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>
99 #include <linux/netfilter.h>
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/sockios.h>
105 #include <linux/atalk.h>
107 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
108 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
109 unsigned long nr_segs, loff_t pos);
110 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
111 unsigned long nr_segs, loff_t pos);
112 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
114 static int sock_close(struct inode *inode, struct file *file);
115 static unsigned int sock_poll(struct file *file,
116 struct poll_table_struct *wait);
117 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
119 static long compat_sock_ioctl(struct file *file,
120 unsigned int cmd, unsigned long arg);
122 static int sock_fasync(int fd, struct file *filp, int on);
123 static ssize_t sock_sendpage(struct file *file, struct page *page,
124 int offset, size_t size, loff_t *ppos, int more);
125 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
126 struct pipe_inode_info *pipe, size_t len,
130 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
131 * in the operation structures but are done directly via the socketcall() multiplexor.
134 static const struct file_operations socket_file_ops = {
135 .owner = THIS_MODULE,
137 .aio_read = sock_aio_read,
138 .aio_write = sock_aio_write,
140 .unlocked_ioctl = sock_ioctl,
142 .compat_ioctl = compat_sock_ioctl,
145 .open = sock_no_open, /* special open code to disallow open via /proc */
146 .release = sock_close,
147 .fasync = sock_fasync,
148 .sendpage = sock_sendpage,
149 .splice_write = generic_splice_sendpage,
150 .splice_read = sock_splice_read,
154 * The protocol list. Each protocol is registered in here.
157 static DEFINE_SPINLOCK(net_family_lock);
158 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
161 * Statistics counters of the socket lists
164 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
168 * Move socket addresses back and forth across the kernel/user
169 * divide and look after the messy bits.
172 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
173 16 for IP, 16 for IPX,
176 must be at least one bigger than
177 the AF_UNIX size (see net/unix/af_unix.c
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
198 if (copy_from_user(kaddr, uaddr, ulen))
200 return audit_sockaddr(ulen, kaddr);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
226 err = get_user(len, ulen);
231 if (len < 0 || len > sizeof(struct sockaddr_storage))
234 if (audit_sockaddr(klen, kaddr))
236 if (copy_to_user(uaddr, kaddr, len))
240 * "fromlen shall refer to the value before truncation.."
243 return __put_user(klen, ulen);
246 static struct kmem_cache *sock_inode_cachep __read_mostly;
248 static struct inode *sock_alloc_inode(struct super_block *sb)
250 struct socket_alloc *ei;
252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
256 if (!ei->socket.wq) {
257 kmem_cache_free(sock_inode_cachep, ei);
260 init_waitqueue_head(&ei->socket.wq->wait);
261 ei->socket.wq->fasync_list = NULL;
263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
269 return &ei->vfs_inode;
273 static void wq_free_rcu(struct rcu_head *head)
275 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
280 static void sock_destroy_inode(struct inode *inode)
282 struct socket_alloc *ei;
284 ei = container_of(inode, struct socket_alloc, vfs_inode);
285 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
286 kmem_cache_free(sock_inode_cachep, ei);
289 static void init_once(void *foo)
291 struct socket_alloc *ei = (struct socket_alloc *)foo;
293 inode_init_once(&ei->vfs_inode);
296 static int init_inodecache(void)
298 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
299 sizeof(struct socket_alloc),
301 (SLAB_HWCACHE_ALIGN |
302 SLAB_RECLAIM_ACCOUNT |
305 if (sock_inode_cachep == NULL)
310 static const struct super_operations sockfs_ops = {
311 .alloc_inode = sock_alloc_inode,
312 .destroy_inode =sock_destroy_inode,
313 .statfs = simple_statfs,
316 static int sockfs_get_sb(struct file_system_type *fs_type,
317 int flags, const char *dev_name, void *data,
318 struct vfsmount *mnt)
320 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
324 static struct vfsmount *sock_mnt __read_mostly;
326 static struct file_system_type sock_fs_type = {
328 .get_sb = sockfs_get_sb,
329 .kill_sb = kill_anon_super,
333 * sockfs_dname() is called from d_path().
335 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
337 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
338 dentry->d_inode->i_ino);
341 static const struct dentry_operations sockfs_dentry_operations = {
342 .d_dname = sockfs_dname,
346 * Obtains the first available file descriptor and sets it up for use.
348 * These functions create file structures and maps them to fd space
349 * of the current process. On success it returns file descriptor
350 * and file struct implicitly stored in sock->file.
351 * Note that another thread may close file descriptor before we return
352 * from this function. We use the fact that now we do not refer
353 * to socket after mapping. If one day we will need it, this
354 * function will increment ref. count on file by 1.
356 * In any case returned fd MAY BE not valid!
357 * This race condition is unavoidable
358 * with shared fd spaces, we cannot solve it inside kernel,
359 * but we take care of internal coherence yet.
362 static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
364 struct qstr name = { .name = "" };
369 fd = get_unused_fd_flags(flags);
370 if (unlikely(fd < 0))
373 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
374 if (unlikely(!path.dentry)) {
378 path.mnt = mntget(sock_mnt);
380 path.dentry->d_op = &sockfs_dentry_operations;
381 d_instantiate(path.dentry, SOCK_INODE(sock));
382 SOCK_INODE(sock)->i_fop = &socket_file_ops;
384 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
386 if (unlikely(!file)) {
387 /* drop dentry, keep inode */
388 atomic_inc(&path.dentry->d_inode->i_count);
395 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
397 file->private_data = sock;
403 int sock_map_fd(struct socket *sock, int flags)
405 struct file *newfile;
406 int fd = sock_alloc_file(sock, &newfile, flags);
409 fd_install(fd, newfile);
414 static struct socket *sock_from_file(struct file *file, int *err)
416 if (file->f_op == &socket_file_ops)
417 return file->private_data; /* set in sock_map_fd */
424 * sockfd_lookup - Go from a file number to its socket slot
426 * @err: pointer to an error code return
428 * The file handle passed in is locked and the socket it is bound
429 * too is returned. If an error occurs the err pointer is overwritten
430 * with a negative errno code and NULL is returned. The function checks
431 * for both invalid handles and passing a handle which is not a socket.
433 * On a success the socket object pointer is returned.
