = fa->fa_next;
1151 if (!sock->fasync_list)
1152 sock_reset_flag(sk, SOCK_FASYNC);
1153 write_unlock_bh(&sk->sk_callback_lock);
1159 release_sock(sock->sk);
1163 /* This function may be called only under socket lock or callback_lock */
1165 int sock_wake_async(struct socket *sock, int how, int band)
1167 if (!sock || !sock->fasync_list)
1170 case SOCK_WAKE_WAITD:
1171 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1174 case SOCK_WAKE_SPACE:
1175 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1180 __kill_fasync(sock->fasync_list, SIGIO, band);
1183 __kill_fasync(sock->fasync_list, SIGURG, band);
1188 static int __sock_create(struct net *net, int family, int type, int protocol,
1189 struct socket **res, int kern)
1192 struct socket *sock;
1193 const struct net_proto_family *pf;
1196 * Check protocol is in range
1198 if (family < 0 || family >= NPROTO)
1199 return -EAFNOSUPPORT;
1200 if (type < 0 || type >= SOCK_MAX)
1205 This uglymoron is moved from INET layer to here to avoid
1206 deadlock in module load.
1208 if (family == PF_INET && type == SOCK_PACKET) {
1212 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1218 err = security_socket_create(family, type, protocol, kern);
1223 * Allocate the socket and allow the family to set things up. if
1224 * the protocol is 0, the family is instructed to select an appropriate
1227 sock = sock_alloc();
1229 if (net_ratelimit())
1230 printk(KERN_WARNING "socket: no more sockets\n");
1231 return -ENFILE; /* Not exactly a match, but its the
1232 closest posix thing */
1237 #ifdef CONFIG_MODULES
1238 /* Attempt to load a protocol module if the find failed.
1240 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1241 * requested real, full-featured networking support upon configuration.
1242 * Otherwise module support will break!
1244 if (net_families[family] == NULL)
1245 request_module("net-pf-%d", family);
1249 pf = rcu_dereference(net_families[family]);
1250 err = -EAFNOSUPPORT;
1255 * We will call the ->create function, that possibly is in a loadable
1256 * module, so we have to bump that loadable module refcnt first.
1258 if (!try_module_get(pf->owner))
1261 /* Now protected by module ref count */
1264 err = pf->create(net, sock, protocol, kern);
1266 goto out_module_put;
1269 * Now to bump the refcnt of the [loadable] module that owns this
1270 * socket at sock_release time we decrement its refcnt.
1272 if (!try_module_get(sock->ops->owner))
1273 goto out_module_busy;
1276 * Now that we're done with the ->create function, the [loadable]
1277 * module can have its refcnt decremented
1279 module_put(pf->owner);
1280 err = security_socket_post_create(sock, family, type, protocol, kern);
1282 goto out_sock_release;
1288 err = -EAFNOSUPPORT;
1291 module_put(pf->owner);
1298 goto out_sock_release;
1301 int sock_create(int family, int type, int protocol, struct socket **res)
1303 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1306 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1308 return __sock_create(&init_net, family, type, protocol, res, 1);
1311 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1314 struct socket *sock;
1317 /* Check the SOCK_* constants for consistency. */
1318 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1319 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1320 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1321 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1323 flags = type & ~SOCK_TYPE_MASK;
1324 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1326 type &= SOCK_TYPE_MASK;
1328 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1329 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1331 retval = sock_create(family, type, protocol, &sock);
1335 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1340 /* It may be already another descriptor 8) Not kernel problem. */
1349 * Create a pair of connected sockets.
1352 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1353 int __user *, usockvec)
1355 struct socket *sock1, *sock2;
1357 struct file *newfile1, *newfile2;
1360 flags = type & ~SOCK_TYPE_MASK;
1361 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1363 type &= SOCK_TYPE_MASK;
1365 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1366 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1369 * Obtain the first socket and check if the underlying protocol
1370 * supports the socketpair call.
1373 err = sock_create(family, type, protocol, &sock1);
1377 err = sock_create(family, type, protocol, &sock2);
1381 err = sock1->ops->socketpair(sock1, sock2);
1383 goto out_release_both;
1385 fd1 = sock_alloc_file(sock1, &newfile1, flags);
1386 if (unlikely(fd1 < 0)) {
1388 goto out_release_both;
1391 fd2 = sock_alloc_file(sock2, &newfile2, flags);
1392 if (unlikely(fd2 < 0)) {
1396 sock_release(sock2);
1400 audit_fd_pair(fd1, fd2);
1401 fd_install(fd1, newfile1);
1402 fd_install(fd2, newfile2);
1403 /* fd1 and fd2 may be already another descriptors.
