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>
90 #include <linux/slab.h>
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
108 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
109 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
113 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
115 static int sock_close(struct inode *inode, struct file *file);
116 static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
118 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
120 static long compat_sock_ioctl(struct file *file,
121 unsigned int cmd, unsigned long arg);
123 static int sock_fasync(int fd, struct file *filp, int on);
124 static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
126 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
127 struct pipe_inode_info *pipe, size_t len,
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
135 static const struct file_operations socket_file_ops = {
136 .owner = THIS_MODULE,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
141 .unlocked_ioctl = sock_ioctl,
143 .compat_ioctl = compat_sock_ioctl,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
151 .splice_read = sock_splice_read,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock);
159 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
162 * Statistics counters of the socket lists
165 static DEFINE_PER_CPU(int, sockets_in_use);
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
190 if (copy_from_user(kaddr, uaddr, ulen))
192 return audit_sockaddr(ulen, kaddr);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 static int move_addr_to_user(struct sockaddr *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
218 err = get_user(len, ulen);
223 if (len < 0 || len > sizeof(struct sockaddr_storage))
226 if (audit_sockaddr(klen, kaddr))
228 if (copy_to_user(uaddr, kaddr, len))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen, ulen);
238 static struct kmem_cache *sock_inode_cachep __read_mostly;
240 static struct inode *sock_alloc_inode(struct super_block *sb)
242 struct socket_alloc *ei;
244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
247 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
248 if (!ei->socket.wq) {
249 kmem_cache_free(sock_inode_cachep, ei);
252 init_waitqueue_head(&ei->socket.wq->wait);
253 ei->socket.wq->fasync_list = NULL;
255 ei->socket.state = SS_UNCONNECTED;
256 ei->socket.flags = 0;
257 ei->socket.ops = NULL;
258 ei->socket.sk = NULL;
259 ei->socket.file = NULL;
261 return &ei->vfs_inode;
265 static void wq_free_rcu(struct rcu_head *head)
267 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
272 static void sock_destroy_inode(struct inode *inode)
274 struct socket_alloc *ei;
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
278 kmem_cache_free(sock_inode_cachep, ei);
281 static void init_once(void *foo)
283 struct socket_alloc *ei = (struct socket_alloc *)foo;
285 inode_init_once(&ei->vfs_inode);
288 static int init_inodecache(void)
290 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
291 sizeof(struct socket_alloc),
293 (SLAB_HWCACHE_ALIGN |
294 SLAB_RECLAIM_ACCOUNT |
297 if (sock_inode_cachep == NULL)
302 static const struct super_operations sockfs_ops = {
303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
308 static int sockfs_get_sb(struct file_system_type *fs_type,
309 int flags, const char *dev_name, void *data,
310 struct vfsmount *mnt)
312 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
316 static struct vfsmount *sock_mnt __read_mostly;
318 static struct file_system_type sock_fs_type = {
320 .get_sb = sockfs_get_sb,
321 .kill_sb = kill_anon_super,
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 const struct dentry_operations sockfs_dentry_operations = {
334 .d_dname = sockfs_dname,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
356 struct qstr name = { .name = "" };
361 fd = get_unused_fd_flags(flags);
362 if (unlikely(fd < 0))
365 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
366 if (unlikely(!path.dentry)) {
370 path.mnt = mntget(sock_mnt);
372 path.dentry->d_op = &sockfs_dentry_operations;
373 d_instantiate(path.dentry, SOCK_INODE(sock));
374 SOCK_INODE(sock)->i_fop = &socket_file_ops;
376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 if (unlikely(!file)) {
379 /* drop dentry, keep inode */
380 atomic_inc(&path.dentry->d_inode->i_count);
387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
389 file->private_data = sock;
395 int sock_map_fd(struct socket *sock, int flags)
397 struct file *newfile;
398 int fd = sock_alloc_file(sock, &newfile, flags);
401 fd_install(fd, newfile);
405 EXPORT_SYMBOL(sock_map_fd);
407 static struct socket *sock_from_file(struct file *file, int *err)
409 if (file->f_op == &socket_file_ops)
410 return file->private_data; /* set in sock_map_fd */
417 * sockfd_lookup - Go from a file number to its socket slot
419 * @err: pointer to an error code return
421 * The file handle passed in is locked and the socket it is bound
422 * too is returned. If an error occurs the err pointer is overwritten
423 * with a negative errno code and NULL is returned. The function checks
424 * for both invalid handles and passing a handle which is not a socket.
426 * On a success the socket object pointer is returned.
429 struct socket *sockfd_lookup(int fd, int *err)
440 sock = sock_from_file(file, err);
445 EXPORT_SYMBOL(sockfd_lookup);
447 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
453 file = fget_light(fd, fput_needed);
455 sock = sock_from_file(file, err);
458 fput_light(file, *fput_needed);
464 * sock_alloc - allocate a socket
466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
471 static struct socket *sock_alloc(void)
476 inode = new_inode(sock_mnt->mnt_sb);
480 sock = SOCKET_I(inode);
482 kmemcheck_annotate_bitfield(sock, type);
483 inode->i_mode = S_IFSOCK | S_IRWXUGO;
484 inode->i_uid = current_fsuid();
485 inode->i_gid = current_fsgid();
487 percpu_add(sockets_in_use, 1);
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
497 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
502 const struct file_operations bad_sock_fops = {
503 .owner = THIS_MODULE,
504 .open = sock_no_open,
505 .llseek = noop_llseek,
509 * sock_release - close a socket
510 * @sock: socket to close
512 * The socket is released from the protocol stack if it has a release
513 * callback, and the inode is then released if the socket is bound to
514 * an inode not a file.