436 struct socket *sockfd_lookup(int fd, int *err)
447 sock = sock_from_file(file, err);
453 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
459 file = fget_light(fd, fput_needed);
461 sock = sock_from_file(file, err);
464 fput_light(file, *fput_needed);
470 * sock_alloc - allocate a socket
472 * Allocate a new inode and socket object. The two are bound together
473 * and initialised. The socket is then returned. If we are out of inodes
477 static struct socket *sock_alloc(void)
482 inode = new_inode(sock_mnt->mnt_sb);
486 sock = SOCKET_I(inode);
488 kmemcheck_annotate_bitfield(sock, type);
489 inode->i_mode = S_IFSOCK | S_IRWXUGO;
490 inode->i_uid = current_fsuid();
491 inode->i_gid = current_fsgid();
493 percpu_add(sockets_in_use, 1);
498 * In theory you can't get an open on this inode, but /proc provides
499 * a back door. Remember to keep it shut otherwise you'll let the
500 * creepy crawlies in.
503 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
508 const struct file_operations bad_sock_fops = {
509 .owner = THIS_MODULE,
510 .open = sock_no_open,
514 * sock_release - close a socket
515 * @sock: socket to close
517 * The socket is released from the protocol stack if it has a release
518 * callback, and the inode is then released if the socket is bound to
519 * an inode not a file.
522 void sock_release(struct socket *sock)
525 struct module *owner = sock->ops->owner;
527 sock->ops->release(sock);
532 if (sock->wq->fasync_list)
533 printk(KERN_ERR "sock_release: fasync list not empty!\n");
535 percpu_sub(sockets_in_use, 1);
537 iput(SOCK_INODE(sock));
543 int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
544 union skb_shared_tx *shtx)
547 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
549 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
553 EXPORT_SYMBOL(sock_tx_timestamp);
555 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
556 struct msghdr *msg, size_t size)
558 struct sock_iocb *si = kiocb_to_siocb(iocb);
566 err = security_socket_sendmsg(sock, msg, size);
570 return sock->ops->sendmsg(iocb, sock, msg, size);
573 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
576 struct sock_iocb siocb;
579 init_sync_kiocb(&iocb, NULL);
580 iocb.private = &siocb;
581 ret = __sock_sendmsg(&iocb, sock, msg, size);
582 if (-EIOCBQUEUED == ret)
583 ret = wait_on_sync_kiocb(&iocb);
587 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
588 struct kvec *vec, size_t num, size_t size)
590 mm_segment_t oldfs = get_fs();
595 * the following is safe, since for compiler definitions of kvec and
596 * iovec are identical, yielding the same in-core layout and alignment
598 msg->msg_iov = (struct iovec *)vec;
599 msg->msg_iovlen = num;
600 result = sock_sendmsg(sock, msg, size);
605 static int ktime2ts(ktime_t kt, struct timespec *ts)
608 *ts = ktime_to_timespec(kt);
616 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
618 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
621 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
622 struct timespec ts[3];
624 struct skb_shared_hwtstamps *shhwtstamps =
627 /* Race occurred between timestamp enabling and packet
628 receiving. Fill in the current time for now. */
629 if (need_software_tstamp && skb->tstamp.tv64 == 0)
630 __net_timestamp(skb);
632 if (need_software_tstamp) {
633 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
635 skb_get_timestamp(skb, &tv);
636 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
639 skb_get_timestampns(skb, &ts[0]);
640 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
641 sizeof(ts[0]), &ts[0]);
646 memset(ts, 0, sizeof(ts));
647 if (skb->tstamp.tv64 &&
648 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
649 skb_get_timestampns(skb, ts + 0);
653 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
654 ktime2ts(shhwtstamps->syststamp, ts + 1))
656 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
657 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
661 put_cmsg(msg, SOL_SOCKET,
662 SCM_TIMESTAMPING, sizeof(ts), &ts);
665 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
667 inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
669 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
670 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
671 sizeof(__u32), &skb->dropcount);
674 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
677 sock_recv_timestamp(msg, sk, skb);
678 sock_recv_drops(msg, sk, skb);
680 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
682 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
683 struct msghdr *msg, size_t size, int flags)
685 struct sock_iocb *si = kiocb_to_siocb(iocb);
693 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
696 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
697 struct msghdr *msg, size_t size, int flags)
699 int err = security_socket_recvmsg(sock, msg, size, flags);
701 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
704 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
705 size_t size, int flags)
708 struct sock_iocb siocb;
711 init_sync_kiocb(&iocb, NULL);
712 iocb.private = &siocb;
713 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
714 if (-EIOCBQUEUED == ret)
715 ret = wait_on_sync_kiocb(&iocb);
719 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
720 size_t size, int flags)
723 struct sock_iocb siocb;
726 init_sync_kiocb(&iocb, NULL);
727 iocb.private = &siocb;
728 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
729 if (-EIOCBQUEUED == ret)
730 ret = wait_on_sync_kiocb(&iocb);
734 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
735 struct kvec *vec, size_t num, size_t size, int flags)
737 mm_segment_t oldfs = get_fs();
742 * the following is safe, since for compiler definitions of kvec and
743 * iovec are identical, yielding the same in-core layout and alignment
745 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
746 result = sock_recvmsg(sock, msg, size, flags);
751 static void sock_aio_dtor(struct kiocb *iocb)
753 kfree(iocb->private);
756 static ssize_t sock_sendpage(struct file *file, struct page *page,
757 int offset, size_t size, loff_t *ppos, int more)
762 sock = file->private_data;
764 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
768 return kernel_sendpage(sock, page, offset, size, flags);
771 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
772 struct pipe_inode_info *pipe, size_t len,
775 struct socket *sock = file->private_data;
777 if (unlikely(!sock->ops->splice_read))
780 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
783 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
784 struct sock_iocb *siocb)
786 if (!is_sync_kiocb(iocb)) {
787 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
790 iocb->ki_dtor = sock_aio_dtor;
794 iocb->private = siocb;
798 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
799 struct file *file, const struct iovec *iov,
800 unsigned long nr_segs)
802 struct socket *sock = file->private_data;
806 for (i = 0; i < nr_segs; i++)
807 size += iov[i].iov_len;
809 msg->msg_name = NULL;
810 msg->msg_namelen = 0;
811 msg->msg_control = NULL;
812 msg->msg_controllen = 0;
813 msg->msg_iov = (struct iovec *)iov;
814 msg->msg_iovlen = nr_segs;
815 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
817 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
820 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
821 unsigned long nr_segs, loff_t pos)
823 struct sock_iocb siocb, *x;
828 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
832 x = alloc_sock_iocb(iocb, &siocb);
835 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
838 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
839 struct file *file, const struct iovec *iov,
840 unsigned long nr_segs)
842 struct socket *sock = file->private_data;
846 for (i = 0; i < nr_segs; i++)
847 size += iov[i].iov_len;
849 msg->msg_name = NULL;
850 msg->msg_namelen = 0;
851 msg->msg_control = NULL;
852 msg->msg_controllen = 0;
853 msg->msg_iov = (struct iovec *)iov;
854 msg->msg_iovlen = nr_segs;
855 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
856 if (sock->type == SOCK_SEQPACKET)
857 msg->msg_flags |= MSG_EOR;
859 return __sock_sendmsg(iocb, sock, msg, size);
862 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
863 unsigned long nr_segs, loff_t pos)
865 struct sock_iocb siocb, *x;
870 x = alloc_sock_iocb(iocb, &siocb);
874 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
878 * Atomic setting of ioctl hooks to avoid race
879 * with module unload.
