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 __rcu *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 BUG_ON(klen > sizeof(struct sockaddr_storage));
219 err = get_user(len, ulen);
227 if (audit_sockaddr(klen, kaddr))
229 if (copy_to_user(uaddr, kaddr, len))
233 * "fromlen shall refer to the value before truncation.."
236 return __put_user(klen, ulen);
239 static struct kmem_cache *sock_inode_cachep __read_mostly;
241 static struct inode *sock_alloc_inode(struct super_block *sb)
243 struct socket_alloc *ei;
244 struct socket_wq *wq;
246 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
249 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
251 kmem_cache_free(sock_inode_cachep, ei);
254 init_waitqueue_head(&wq->wait);
255 wq->fasync_list = NULL;
256 RCU_INIT_POINTER(ei->socket.wq, wq);
258 ei->socket.state = SS_UNCONNECTED;
259 ei->socket.flags = 0;
260 ei->socket.ops = NULL;
261 ei->socket.sk = NULL;
262 ei->socket.file = NULL;
264 return &ei->vfs_inode;
267 static void sock_destroy_inode(struct inode *inode)
269 struct socket_alloc *ei;
270 struct socket_wq *wq;
272 ei = container_of(inode, struct socket_alloc, vfs_inode);
273 wq = rcu_dereference_protected(ei->socket.wq, 1);
275 kmem_cache_free(sock_inode_cachep, ei);
278 static void init_once(void *foo)
280 struct socket_alloc *ei = (struct socket_alloc *)foo;
282 inode_init_once(&ei->vfs_inode);
285 static int init_inodecache(void)
287 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
288 sizeof(struct socket_alloc),
290 (SLAB_HWCACHE_ALIGN |
291 SLAB_RECLAIM_ACCOUNT |
294 if (sock_inode_cachep == NULL)
299 static const struct super_operations sockfs_ops = {
300 .alloc_inode = sock_alloc_inode,
301 .destroy_inode = sock_destroy_inode,
302 .statfs = simple_statfs,
306 * sockfs_dname() is called from d_path().
308 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
310 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
311 dentry->d_inode->i_ino);
314 static const struct dentry_operations sockfs_dentry_operations = {
315 .d_dname = sockfs_dname,
318 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
319 int flags, const char *dev_name, void *data)
321 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
322 &sockfs_dentry_operations, SOCKFS_MAGIC);
325 static struct vfsmount *sock_mnt __read_mostly;
327 static struct file_system_type sock_fs_type = {
329 .mount = sockfs_mount,
330 .kill_sb = kill_anon_super,
334 * Obtains the first available file descriptor and sets it up for use.
336 * These functions create file structures and maps them to fd space
337 * of the current process. On success it returns file descriptor
338 * and file struct implicitly stored in sock->file.
339 * Note that another thread may close file descriptor before we return
340 * from this function. We use the fact that now we do not refer
341 * to socket after mapping. If one day we will need it, this
342 * function will increment ref. count on file by 1.
344 * In any case returned fd MAY BE not valid!
345 * This race condition is unavoidable
346 * with shared fd spaces, we cannot solve it inside kernel,
347 * but we take care of internal coherence yet.
350 static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
352 struct qstr name = { .name = "" };
357 fd = get_unused_fd_flags(flags);
358 if (unlikely(fd < 0))
361 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
362 if (unlikely(!path.dentry)) {
366 path.mnt = mntget(sock_mnt);
368 d_instantiate(path.dentry, SOCK_INODE(sock));
369 SOCK_INODE(sock)->i_fop = &socket_file_ops;
371 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
373 if (unlikely(!file)) {
374 /* drop dentry, keep inode */
375 ihold(path.dentry->d_inode);
382 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
384 file->private_data = sock;
390 int sock_map_fd(struct socket *sock, int flags)
392 struct file *newfile;
393 int fd = sock_alloc_file(sock, &newfile, flags);
396 fd_install(fd, newfile);
400 EXPORT_SYMBOL(sock_map_fd);
402 static struct socket *sock_from_file(struct file *file, int *err)
404 if (file->f_op == &socket_file_ops)
405 return file->private_data; /* set in sock_map_fd */
412 * sockfd_lookup - Go from a file number to its socket slot
414 * @err: pointer to an error code return
416 * The file handle passed in is locked and the socket it is bound
417 * too is returned. If an error occurs the err pointer is overwritten
418 * with a negative errno code and NULL is returned. The function checks
419 * for both invalid handles and passing a handle which is not a socket.
421 * On a success the socket object pointer is returned.
424 struct socket *sockfd_lookup(int fd, int *err)
435 sock = sock_from_file(file, err);
440 EXPORT_SYMBOL(sockfd_lookup);
442 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
448 file = fget_light(fd, fput_needed);
450 sock = sock_from_file(file, err);
453 fput_light(file, *fput_needed);
459 * sock_alloc - allocate a socket
461 * Allocate a new inode and socket object. The two are bound together
462 * and initialised. The socket is then returned. If we are out of inodes
466 static struct socket *sock_alloc(void)
471 inode = new_inode_pseudo(sock_mnt->mnt_sb);
475 sock = SOCKET_I(inode);
477 kmemcheck_annotate_bitfield(sock, type);
478 inode->i_ino = get_next_ino();
479 inode->i_mode = S_IFSOCK | S_IRWXUGO;
480 inode->i_uid = current_fsuid();
481 inode->i_gid = current_fsgid();
483 percpu_add(sockets_in_use, 1);
488 * In theory you can't get an open on this inode, but /proc provides
489 * a back door. Remember to keep it shut otherwise you'll let the
490 * creepy crawlies in.
493 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
498 const struct file_operations bad_sock_fops = {
499 .owner = THIS_MODULE,
500 .open = sock_no_open,
501 .llseek = noop_llseek,
505 * sock_release - close a socket
506 * @sock: socket to close
508 * The socket is released from the protocol stack if it has a release
509 * callback, and the inode is then released if the socket is bound to
510 * an inode not a file.