517 void sock_release(struct socket *sock)
520 struct module *owner = sock->ops->owner;
522 sock->ops->release(sock);
527 if (sock->wq->fasync_list)
528 printk(KERN_ERR "sock_release: fasync list not empty!\n");
530 percpu_sub(sockets_in_use, 1);
532 iput(SOCK_INODE(sock));
537 EXPORT_SYMBOL(sock_release);
539 int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
542 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
543 *tx_flags |= SKBTX_HW_TSTAMP;
544 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
545 *tx_flags |= SKBTX_SW_TSTAMP;
548 EXPORT_SYMBOL(sock_tx_timestamp);
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);
556 sock_update_classid(sock->sk);
563 err = security_socket_sendmsg(sock, msg, size);
567 return sock->ops->sendmsg(iocb, sock, msg, size);
570 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
573 struct sock_iocb siocb;
576 init_sync_kiocb(&iocb, NULL);
577 iocb.private = &siocb;
578 ret = __sock_sendmsg(&iocb, sock, msg, size);
579 if (-EIOCBQUEUED == ret)
580 ret = wait_on_sync_kiocb(&iocb);
583 EXPORT_SYMBOL(sock_sendmsg);
585 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
586 struct kvec *vec, size_t num, size_t size)
588 mm_segment_t oldfs = get_fs();
593 * the following is safe, since for compiler definitions of kvec and
594 * iovec are identical, yielding the same in-core layout and alignment
596 msg->msg_iov = (struct iovec *)vec;
597 msg->msg_iovlen = num;
598 result = sock_sendmsg(sock, msg, size);
602 EXPORT_SYMBOL(kernel_sendmsg);
604 static int ktime2ts(ktime_t kt, struct timespec *ts)
607 *ts = ktime_to_timespec(kt);
615 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
617 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
620 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
621 struct timespec ts[3];
623 struct skb_shared_hwtstamps *shhwtstamps =
626 /* Race occurred between timestamp enabling and packet
627 receiving. Fill in the current time for now. */
628 if (need_software_tstamp && skb->tstamp.tv64 == 0)
629 __net_timestamp(skb);
631 if (need_software_tstamp) {
632 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
634 skb_get_timestamp(skb, &tv);
635 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
638 skb_get_timestampns(skb, &ts[0]);
639 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
640 sizeof(ts[0]), &ts[0]);
645 memset(ts, 0, sizeof(ts));
646 if (skb->tstamp.tv64 &&
647 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
648 skb_get_timestampns(skb, ts + 0);
652 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
653 ktime2ts(shhwtstamps->syststamp, ts + 1))
655 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
656 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
660 put_cmsg(msg, SOL_SOCKET,
661 SCM_TIMESTAMPING, sizeof(ts), &ts);
663 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
665 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
668 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
669 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
670 sizeof(__u32), &skb->dropcount);
673 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
676 sock_recv_timestamp(msg, sk, skb);
677 sock_recv_drops(msg, sk, skb);
679 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
681 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
682 struct msghdr *msg, size_t size, int flags)
684 struct sock_iocb *si = kiocb_to_siocb(iocb);
686 sock_update_classid(sock->sk);
694 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
697 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
698 struct msghdr *msg, size_t size, int flags)
700 int err = security_socket_recvmsg(sock, msg, size, flags);
702 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
705 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
706 size_t size, int flags)
709 struct sock_iocb siocb;
712 init_sync_kiocb(&iocb, NULL);
713 iocb.private = &siocb;
714 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
715 if (-EIOCBQUEUED == ret)
716 ret = wait_on_sync_kiocb(&iocb);
719 EXPORT_SYMBOL(sock_recvmsg);
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);
752 EXPORT_SYMBOL(kernel_recvmsg);
754 static void sock_aio_dtor(struct kiocb *iocb)
756 kfree(iocb->private);
759 static ssize_t sock_sendpage(struct file *file, struct page *page,
760 int offset, size_t size, loff_t *ppos, int more)
765 sock = file->private_data;
767 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
771 return kernel_sendpage(sock, page, offset, size, flags);
774 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
775 struct pipe_inode_info *pipe, size_t len,
778 struct socket *sock = file->private_data;
780 if (unlikely(!sock->ops->splice_read))
783 sock_update_classid(sock->sk);
785 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
788 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
789 struct sock_iocb *siocb)
791 if (!is_sync_kiocb(iocb)) {
792 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
795 iocb->ki_dtor = sock_aio_dtor;
799 iocb->private = siocb;
803 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
804 struct file *file, const struct iovec *iov,
805 unsigned long nr_segs)
807 struct socket *sock = file->private_data;
811 for (i = 0; i < nr_segs; i++)
812 size += iov[i].iov_len;
814 msg->msg_name = NULL;
815 msg->msg_namelen = 0;
816 msg->msg_control = NULL;
817 msg->msg_controllen = 0;
818 msg->msg_iov = (struct iovec *)iov;
819 msg->msg_iovlen = nr_segs;
820 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
822 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
825 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
826 unsigned long nr_segs, loff_t pos)
828 struct sock_iocb siocb, *x;
833 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
837 x = alloc_sock_iocb(iocb, &siocb);
840 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
843 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
844 struct file *file, const struct iovec *iov,
845 unsigned long nr_segs)
847 struct socket *sock = file->private_data;
851 for (i = 0; i < nr_segs; i++)
852 size += iov[i].iov_len;
854 msg->msg_name = NULL;
855 msg->msg_namelen = 0;
856 msg->msg_control = NULL;
857 msg->msg_controllen = 0;
858 msg->msg_iov = (struct iovec *)iov;
859 msg->msg_iovlen = nr_segs;
860 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
861 if (sock->type == SOCK_SEQPACKET)
862 msg->msg_flags |= MSG_EOR;
864 return __sock_sendmsg(iocb, sock, msg, size);
867 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
868 unsigned long nr_segs, loff_t pos)
870 struct sock_iocb siocb, *x;
875 x = alloc_sock_iocb(iocb, &siocb);
879 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
883 * Atomic setting of ioctl hooks to avoid race
884 * with module unload.