882 static DEFINE_MUTEX(br_ioctl_mutex);
883 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
885 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
887 mutex_lock(&br_ioctl_mutex);
888 br_ioctl_hook = hook;
889 mutex_unlock(&br_ioctl_mutex);
892 EXPORT_SYMBOL(brioctl_set);
894 static DEFINE_MUTEX(vlan_ioctl_mutex);
895 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
897 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
899 mutex_lock(&vlan_ioctl_mutex);
900 vlan_ioctl_hook = hook;
901 mutex_unlock(&vlan_ioctl_mutex);
904 EXPORT_SYMBOL(vlan_ioctl_set);
906 static DEFINE_MUTEX(dlci_ioctl_mutex);
907 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
909 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
911 mutex_lock(&dlci_ioctl_mutex);
912 dlci_ioctl_hook = hook;
913 mutex_unlock(&dlci_ioctl_mutex);
916 EXPORT_SYMBOL(dlci_ioctl_set);
918 static long sock_do_ioctl(struct net *net, struct socket *sock,
919 unsigned int cmd, unsigned long arg)
922 void __user *argp = (void __user *)arg;
924 err = sock->ops->ioctl(sock, cmd, arg);
927 * If this ioctl is unknown try to hand it down
930 if (err == -ENOIOCTLCMD)
931 err = dev_ioctl(net, cmd, argp);
937 * With an ioctl, arg may well be a user mode pointer, but we don't know
938 * what to do with it - that's up to the protocol still.
941 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
945 void __user *argp = (void __user *)arg;
949 sock = file->private_data;
952 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
953 err = dev_ioctl(net, cmd, argp);
955 #ifdef CONFIG_WEXT_CORE
956 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
957 err = dev_ioctl(net, cmd, argp);
964 if (get_user(pid, (int __user *)argp))
966 err = f_setown(sock->file, pid, 1);
970 err = put_user(f_getown(sock->file),
979 request_module("bridge");
981 mutex_lock(&br_ioctl_mutex);
983 err = br_ioctl_hook(net, cmd, argp);
984 mutex_unlock(&br_ioctl_mutex);
989 if (!vlan_ioctl_hook)
990 request_module("8021q");
992 mutex_lock(&vlan_ioctl_mutex);
994 err = vlan_ioctl_hook(net, argp);
995 mutex_unlock(&vlan_ioctl_mutex);
1000 if (!dlci_ioctl_hook)
1001 request_module("dlci");
1003 mutex_lock(&dlci_ioctl_mutex);
1004 if (dlci_ioctl_hook)
1005 err = dlci_ioctl_hook(cmd, argp);
1006 mutex_unlock(&dlci_ioctl_mutex);
1009 err = sock_do_ioctl(net, sock, cmd, arg);
1015 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1018 struct socket *sock = NULL;
1020 err = security_socket_create(family, type, protocol, 1);
1024 sock = sock_alloc();
1031 err = security_socket_post_create(sock, family, type, protocol, 1);
1044 /* No kernel lock held - perfect */
1045 static unsigned int sock_poll(struct file *file, poll_table *wait)
1047 struct socket *sock;
1050 * We can't return errors to poll, so it's either yes or no.
1052 sock = file->private_data;
1053 return sock->ops->poll(file, sock, wait);
1056 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1058 struct socket *sock = file->private_data;
1060 return sock->ops->mmap(file, sock, vma);
1063 static int sock_close(struct inode *inode, struct file *filp)
1066 * It was possible the inode is NULL we were
1067 * closing an unfinished socket.
1071 printk(KERN_DEBUG "sock_close: NULL inode\n");
1074 sock_release(SOCKET_I(inode));
1079 * Update the socket async list
1081 * Fasync_list locking strategy.
1083 * 1. fasync_list is modified only under process context socket lock
1084 * i.e. under semaphore.
1085 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1086 * or under socket lock
1089 static int sock_fasync(int fd, struct file *filp, int on)
1091 struct socket *sock = filp->private_data;
1092 struct sock *sk = sock->sk;
1099 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1101 if (!sock->wq->fasync_list)
1102 sock_reset_flag(sk, SOCK_FASYNC);
1104 sock_set_flag(sk, SOCK_FASYNC);
1110 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1112 int sock_wake_async(struct socket *sock, int how, int band)
1114 struct socket_wq *wq;
1119 wq = rcu_dereference(sock->wq);
1120 if (!wq || !wq->fasync_list) {
1125 case SOCK_WAKE_WAITD:
1126 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1129 case SOCK_WAKE_SPACE:
1130 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1135 kill_fasync(&wq->fasync_list, SIGIO, band);
1138 kill_fasync(&wq->fasync_list, SIGURG, band);
1144 static int __sock_create(struct net *net, int family, int type, int protocol,
1145 struct socket **res, int kern)
1148 struct socket *sock;
1149 const struct net_proto_family *pf;
1152 * Check protocol is in range
1154 if (family < 0 || family >= NPROTO)
1155 return -EAFNOSUPPORT;
1156 if (type < 0 || type >= SOCK_MAX)
1161 This uglymoron is moved from INET layer to here to avoid
1162 deadlock in module load.
1164 if (family == PF_INET && type == SOCK_PACKET) {
1168 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1174 err = security_socket_create(family, type, protocol, kern);
1179 * Allocate the socket and allow the family to set things up. if
1180 * the protocol is 0, the family is instructed to select an appropriate
1183 sock = sock_alloc();
1185 if (net_ratelimit())
1186 printk(KERN_WARNING "socket: no more sockets\n");
1187 return -ENFILE; /* Not exactly a match, but its the
1188 closest posix thing */
1193 #ifdef CONFIG_MODULES
1194 /* Attempt to load a protocol module if the find failed.
1196 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1197 * requested real, full-featured networking support upon configuration.
1198 * Otherwise module support will break!