513 void sock_release(struct socket *sock)
516 struct module *owner = sock->ops->owner;
518 sock->ops->release(sock);
523 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
524 printk(KERN_ERR "sock_release: fasync list not empty!\n");
526 percpu_sub(sockets_in_use, 1);
528 iput(SOCK_INODE(sock));
533 EXPORT_SYMBOL(sock_release);
535 int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
538 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
539 *tx_flags |= SKBTX_HW_TSTAMP;
540 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
541 *tx_flags |= SKBTX_SW_TSTAMP;
544 EXPORT_SYMBOL(sock_tx_timestamp);
546 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
547 struct msghdr *msg, size_t size)
549 struct sock_iocb *si = kiocb_to_siocb(iocb);
551 sock_update_classid(sock->sk);
558 return sock->ops->sendmsg(iocb, sock, msg, size);
561 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
562 struct msghdr *msg, size_t size)
564 int err = security_socket_sendmsg(sock, msg, size);
566 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
569 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
572 struct sock_iocb siocb;
575 init_sync_kiocb(&iocb, NULL);
576 iocb.private = &siocb;
577 ret = __sock_sendmsg(&iocb, sock, msg, size);
578 if (-EIOCBQUEUED == ret)
579 ret = wait_on_sync_kiocb(&iocb);
582 EXPORT_SYMBOL(sock_sendmsg);
584 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
587 struct sock_iocb siocb;
590 init_sync_kiocb(&iocb, NULL);
591 iocb.private = &siocb;
592 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
593 if (-EIOCBQUEUED == ret)
594 ret = wait_on_sync_kiocb(&iocb);
598 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
599 struct kvec *vec, size_t num, size_t size)
601 mm_segment_t oldfs = get_fs();
606 * the following is safe, since for compiler definitions of kvec and
607 * iovec are identical, yielding the same in-core layout and alignment
609 msg->msg_iov = (struct iovec *)vec;
610 msg->msg_iovlen = num;
611 result = sock_sendmsg(sock, msg, size);
615 EXPORT_SYMBOL(kernel_sendmsg);
617 static int ktime2ts(ktime_t kt, struct timespec *ts)
620 *ts = ktime_to_timespec(kt);
628 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
630 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
633 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
634 struct timespec ts[3];
636 struct skb_shared_hwtstamps *shhwtstamps =
639 /* Race occurred between timestamp enabling and packet
640 receiving. Fill in the current time for now. */
641 if (need_software_tstamp && skb->tstamp.tv64 == 0)
642 __net_timestamp(skb);
644 if (need_software_tstamp) {
645 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
647 skb_get_timestamp(skb, &tv);
648 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
651 skb_get_timestampns(skb, &ts[0]);
652 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
653 sizeof(ts[0]), &ts[0]);
658 memset(ts, 0, sizeof(ts));
659 if (skb->tstamp.tv64 &&
660 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
661 skb_get_timestampns(skb, ts + 0);
665 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
666 ktime2ts(shhwtstamps->syststamp, ts + 1))
668 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
669 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
673 put_cmsg(msg, SOL_SOCKET,
674 SCM_TIMESTAMPING, sizeof(ts), &ts);
676 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
678 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
681 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
682 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
683 sizeof(__u32), &skb->dropcount);
686 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
689 sock_recv_timestamp(msg, sk, skb);
690 sock_recv_drops(msg, sk, skb);
692 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
694 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
695 struct msghdr *msg, size_t size, int flags)
697 struct sock_iocb *si = kiocb_to_siocb(iocb);
699 sock_update_classid(sock->sk);
707 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
710 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
711 struct msghdr *msg, size_t size, int flags)
713 int err = security_socket_recvmsg(sock, msg, size, flags);
715 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
718 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
719 size_t size, int flags)
722 struct sock_iocb siocb;
725 init_sync_kiocb(&iocb, NULL);
726 iocb.private = &siocb;
727 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
728 if (-EIOCBQUEUED == ret)
729 ret = wait_on_sync_kiocb(&iocb);
732 EXPORT_SYMBOL(sock_recvmsg);
734 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
735 size_t size, int flags)
738 struct sock_iocb siocb;
741 init_sync_kiocb(&iocb, NULL);
742 iocb.private = &siocb;
743 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
744 if (-EIOCBQUEUED == ret)
745 ret = wait_on_sync_kiocb(&iocb);
750 * kernel_recvmsg - Receive a message from a socket (kernel space)
751 * @sock: The socket to receive the message from
752 * @msg: Received message
753 * @vec: Input s/g array for message data
754 * @num: Size of input s/g array
755 * @size: Number of bytes to read
756 * @flags: Message flags (MSG_DONTWAIT, etc...)
758 * On return the msg structure contains the scatter/gather array passed in the
759 * vec argument. The array is modified so that it consists of the unfilled
760 * portion of the original array.
762 * The returned value is the total number of bytes received, or an error.
764 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
765 struct kvec *vec, size_t num, size_t size, int flags)
767 mm_segment_t oldfs = get_fs();
772 * the following is safe, since for compiler definitions of kvec and
773 * iovec are identical, yielding the same in-core layout and alignment
775 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
776 result = sock_recvmsg(sock, msg, size, flags);
780 EXPORT_SYMBOL(kernel_recvmsg);
782 static void sock_aio_dtor(struct kiocb *iocb)
784 kfree(iocb->private);
787 static ssize_t sock_sendpage(struct file *file, struct page *page,
788 int offset, size_t size, loff_t *ppos, int more)
793 sock = file->private_data;
795 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
796 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
799 return kernel_sendpage(sock, page, offset, size, flags);
802 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
803 struct pipe_inode_info *pipe, size_t len,
806 struct socket *sock = file->private_data;
808 if (unlikely(!sock->ops->splice_read))
811 sock_update_classid(sock->sk);
813 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
816 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
817 struct sock_iocb *siocb)
819 if (!is_sync_kiocb(iocb)) {
820 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
823 iocb->ki_dtor = sock_aio_dtor;
827 iocb->private = siocb;
831 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
832 struct file *file, const struct iovec *iov,
833 unsigned long nr_segs)
835 struct socket *sock = file->private_data;
839 for (i = 0; i < nr_segs; i++)
840 size += iov[i].iov_len;
842 msg->msg_name = NULL;
843 msg->msg_namelen = 0;
844 msg->msg_control = NULL;
845 msg->msg_controllen = 0;
846 msg->msg_iov = (struct iovec *)iov;
847 msg->msg_iovlen = nr_segs;
848 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
850 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
853 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
854 unsigned long nr_segs, loff_t pos)
856 struct sock_iocb siocb, *x;
861 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
865 x = alloc_sock_iocb(iocb, &siocb);
868 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
871 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
872 struct file *file, const struct iovec *iov,
873 unsigned long nr_segs)
875 struct socket *sock = file->private_data;
879 for (i = 0; i < nr_segs; i++)
880 size += iov[i].iov_len;
882 msg->msg_name = NULL;
883 msg->msg_namelen = 0;
884 msg->msg_control = NULL;
885 msg->msg_controllen = 0;
886 msg->msg_iov = (struct iovec *)iov;
887 msg->msg_iovlen = nr_segs;
888 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
889 if (sock->type == SOCK_SEQPACKET)
890 msg->msg_flags |= MSG_EOR;
892 return __sock_sendmsg(iocb, sock, msg, size);
895 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
896 unsigned long nr_segs, loff_t pos)
898 struct sock_iocb siocb, *x;
903 x = alloc_sock_iocb(iocb, &siocb);
907 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
911 * Atomic setting of ioctl hooks to avoid race
912 * with module unload.