1404 * Not kernel problem.
1407 err = put_user(fd1, &usockvec[0]);
1409 err = put_user(fd2, &usockvec[1]);
1418 sock_release(sock2);
1420 sock_release(sock1);
1426 * Bind a name to a socket. Nothing much to do here since it's
1427 * the protocol's responsibility to handle the local address.
1429 * We move the socket address to kernel space before we call
1430 * the protocol layer (having also checked the address is ok).
1433 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1435 struct socket *sock;
1436 struct sockaddr_storage address;
1437 int err, fput_needed;
1439 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1441 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1443 err = security_socket_bind(sock,
1444 (struct sockaddr *)&address,
1447 err = sock->ops->bind(sock,
1451 fput_light(sock->file, fput_needed);
1457 * Perform a listen. Basically, we allow the protocol to do anything
1458 * necessary for a listen, and if that works, we mark the socket as
1459 * ready for listening.
1462 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1464 struct socket *sock;
1465 int err, fput_needed;
1468 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1470 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1471 if ((unsigned)backlog > somaxconn)
1472 backlog = somaxconn;
1474 err = security_socket_listen(sock, backlog);
1476 err = sock->ops->listen(sock, backlog);
1478 fput_light(sock->file, fput_needed);
1484 * For accept, we attempt to create a new socket, set up the link
1485 * with the client, wake up the client, then return the new
1486 * connected fd. We collect the address of the connector in kernel
1487 * space and move it to user at the very end. This is unclean because
1488 * we open the socket then return an error.
1490 * 1003.1g adds the ability to recvmsg() to query connection pending
1491 * status to recvmsg. We need to add that support in a way thats
1492 * clean when we restucture accept also.
1495 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1496 int __user *, upeer_addrlen, int, flags)
1498 struct socket *sock, *newsock;
1499 struct file *newfile;
1500 int err, len, newfd, fput_needed;
1501 struct sockaddr_storage address;
1503 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1506 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1507 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1509 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1514 if (!(newsock = sock_alloc()))
1517 newsock->type = sock->type;
1518 newsock->ops = sock->ops;
1521 * We don't need try_module_get here, as the listening socket (sock)
1522 * has the protocol module (sock->ops->owner) held.
1524 __module_get(newsock->ops->owner);
1526 newfd = sock_alloc_file(newsock, &newfile, flags);
1527 if (unlikely(newfd < 0)) {
1529 sock_release(newsock);
1533 err = security_socket_accept(sock, newsock);
1537 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1541 if (upeer_sockaddr) {
1542 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1544 err = -ECONNABORTED;
1547 err = move_addr_to_user((struct sockaddr *)&address,
1548 len, upeer_sockaddr, upeer_addrlen);
1553 /* File flags are not inherited via accept() unlike another OSes. */
1555 fd_install(newfd, newfile);
1559 fput_light(sock->file, fput_needed);
1564 put_unused_fd(newfd);
1568 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1569 int __user *, upeer_addrlen)
1571 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1575 * Attempt to connect to a socket with the server address. The address
1576 * is in user space so we verify it is OK and move it to kernel space.
1578 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1581 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1582 * other SEQPACKET protocols that take time to connect() as it doesn't
1583 * include the -EINPROGRESS status for such sockets.
1586 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1589 struct socket *sock;
1590 struct sockaddr_storage address;
1591 int err, fput_needed;
1593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1596 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1601 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1605 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1606 sock->file->f_flags);
1608 fput_light(sock->file, fput_needed);
1614 * Get the local address ('name') of a socket object. Move the obtained
1615 * name to user space.
1618 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1619 int __user *, usockaddr_len)
1621 struct socket *sock;
1622 struct sockaddr_storage address;
1623 int len, err, fput_needed;
1625 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1629 err = security_socket_getsockname(sock);
1633 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1636 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1639 fput_light(sock->file, fput_needed);
1645 * Get the remote address ('name') of a socket object. Move the obtained
1646 * name to user space.