1200 if (net_families[family] == NULL)
1201 request_module("net-pf-%d", family);
1205 pf = rcu_dereference(net_families[family]);
1206 err = -EAFNOSUPPORT;
1211 * We will call the ->create function, that possibly is in a loadable
1212 * module, so we have to bump that loadable module refcnt first.
1214 if (!try_module_get(pf->owner))
1217 /* Now protected by module ref count */
1220 err = pf->create(net, sock, protocol, kern);
1222 goto out_module_put;
1225 * Now to bump the refcnt of the [loadable] module that owns this
1226 * socket at sock_release time we decrement its refcnt.
1228 if (!try_module_get(sock->ops->owner))
1229 goto out_module_busy;
1232 * Now that we're done with the ->create function, the [loadable]
1233 * module can have its refcnt decremented
1235 module_put(pf->owner);
1236 err = security_socket_post_create(sock, family, type, protocol, kern);
1238 goto out_sock_release;
1244 err = -EAFNOSUPPORT;
1247 module_put(pf->owner);
1254 goto out_sock_release;
1257 int sock_create(int family, int type, int protocol, struct socket **res)
1259 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1262 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1264 return __sock_create(&init_net, family, type, protocol, res, 1);
1267 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1270 struct socket *sock;
1273 /* Check the SOCK_* constants for consistency. */
1274 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1275 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1276 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1277 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1279 flags = type & ~SOCK_TYPE_MASK;
1280 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1282 type &= SOCK_TYPE_MASK;
1284 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1285 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1287 retval = sock_create(family, type, protocol, &sock);
1291 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1296 /* It may be already another descriptor 8) Not kernel problem. */
1305 * Create a pair of connected sockets.
1308 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1309 int __user *, usockvec)
1311 struct socket *sock1, *sock2;
1313 struct file *newfile1, *newfile2;
1316 flags = type & ~SOCK_TYPE_MASK;
1317 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1319 type &= SOCK_TYPE_MASK;
1321 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1322 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1325 * Obtain the first socket and check if the underlying protocol
1326 * supports the socketpair call.
1329 err = sock_create(family, type, protocol, &sock1);
1333 err = sock_create(family, type, protocol, &sock2);
1337 err = sock1->ops->socketpair(sock1, sock2);
1339 goto out_release_both;
1341 fd1 = sock_alloc_file(sock1, &newfile1, flags);
1342 if (unlikely(fd1 < 0)) {
1344 goto out_release_both;
1347 fd2 = sock_alloc_file(sock2, &newfile2, flags);
1348 if (unlikely(fd2 < 0)) {
1352 sock_release(sock2);
1356 audit_fd_pair(fd1, fd2);
1357 fd_install(fd1, newfile1);
1358 fd_install(fd2, newfile2);
1359 /* fd1 and fd2 may be already another descriptors.
1360 * Not kernel problem.
1363 err = put_user(fd1, &usockvec[0]);
1365 err = put_user(fd2, &usockvec[1]);
1374 sock_release(sock2);
1376 sock_release(sock1);
1382 * Bind a name to a socket. Nothing much to do here since it's
1383 * the protocol's responsibility to handle the local address.
1385 * We move the socket address to kernel space before we call
1386 * the protocol layer (having also checked the address is ok).
1389 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1391 struct socket *sock;
1392 struct sockaddr_storage address;
1393 int err, fput_needed;
1395 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1397 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1399 err = security_socket_bind(sock,
1400 (struct sockaddr *)&address,
1403 err = sock->ops->bind(sock,
1407 fput_light(sock->file, fput_needed);
1413 * Perform a listen. Basically, we allow the protocol to do anything
1414 * necessary for a listen, and if that works, we mark the socket as
1415 * ready for listening.
1418 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1420 struct socket *sock;
1421 int err, fput_needed;
1424 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1426 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1427 if ((unsigned)backlog > somaxconn)
1428 backlog = somaxconn;
1430 err = security_socket_listen(sock, backlog);
1432 err = sock->ops->listen(sock, backlog);
1434 fput_light(sock->file, fput_needed);
1440 * For accept, we attempt to create a new socket, set up the link
1441 * with the client, wake up the client, then return the new
1442 * connected fd. We collect the address of the connector in kernel
1443 * space and move it to user at the very end. This is unclean because
1444 * we open the socket then return an error.
1446 * 1003.1g adds the ability to recvmsg() to query connection pending
1447 * status to recvmsg. We need to add that support in a way thats
1448 * clean when we restucture accept also.
1451 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1452 int __user *, upeer_addrlen, int, flags)
1454 struct socket *sock, *newsock;
1455 struct file *newfile;
1456 int err, len, newfd, fput_needed;
1457 struct sockaddr_storage address;
1459 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1462 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1463 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1465 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1470 if (!(newsock = sock_alloc()))
1473 newsock->type = sock->type;
1474 newsock->ops = sock->ops;
1477 * We don't need try_module_get here, as the listening socket (sock)
1478 * has the protocol module (sock->ops->owner) held.
1480 __module_get(newsock->ops->owner);
1482 newfd = sock_alloc_file(newsock, &newfile, flags);
1483 if (unlikely(newfd < 0)) {
1485 sock_release(newsock);
1489 err = security_socket_accept(sock, newsock);
1493 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1497 if (upeer_sockaddr) {
1498 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1500 err = -ECONNABORTED;
1503 err = move_addr_to_user((struct sockaddr *)&address,
1504 len, upeer_sockaddr, upeer_addrlen);
1509 /* File flags are not inherited via accept() unlike another OSes. */
1511 fd_install(newfd, newfile);
1515 fput_light(sock->file, fput_needed);
1520 put_unused_fd(newfd);
1524 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1525 int __user *, upeer_addrlen)
1527 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1531 * Attempt to connect to a socket with the server address. The address
1532 * is in user space so we verify it is OK and move it to kernel space.
1534 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1537 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1538 * other SEQPACKET protocols that take time to connect() as it doesn't
1539 * include the -EINPROGRESS status for such sockets.
1542 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1545 struct socket *sock;
1546 struct sockaddr_storage address;
1547 int err, fput_needed;
1549 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1552 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1557 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1561 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1562 sock->file->f_flags);
1564 fput_light(sock->file, fput_needed);
1570 * Get the local address ('name') of a socket object. Move the obtained
1571 * name to user space.
1574 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1575 int __user *, usockaddr_len)
1577 struct socket *sock;
1578 struct sockaddr_storage address;
1579 int len, err, fput_needed;
1581 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1585 err = security_socket_getsockname(sock);
1589 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1592 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1595 fput_light(sock->file, fput_needed);
1601 * Get the remote address ('name') of a socket object. Move the obtained
1602 * name to user space.