915 static DEFINE_MUTEX(br_ioctl_mutex);
916 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
918 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
920 mutex_lock(&br_ioctl_mutex);
921 br_ioctl_hook = hook;
922 mutex_unlock(&br_ioctl_mutex);
924 EXPORT_SYMBOL(brioctl_set);
926 static DEFINE_MUTEX(vlan_ioctl_mutex);
927 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
929 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
931 mutex_lock(&vlan_ioctl_mutex);
932 vlan_ioctl_hook = hook;
933 mutex_unlock(&vlan_ioctl_mutex);
935 EXPORT_SYMBOL(vlan_ioctl_set);
937 static DEFINE_MUTEX(dlci_ioctl_mutex);
938 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
940 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
942 mutex_lock(&dlci_ioctl_mutex);
943 dlci_ioctl_hook = hook;
944 mutex_unlock(&dlci_ioctl_mutex);
946 EXPORT_SYMBOL(dlci_ioctl_set);
948 static long sock_do_ioctl(struct net *net, struct socket *sock,
949 unsigned int cmd, unsigned long arg)
952 void __user *argp = (void __user *)arg;
954 err = sock->ops->ioctl(sock, cmd, arg);
957 * If this ioctl is unknown try to hand it down
960 if (err == -ENOIOCTLCMD)
961 err = dev_ioctl(net, cmd, argp);
967 * With an ioctl, arg may well be a user mode pointer, but we don't know
968 * what to do with it - that's up to the protocol still.
971 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
975 void __user *argp = (void __user *)arg;
979 sock = file->private_data;
982 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
983 err = dev_ioctl(net, cmd, argp);
985 #ifdef CONFIG_WEXT_CORE
986 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
987 err = dev_ioctl(net, cmd, argp);
994 if (get_user(pid, (int __user *)argp))
996 err = f_setown(sock->file, pid, 1);
1000 err = put_user(f_getown(sock->file),
1001 (int __user *)argp);
1009 request_module("bridge");
1011 mutex_lock(&br_ioctl_mutex);
1013 err = br_ioctl_hook(net, cmd, argp);
1014 mutex_unlock(&br_ioctl_mutex);
1019 if (!vlan_ioctl_hook)
1020 request_module("8021q");
1022 mutex_lock(&vlan_ioctl_mutex);
1023 if (vlan_ioctl_hook)
1024 err = vlan_ioctl_hook(net, argp);
1025 mutex_unlock(&vlan_ioctl_mutex);
1030 if (!dlci_ioctl_hook)
1031 request_module("dlci");
1033 mutex_lock(&dlci_ioctl_mutex);
1034 if (dlci_ioctl_hook)
1035 err = dlci_ioctl_hook(cmd, argp);
1036 mutex_unlock(&dlci_ioctl_mutex);
1039 err = sock_do_ioctl(net, sock, cmd, arg);
1045 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1048 struct socket *sock = NULL;
1050 err = security_socket_create(family, type, protocol, 1);
1054 sock = sock_alloc();
1061 err = security_socket_post_create(sock, family, type, protocol, 1);
1073 EXPORT_SYMBOL(sock_create_lite);
1075 /* No kernel lock held - perfect */
1076 static unsigned int sock_poll(struct file *file, poll_table *wait)
1078 struct socket *sock;
1081 * We can't return errors to poll, so it's either yes or no.
1083 sock = file->private_data;
1084 return sock->ops->poll(file, sock, wait);
1087 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1089 struct socket *sock = file->private_data;
1091 return sock->ops->mmap(file, sock, vma);
1094 static int sock_close(struct inode *inode, struct file *filp)
1097 * It was possible the inode is NULL we were
1098 * closing an unfinished socket.
1102 printk(KERN_DEBUG "sock_close: NULL inode\n");
1105 sock_release(SOCKET_I(inode));
1110 * Update the socket async list
1112 * Fasync_list locking strategy.
1114 * 1. fasync_list is modified only under process context socket lock
1115 * i.e. under semaphore.
1116 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1117 * or under socket lock
1120 static int sock_fasync(int fd, struct file *filp, int on)
1122 struct socket *sock = filp->private_data;
1123 struct sock *sk = sock->sk;
1124 struct socket_wq *wq;
1130 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1131 fasync_helper(fd, filp, on, &wq->fasync_list);
1133 if (!wq->fasync_list)
1134 sock_reset_flag(sk, SOCK_FASYNC);
1136 sock_set_flag(sk, SOCK_FASYNC);
1142 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1144 int sock_wake_async(struct socket *sock, int how, int band)
1146 struct socket_wq *wq;
1151 wq = rcu_dereference(sock->wq);
1152 if (!wq || !wq->fasync_list) {
1157 case SOCK_WAKE_WAITD:
1158 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1161 case SOCK_WAKE_SPACE:
1162 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1167 kill_fasync(&wq->fasync_list, SIGIO, band);
1170 kill_fasync(&wq->fasync_list, SIGURG, band);
1175 EXPORT_SYMBOL(sock_wake_async);
1177 int __sock_create(struct net *net, int family, int type, int protocol,
1178 struct socket **res, int kern)
1181 struct socket *sock;
1182 const struct net_proto_family *pf;
1185 * Check protocol is in range
1187 if (family < 0 || family >= NPROTO)
1188 return -EAFNOSUPPORT;
1189 if (type < 0 || type >= SOCK_MAX)
1194 This uglymoron is moved from INET layer to here to avoid
1195 deadlock in module load.