1649 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1650 int __user *, usockaddr_len)
1652 struct socket *sock;
1653 struct sockaddr_storage address;
1654 int len, err, fput_needed;
1656 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1658 err = security_socket_getpeername(sock);
1660 fput_light(sock->file, fput_needed);
1665 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1668 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1670 fput_light(sock->file, fput_needed);
1676 * Send a datagram to a given address. We move the address into kernel
1677 * space and check the user space data area is readable before invoking
1681 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1682 unsigned, flags, struct sockaddr __user *, addr,
1685 struct socket *sock;
1686 struct sockaddr_storage address;
1692 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1696 iov.iov_base = buff;
1698 msg.msg_name = NULL;
1701 msg.msg_control = NULL;
1702 msg.msg_controllen = 0;
1703 msg.msg_namelen = 0;
1705 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1708 msg.msg_name = (struct sockaddr *)&address;
1709 msg.msg_namelen = addr_len;
1711 if (sock->file->f_flags & O_NONBLOCK)
1712 flags |= MSG_DONTWAIT;
1713 msg.msg_flags = flags;
1714 err = sock_sendmsg(sock, &msg, len);
1717 fput_light(sock->file, fput_needed);
1723 * Send a datagram down a socket.
1726 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1729 return sys_sendto(fd, buff, len, flags, NULL, 0);
1733 * Receive a frame from the socket and optionally record the address of the
1734 * sender. We verify the buffers are writable and if needed move the
1735 * sender address from kernel to user space.
1738 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1739 unsigned, flags, struct sockaddr __user *, addr,
1740 int __user *, addr_len)
1742 struct socket *sock;
1745 struct sockaddr_storage address;
1749 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1753 msg.msg_control = NULL;
1754 msg.msg_controllen = 0;
1758 iov.iov_base = ubuf;
1759 msg.msg_name = (struct sockaddr *)&address;
1760 msg.msg_namelen = sizeof(address);
1761 if (sock->file->f_flags & O_NONBLOCK)
1762 flags |= MSG_DONTWAIT;
1763 err = sock_recvmsg(sock, &msg, size, flags);
1765 if (err >= 0 && addr != NULL) {
1766 err2 = move_addr_to_user((struct sockaddr *)&address,
1767 msg.msg_namelen, addr, addr_len);
1772 fput_light(sock->file, fput_needed);
1778 * Receive a datagram from a socket.
1781 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1784 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1788 * Set a socket option. Because we don't know the option lengths we have
1789 * to pass the user mode parameter for the protocols to sort out.
1792 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1793 char __user *, optval, int, optlen)
1795 int err, fput_needed;
1796 struct socket *sock;
1801 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1803 err = security_socket_setsockopt(sock, level, optname);
1807 if (level == SOL_SOCKET)
1809 sock_setsockopt(sock, level, optname, optval,
1813 sock->ops->setsockopt(sock, level, optname, optval,
1816 fput_light(sock->file, fput_needed);
1822 * Get a socket option. Because we don't know the option lengths we have
1823 * to pass a user mode parameter for the protocols to sort out.
1826 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1827 char __user *, optval, int __user *, optlen)
1829 int err, fput_needed;
1830 struct socket *sock;
1832 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1834 err = security_socket_getsockopt(sock, level, optname);
1838 if (level == SOL_SOCKET)
1840 sock_getsockopt(sock, level, optname, optval,
1844 sock->ops->getsockopt(sock, level, optname, optval,
1847 fput_light(sock->file, fput_needed);
1853 * Shutdown a socket.
1856 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1858 int err, fput_needed;
1859 struct socket *sock;
1861 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1863 err = security_socket_shutdown(sock, how);
1865 err = sock->ops->shutdown(sock, how);
1866 fput_light(sock->file, fput_needed);
1871 /* A couple of helpful macros for getting the address of the 32/64 bit
1872 * fields which are the same type (int / unsigned) on our platforms.