1605 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1606 int __user *, usockaddr_len)
1608 struct socket *sock;
1609 struct sockaddr_storage address;
1610 int len, err, fput_needed;
1612 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1614 err = security_socket_getpeername(sock);
1616 fput_light(sock->file, fput_needed);
1621 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1624 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1626 fput_light(sock->file, fput_needed);
1632 * Send a datagram to a given address. We move the address into kernel
1633 * space and check the user space data area is readable before invoking
1637 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1638 unsigned, flags, struct sockaddr __user *, addr,
1641 struct socket *sock;
1642 struct sockaddr_storage address;
1648 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1652 iov.iov_base = buff;
1654 msg.msg_name = NULL;
1657 msg.msg_control = NULL;
1658 msg.msg_controllen = 0;
1659 msg.msg_namelen = 0;
1661 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1664 msg.msg_name = (struct sockaddr *)&address;
1665 msg.msg_namelen = addr_len;
1667 if (sock->file->f_flags & O_NONBLOCK)
1668 flags |= MSG_DONTWAIT;
1669 msg.msg_flags = flags;
1670 err = sock_sendmsg(sock, &msg, len);
1673 fput_light(sock->file, fput_needed);
1679 * Send a datagram down a socket.
1682 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1685 return sys_sendto(fd, buff, len, flags, NULL, 0);
1689 * Receive a frame from the socket and optionally record the address of the
1690 * sender. We verify the buffers are writable and if needed move the
1691 * sender address from kernel to user space.
1694 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1695 unsigned, flags, struct sockaddr __user *, addr,
1696 int __user *, addr_len)
1698 struct socket *sock;
1701 struct sockaddr_storage address;
1705 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1709 msg.msg_control = NULL;
1710 msg.msg_controllen = 0;
1714 iov.iov_base = ubuf;
1715 msg.msg_name = (struct sockaddr *)&address;
1716 msg.msg_namelen = sizeof(address);
1717 if (sock->file->f_flags & O_NONBLOCK)
1718 flags |= MSG_DONTWAIT;
1719 err = sock_recvmsg(sock, &msg, size, flags);
1721 if (err >= 0 && addr != NULL) {
1722 err2 = move_addr_to_user((struct sockaddr *)&address,
1723 msg.msg_namelen, addr, addr_len);
1728 fput_light(sock->file, fput_needed);
1734 * Receive a datagram from a socket.
1737 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1740 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1744 * Set a socket option. Because we don't know the option lengths we have
1745 * to pass the user mode parameter for the protocols to sort out.
1748 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1749 char __user *, optval, int, optlen)
1751 int err, fput_needed;
1752 struct socket *sock;
1757 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1759 err = security_socket_setsockopt(sock, level, optname);
1763 if (level == SOL_SOCKET)
1765 sock_setsockopt(sock, level, optname, optval,
1769 sock->ops->setsockopt(sock, level, optname, optval,
1772 fput_light(sock->file, fput_needed);
1778 * Get a socket option. Because we don't know the option lengths we have
1779 * to pass a user mode parameter for the protocols to sort out.
1782 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1783 char __user *, optval, int __user *, optlen)
1785 int err, fput_needed;
1786 struct socket *sock;
1788 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1790 err = security_socket_getsockopt(sock, level, optname);
1794 if (level == SOL_SOCKET)
1796 sock_getsockopt(sock, level, optname, optval,
1800 sock->ops->getsockopt(sock, level, optname, optval,
1803 fput_light(sock->file, fput_needed);
1809 * Shutdown a socket.
1812 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1814 int err, fput_needed;
1815 struct socket *sock;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1819 err = security_socket_shutdown(sock, how);
1821 err = sock->ops->shutdown(sock, how);
1822 fput_light(sock->file, fput_needed);
1827 /* A couple of helpful macros for getting the address of the 32/64 bit
1828 * fields which are the same type (int / unsigned) on our platforms.
1830 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1831 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1832 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1835 * BSD sendmsg interface
1838 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1840 struct compat_msghdr __user *msg_compat =
1841 (struct compat_msghdr __user *)msg;
1842 struct socket *sock;
1843 struct sockaddr_storage address;
1844 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1845 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1846 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1847 /* 20 is size of ipv6_pktinfo */
1848 unsigned char *ctl_buf = ctl;
1849 struct msghdr msg_sys;
1850 int err, ctl_len, iov_size, total_len;
1854 if (MSG_CMSG_COMPAT & flags) {
1855 if (get_compat_msghdr(&msg_sys, msg_compat))
1858 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1861 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1865 /* do not move before msg_sys is valid */
1867 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1870 /* Check whether to allocate the iovec area */
1872 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1873 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1874 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1879 /* This will also move the address data into kernel space */
1880 if (MSG_CMSG_COMPAT & flags) {
1881 err = verify_compat_iovec(&msg_sys, iov,
1882 (struct sockaddr *)&address,
1885 err = verify_iovec(&msg_sys, iov,
1886 (struct sockaddr *)&address,
1894 if (msg_sys.msg_controllen > INT_MAX)
1896 ctl_len = msg_sys.msg_controllen;
1897 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1899 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1903 ctl_buf = msg_sys.msg_control;
1904 ctl_len = msg_sys.msg_controllen;
1905 } else if (ctl_len) {
1906 if (ctl_len > sizeof(ctl)) {
1907 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1908 if (ctl_buf == NULL)
1913 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1914 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1915 * checking falls down on this.