887 static DEFINE_MUTEX(br_ioctl_mutex);
888 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
890 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
892 mutex_lock(&br_ioctl_mutex);
893 br_ioctl_hook = hook;
894 mutex_unlock(&br_ioctl_mutex);
896 EXPORT_SYMBOL(brioctl_set);
898 static DEFINE_MUTEX(vlan_ioctl_mutex);
899 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
901 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
903 mutex_lock(&vlan_ioctl_mutex);
904 vlan_ioctl_hook = hook;
905 mutex_unlock(&vlan_ioctl_mutex);
907 EXPORT_SYMBOL(vlan_ioctl_set);
909 static DEFINE_MUTEX(dlci_ioctl_mutex);
910 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
912 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
914 mutex_lock(&dlci_ioctl_mutex);
915 dlci_ioctl_hook = hook;
916 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);
1045 EXPORT_SYMBOL(sock_create_lite);
1047 /* No kernel lock held - perfect */
1048 static unsigned int sock_poll(struct file *file, poll_table *wait)
1050 struct socket *sock;
1053 * We can't return errors to poll, so it's either yes or no.
1055 sock = file->private_data;
1056 return sock->ops->poll(file, sock, wait);
1059 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1061 struct socket *sock = file->private_data;
1063 return sock->ops->mmap(file, sock, vma);
1066 static int sock_close(struct inode *inode, struct file *filp)
1069 * It was possible the inode is NULL we were
1070 * closing an unfinished socket.
1074 printk(KERN_DEBUG "sock_close: NULL inode\n");
1077 sock_release(SOCKET_I(inode));
1082 * Update the socket async list
1084 * Fasync_list locking strategy.
1086 * 1. fasync_list is modified only under process context socket lock
1087 * i.e. under semaphore.
1088 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1089 * or under socket lock
1092 static int sock_fasync(int fd, struct file *filp, int on)
1094 struct socket *sock = filp->private_data;
1095 struct sock *sk = sock->sk;
1102 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1104 if (!sock->wq->fasync_list)
1105 sock_reset_flag(sk, SOCK_FASYNC);
1107 sock_set_flag(sk, SOCK_FASYNC);
1113 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1115 int sock_wake_async(struct socket *sock, int how, int band)
1117 struct socket_wq *wq;
1122 wq = rcu_dereference(sock->wq);
1123 if (!wq || !wq->fasync_list) {
1128 case SOCK_WAKE_WAITD:
1129 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1132 case SOCK_WAKE_SPACE:
1133 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1138 kill_fasync(&wq->fasync_list, SIGIO, band);
1141 kill_fasync(&wq->fasync_list, SIGURG, band);
1146 EXPORT_SYMBOL(sock_wake_async);
1148 static int __sock_create(struct net *net, int family, int type, int protocol,
1149 struct socket **res, int kern)
1152 struct socket *sock;
1153 const struct net_proto_family *pf;
1156 * Check protocol is in range
1158 if (family < 0 || family >= NPROTO)
1159 return -EAFNOSUPPORT;
1160 if (type < 0 || type >= SOCK_MAX)
1165 This uglymoron is moved from INET layer to here to avoid
1166 deadlock in module load.
1168 if (family == PF_INET && type == SOCK_PACKET) {
1172 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1178 err = security_socket_create(family, type, protocol, kern);
1183 * Allocate the socket and allow the family to set things up. if
1184 * the protocol is 0, the family is instructed to select an appropriate
1187 sock = sock_alloc();
1189 if (net_ratelimit())
1190 printk(KERN_WARNING "socket: no more sockets\n");
1191 return -ENFILE; /* Not exactly a match, but its the
1192 closest posix thing */
1197 #ifdef CONFIG_MODULES
1198 /* Attempt to load a protocol module if the find failed.
1200 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1201 * requested real, full-featured networking support upon configuration.
1202 * Otherwise module support will break!
1204 if (net_families[family] == NULL)
1205 request_module("net-pf-%d", family);
1209 pf = rcu_dereference(net_families[family]);
1210 err = -EAFNOSUPPORT;
1215 * We will call the ->create function, that possibly is in a loadable
1216 * module, so we have to bump that loadable module refcnt first.
1218 if (!try_module_get(pf->owner))
1221 /* Now protected by module ref count */
1224 err = pf->create(net, sock, protocol, kern);
1226 goto out_module_put;
1229 * Now to bump the refcnt of the [loadable] module that owns this
1230 * socket at sock_release time we decrement its refcnt.
1232 if (!try_module_get(sock->ops->owner))
1233 goto out_module_busy;
1236 * Now that we're done with the ->create function, the [loadable]
1237 * module can have its refcnt decremented
1239 module_put(pf->owner);
1240 err = security_socket_post_create(sock, family, type, protocol, kern);
1242 goto out_sock_release;
1248 err = -EAFNOSUPPORT;
1251 module_put(pf->owner);
1258 goto out_sock_release;
1261 int sock_create(int family, int type, int protocol, struct socket **res)
1263 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1265 EXPORT_SYMBOL(sock_create);
1267 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1269 return __sock_create(&init_net, family, type, protocol, res, 1);
1271 EXPORT_SYMBOL(sock_create_kern);
1273 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1276 struct socket *sock;
1279 /* Check the SOCK_* constants for consistency. */
1280 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1281 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1282 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1283 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1285 flags = type & ~SOCK_TYPE_MASK;
1286 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1288 type &= SOCK_TYPE_MASK;
1290 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1291 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1293 retval = sock_create(family, type, protocol, &sock);
1297 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1302 /* It may be already another descriptor 8) Not kernel problem. */
1311 * Create a pair of connected sockets.
1314 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1315 int __user *, usockvec)
1317 struct socket *sock1, *sock2;
1319 struct file *newfile1, *newfile2;
1322 flags = type & ~SOCK_TYPE_MASK;
1323 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1325 type &= SOCK_TYPE_MASK;
1327 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1328 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1331 * Obtain the first socket and check if the underlying protocol
1332 * supports the socketpair call.