1197 if (family == PF_INET && type == SOCK_PACKET) {
1201 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1207 err = security_socket_create(family, type, protocol, kern);
1212 * Allocate the socket and allow the family to set things up. if
1213 * the protocol is 0, the family is instructed to select an appropriate
1216 sock = sock_alloc();
1218 if (net_ratelimit())
1219 printk(KERN_WARNING "socket: no more sockets\n");
1220 return -ENFILE; /* Not exactly a match, but its the
1221 closest posix thing */
1226 #ifdef CONFIG_MODULES
1227 /* Attempt to load a protocol module if the find failed.
1229 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1230 * requested real, full-featured networking support upon configuration.
1231 * Otherwise module support will break!
1233 if (rcu_access_pointer(net_families[family]) == NULL)
1234 request_module("net-pf-%d", family);
1238 pf = rcu_dereference(net_families[family]);
1239 err = -EAFNOSUPPORT;
1244 * We will call the ->create function, that possibly is in a loadable
1245 * module, so we have to bump that loadable module refcnt first.
1247 if (!try_module_get(pf->owner))
1250 /* Now protected by module ref count */
1253 err = pf->create(net, sock, protocol, kern);
1255 goto out_module_put;
1258 * Now to bump the refcnt of the [loadable] module that owns this
1259 * socket at sock_release time we decrement its refcnt.
1261 if (!try_module_get(sock->ops->owner))
1262 goto out_module_busy;
1265 * Now that we're done with the ->create function, the [loadable]
1266 * module can have its refcnt decremented
1268 module_put(pf->owner);
1269 err = security_socket_post_create(sock, family, type, protocol, kern);
1271 goto out_sock_release;
1277 err = -EAFNOSUPPORT;
1280 module_put(pf->owner);
1287 goto out_sock_release;
1289 EXPORT_SYMBOL(__sock_create);
1291 int sock_create(int family, int type, int protocol, struct socket **res)
1293 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1295 EXPORT_SYMBOL(sock_create);
1297 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1299 return __sock_create(&init_net, family, type, protocol, res, 1);
1301 EXPORT_SYMBOL(sock_create_kern);
1303 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1306 struct socket *sock;
1309 /* Check the SOCK_* constants for consistency. */
1310 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1311 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1312 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1313 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1315 flags = type & ~SOCK_TYPE_MASK;
1316 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1318 type &= SOCK_TYPE_MASK;
1320 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1321 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1323 retval = sock_create(family, type, protocol, &sock);
1327 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1332 /* It may be already another descriptor 8) Not kernel problem. */
1341 * Create a pair of connected sockets.
1344 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1345 int __user *, usockvec)
1347 struct socket *sock1, *sock2;
1349 struct file *newfile1, *newfile2;
1352 flags = type & ~SOCK_TYPE_MASK;
1353 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1355 type &= SOCK_TYPE_MASK;
1357 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1358 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1361 * Obtain the first socket and check if the underlying protocol
1362 * supports the socketpair call.
1365 err = sock_create(family, type, protocol, &sock1);
1369 err = sock_create(family, type, protocol, &sock2);
1373 err = sock1->ops->socketpair(sock1, sock2);
1375 goto out_release_both;
1377 fd1 = sock_alloc_file(sock1, &newfile1, flags);
1378 if (unlikely(fd1 < 0)) {
1380 goto out_release_both;
1383 fd2 = sock_alloc_file(sock2, &newfile2, flags);
1384 if (unlikely(fd2 < 0)) {
1388 sock_release(sock2);
1392 audit_fd_pair(fd1, fd2);
1393 fd_install(fd1, newfile1);
1394 fd_install(fd2, newfile2);
1395 /* fd1 and fd2 may be already another descriptors.
1396 * Not kernel problem.
1399 err = put_user(fd1, &usockvec[0]);
1401 err = put_user(fd2, &usockvec[1]);
1410 sock_release(sock2);
1412 sock_release(sock1);
1418 * Bind a name to a socket. Nothing much to do here since it's
1419 * the protocol's responsibility to handle the local address.
1421 * We move the socket address to kernel space before we call
1422 * the protocol layer (having also checked the address is ok).
1425 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1427 struct socket *sock;
1428 struct sockaddr_storage address;
1429 int err, fput_needed;
1431 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1433 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1435 err = security_socket_bind(sock,
1436 (struct sockaddr *)&address,
1439 err = sock->ops->bind(sock,
1443 fput_light(sock->file, fput_needed);
1449 * Perform a listen. Basically, we allow the protocol to do anything
1450 * necessary for a listen, and if that works, we mark the socket as
1451 * ready for listening.
1454 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1456 struct socket *sock;
1457 int err, fput_needed;
1460 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1462 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1463 if ((unsigned)backlog > somaxconn)
1464 backlog = somaxconn;
1466 err = security_socket_listen(sock, backlog);
1468 err = sock->ops->listen(sock, backlog);
1470 fput_light(sock->file, fput_needed);
1476 * For accept, we attempt to create a new socket, set up the link
1477 * with the client, wake up the client, then return the new
1478 * connected fd. We collect the address of the connector in kernel
1479 * space and move it to user at the very end. This is unclean because
1480 * we open the socket then return an error.
1482 * 1003.1g adds the ability to recvmsg() to query connection pending
1483 * status to recvmsg. We need to add that support in a way thats
1484 * clean when we restucture accept also.
1487 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1488 int __user *, upeer_addrlen, int, flags)
1490 struct socket *sock, *newsock;
1491 struct file *newfile;
1492 int err, len, newfd, fput_needed;
1493 struct sockaddr_storage address;
1495 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1498 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1499 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1501 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1506 newsock = sock_alloc();
1510 newsock->type = sock->type;
1511 newsock->ops = sock->ops;
1514 * We don't need try_module_get here, as the listening socket (sock)
1515 * has the protocol module (sock->ops->owner) held.