1917 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1920 msg_sys.msg_control = ctl_buf;
1922 msg_sys.msg_flags = flags;
1924 if (sock->file->f_flags & O_NONBLOCK)
1925 msg_sys.msg_flags |= MSG_DONTWAIT;
1926 err = sock_sendmsg(sock, &msg_sys, total_len);
1930 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1932 if (iov != iovstack)
1933 sock_kfree_s(sock->sk, iov, iov_size);
1935 fput_light(sock->file, fput_needed);
1940 static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1941 struct msghdr *msg_sys, unsigned flags, int nosec)
1943 struct compat_msghdr __user *msg_compat =
1944 (struct compat_msghdr __user *)msg;
1945 struct iovec iovstack[UIO_FASTIOV];
1946 struct iovec *iov = iovstack;
1947 unsigned long cmsg_ptr;
1948 int err, iov_size, total_len, len;
1950 /* kernel mode address */
1951 struct sockaddr_storage addr;
1953 /* user mode address pointers */
1954 struct sockaddr __user *uaddr;
1955 int __user *uaddr_len;
1957 if (MSG_CMSG_COMPAT & flags) {
1958 if (get_compat_msghdr(msg_sys, msg_compat))
1961 else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1965 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1968 /* Check whether to allocate the iovec area */
1970 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1971 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1972 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1978 * Save the user-mode address (verify_iovec will change the
1979 * kernel msghdr to use the kernel address space)
1982 uaddr = (__force void __user *)msg_sys->msg_name;
1983 uaddr_len = COMPAT_NAMELEN(msg);
1984 if (MSG_CMSG_COMPAT & flags) {
1985 err = verify_compat_iovec(msg_sys, iov,
1986 (struct sockaddr *)&addr,
1989 err = verify_iovec(msg_sys, iov,
1990 (struct sockaddr *)&addr,
1996 cmsg_ptr = (unsigned long)msg_sys->msg_control;
1997 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
1999 if (sock->file->f_flags & O_NONBLOCK)
2000 flags |= MSG_DONTWAIT;
2001 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2007 if (uaddr != NULL) {
2008 err = move_addr_to_user((struct sockaddr *)&addr,
2009 msg_sys->msg_namelen, uaddr,
2014 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2018 if (MSG_CMSG_COMPAT & flags)
2019 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2020 &msg_compat->msg_controllen);
2022 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2023 &msg->msg_controllen);
2029 if (iov != iovstack)
2030 sock_kfree_s(sock->sk, iov, iov_size);
2036 * BSD recvmsg interface
2039 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2040 unsigned int, flags)
2042 int fput_needed, err;
2043 struct msghdr msg_sys;
2044 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2049 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2051 fput_light(sock->file, fput_needed);
2057 * Linux recvmmsg interface
2060 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2061 unsigned int flags, struct timespec *timeout)
2063 int fput_needed, err, datagrams;
2064 struct socket *sock;
2065 struct mmsghdr __user *entry;
2066 struct compat_mmsghdr __user *compat_entry;
2067 struct msghdr msg_sys;
2068 struct timespec end_time;
2071 poll_select_set_timeout(&end_time, timeout->tv_sec,
2077 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2081 err = sock_error(sock->sk);
2086 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2088 while (datagrams < vlen) {
2090 * No need to ask LSM for more than the first datagram.
2092 if (MSG_CMSG_COMPAT & flags) {
2093 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2094 &msg_sys, flags, datagrams);
2097 err = __put_user(err, &compat_entry->msg_len);
2100 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2101 &msg_sys, flags, datagrams);
2104 err = put_user(err, &entry->msg_len);
2112 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2113 if (flags & MSG_WAITFORONE)
2114 flags |= MSG_DONTWAIT;
2117 ktime_get_ts(timeout);
2118 *timeout = timespec_sub(end_time, *timeout);
2119 if (timeout->tv_sec < 0) {
2120 timeout->tv_sec = timeout->tv_nsec = 0;
2124 /* Timeout, return less than vlen datagrams */
2125 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2129 /* Out of band data, return right away */
2130 if (msg_sys.msg_flags & MSG_OOB)
2135 fput_light(sock->file, fput_needed);
2140 if (datagrams != 0) {
2142 * We may return less entries than requested (vlen) if the
2143 * sock is non block and there aren't enough datagrams...
2145 if (err != -EAGAIN) {
2147 * ... or if recvmsg returns an error after we
2148 * received some datagrams, where we record the
2149 * error to return on the next call or if the
2150 * app asks about it using getsockopt(SO_ERROR).
2152 sock->sk->sk_err = -err;
2161 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2162 unsigned int, vlen, unsigned int, flags,
2163 struct timespec __user *, timeout)
2166 struct timespec timeout_sys;
2169 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2171 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2174 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2176 if (datagrams > 0 &&
2177 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2178 datagrams = -EFAULT;
2183 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2184 /* Argument list sizes for sys_socketcall */
2185 #define AL(x) ((x) * sizeof(unsigned long))
2186 static const unsigned char nargs[20] = {
2187 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2188 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2189 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2196 * System call vectors.
2198 * Argument checking cleaned up. Saved 20% in size.
2199 * This function doesn't need to set the kernel lock because
2200 * it is set by the callees.
2203 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2206 unsigned long a0, a1;
2210 if (call < 1 || call > SYS_RECVMMSG)
2214 if (len > sizeof(a))
2217 /* copy_from_user should be SMP safe. */
2218 if (copy_from_user(a, args, len))
2221 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2228 err = sys_socket(a0, a1, a[2]);
2231 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2234 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2237 err = sys_listen(a0, a1);
2240 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2241 (int __user *)a[2], 0);
2243 case SYS_GETSOCKNAME:
2245 sys_getsockname(a0, (struct sockaddr __user *)a1,
2246 (int __user *)a[2]);
2248 case SYS_GETPEERNAME:
2250 sys_getpeername(a0, (struct sockaddr __user *)a1,
2251 (int __user *)a[2]);
2253 case SYS_SOCKETPAIR:
2254 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2257 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2260 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2261 (struct sockaddr __user *)a[4], a[5]);
2264 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2267 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2268 (struct sockaddr __user *)a[4],
2269 (int __user *)a[5]);
2272 err = sys_shutdown(a0, a1);
2274 case SYS_SETSOCKOPT:
2275 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2277 case SYS_GETSOCKOPT:
2279 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2280 (int __user *)a[4]);
2283 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2286 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2289 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2290 (struct timespec __user *)a[4]);
2293 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2294 (int __user *)a[2], a[3]);
2303 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2306 * sock_register - add a socket protocol handler
2307 * @ops: description of protocol
2309 * This function is called by a protocol handler that wants to
2310 * advertise its address family, and have it linked into the
2311 * socket interface. The value ops->family coresponds to the
2312 * socket system call protocol family.
2314 int sock_register(const struct net_proto_family *ops)
2318 if (ops->family >= NPROTO) {
2319 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2324 spin_lock(&net_family_lock);
2325 if (net_families[ops->family])
2328 net_families[ops->family] = ops;
2331 spin_unlock(&net_family_lock);
2333 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2338 * sock_unregister - remove a protocol handler
2339 * @family: protocol family to remove
2341 * This function is called by a protocol handler that wants to
2342 * remove its address family, and have it unlinked from the
2343 * new socket creation.
2345 * If protocol handler is a module, then it can use module reference
2346 * counts to protect against new references. If protocol handler is not
2347 * a module then it needs to provide its own protection in
2348 * the ops->create routine.