1874 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1875 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1876 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1879 * BSD sendmsg interface
1882 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1884 struct compat_msghdr __user *msg_compat =
1885 (struct compat_msghdr __user *)msg;
1886 struct socket *sock;
1887 struct sockaddr_storage address;
1888 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1889 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1890 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1891 /* 20 is size of ipv6_pktinfo */
1892 unsigned char *ctl_buf = ctl;
1893 struct msghdr msg_sys;
1894 int err, ctl_len, iov_size, total_len;
1898 if (MSG_CMSG_COMPAT & flags) {
1899 if (get_compat_msghdr(&msg_sys, msg_compat))
1902 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1905 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1909 /* do not move before msg_sys is valid */
1911 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1914 /* Check whether to allocate the iovec area */
1916 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1917 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1918 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1923 /* This will also move the address data into kernel space */
1924 if (MSG_CMSG_COMPAT & flags) {
1925 err = verify_compat_iovec(&msg_sys, iov,
1926 (struct sockaddr *)&address,
1929 err = verify_iovec(&msg_sys, iov,
1930 (struct sockaddr *)&address,
1938 if (msg_sys.msg_controllen > INT_MAX)
1940 ctl_len = msg_sys.msg_controllen;
1941 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1943 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1947 ctl_buf = msg_sys.msg_control;
1948 ctl_len = msg_sys.msg_controllen;
1949 } else if (ctl_len) {
1950 if (ctl_len > sizeof(ctl)) {
1951 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1952 if (ctl_buf == NULL)
1957 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1958 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1959 * checking falls down on this.
1961 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1964 msg_sys.msg_control = ctl_buf;
1966 msg_sys.msg_flags = flags;
1968 if (sock->file->f_flags & O_NONBLOCK)
1969 msg_sys.msg_flags |= MSG_DONTWAIT;
1970 err = sock_sendmsg(sock, &msg_sys, total_len);
1974 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1976 if (iov != iovstack)
1977 sock_kfree_s(sock->sk, iov, iov_size);
1979 fput_light(sock->file, fput_needed);
1984 static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1985 struct msghdr *msg_sys, unsigned flags, int nosec)
1987 struct compat_msghdr __user *msg_compat =
1988 (struct compat_msghdr __user *)msg;
1989 struct iovec iovstack[UIO_FASTIOV];
1990 struct iovec *iov = iovstack;
1991 unsigned long cmsg_ptr;
1992 int err, iov_size, total_len, len;
1994 /* kernel mode address */
1995 struct sockaddr_storage addr;
1997 /* user mode address pointers */
1998 struct sockaddr __user *uaddr;
1999 int __user *uaddr_len;
2001 if (MSG_CMSG_COMPAT & flags) {
2002 if (get_compat_msghdr(msg_sys, msg_compat))
2005 else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2009 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2012 /* Check whether to allocate the iovec area */
2014 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2015 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2016 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2022 * Save the user-mode address (verify_iovec will change the
2023 * kernel msghdr to use the kernel address space)
2026 uaddr = (__force void __user *)msg_sys->msg_name;
2027 uaddr_len = COMPAT_NAMELEN(msg);
2028 if (MSG_CMSG_COMPAT & flags) {
2029 err = verify_compat_iovec(msg_sys, iov,
2030 (struct sockaddr *)&addr,
2033 err = verify_iovec(msg_sys, iov,
2034 (struct sockaddr *)&addr,
2040 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2041 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2043 if (sock->file->f_flags & O_NONBLOCK)
2044 flags |= MSG_DONTWAIT;
2045 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2051 if (uaddr != NULL) {
2052 err = move_addr_to_user((struct sockaddr *)&addr,
2053 msg_sys->msg_namelen, uaddr,
2058 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2062 if (MSG_CMSG_COMPAT & flags)
2063 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2064 &msg_compat->msg_controllen);
2066 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2067 &msg->msg_controllen);
2073 if (iov != iovstack)
2074 sock_kfree_s(sock->sk, iov, iov_size);
2080 * BSD recvmsg interface
2083 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2084 unsigned int, flags)
2086 int fput_needed, err;
2087 struct msghdr msg_sys;
2088 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2093 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2095 fput_light(sock->file, fput_needed);
2101 * Linux recvmmsg interface
2104 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2105 unsigned int flags, struct timespec *timeout)
2107 int fput_needed, err, datagrams;
2108 struct socket *sock;
2109 struct mmsghdr __user *entry;
2110 struct compat_mmsghdr __user *compat_entry;
2111 struct msghdr msg_sys;
2112 struct timespec end_time;
2115 poll_select_set_timeout(&end_time, timeout->tv_sec,
2121 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2125 err = sock_error(sock->sk);
2130 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2132 while (datagrams < vlen) {
2134 * No need to ask LSM for more than the first datagram.