1335 err = sock_create(family, type, protocol, &sock1);
1339 err = sock_create(family, type, protocol, &sock2);
1343 err = sock1->ops->socketpair(sock1, sock2);
1345 goto out_release_both;
1347 fd1 = sock_alloc_file(sock1, &newfile1, flags);
1348 if (unlikely(fd1 < 0)) {
1350 goto out_release_both;
1353 fd2 = sock_alloc_file(sock2, &newfile2, flags);
1354 if (unlikely(fd2 < 0)) {
1358 sock_release(sock2);
1362 audit_fd_pair(fd1, fd2);
1363 fd_install(fd1, newfile1);
1364 fd_install(fd2, newfile2);
1365 /* fd1 and fd2 may be already another descriptors.
1366 * Not kernel problem.
1369 err = put_user(fd1, &usockvec[0]);
1371 err = put_user(fd2, &usockvec[1]);
1380 sock_release(sock2);
1382 sock_release(sock1);
1388 * Bind a name to a socket. Nothing much to do here since it's
1389 * the protocol's responsibility to handle the local address.
1391 * We move the socket address to kernel space before we call
1392 * the protocol layer (having also checked the address is ok).
1395 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1397 struct socket *sock;
1398 struct sockaddr_storage address;
1399 int err, fput_needed;
1401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1403 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1405 err = security_socket_bind(sock,
1406 (struct sockaddr *)&address,
1409 err = sock->ops->bind(sock,
1413 fput_light(sock->file, fput_needed);
1419 * Perform a listen. Basically, we allow the protocol to do anything
1420 * necessary for a listen, and if that works, we mark the socket as
1421 * ready for listening.
1424 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1426 struct socket *sock;
1427 int err, fput_needed;
1430 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1432 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1433 if ((unsigned)backlog > somaxconn)
1434 backlog = somaxconn;
1436 err = security_socket_listen(sock, backlog);
1438 err = sock->ops->listen(sock, backlog);
1440 fput_light(sock->file, fput_needed);
1446 * For accept, we attempt to create a new socket, set up the link
1447 * with the client, wake up the client, then return the new
1448 * connected fd. We collect the address of the connector in kernel
1449 * space and move it to user at the very end. This is unclean because
1450 * we open the socket then return an error.
1452 * 1003.1g adds the ability to recvmsg() to query connection pending
1453 * status to recvmsg. We need to add that support in a way thats
1454 * clean when we restucture accept also.
1457 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1458 int __user *, upeer_addrlen, int, flags)
1460 struct socket *sock, *newsock;
1461 struct file *newfile;
1462 int err, len, newfd, fput_needed;
1463 struct sockaddr_storage address;
1465 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1468 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1469 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1471 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1476 newsock = sock_alloc();
1480 newsock->type = sock->type;
1481 newsock->ops = sock->ops;
1484 * We don't need try_module_get here, as the listening socket (sock)
1485 * has the protocol module (sock->ops->owner) held.
1487 __module_get(newsock->ops->owner);
1489 newfd = sock_alloc_file(newsock, &newfile, flags);
1490 if (unlikely(newfd < 0)) {
1492 sock_release(newsock);
1496 err = security_socket_accept(sock, newsock);
1500 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1504 if (upeer_sockaddr) {
1505 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1507 err = -ECONNABORTED;
1510 err = move_addr_to_user((struct sockaddr *)&address,
1511 len, upeer_sockaddr, upeer_addrlen);
1516 /* File flags are not inherited via accept() unlike another OSes. */
1518 fd_install(newfd, newfile);
1522 fput_light(sock->file, fput_needed);
1527 put_unused_fd(newfd);
1531 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1532 int __user *, upeer_addrlen)
1534 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1538 * Attempt to connect to a socket with the server address. The address
1539 * is in user space so we verify it is OK and move it to kernel space.
1541 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1544 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1545 * other SEQPACKET protocols that take time to connect() as it doesn't
1546 * include the -EINPROGRESS status for such sockets.
1549 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1552 struct socket *sock;
1553 struct sockaddr_storage address;
1554 int err, fput_needed;
1556 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1559 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1564 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1568 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1569 sock->file->f_flags);
1571 fput_light(sock->file, fput_needed);
1577 * Get the local address ('name') of a socket object. Move the obtained
1578 * name to user space.
1581 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1582 int __user *, usockaddr_len)
1584 struct socket *sock;
1585 struct sockaddr_storage address;
1586 int len, err, fput_needed;
1588 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1592 err = security_socket_getsockname(sock);
1596 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1599 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1602 fput_light(sock->file, fput_needed);
1608 * Get the remote address ('name') of a socket object. Move the obtained
1609 * name to user space.
1612 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1613 int __user *, usockaddr_len)
1615 struct socket *sock;
1616 struct sockaddr_storage address;
1617 int len, err, fput_needed;
1619 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1621 err = security_socket_getpeername(sock);
1623 fput_light(sock->file, fput_needed);
1628 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1631 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1633 fput_light(sock->file, fput_needed);
1639 * Send a datagram to a given address. We move the address into kernel
1640 * space and check the user space data area is readable before invoking
1644 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1645 unsigned, flags, struct sockaddr __user *, addr,
1648 struct socket *sock;
1649 struct sockaddr_storage address;
1655 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1659 iov.iov_base = buff;
1661 msg.msg_name = NULL;
1664 msg.msg_control = NULL;
1665 msg.msg_controllen = 0;
1666 msg.msg_namelen = 0;
1668 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1671 msg.msg_name = (struct sockaddr *)&address;
1672 msg.msg_namelen = addr_len;
1674 if (sock->file->f_flags & O_NONBLOCK)
1675 flags |= MSG_DONTWAIT;
1676 msg.msg_flags = flags;
1677 err = sock_sendmsg(sock, &msg, len);
1680 fput_light(sock->file, fput_needed);
1686 * Send a datagram down a socket.
1689 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1692 return sys_sendto(fd, buff, len, flags, NULL, 0);
1696 * Receive a frame from the socket and optionally record the address of the
1697 * sender. We verify the buffers are writable and if needed move the
1698 * sender address from kernel to user space.