2350 void sock_unregister(int family)
2352 BUG_ON(family < 0 || family >= NPROTO);
2354 spin_lock(&net_family_lock);
2355 net_families[family] = NULL;
2356 spin_unlock(&net_family_lock);
2360 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2363 static int __init sock_init(void)
2366 * Initialize sock SLAB cache.
2372 * Initialize skbuff SLAB cache
2377 * Initialize the protocols module.
2381 register_filesystem(&sock_fs_type);
2382 sock_mnt = kern_mount(&sock_fs_type);
2384 /* The real protocol initialization is performed in later initcalls.
2387 #ifdef CONFIG_NETFILTER
2394 core_initcall(sock_init); /* early initcall */
2396 #ifdef CONFIG_PROC_FS
2397 void socket_seq_show(struct seq_file *seq)
2402 for_each_possible_cpu(cpu)
2403 counter += per_cpu(sockets_in_use, cpu);
2405 /* It can be negative, by the way. 8) */
2409 seq_printf(seq, "sockets: used %d\n", counter);
2411 #endif /* CONFIG_PROC_FS */
2413 #ifdef CONFIG_COMPAT
2414 static int do_siocgstamp(struct net *net, struct socket *sock,
2415 unsigned int cmd, struct compat_timeval __user *up)
2417 mm_segment_t old_fs = get_fs();
2422 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2425 err = put_user(ktv.tv_sec, &up->tv_sec);
2426 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2431 static int do_siocgstampns(struct net *net, struct socket *sock,
2432 unsigned int cmd, struct compat_timespec __user *up)
2434 mm_segment_t old_fs = get_fs();
2435 struct timespec kts;
2439 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2442 err = put_user(kts.tv_sec, &up->tv_sec);
2443 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2448 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2450 struct ifreq __user *uifr;
2453 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2454 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2457 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2461 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2467 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2469 struct compat_ifconf ifc32;
2471 struct ifconf __user *uifc;
2472 struct compat_ifreq __user *ifr32;
2473 struct ifreq __user *ifr;
2477 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2480 if (ifc32.ifcbuf == 0) {
2484 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2486 size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
2487 sizeof (struct ifreq);
2488 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2490 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2491 ifr32 = compat_ptr(ifc32.ifcbuf);
2492 for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2493 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2499 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2502 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2506 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2510 ifr32 = compat_ptr(ifc32.ifcbuf);
2512 i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2513 i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2514 if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
2520 if (ifc32.ifcbuf == 0) {
2521 /* Translate from 64-bit structure multiple to
2525 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2530 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2536 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2538 struct ifreq __user *ifr;
2542 ifr = compat_alloc_user_space(sizeof(*ifr));
2544 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2547 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2550 datap = compat_ptr(data);
2551 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2554 return dev_ioctl(net, SIOCETHTOOL, ifr);
2557 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2560 compat_uptr_t uptr32;
2561 struct ifreq __user *uifr;
2563 uifr = compat_alloc_user_space(sizeof (*uifr));
2564 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2567 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2570 uptr = compat_ptr(uptr32);
2572 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2575 return dev_ioctl(net, SIOCWANDEV, uifr);
2578 static int bond_ioctl(struct net *net, unsigned int cmd,
2579 struct compat_ifreq __user *ifr32)
2582 struct ifreq __user *uifr;
2583 mm_segment_t old_fs;
2589 case SIOCBONDENSLAVE:
2590 case SIOCBONDRELEASE:
2591 case SIOCBONDSETHWADDR:
2592 case SIOCBONDCHANGEACTIVE:
2593 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2598 err = dev_ioctl(net, cmd, &kifr);
2602 case SIOCBONDSLAVEINFOQUERY:
2603 case SIOCBONDINFOQUERY:
2604 uifr = compat_alloc_user_space(sizeof(*uifr));
2605 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2608 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2611 datap = compat_ptr(data);
2612 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2615 return dev_ioctl(net, cmd, uifr);
2621 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2622 struct compat_ifreq __user *u_ifreq32)
2624 struct ifreq __user *u_ifreq64;
2625 char tmp_buf[IFNAMSIZ];
2626 void __user *data64;
2629 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2632 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2634 data64 = compat_ptr(data32);
2636 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2638 /* Don't check these user accesses, just let that get trapped
2639 * in the ioctl handler instead.
2641 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2644 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2647 return dev_ioctl(net, cmd, u_ifreq64);
2650 static int dev_ifsioc(struct net *net, struct socket *sock,
2651 unsigned int cmd, struct compat_ifreq __user *uifr32)
2653 struct ifreq __user *uifr;
2656 uifr = compat_alloc_user_space(sizeof(*uifr));
2657 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2660 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2671 case SIOCGIFBRDADDR:
2672 case SIOCGIFDSTADDR:
2673 case SIOCGIFNETMASK:
2678 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2686 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2687 struct compat_ifreq __user *uifr32)
2690 struct compat_ifmap __user *uifmap32;
2691 mm_segment_t old_fs;
2694 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2695 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2696 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2697 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2698 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2699 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2700 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2701 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2707 err = dev_ioctl(net, cmd, (void __user *)&ifr);
2710 if (cmd == SIOCGIFMAP && !err) {
2711 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2712 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2713 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2714 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2715 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2716 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2717 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2724 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2727 compat_uptr_t uptr32;
2728 struct ifreq __user *uifr;
2730 uifr = compat_alloc_user_space(sizeof (*uifr));
2731 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2734 if (get_user(uptr32, &uifr32->ifr_data))