1517 __module_get(newsock->ops->owner);
1519 newfd = sock_alloc_file(newsock, &newfile, flags);
1520 if (unlikely(newfd < 0)) {
1522 sock_release(newsock);
1526 err = security_socket_accept(sock, newsock);
1530 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1534 if (upeer_sockaddr) {
1535 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1537 err = -ECONNABORTED;
1540 err = move_addr_to_user((struct sockaddr *)&address,
1541 len, upeer_sockaddr, upeer_addrlen);
1546 /* File flags are not inherited via accept() unlike another OSes. */
1548 fd_install(newfd, newfile);
1552 fput_light(sock->file, fput_needed);
1557 put_unused_fd(newfd);
1561 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1562 int __user *, upeer_addrlen)
1564 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1568 * Attempt to connect to a socket with the server address. The address
1569 * is in user space so we verify it is OK and move it to kernel space.
1571 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1574 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1575 * other SEQPACKET protocols that take time to connect() as it doesn't
1576 * include the -EINPROGRESS status for such sockets.
1579 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1582 struct socket *sock;
1583 struct sockaddr_storage address;
1584 int err, fput_needed;
1586 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1589 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1594 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1598 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1599 sock->file->f_flags);
1601 fput_light(sock->file, fput_needed);
1607 * Get the local address ('name') of a socket object. Move the obtained
1608 * name to user space.
1611 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1612 int __user *, usockaddr_len)
1614 struct socket *sock;
1615 struct sockaddr_storage address;
1616 int len, err, fput_needed;
1618 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1622 err = security_socket_getsockname(sock);
1626 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1629 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1632 fput_light(sock->file, fput_needed);
1638 * Get the remote address ('name') of a socket object. Move the obtained
1639 * name to user space.
1642 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1643 int __user *, usockaddr_len)
1645 struct socket *sock;
1646 struct sockaddr_storage address;
1647 int len, err, fput_needed;
1649 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1651 err = security_socket_getpeername(sock);
1653 fput_light(sock->file, fput_needed);
1658 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1661 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1663 fput_light(sock->file, fput_needed);
1669 * Send a datagram to a given address. We move the address into kernel
1670 * space and check the user space data area is readable before invoking
1674 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1675 unsigned, flags, struct sockaddr __user *, addr,
1678 struct socket *sock;
1679 struct sockaddr_storage address;
1687 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1691 iov.iov_base = buff;
1693 msg.msg_name = NULL;
1696 msg.msg_control = NULL;
1697 msg.msg_controllen = 0;
1698 msg.msg_namelen = 0;
1700 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1703 msg.msg_name = (struct sockaddr *)&address;
1704 msg.msg_namelen = addr_len;
1706 if (sock->file->f_flags & O_NONBLOCK)
1707 flags |= MSG_DONTWAIT;
1708 msg.msg_flags = flags;
1709 err = sock_sendmsg(sock, &msg, len);
1712 fput_light(sock->file, fput_needed);
1718 * Send a datagram down a socket.
1721 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1724 return sys_sendto(fd, buff, len, flags, NULL, 0);
1728 * Receive a frame from the socket and optionally record the address of the
1729 * sender. We verify the buffers are writable and if needed move the
1730 * sender address from kernel to user space.
1733 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1734 unsigned, flags, struct sockaddr __user *, addr,
1735 int __user *, addr_len)
1737 struct socket *sock;
1740 struct sockaddr_storage address;
1746 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1750 msg.msg_control = NULL;
1751 msg.msg_controllen = 0;
1755 iov.iov_base = ubuf;
1756 /* Save some cycles and don't copy the address if not needed */
1757 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1758 /* We assume all kernel code knows the size of sockaddr_storage */
1759 msg.msg_namelen = 0;
1760 if (sock->file->f_flags & O_NONBLOCK)
1761 flags |= MSG_DONTWAIT;
1762 err = sock_recvmsg(sock, &msg, size, flags);
1764 if (err >= 0 && addr != NULL) {
1765 err2 = move_addr_to_user((struct sockaddr *)&address,
1766 msg.msg_namelen, addr, addr_len);
1771 fput_light(sock->file, fput_needed);
1777 * Receive a datagram from a socket.
1780 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1783 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1787 * Set a socket option. Because we don't know the option lengths we have
1788 * to pass the user mode parameter for the protocols to sort out.
1791 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1792 char __user *, optval, int, optlen)
1794 int err, fput_needed;
1795 struct socket *sock;
1800 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1802 err = security_socket_setsockopt(sock, level, optname);
1806 if (level == SOL_SOCKET)
1808 sock_setsockopt(sock, level, optname, optval,
1812 sock->ops->setsockopt(sock, level, optname, optval,
1815 fput_light(sock->file, fput_needed);
1821 * Get a socket option. Because we don't know the option lengths we have
1822 * to pass a user mode parameter for the protocols to sort out.
1825 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1826 char __user *, optval, int __user *, optlen)
1828 int err, fput_needed;
1829 struct socket *sock;
1831 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1833 err = security_socket_getsockopt(sock, level, optname);
1837 if (level == SOL_SOCKET)
1839 sock_getsockopt(sock, level, optname, optval,
1843 sock->ops->getsockopt(sock, level, optname, optval,
1846 fput_light(sock->file, fput_needed);
1852 * Shutdown a socket.
1855 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1857 int err, fput_needed;
1858 struct socket *sock;
1860 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1862 err = security_socket_shutdown(sock, how);
1864 err = sock->ops->shutdown(sock, how);
1865 fput_light(sock->file, fput_needed);
1870 /* A couple of helpful macros for getting the address of the 32/64 bit
1871 * fields which are the same type (int / unsigned) on our platforms.