2136 if (MSG_CMSG_COMPAT & flags) {
2137 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2138 &msg_sys, flags, datagrams);
2141 err = __put_user(err, &compat_entry->msg_len);
2144 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2145 &msg_sys, flags, datagrams);
2148 err = put_user(err, &entry->msg_len);
2157 ktime_get_ts(timeout);
2158 *timeout = timespec_sub(end_time, *timeout);
2159 if (timeout->tv_sec < 0) {
2160 timeout->tv_sec = timeout->tv_nsec = 0;
2164 /* Timeout, return less than vlen datagrams */
2165 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2169 /* Out of band data, return right away */
2170 if (msg_sys.msg_flags & MSG_OOB)
2175 fput_light(sock->file, fput_needed);
2180 if (datagrams != 0) {
2182 * We may return less entries than requested (vlen) if the
2183 * sock is non block and there aren't enough datagrams...
2185 if (err != -EAGAIN) {
2187 * ... or if recvmsg returns an error after we
2188 * received some datagrams, where we record the
2189 * error to return on the next call or if the
2190 * app asks about it using getsockopt(SO_ERROR).
2192 sock->sk->sk_err = -err;
2201 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2202 unsigned int, vlen, unsigned int, flags,
2203 struct timespec __user *, timeout)
2206 struct timespec timeout_sys;
2209 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2211 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2214 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2216 if (datagrams > 0 &&
2217 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2218 datagrams = -EFAULT;
2223 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2224 /* Argument list sizes for sys_socketcall */
2225 #define AL(x) ((x) * sizeof(unsigned long))
2226 static const unsigned char nargs[20] = {
2227 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2228 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2229 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2236 * System call vectors.
2238 * Argument checking cleaned up. Saved 20% in size.
2239 * This function doesn't need to set the kernel lock because
2240 * it is set by the callees.
2243 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2246 unsigned long a0, a1;
2250 if (call < 1 || call > SYS_RECVMMSG)
2254 if (len > sizeof(a))
2257 /* copy_from_user should be SMP safe. */
2258 if (copy_from_user(a, args, len))
2261 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2268 err = sys_socket(a0, a1, a[2]);
2271 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2274 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2277 err = sys_listen(a0, a1);
2280 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2281 (int __user *)a[2], 0);
2283 case SYS_GETSOCKNAME:
2285 sys_getsockname(a0, (struct sockaddr __user *)a1,
2286 (int __user *)a[2]);
2288 case SYS_GETPEERNAME:
2290 sys_getpeername(a0, (struct sockaddr __user *)a1,
2291 (int __user *)a[2]);
2293 case SYS_SOCKETPAIR:
2294 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2297 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2300 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2301 (struct sockaddr __user *)a[4], a[5]);
2304 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2307 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2308 (struct sockaddr __user *)a[4],
2309 (int __user *)a[5]);
2312 err = sys_shutdown(a0, a1);
2314 case SYS_SETSOCKOPT:
2315 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2317 case SYS_GETSOCKOPT:
2319 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2320 (int __user *)a[4]);
2323 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2326 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2329 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2330 (struct timespec __user *)a[4]);
2333 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2334 (int __user *)a[2], a[3]);
2343 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2346 * sock_register - add a socket protocol handler
2347 * @ops: description of protocol
2349 * This function is called by a protocol handler that wants to
2350 * advertise its address family, and have it linked into the
2351 * socket interface. The value ops->family coresponds to the
2352 * socket system call protocol family.
2354 int sock_register(const struct net_proto_family *ops)
2358 if (ops->family >= NPROTO) {
2359 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2364 spin_lock(&net_family_lock);
2365 if (net_families[ops->family])
2368 net_families[ops->family] = ops;
2371 spin_unlock(&net_family_lock);
2373 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2378 * sock_unregister - remove a protocol handler
2379 * @family: protocol family to remove
2381 * This function is called by a protocol handler that wants to
2382 * remove its address family, and have it unlinked from the
2383 * new socket creation.
2385 * If protocol handler is a module, then it can use module reference
2386 * counts to protect against new references. If protocol handler is not
2387 * a module then it needs to provide its own protection in
2388 * the ops->create routine.
2390 void sock_unregister(int family)
2392 BUG_ON(family < 0 || family >= NPROTO);
2394 spin_lock(&net_family_lock);
2395 net_families[family] = NULL;
2396 spin_unlock(&net_family_lock);
2400 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2403 static int __init sock_init(void)
2406 * Initialize sock SLAB cache.