1701 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1702 unsigned, flags, struct sockaddr __user *, addr,
1703 int __user *, addr_len)
1705 struct socket *sock;
1708 struct sockaddr_storage address;
1712 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1716 msg.msg_control = NULL;
1717 msg.msg_controllen = 0;
1721 iov.iov_base = ubuf;
1722 msg.msg_name = (struct sockaddr *)&address;
1723 msg.msg_namelen = sizeof(address);
1724 if (sock->file->f_flags & O_NONBLOCK)
1725 flags |= MSG_DONTWAIT;
1726 err = sock_recvmsg(sock, &msg, size, flags);
1728 if (err >= 0 && addr != NULL) {
1729 err2 = move_addr_to_user((struct sockaddr *)&address,
1730 msg.msg_namelen, addr, addr_len);
1735 fput_light(sock->file, fput_needed);
1741 * Receive a datagram from a socket.
1744 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1747 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1751 * Set a socket option. Because we don't know the option lengths we have
1752 * to pass the user mode parameter for the protocols to sort out.
1755 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1756 char __user *, optval, int, optlen)
1758 int err, fput_needed;
1759 struct socket *sock;
1764 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1766 err = security_socket_setsockopt(sock, level, optname);
1770 if (level == SOL_SOCKET)
1772 sock_setsockopt(sock, level, optname, optval,
1776 sock->ops->setsockopt(sock, level, optname, optval,
1779 fput_light(sock->file, fput_needed);
1785 * Get a socket option. Because we don't know the option lengths we have
1786 * to pass a user mode parameter for the protocols to sort out.
1789 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1790 char __user *, optval, int __user *, optlen)
1792 int err, fput_needed;
1793 struct socket *sock;
1795 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1797 err = security_socket_getsockopt(sock, level, optname);
1801 if (level == SOL_SOCKET)
1803 sock_getsockopt(sock, level, optname, optval,
1807 sock->ops->getsockopt(sock, level, optname, optval,
1810 fput_light(sock->file, fput_needed);
1816 * Shutdown a socket.
1819 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1821 int err, fput_needed;
1822 struct socket *sock;
1824 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1826 err = security_socket_shutdown(sock, how);
1828 err = sock->ops->shutdown(sock, how);
1829 fput_light(sock->file, fput_needed);
1834 /* A couple of helpful macros for getting the address of the 32/64 bit
1835 * fields which are the same type (int / unsigned) on our platforms.
1837 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1838 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1839 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1842 * BSD sendmsg interface
1845 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1847 struct compat_msghdr __user *msg_compat =
1848 (struct compat_msghdr __user *)msg;
1849 struct socket *sock;
1850 struct sockaddr_storage address;
1851 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1852 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1853 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1854 /* 20 is size of ipv6_pktinfo */
1855 unsigned char *ctl_buf = ctl;
1856 struct msghdr msg_sys;
1857 int err, ctl_len, iov_size, total_len;
1861 if (MSG_CMSG_COMPAT & flags) {
1862 if (get_compat_msghdr(&msg_sys, msg_compat))
1864 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1867 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1871 /* do not move before msg_sys is valid */
1873 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1876 /* Check whether to allocate the iovec area */
1878 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1879 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1880 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1885 /* This will also move the address data into kernel space */
1886 if (MSG_CMSG_COMPAT & flags) {
1887 err = verify_compat_iovec(&msg_sys, iov,
1888 (struct sockaddr *)&address,
1891 err = verify_iovec(&msg_sys, iov,
1892 (struct sockaddr *)&address,
1900 if (msg_sys.msg_controllen > INT_MAX)
1902 ctl_len = msg_sys.msg_controllen;
1903 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1905 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1909 ctl_buf = msg_sys.msg_control;
1910 ctl_len = msg_sys.msg_controllen;
1911 } else if (ctl_len) {
1912 if (ctl_len > sizeof(ctl)) {
1913 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1914 if (ctl_buf == NULL)
1919 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1920 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1921 * checking falls down on this.
1923 if (copy_from_user(ctl_buf,
1924 (void __user __force *)msg_sys.msg_control,
1927 msg_sys.msg_control = ctl_buf;
1929 msg_sys.msg_flags = flags;
1931 if (sock->file->f_flags & O_NONBLOCK)
1932 msg_sys.msg_flags |= MSG_DONTWAIT;
1933 err = sock_sendmsg(sock, &msg_sys, total_len);
1937 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1939 if (iov != iovstack)
1940 sock_kfree_s(sock->sk, iov, iov_size);
1942 fput_light(sock->file, fput_needed);
1947 static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1948 struct msghdr *msg_sys, unsigned flags, int nosec)
1950 struct compat_msghdr __user *msg_compat =
1951 (struct compat_msghdr __user *)msg;
1952 struct iovec iovstack[UIO_FASTIOV];
1953 struct iovec *iov = iovstack;
1954 unsigned long cmsg_ptr;
1955 int err, iov_size, total_len, len;
1957 /* kernel mode address */
1958 struct sockaddr_storage addr;
1960 /* user mode address pointers */
1961 struct sockaddr __user *uaddr;
1962 int __user *uaddr_len;
1964 if (MSG_CMSG_COMPAT & flags) {
1965 if (get_compat_msghdr(msg_sys, msg_compat))
1967 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1971 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1974 /* Check whether to allocate the iovec area */
1976 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1977 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1978 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1984 * Save the user-mode address (verify_iovec will change the
1985 * kernel msghdr to use the kernel address space)
1988 uaddr = (__force void __user *)msg_sys->msg_name;
1989 uaddr_len = COMPAT_NAMELEN(msg);
1990 if (MSG_CMSG_COMPAT & flags) {
1991 err = verify_compat_iovec(msg_sys, iov,
1992 (struct sockaddr *)&addr,
1995 err = verify_iovec(msg_sys, iov,
1996 (struct sockaddr *)&addr,
2002 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2003 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2005 if (sock->file->f_flags & O_NONBLOCK)
2006 flags |= MSG_DONTWAIT;
2007 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2013 if (uaddr != NULL) {
2014 err = move_addr_to_user((struct sockaddr *)&addr,
2015 msg_sys->msg_namelen, uaddr,
2020 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2024 if (MSG_CMSG_COMPAT & flags)
2025 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2026 &msg_compat->msg_controllen);
2028 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2029 &msg->msg_controllen);
2035 if (iov != iovstack)
2036 sock_kfree_s(sock->sk, iov, iov_size);
2042 * BSD recvmsg interface
2045 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2046 unsigned int, flags)
2048 int fput_needed, err;
2049 struct msghdr msg_sys;
2050 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2055 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2057 fput_light(sock->file, fput_needed);
2063 * Linux recvmmsg interface
2066 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2067 unsigned int flags, struct timespec *timeout)
2069 int fput_needed, err, datagrams;
2070 struct socket *sock;
2071 struct mmsghdr __user *entry;
2072 struct compat_mmsghdr __user *compat_entry;
2073 struct msghdr msg_sys;
2074 struct timespec end_time;
2077 poll_select_set_timeout(&end_time, timeout->tv_sec,
2083 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2087 err = sock_error(sock->sk);
2092 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2094 while (datagrams < vlen) {
2096 * No need to ask LSM for more than the first datagram.