1873 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1874 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1875 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1877 struct used_address {
1878 struct sockaddr_storage name;
1879 unsigned int name_len;
1882 static int copy_msghdr_from_user(struct msghdr *kmsg,
1883 struct msghdr __user *umsg)
1885 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1888 if (kmsg->msg_name == NULL)
1889 kmsg->msg_namelen = 0;
1891 if (kmsg->msg_namelen < 0)
1894 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1895 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1899 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1900 struct msghdr *msg_sys, unsigned flags,
1901 struct used_address *used_address)
1903 struct compat_msghdr __user *msg_compat =
1904 (struct compat_msghdr __user *)msg;
1905 struct sockaddr_storage address;
1906 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1907 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1908 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1909 /* 20 is size of ipv6_pktinfo */
1910 unsigned char *ctl_buf = ctl;
1911 int err, ctl_len, iov_size, total_len;
1914 if (MSG_CMSG_COMPAT & flags)
1915 err = get_compat_msghdr(msg_sys, msg_compat);
1917 err = copy_msghdr_from_user(msg_sys, msg);
1921 /* do not move before msg_sys is valid */
1923 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1926 /* Check whether to allocate the iovec area */
1928 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1929 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1930 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1935 /* This will also move the address data into kernel space */
1936 if (MSG_CMSG_COMPAT & flags) {
1937 err = verify_compat_iovec(msg_sys, iov,
1938 (struct sockaddr *)&address,
1941 err = verify_iovec(msg_sys, iov,
1942 (struct sockaddr *)&address,
1950 if (msg_sys->msg_controllen > INT_MAX)
1952 ctl_len = msg_sys->msg_controllen;
1953 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1955 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1959 ctl_buf = msg_sys->msg_control;
1960 ctl_len = msg_sys->msg_controllen;
1961 } else if (ctl_len) {
1962 if (ctl_len > sizeof(ctl)) {
1963 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1964 if (ctl_buf == NULL)
1969 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1970 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1971 * checking falls down on this.
1973 if (copy_from_user(ctl_buf,
1974 (void __user __force *)msg_sys->msg_control,
1977 msg_sys->msg_control = ctl_buf;
1979 msg_sys->msg_flags = flags;
1981 if (sock->file->f_flags & O_NONBLOCK)
1982 msg_sys->msg_flags |= MSG_DONTWAIT;
1984 * If this is sendmmsg() and current destination address is same as
1985 * previously succeeded address, omit asking LSM's decision.
1986 * used_address->name_len is initialized to UINT_MAX so that the first
1987 * destination address never matches.
1989 if (used_address && msg_sys->msg_name &&
1990 used_address->name_len == msg_sys->msg_namelen &&
1991 !memcmp(&used_address->name, msg_sys->msg_name,
1992 used_address->name_len)) {
1993 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1996 err = sock_sendmsg(sock, msg_sys, total_len);
1998 * If this is sendmmsg() and sending to current destination address was
1999 * successful, remember it.
2001 if (used_address && err >= 0) {
2002 used_address->name_len = msg_sys->msg_namelen;
2003 if (msg_sys->msg_name)
2004 memcpy(&used_address->name, msg_sys->msg_name,
2005 used_address->name_len);
2010 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2012 if (iov != iovstack)
2013 sock_kfree_s(sock->sk, iov, iov_size);
2019 * BSD sendmsg interface
2022 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2024 int fput_needed, err;
2025 struct msghdr msg_sys;
2026 struct socket *sock;
2028 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2032 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2034 fput_light(sock->file, fput_needed);
2039 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2041 if (flags & MSG_CMSG_COMPAT)
2043 return __sys_sendmsg(fd, msg, flags);
2047 * Linux sendmmsg interface
2050 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2053 int fput_needed, err, datagrams;
2054 struct socket *sock;
2055 struct mmsghdr __user *entry;
2056 struct compat_mmsghdr __user *compat_entry;
2057 struct msghdr msg_sys;
2058 struct used_address used_address;
2060 if (vlen > UIO_MAXIOV)
2065 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2069 used_address.name_len = UINT_MAX;
2071 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2074 while (datagrams < vlen) {
2075 if (MSG_CMSG_COMPAT & flags) {
2076 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2077 &msg_sys, flags, &used_address);
2080 err = __put_user(err, &compat_entry->msg_len);
2083 err = ___sys_sendmsg(sock,
2084 (struct msghdr __user *)entry,
2085 &msg_sys, flags, &used_address);
2088 err = put_user(err, &entry->msg_len);
2097 fput_light(sock->file, fput_needed);
2099 /* We only return an error if no datagrams were able to be sent */
2106 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2107 unsigned int, vlen, unsigned int, flags)
2109 if (flags & MSG_CMSG_COMPAT)
2111 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2114 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2115 struct msghdr *msg_sys, unsigned flags, int nosec)
2117 struct compat_msghdr __user *msg_compat =
2118 (struct compat_msghdr __user *)msg;
2119 struct iovec iovstack[UIO_FASTIOV];
2120 struct iovec *iov = iovstack;
2121 unsigned long cmsg_ptr;
2122 int err, iov_size, total_len, len;
2124 /* kernel mode address */
2125 struct sockaddr_storage addr;
2127 /* user mode address pointers */
2128 struct sockaddr __user *uaddr;
2129 int __user *uaddr_len;
2131 if (MSG_CMSG_COMPAT & flags)
2132 err = get_compat_msghdr(msg_sys, msg_compat);
2134 err = copy_msghdr_from_user(msg_sys, msg);
2139 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2142 /* Check whether to allocate the iovec area */
2144 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2145 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2146 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2151 /* Save the user-mode address (verify_iovec will change the
2152 * kernel msghdr to use the kernel address space)
2154 uaddr = (__force void __user *)msg_sys->msg_name;
2155 uaddr_len = COMPAT_NAMELEN(msg);
2156 if (MSG_CMSG_COMPAT & flags)
2157 err = verify_compat_iovec(msg_sys, iov,
2158 (struct sockaddr *)&addr,
2161 err = verify_iovec(msg_sys, iov,
2162 (struct sockaddr *)&addr,
2168 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2169 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2171 /* We assume all kernel code knows the size of sockaddr_storage */
2172 msg_sys->msg_namelen = 0;
2174 if (sock->file->f_flags & O_NONBLOCK)
2175 flags |= MSG_DONTWAIT;
2176 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2182 if (uaddr != NULL) {
2183 err = move_addr_to_user((struct sockaddr *)&addr,
2184 msg_sys->msg_namelen, uaddr,
2189 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2193 if (MSG_CMSG_COMPAT & flags)
2194 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2195 &msg_compat->msg_controllen);
2197 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2198 &msg->msg_controllen);
2204 if (iov != iovstack)
2205 sock_kfree_s(sock->sk, iov, iov_size);
2211 * BSD recvmsg interface
2214 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2216 int fput_needed, err;
2217 struct msghdr msg_sys;
2218 struct socket *sock;
2220 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2224 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2226 fput_light(sock->file, fput_needed);
2231 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2232 unsigned int, flags)
2234 if (flags & MSG_CMSG_COMPAT)
2236 return __sys_recvmsg(fd, msg, flags);
2240 * Linux recvmmsg interface
2243 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2244 unsigned int flags, struct timespec *timeout)
2246 int fput_needed, err, datagrams;
2247 struct socket *sock;
2248 struct mmsghdr __user *entry;
2249 struct compat_mmsghdr __user *compat_entry;
2250 struct msghdr msg_sys;
2251 struct timespec end_time;
2254 poll_select_set_timeout(&end_time, timeout->tv_sec,
2260 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2264 err = sock_error(sock->sk);
2269 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2271 while (datagrams < vlen) {
2273 * No need to ask LSM for more than the first datagram.