2412 * Initialize skbuff SLAB cache
2417 * Initialize the protocols module.
2421 register_filesystem(&sock_fs_type);
2422 sock_mnt = kern_mount(&sock_fs_type);
2424 /* The real protocol initialization is performed in later initcalls.
2427 #ifdef CONFIG_NETFILTER
2434 core_initcall(sock_init); /* early initcall */
2436 #ifdef CONFIG_PROC_FS
2437 void socket_seq_show(struct seq_file *seq)
2442 for_each_possible_cpu(cpu)
2443 counter += per_cpu(sockets_in_use, cpu);
2445 /* It can be negative, by the way. 8) */
2449 seq_printf(seq, "sockets: used %d\n", counter);
2451 #endif /* CONFIG_PROC_FS */
2453 #ifdef CONFIG_COMPAT
2454 static int do_siocgstamp(struct net *net, struct socket *sock,
2455 unsigned int cmd, struct compat_timeval __user *up)
2457 mm_segment_t old_fs = get_fs();
2462 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2465 err = put_user(ktv.tv_sec, &up->tv_sec);
2466 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2471 static int do_siocgstampns(struct net *net, struct socket *sock,
2472 unsigned int cmd, struct compat_timespec __user *up)
2474 mm_segment_t old_fs = get_fs();
2475 struct timespec kts;
2479 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2482 err = put_user(kts.tv_sec, &up->tv_sec);
2483 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2488 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2490 struct ifreq __user *uifr;
2493 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2494 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2497 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2501 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2507 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2509 struct compat_ifconf ifc32;
2511 struct ifconf __user *uifc;
2512 struct compat_ifreq __user *ifr32;
2513 struct ifreq __user *ifr;
2517 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2520 if (ifc32.ifcbuf == 0) {
2524 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2526 size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
2527 sizeof (struct ifreq);
2528 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2530 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2531 ifr32 = compat_ptr(ifc32.ifcbuf);
2532 for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2533 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2539 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2542 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2546 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2550 ifr32 = compat_ptr(ifc32.ifcbuf);
2552 i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2553 i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2554 if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
2560 if (ifc32.ifcbuf == 0) {
2561 /* Translate from 64-bit structure multiple to
2565 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2570 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2576 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2578 struct ifreq __user *ifr;
2582 ifr = compat_alloc_user_space(sizeof(*ifr));
2584 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2587 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2590 datap = compat_ptr(data);
2591 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2594 return dev_ioctl(net, SIOCETHTOOL, ifr);
2597 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2600 compat_uptr_t uptr32;
2601 struct ifreq __user *uifr;
2603 uifr = compat_alloc_user_space(sizeof (*uifr));
2604 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2607 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2610 uptr = compat_ptr(uptr32);
2612 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2615 return dev_ioctl(net, SIOCWANDEV, uifr);
2618 static int bond_ioctl(struct net *net, unsigned int cmd,
2619 struct compat_ifreq __user *ifr32)
2622 struct ifreq __user *uifr;
2623 mm_segment_t old_fs;
2629 case SIOCBONDENSLAVE:
2630 case SIOCBONDRELEASE:
2631 case SIOCBONDSETHWADDR:
2632 case SIOCBONDCHANGEACTIVE:
2633 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2638 err = dev_ioctl(net, cmd, &kifr);
2642 case SIOCBONDSLAVEINFOQUERY:
2643 case SIOCBONDINFOQUERY:
2644 uifr = compat_alloc_user_space(sizeof(*uifr));
2645 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2648 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2651 datap = compat_ptr(data);
2652 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2655 return dev_ioctl(net, cmd, uifr);
2661 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2662 struct compat_ifreq __user *u_ifreq32)
2664 struct ifreq __user *u_ifreq64;
2665 char tmp_buf[IFNAMSIZ];
2666 void __user *data64;
2669 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2672 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2674 data64 = compat_ptr(data32);
2676 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2678 /* Don't check these user accesses, just let that get trapped
2679 * in the ioctl handler instead.