2737 uptr = compat_ptr(uptr32);
2739 if (put_user(uptr, &uifr->ifr_data))
2742 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2747 struct sockaddr rt_dst; /* target address */
2748 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2749 struct sockaddr rt_genmask; /* target network mask (IP) */
2750 unsigned short rt_flags;
2753 unsigned char rt_tos;
2754 unsigned char rt_class;
2756 short rt_metric; /* +1 for binary compatibility! */
2757 /* char * */ u32 rt_dev; /* forcing the device at add */
2758 u32 rt_mtu; /* per route MTU/Window */
2759 u32 rt_window; /* Window clamping */
2760 unsigned short rt_irtt; /* Initial RTT */
2763 struct in6_rtmsg32 {
2764 struct in6_addr rtmsg_dst;
2765 struct in6_addr rtmsg_src;
2766 struct in6_addr rtmsg_gateway;
2776 static int routing_ioctl(struct net *net, struct socket *sock,
2777 unsigned int cmd, void __user *argp)
2781 struct in6_rtmsg r6;
2785 mm_segment_t old_fs = get_fs();
2787 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2788 struct in6_rtmsg32 __user *ur6 = argp;
2789 ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2790 3 * sizeof(struct in6_addr));
2791 ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2792 ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2793 ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2794 ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2795 ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2796 ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2797 ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2801 struct rtentry32 __user *ur4 = argp;
2802 ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2803 3 * sizeof(struct sockaddr));
2804 ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2805 ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2806 ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2807 ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2808 ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2809 ret |= __get_user (rtdev, &(ur4->rt_dev));
2811 ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2812 r4.rt_dev = devname; devname[15] = 0;
2825 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2832 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2833 * for some operations; this forces use of the newer bridge-utils that
2834 * use compatiable ioctls
2836 static int old_bridge_ioctl(compat_ulong_t __user *argp)
2840 if (get_user(tmp, argp))
2842 if (tmp == BRCTL_GET_VERSION)
2843 return BRCTL_VERSION + 1;
2847 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2848 unsigned int cmd, unsigned long arg)
2850 void __user *argp = compat_ptr(arg);
2851 struct sock *sk = sock->sk;
2852 struct net *net = sock_net(sk);
2854 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2855 return siocdevprivate_ioctl(net, cmd, argp);
2860 return old_bridge_ioctl(argp);
2862 return dev_ifname32(net, argp);
2864 return dev_ifconf(net, argp);
2866 return ethtool_ioctl(net, argp);
2868 return compat_siocwandev(net, argp);
2871 return compat_sioc_ifmap(net, cmd, argp);
2872 case SIOCBONDENSLAVE:
2873 case SIOCBONDRELEASE:
2874 case SIOCBONDSETHWADDR:
2875 case SIOCBONDSLAVEINFOQUERY:
2876 case SIOCBONDINFOQUERY:
2877 case SIOCBONDCHANGEACTIVE:
2878 return bond_ioctl(net, cmd, argp);
2881 return routing_ioctl(net, sock, cmd, argp);
2883 return do_siocgstamp(net, sock, cmd, argp);
2885 return do_siocgstampns(net, sock, cmd, argp);
2887 return compat_siocshwtstamp(net, argp);
2899 return sock_ioctl(file, cmd, arg);
2916 case SIOCSIFHWBROADCAST:
2918 case SIOCGIFBRDADDR:
2919 case SIOCSIFBRDADDR:
2920 case SIOCGIFDSTADDR:
2921 case SIOCSIFDSTADDR:
2922 case SIOCGIFNETMASK:
2923 case SIOCSIFNETMASK:
2934 return dev_ifsioc(net, sock, cmd, argp);
2940 return sock_do_ioctl(net, sock, cmd, arg);
2943 /* Prevent warning from compat_sys_ioctl, these always
2944 * result in -EINVAL in the native case anyway. */
2957 return -ENOIOCTLCMD;
2960 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2963 struct socket *sock = file->private_data;
2964 int ret = -ENOIOCTLCMD;
2971 if (sock->ops->compat_ioctl)
2972 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2974 if (ret == -ENOIOCTLCMD &&
2975 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2976 ret = compat_wext_handle_ioctl(net, cmd, arg);
2978 if (ret == -ENOIOCTLCMD)
2979 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2985 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2987 return sock->ops->bind(sock, addr, addrlen);
2990 int kernel_listen(struct socket *sock, int backlog)
2992 return sock->ops->listen(sock, backlog);
2995 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2997 struct sock *sk = sock->sk;
3000 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3005 err = sock->ops->accept(sock, *newsock, flags);
3007 sock_release(*newsock);
3012 (*newsock)->ops = sock->ops;
3013 __module_get((*newsock)->ops->owner);
3019 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3022 return sock->ops->connect(sock, addr, addrlen, flags);
3025 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3028 return sock->ops->getname(sock, addr, addrlen, 0);
3031 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3034 return sock->ops->getname(sock, addr, addrlen, 1);
3037 int kernel_getsockopt(struct socket *sock, int level, int optname,
3038 char *optval, int *optlen)
3040 mm_segment_t oldfs = get_fs();
3044 if (level == SOL_SOCKET)
3045 err = sock_getsockopt(sock, level, optname, optval, optlen);
3047 err = sock->ops->getsockopt(sock, level, optname, optval,
3053 int kernel_setsockopt(struct socket *sock, int level, int optname,
3054 char *optval, unsigned int optlen)
3056 mm_segment_t oldfs = get_fs();
3060 if (level == SOL_SOCKET)
3061 err = sock_setsockopt(sock, level, optname, optval, optlen);
3063 err = sock->ops->setsockopt(sock, level, optname, optval,
3069 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3070 size_t size, int flags)
3072 if (sock->ops->sendpage)
3073 return sock->ops->sendpage(sock, page, offset, size, flags);
3075 return sock_no_sendpage(sock, page, offset, size, flags);
3078 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3080 mm_segment_t oldfs = get_fs();
3084 err = sock->ops->ioctl(sock, cmd, arg);
3090 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3092 return sock->ops->shutdown(sock, how);
3095 EXPORT_SYMBOL(sock_create);
3096 EXPORT_SYMBOL(sock_create_kern);
3097 EXPORT_SYMBOL(sock_create_lite);
3098 EXPORT_SYMBOL(sock_map_fd);
3099 EXPORT_SYMBOL(sock_recvmsg);
3100 EXPORT_SYMBOL(sock_register);
3101 EXPORT_SYMBOL(sock_release);
3102 EXPORT_SYMBOL(sock_sendmsg);
3103 EXPORT_SYMBOL(sock_unregister);
3104 EXPORT_SYMBOL(sock_wake_async);
3105 EXPORT_SYMBOL(sockfd_lookup);
3106 EXPORT_SYMBOL(kernel_sendmsg);
3107 EXPORT_SYMBOL(kernel_recvmsg);
3108 EXPORT_SYMBOL(kernel_bind);
3109 EXPORT_SYMBOL(kernel_listen);
3110 EXPORT_SYMBOL(kernel_accept);
3111 EXPORT_SYMBOL(kernel_connect);
3112 EXPORT_SYMBOL(kernel_getsockname);
3113 EXPORT_SYMBOL(kernel_getpeername);
3114 EXPORT_SYMBOL(kernel_getsockopt);
3115 EXPORT_SYMBOL(kernel_setsockopt);
3116 EXPORT_SYMBOL(kernel_sendpage);
3117 EXPORT_SYMBOL(kernel_sock_ioctl);
3118 EXPORT_SYMBOL(kernel_sock_shutdown);