2275 if (MSG_CMSG_COMPAT & flags) {
2276 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2277 &msg_sys, flags & ~MSG_WAITFORONE,
2281 err = __put_user(err, &compat_entry->msg_len);
2284 err = ___sys_recvmsg(sock,
2285 (struct msghdr __user *)entry,
2286 &msg_sys, flags & ~MSG_WAITFORONE,
2290 err = put_user(err, &entry->msg_len);
2298 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2299 if (flags & MSG_WAITFORONE)
2300 flags |= MSG_DONTWAIT;
2303 ktime_get_ts(timeout);
2304 *timeout = timespec_sub(end_time, *timeout);
2305 if (timeout->tv_sec < 0) {
2306 timeout->tv_sec = timeout->tv_nsec = 0;
2310 /* Timeout, return less than vlen datagrams */
2311 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2315 /* Out of band data, return right away */
2316 if (msg_sys.msg_flags & MSG_OOB)
2323 if (datagrams == 0) {
2329 * We may return less entries than requested (vlen) if the
2330 * sock is non block and there aren't enough datagrams...
2332 if (err != -EAGAIN) {
2334 * ... or if recvmsg returns an error after we
2335 * received some datagrams, where we record the
2336 * error to return on the next call or if the
2337 * app asks about it using getsockopt(SO_ERROR).
2339 sock->sk->sk_err = -err;
2342 fput_light(sock->file, fput_needed);
2347 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2348 unsigned int, vlen, unsigned int, flags,
2349 struct timespec __user *, timeout)
2352 struct timespec timeout_sys;
2354 if (flags & MSG_CMSG_COMPAT)
2358 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2360 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2363 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2365 if (datagrams > 0 &&
2366 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2367 datagrams = -EFAULT;
2372 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2373 /* Argument list sizes for sys_socketcall */
2374 #define AL(x) ((x) * sizeof(unsigned long))
2375 static const unsigned char nargs[21] = {
2376 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2377 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2378 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2385 * System call vectors.
2387 * Argument checking cleaned up. Saved 20% in size.
2388 * This function doesn't need to set the kernel lock because
2389 * it is set by the callees.
2392 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2395 unsigned long a0, a1;
2399 if (call < 1 || call > SYS_SENDMMSG)
2403 if (len > sizeof(a))
2406 /* copy_from_user should be SMP safe. */
2407 if (copy_from_user(a, args, len))
2410 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2417 err = sys_socket(a0, a1, a[2]);
2420 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2423 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2426 err = sys_listen(a0, a1);
2429 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2430 (int __user *)a[2], 0);
2432 case SYS_GETSOCKNAME:
2434 sys_getsockname(a0, (struct sockaddr __user *)a1,
2435 (int __user *)a[2]);
2437 case SYS_GETPEERNAME:
2439 sys_getpeername(a0, (struct sockaddr __user *)a1,
2440 (int __user *)a[2]);
2442 case SYS_SOCKETPAIR:
2443 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2446 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2449 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2450 (struct sockaddr __user *)a[4], a[5]);
2453 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2456 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2457 (struct sockaddr __user *)a[4],
2458 (int __user *)a[5]);
2461 err = sys_shutdown(a0, a1);
2463 case SYS_SETSOCKOPT:
2464 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2466 case SYS_GETSOCKOPT:
2468 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2469 (int __user *)a[4]);
2472 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2475 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2478 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2481 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2482 (struct timespec __user *)a[4]);
2485 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2486 (int __user *)a[2], a[3]);
2495 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2498 * sock_register - add a socket protocol handler
2499 * @ops: description of protocol
2501 * This function is called by a protocol handler that wants to
2502 * advertise its address family, and have it linked into the
2503 * socket interface. The value ops->family coresponds to the
2504 * socket system call protocol family.
2506 int sock_register(const struct net_proto_family *ops)
2510 if (ops->family >= NPROTO) {
2511 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2516 spin_lock(&net_family_lock);
2517 if (rcu_dereference_protected(net_families[ops->family],
2518 lockdep_is_held(&net_family_lock)))
2521 rcu_assign_pointer(net_families[ops->family], ops);
2524 spin_unlock(&net_family_lock);
2526 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2529 EXPORT_SYMBOL(sock_register);
2532 * sock_unregister - remove a protocol handler
2533 * @family: protocol family to remove
2535 * This function is called by a protocol handler that wants to
2536 * remove its address family, and have it unlinked from the
2537 * new socket creation.
2539 * If protocol handler is a module, then it can use module reference
2540 * counts to protect against new references. If protocol handler is not
2541 * a module then it needs to provide its own protection in
2542 * the ops->create routine.
2544 void sock_unregister(int family)
2546 BUG_ON(family < 0 || family >= NPROTO);
2548 spin_lock(&net_family_lock);
2549 RCU_INIT_POINTER(net_families[family], NULL);
2550 spin_unlock(&net_family_lock);
2554 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2556 EXPORT_SYMBOL(sock_unregister);
2558 static int __init sock_init(void)
2563 * Initialize sock SLAB cache.