2098 if (MSG_CMSG_COMPAT & flags) {
2099 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2100 &msg_sys, flags, datagrams);
2103 err = __put_user(err, &compat_entry->msg_len);
2106 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2107 &msg_sys, flags, datagrams);
2110 err = put_user(err, &entry->msg_len);
2118 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2119 if (flags & MSG_WAITFORONE)
2120 flags |= MSG_DONTWAIT;
2123 ktime_get_ts(timeout);
2124 *timeout = timespec_sub(end_time, *timeout);
2125 if (timeout->tv_sec < 0) {
2126 timeout->tv_sec = timeout->tv_nsec = 0;
2130 /* Timeout, return less than vlen datagrams */
2131 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2135 /* Out of band data, return right away */
2136 if (msg_sys.msg_flags & MSG_OOB)
2141 fput_light(sock->file, fput_needed);
2146 if (datagrams != 0) {
2148 * We may return less entries than requested (vlen) if the
2149 * sock is non block and there aren't enough datagrams...
2151 if (err != -EAGAIN) {
2153 * ... or if recvmsg returns an error after we
2154 * received some datagrams, where we record the
2155 * error to return on the next call or if the
2156 * app asks about it using getsockopt(SO_ERROR).
2158 sock->sk->sk_err = -err;
2167 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2168 unsigned int, vlen, unsigned int, flags,
2169 struct timespec __user *, timeout)
2172 struct timespec timeout_sys;
2175 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2177 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2180 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2182 if (datagrams > 0 &&
2183 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2184 datagrams = -EFAULT;
2189 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2190 /* Argument list sizes for sys_socketcall */
2191 #define AL(x) ((x) * sizeof(unsigned long))
2192 static const unsigned char nargs[20] = {
2193 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2194 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2195 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2202 * System call vectors.
2204 * Argument checking cleaned up. Saved 20% in size.
2205 * This function doesn't need to set the kernel lock because
2206 * it is set by the callees.
2209 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2212 unsigned long a0, a1;
2216 if (call < 1 || call > SYS_RECVMMSG)
2220 if (len > sizeof(a))
2223 /* copy_from_user should be SMP safe. */
2224 if (copy_from_user(a, args, len))
2227 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2234 err = sys_socket(a0, a1, a[2]);
2237 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2240 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2243 err = sys_listen(a0, a1);
2246 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2247 (int __user *)a[2], 0);
2249 case SYS_GETSOCKNAME:
2251 sys_getsockname(a0, (struct sockaddr __user *)a1,
2252 (int __user *)a[2]);
2254 case SYS_GETPEERNAME:
2256 sys_getpeername(a0, (struct sockaddr __user *)a1,
2257 (int __user *)a[2]);
2259 case SYS_SOCKETPAIR:
2260 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2263 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2266 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2267 (struct sockaddr __user *)a[4], a[5]);
2270 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2273 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2274 (struct sockaddr __user *)a[4],
2275 (int __user *)a[5]);
2278 err = sys_shutdown(a0, a1);
2280 case SYS_SETSOCKOPT:
2281 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2283 case SYS_GETSOCKOPT:
2285 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2286 (int __user *)a[4]);
2289 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2292 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2295 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2296 (struct timespec __user *)a[4]);
2299 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2300 (int __user *)a[2], a[3]);
2309 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2312 * sock_register - add a socket protocol handler
2313 * @ops: description of protocol
2315 * This function is called by a protocol handler that wants to
2316 * advertise its address family, and have it linked into the
2317 * socket interface. The value ops->family coresponds to the
2318 * socket system call protocol family.
2320 int sock_register(const struct net_proto_family *ops)
2324 if (ops->family >= NPROTO) {
2325 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2330 spin_lock(&net_family_lock);
2331 if (net_families[ops->family])
2334 net_families[ops->family] = ops;
2337 spin_unlock(&net_family_lock);
2339 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2342 EXPORT_SYMBOL(sock_register);
2345 * sock_unregister - remove a protocol handler
2346 * @family: protocol family to remove
2348 * This function is called by a protocol handler that wants to
2349 * remove its address family, and have it unlinked from the
2350 * new socket creation.
2352 * If protocol handler is a module, then it can use module reference
2353 * counts to protect against new references. If protocol handler is not
2354 * a module then it needs to provide its own protection in
2355 * the ops->create routine.
2357 void sock_unregister(int family)
2359 BUG_ON(family < 0 || family >= NPROTO);
2361 spin_lock(&net_family_lock);
2362 net_families[family] = NULL;
2363 spin_unlock(&net_family_lock);
2367 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2369 EXPORT_SYMBOL(sock_unregister);
2371 static int __init sock_init(void)
2374 * Initialize sock SLAB cache.
2380 * Initialize skbuff SLAB cache
2385 * Initialize the protocols module.