2681 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2684 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2687 return dev_ioctl(net, cmd, u_ifreq64);
2690 static int dev_ifsioc(struct net *net, struct socket *sock,
2691 unsigned int cmd, struct compat_ifreq __user *uifr32)
2693 struct ifreq __user *uifr;
2696 uifr = compat_alloc_user_space(sizeof(*uifr));
2697 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2700 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2711 case SIOCGIFBRDADDR:
2712 case SIOCGIFDSTADDR:
2713 case SIOCGIFNETMASK:
2718 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2726 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2727 struct compat_ifreq __user *uifr32)
2730 struct compat_ifmap __user *uifmap32;
2731 mm_segment_t old_fs;
2734 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2735 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2736 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2737 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2738 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2739 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2740 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2741 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2747 err = dev_ioctl(net, cmd, (void __user *)&ifr);
2750 if (cmd == SIOCGIFMAP && !err) {
2751 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2752 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2753 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2754 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2755 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2756 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2757 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2764 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2767 compat_uptr_t uptr32;
2768 struct ifreq __user *uifr;
2770 uifr = compat_alloc_user_space(sizeof (*uifr));
2771 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2774 if (get_user(uptr32, &uifr32->ifr_data))
2777 uptr = compat_ptr(uptr32);
2779 if (put_user(uptr, &uifr->ifr_data))
2782 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2787 struct sockaddr rt_dst; /* target address */
2788 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2789 struct sockaddr rt_genmask; /* target network mask (IP) */
2790 unsigned short rt_flags;
2793 unsigned char rt_tos;
2794 unsigned char rt_class;
2796 short rt_metric; /* +1 for binary compatibility! */
2797 /* char * */ u32 rt_dev; /* forcing the device at add */
2798 u32 rt_mtu; /* per route MTU/Window */
2799 u32 rt_window; /* Window clamping */
2800 unsigned short rt_irtt; /* Initial RTT */
2803 struct in6_rtmsg32 {
2804 struct in6_addr rtmsg_dst;
2805 struct in6_addr rtmsg_src;
2806 struct in6_addr rtmsg_gateway;
2816 static int routing_ioctl(struct net *net, struct socket *sock,
2817 unsigned int cmd, void __user *argp)
2821 struct in6_rtmsg r6;
2825 mm_segment_t old_fs = get_fs();
2827 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2828 struct in6_rtmsg32 __user *ur6 = argp;
2829 ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2830 3 * sizeof(struct in6_addr));
2831 ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2832 ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2833 ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2834 ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2835 ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2836 ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2837 ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2841 struct rtentry32 __user *ur4 = argp;
2842 ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2843 3 * sizeof(struct sockaddr));
2844 ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2845 ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2846 ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2847 ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2848 ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2849 ret |= __get_user (rtdev, &(ur4->rt_dev));
2851 ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2852 r4.rt_dev = devname; devname[15] = 0;
2865 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2872 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2873 * for some operations; this forces use of the newer bridge-utils that
2874 * use compatiable ioctls
2876 static int old_bridge_ioctl(compat_ulong_t __user *argp)
2880 if (get_user(tmp, argp))
2882 if (tmp == BRCTL_GET_VERSION)
2883 return BRCTL_VERSION + 1;
2887 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2888 unsigned int cmd, unsigned long arg)
2890 void __user *argp = compat_ptr(arg);
2891 struct sock *sk = sock->sk;
2892 struct net *net = sock_net(sk);
2894 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2895 return siocdevprivate_ioctl(net, cmd, argp);
2900 return old_bridge_ioctl(argp);
2902 return dev_ifname32(net, argp);
2904 return dev_ifconf(net, argp);
2906 return ethtool_ioctl(net, argp);
2908 return compat_siocwandev(net, argp);
2911 return compat_sioc_ifmap(net, cmd, argp);
2912 case SIOCBONDENSLAVE:
2913 case SIOCBONDRELEASE:
2914 case SIOCBONDSETHWADDR:
2915 case SIOCBONDSLAVEINFOQUERY:
2916 case SIOCBONDINFOQUERY:
2917 case SIOCBONDCHANGEACTIVE:
2918 return bond_ioctl(net, cmd, argp);
2921 return routing_ioctl(net, sock, cmd, argp);
2923 return do_siocgstamp(net, sock, cmd, argp);
2925 return do_siocgstampns(net, sock, cmd, argp);
2927 return compat_siocshwtstamp(net, argp);
2939 return sock_ioctl(file, cmd, arg);
2956 case SIOCSIFHWBROADCAST:
2958 case SIOCGIFBRDADDR:
2959 case SIOCSIFBRDADDR:
2960 case SIOCGIFDSTADDR:
2961 case SIOCSIFDSTADDR:
2962 case SIOCGIFNETMASK:
2963 case SIOCSIFNETMASK:
2974 return dev_ifsioc(net, sock, cmd, argp);
2980 return sock_do_ioctl(net, sock, cmd, arg);
2983 /* Prevent warning from compat_sys_ioctl, these always
2984 * result in -EINVAL in the native case anyway. */
2997 return -ENOIOCTLCMD;
3000 static long compat_sock_ioctl(struct file *file, unsigned cmd,
3003 struct socket *sock = file->private_data;
3004 int ret = -ENOIOCTLCMD;
3011 if (sock->ops->compat_ioctl)
3012 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3014 if (ret == -ENOIOCTLCMD &&
3015 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3016 ret = compat_wext_handle_ioctl(net, cmd, arg);
3018 if (ret == -ENOIOCTLCMD)
3019 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3025 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3027 return sock->ops->bind(sock, addr, addrlen);
3030 int kernel_listen(struct socket *sock, int backlog)
3032 return sock->ops->listen(sock, backlog);
3035 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3037 struct sock *sk = sock->sk;
3040 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3045 err = sock->ops->accept(sock, *newsock, flags);
3047 sock_release(*newsock);
3052 (*newsock)->ops = sock->ops;
3053 __module_get((*newsock)->ops->owner);
3059 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3062 return sock->ops->connect(sock, addr, addrlen, flags);
3065 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3068 return sock->ops->getname(sock, addr, addrlen, 0);
3071 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3074 return sock->ops->getname(sock, addr, addrlen, 1);
3077 int kernel_getsockopt(struct socket *sock, int level, int optname,
3078 char *optval, int *optlen)
3080 mm_segment_t oldfs = get_fs();
3084 if (level == SOL_SOCKET)
3085 err = sock_getsockopt(sock, level, optname, optval, optlen);
3087 err = sock->ops->getsockopt(sock, level, optname, optval,
3093 int kernel_setsockopt(struct socket *sock, int level, int optname,
3094 char *optval, unsigned int optlen)
3096 mm_segment_t oldfs = get_fs();
3100 if (level == SOL_SOCKET)
3101 err = sock_setsockopt(sock, level, optname, optval, optlen);
3103 err = sock->ops->setsockopt(sock, level, optname, optval,
3109 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3110 size_t size, int flags)
3112 if (sock->ops->sendpage)
3113 return sock->ops->sendpage(sock, page, offset, size, flags);
3115 return sock_no_sendpage(sock, page, offset, size, flags);
3118 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3120 mm_segment_t oldfs = get_fs();
3124 err = sock->ops->ioctl(sock, cmd, arg);
3130 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3132 return sock->ops->shutdown(sock, how);
3135 EXPORT_SYMBOL(sock_create);
3136 EXPORT_SYMBOL(sock_create_kern);
3137 EXPORT_SYMBOL(sock_create_lite);
3138 EXPORT_SYMBOL(sock_map_fd);
3139 EXPORT_SYMBOL(sock_recvmsg);
3140 EXPORT_SYMBOL(sock_register);
3141 EXPORT_SYMBOL(sock_release);
3142 EXPORT_SYMBOL(sock_sendmsg);
3143 EXPORT_SYMBOL(sock_unregister);
3144 EXPORT_SYMBOL(sock_wake_async);
3145 EXPORT_SYMBOL(sockfd_lookup);
3146 EXPORT_SYMBOL(kernel_sendmsg);
3147 EXPORT_SYMBOL(kernel_recvmsg);
3148 EXPORT_SYMBOL(kernel_bind);
3149 EXPORT_SYMBOL(kernel_listen);
3150 EXPORT_SYMBOL(kernel_accept);
3151 EXPORT_SYMBOL(kernel_connect);
3152 EXPORT_SYMBOL(kernel_getsockname);
3153 EXPORT_SYMBOL(kernel_getpeername);
3154 EXPORT_SYMBOL(kernel_getsockopt);
3155 EXPORT_SYMBOL(kernel_setsockopt);
3156 EXPORT_SYMBOL(kernel_sendpage);
3157 EXPORT_SYMBOL(kernel_sock_ioctl);
3158 EXPORT_SYMBOL(kernel_sock_shutdown);