2389 register_filesystem(&sock_fs_type);
2390 sock_mnt = kern_mount(&sock_fs_type);
2392 /* The real protocol initialization is performed in later initcalls.
2395 #ifdef CONFIG_NETFILTER
2399 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2400 skb_timestamping_init();
2406 core_initcall(sock_init); /* early initcall */
2408 #ifdef CONFIG_PROC_FS
2409 void socket_seq_show(struct seq_file *seq)
2414 for_each_possible_cpu(cpu)
2415 counter += per_cpu(sockets_in_use, cpu);
2417 /* It can be negative, by the way. 8) */
2421 seq_printf(seq, "sockets: used %d\n", counter);
2423 #endif /* CONFIG_PROC_FS */
2425 #ifdef CONFIG_COMPAT
2426 static int do_siocgstamp(struct net *net, struct socket *sock,
2427 unsigned int cmd, struct compat_timeval __user *up)
2429 mm_segment_t old_fs = get_fs();
2434 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2437 err = put_user(ktv.tv_sec, &up->tv_sec);
2438 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2443 static int do_siocgstampns(struct net *net, struct socket *sock,
2444 unsigned int cmd, struct compat_timespec __user *up)
2446 mm_segment_t old_fs = get_fs();
2447 struct timespec kts;
2451 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2454 err = put_user(kts.tv_sec, &up->tv_sec);
2455 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2460 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2462 struct ifreq __user *uifr;
2465 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2466 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2469 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2473 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2479 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2481 struct compat_ifconf ifc32;
2483 struct ifconf __user *uifc;
2484 struct compat_ifreq __user *ifr32;
2485 struct ifreq __user *ifr;
2489 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2492 if (ifc32.ifcbuf == 0) {
2496 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2498 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2499 sizeof(struct ifreq);
2500 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2502 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2503 ifr32 = compat_ptr(ifc32.ifcbuf);
2504 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2505 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2511 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2514 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2518 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2522 ifr32 = compat_ptr(ifc32.ifcbuf);
2524 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2525 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2526 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2532 if (ifc32.ifcbuf == 0) {
2533 /* Translate from 64-bit structure multiple to
2537 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2542 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2548 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2550 struct ifreq __user *ifr;
2554 ifr = compat_alloc_user_space(sizeof(*ifr));
2556 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2559 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2562 datap = compat_ptr(data);
2563 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2566 return dev_ioctl(net, SIOCETHTOOL, ifr);
2569 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2572 compat_uptr_t uptr32;
2573 struct ifreq __user *uifr;
2575 uifr = compat_alloc_user_space(sizeof(*uifr));
2576 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2579 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2582 uptr = compat_ptr(uptr32);
2584 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2587 return dev_ioctl(net, SIOCWANDEV, uifr);
2590 static int bond_ioctl(struct net *net, unsigned int cmd,
2591 struct compat_ifreq __user *ifr32)
2594 struct ifreq __user *uifr;
2595 mm_segment_t old_fs;
2601 case SIOCBONDENSLAVE:
2602 case SIOCBONDRELEASE:
2603 case SIOCBONDSETHWADDR:
2604 case SIOCBONDCHANGEACTIVE:
2605 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2610 err = dev_ioctl(net, cmd, &kifr);
2614 case SIOCBONDSLAVEINFOQUERY:
2615 case SIOCBONDINFOQUERY:
2616 uifr = compat_alloc_user_space(sizeof(*uifr));
2617 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2620 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2623 datap = compat_ptr(data);
2624 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2627 return dev_ioctl(net, cmd, uifr);
2633 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2634 struct compat_ifreq __user *u_ifreq32)
2636 struct ifreq __user *u_ifreq64;
2637 char tmp_buf[IFNAMSIZ];
2638 void __user *data64;
2641 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2644 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2646 data64 = compat_ptr(data32);
2648 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2650 /* Don't check these user accesses, just let that get trapped
2651 * in the ioctl handler instead.
2653 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2656 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2659 return dev_ioctl(net, cmd, u_ifreq64);
2662 static int dev_ifsioc(struct net *net, struct socket *sock,
2663 unsigned int cmd, struct compat_ifreq __user *uifr32)
2665 struct ifreq __user *uifr;
2668 uifr = compat_alloc_user_space(sizeof(*uifr));
2669 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2672 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2683 case SIOCGIFBRDADDR:
2684 case SIOCGIFDSTADDR:
2685 case SIOCGIFNETMASK:
2690 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2698 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2699 struct compat_ifreq __user *uifr32)
2702 struct compat_ifmap __user *uifmap32;
2703 mm_segment_t old_fs;
2706 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2707 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2708 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2709 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2710 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2711 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2712 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2713 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2719 err = dev_ioctl(net, cmd, (void __user *)&ifr);
2722 if (cmd == SIOCGIFMAP && !err) {
2723 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2724 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2725 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2726 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2727 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2728 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2729 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2736 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2739 compat_uptr_t uptr32;
2740 struct ifreq __user *uifr;
2742 uifr = compat_alloc_user_space(sizeof(*uifr));
2743 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2746 if (get_user(uptr32, &uifr32->ifr_data))
2749 uptr = compat_ptr(uptr32);
2751 if (put_user(uptr, &uifr->ifr_data))
2754 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2759 struct sockaddr rt_dst; /* target address */
2760 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2761 struct sockaddr rt_genmask; /* target network mask (IP) */
2762 unsigned short rt_flags;
2765 unsigned char rt_tos;
2766 unsigned char rt_class;
2768 short rt_metric; /* +1 for binary compatibility! */
2769 /* char * */ u32 rt_dev; /* forcing the device at add */
2770 u32 rt_mtu; /* per route MTU/Window */
2771 u32 rt_window; /* Window clamping */
2772 unsigned short rt_irtt; /* Initial RTT */
2775 struct in6_rtmsg32 {
2776 struct in6_addr rtmsg_dst;
2777 struct in6_addr rtmsg_src;
2778 struct in6_addr rtmsg_gateway;
2788 static int routing_ioctl(struct net *net, struct socket *sock,
2789 unsigned int cmd, void __user *argp)
2793 struct in6_rtmsg r6;
2797 mm_segment_t old_fs = get_fs();
2799 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2800 struct in6_rtmsg32 __user *ur6 = argp;
2801 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2802 3 * sizeof(struct in6_addr));
2803 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2804 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2805 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2806 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2807 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2808 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2809 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2813 struct rtentry32 __user *ur4 = argp;
2814 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2815 3 * sizeof(struct sockaddr));
2816 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2817 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2818 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2819 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2820 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2821 ret |= __get_user(rtdev, &(ur4->rt_dev));
2823 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
2824 r4.rt_dev = devname; devname[15] = 0;
2837 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2844 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2845 * for some operations; this forces use of the newer bridge-utils that
2846 * use compatiable ioctls
2848 static int old_bridge_ioctl(compat_ulong_t __user *argp)
2852 if (get_user(tmp, argp))
2854 if (tmp == BRCTL_GET_VERSION)
2855 return BRCTL_VERSION + 1;
2859 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2860 unsigned int cmd, unsigned long arg)
2862 void __user *argp = compat_ptr(arg);
2863 struct sock *sk = sock->sk;
2864 struct net *net = sock_net(sk);
2866 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2867 return siocdevprivate_ioctl(net, cmd, argp);
2872 return old_bridge_ioctl(argp);
2874 return dev_ifname32(net, argp);
2876 return dev_ifconf(net, argp);
2878 return ethtool_ioctl(net, argp);
2880 return compat_siocwandev(net, argp);
2883 return compat_sioc_ifmap(net, cmd, argp);
2884 case SIOCBONDENSLAVE:
2885 case SIOCBONDRELEASE:
2886 case SIOCBONDSETHWADDR:
2887 case SIOCBONDSLAVEINFOQUERY:
2888 case SIOCBONDINFOQUERY:
2889 case SIOCBONDCHANGEACTIVE:
2890 return bond_ioctl(net, cmd, argp);
2893 return routing_ioctl(net, sock, cmd, argp);
2895 return do_siocgstamp(net, sock, cmd, argp);
2897 return do_siocgstampns(net, sock, cmd, argp);
2899 return compat_siocshwtstamp(net, argp);
2911 return sock_ioctl(file, cmd, arg);
2928 case SIOCSIFHWBROADCAST:
2930 case SIOCGIFBRDADDR:
2931 case SIOCSIFBRDADDR:
2932 case SIOCGIFDSTADDR:
2933 case SIOCSIFDSTADDR:
2934 case SIOCGIFNETMASK:
2935 case SIOCSIFNETMASK:
2946 return dev_ifsioc(net, sock, cmd, argp);
2952 return sock_do_ioctl(net, sock, cmd, arg);
2955 /* Prevent warning from compat_sys_ioctl, these always
2956 * result in -EINVAL in the native case anyway. */
2969 return -ENOIOCTLCMD;
2972 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2975 struct socket *sock = file->private_data;
2976 int ret = -ENOIOCTLCMD;
2983 if (sock->ops->compat_ioctl)
2984 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2986 if (ret == -ENOIOCTLCMD &&
2987 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2988 ret = compat_wext_handle_ioctl(net, cmd, arg);
2990 if (ret == -ENOIOCTLCMD)
2991 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2997 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2999 return sock->ops->bind(sock, addr, addrlen);
3001 EXPORT_SYMBOL(kernel_bind);
3003 int kernel_listen(struct socket *sock, int backlog)
3005 return sock->ops->listen(sock, backlog);
3007 EXPORT_SYMBOL(kernel_listen);
3009 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3011 struct sock *sk = sock->sk;
3014 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3019 err = sock->ops->accept(sock, *newsock, flags);
3021 sock_release(*newsock);
3026 (*newsock)->ops = sock->ops;
3027 __module_get((*newsock)->ops->owner);
3032 EXPORT_SYMBOL(kernel_accept);
3034 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3037 return sock->ops->connect(sock, addr, addrlen, flags);
3039 EXPORT_SYMBOL(kernel_connect);
3041 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3044 return sock->ops->getname(sock, addr, addrlen, 0);
3046 EXPORT_SYMBOL(kernel_getsockname);
3048 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3051 return sock->ops->getname(sock, addr, addrlen, 1);
3053 EXPORT_SYMBOL(kernel_getpeername);
3055 int kernel_getsockopt(struct socket *sock, int level, int optname,
3056 char *optval, int *optlen)
3058 mm_segment_t oldfs = get_fs();
3059 char __user *uoptval;
3060 int __user *uoptlen;
3063 uoptval = (char __user __force *) optval;
3064 uoptlen = (int __user __force *) optlen;
3067 if (level == SOL_SOCKET)
3068 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3070 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3075 EXPORT_SYMBOL(kernel_getsockopt);
3077 int kernel_setsockopt(struct socket *sock, int level, int optname,
3078 char *optval, unsigned int optlen)
3080 mm_segment_t oldfs = get_fs();
3081 char __user *uoptval;
3084 uoptval = (char __user __force *) optval;
3087 if (level == SOL_SOCKET)
3088 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3090 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3095 EXPORT_SYMBOL(kernel_setsockopt);
3097 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3098 size_t size, int flags)
3100 sock_update_classid(sock->sk);
3102 if (sock->ops->sendpage)
3103 return sock->ops->sendpage(sock, page, offset, size, flags);
3105 return sock_no_sendpage(sock, page, offset, size, flags);
3107 EXPORT_SYMBOL(kernel_sendpage);
3109 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3111 mm_segment_t oldfs = get_fs();
3115 err = sock->ops->ioctl(sock, cmd, arg);
3120 EXPORT_SYMBOL(kernel_sock_ioctl);
3122 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3124 return sock->ops->shutdown(sock, how);
3126 EXPORT_SYMBOL(kernel_sock_shutdown);