2569 * Initialize skbuff SLAB cache
2574 * Initialize the protocols module.
2579 err = register_filesystem(&sock_fs_type);
2582 sock_mnt = kern_mount(&sock_fs_type);
2583 if (IS_ERR(sock_mnt)) {
2584 err = PTR_ERR(sock_mnt);
2588 /* The real protocol initialization is performed in later initcalls.
2591 #ifdef CONFIG_NETFILTER
2595 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2596 skb_timestamping_init();
2603 unregister_filesystem(&sock_fs_type);
2608 core_initcall(sock_init); /* early initcall */
2610 #ifdef CONFIG_PROC_FS
2611 void socket_seq_show(struct seq_file *seq)
2616 for_each_possible_cpu(cpu)
2617 counter += per_cpu(sockets_in_use, cpu);
2619 /* It can be negative, by the way. 8) */
2623 seq_printf(seq, "sockets: used %d\n", counter);
2625 #endif /* CONFIG_PROC_FS */
2627 #ifdef CONFIG_COMPAT
2628 static int do_siocgstamp(struct net *net, struct socket *sock,
2629 unsigned int cmd, struct compat_timeval __user *up)
2631 mm_segment_t old_fs = get_fs();
2636 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2639 err = put_user(ktv.tv_sec, &up->tv_sec);
2640 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2645 static int do_siocgstampns(struct net *net, struct socket *sock,
2646 unsigned int cmd, struct compat_timespec __user *up)
2648 mm_segment_t old_fs = get_fs();
2649 struct timespec kts;
2653 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2656 err = put_user(kts.tv_sec, &up->tv_sec);
2657 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2662 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2664 struct ifreq __user *uifr;
2667 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2668 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2671 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2675 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2681 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2683 struct compat_ifconf ifc32;
2685 struct ifconf __user *uifc;
2686 struct compat_ifreq __user *ifr32;
2687 struct ifreq __user *ifr;
2691 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2694 memset(&ifc, 0, sizeof(ifc));
2695 if (ifc32.ifcbuf == 0) {
2699 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2701 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2702 sizeof(struct ifreq);
2703 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2705 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2706 ifr32 = compat_ptr(ifc32.ifcbuf);
2707 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2708 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2714 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2717 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2721 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2725 ifr32 = compat_ptr(ifc32.ifcbuf);
2727 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2728 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2729 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2735 if (ifc32.ifcbuf == 0) {
2736 /* Translate from 64-bit structure multiple to
2740 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2745 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2751 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2753 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2754 bool convert_in = false, convert_out = false;
2755 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2756 struct ethtool_rxnfc __user *rxnfc;
2757 struct ifreq __user *ifr;
2758 u32 rule_cnt = 0, actual_rule_cnt;
2763 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2766 compat_rxnfc = compat_ptr(data);
2768 if (get_user(ethcmd, &compat_rxnfc->cmd))
2771 /* Most ethtool structures are defined without padding.
2772 * Unfortunately struct ethtool_rxnfc is an exception.
2777 case ETHTOOL_GRXCLSRLALL:
2778 /* Buffer size is variable */
2779 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2781 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2783 buf_size += rule_cnt * sizeof(u32);
2785 case ETHTOOL_GRXRINGS:
2786 case ETHTOOL_GRXCLSRLCNT:
2787 case ETHTOOL_GRXCLSRULE:
2790 case ETHTOOL_SRXCLSRLDEL:
2791 case ETHTOOL_SRXCLSRLINS:
2792 buf_size += sizeof(struct ethtool_rxnfc);
2797 ifr = compat_alloc_user_space(buf_size);
2798 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2800 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2803 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2804 &ifr->ifr_ifru.ifru_data))
2808 /* We expect there to be holes between fs.m_ext and
2809 * fs.ring_cookie and at the end of fs, but nowhere else.
2811 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2812 sizeof(compat_rxnfc->fs.m_ext) !=
2813 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2814 sizeof(rxnfc->fs.m_ext));
2816 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2817 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2818 offsetof(struct ethtool_rxnfc, fs.location) -
2819 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2821 if (copy_in_user(rxnfc, compat_rxnfc,
2822 (void *)(&rxnfc->fs.m_ext + 1) -
2824 copy_in_user(&rxnfc->fs.ring_cookie,
2825 &compat_rxnfc->fs.ring_cookie,
2826 (void *)(&rxnfc->fs.location + 1) -
2827 (void *)&rxnfc->fs.ring_cookie) ||
2828 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2829 sizeof(rxnfc->rule_cnt)))
2833 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2838 if (copy_in_user(compat_rxnfc, rxnfc,
2839 (const void *)(&rxnfc->fs.m_ext + 1) -
2840 (const void *)rxnfc) ||
2841 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2842 &rxnfc->fs.ring_cookie,
2843 (const void *)(&rxnfc->fs.location + 1) -
2844 (const void *)&rxnfc->fs.ring_cookie) ||
2845 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2846 sizeof(rxnfc->rule_cnt)))
2849 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2850 /* As an optimisation, we only copy the actual
2851 * number of rules that the underlying
2852 * function returned. Since Mallory might
2853 * change the rule count in user memory, we
2854 * check that it is less than the rule count
2855 * originally given (as the user buffer size),
2856 * which has been range-checked.
2858 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2860 if (actual_rule_cnt < rule_cnt)
2861 rule_cnt = actual_rule_cnt;
2862 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2863 &rxnfc->rule_locs[0],
2864 rule_cnt * sizeof(u32)))
2872 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2875 compat_uptr_t uptr32;
2876 struct ifreq __user *uifr;
2878 uifr = compat_alloc_user_space(sizeof(*uifr));
2879 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2882 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2885 uptr = compat_ptr(uptr32);
2887 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2890 return dev_ioctl(net, SIOCWANDEV, uifr);
2893 static int bond_ioctl(struct net *net, unsigned int cmd,
2894 struct compat_ifreq __user *ifr32)
2897 struct ifreq __user *uifr;
2898 mm_segment_t old_fs;
2904 case SIOCBONDENSLAVE:
2905 case SIOCBONDRELEASE:
2906 case SIOCBONDSETHWADDR:
2907 case SIOCBONDCHANGEACTIVE:
2908 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2913 err = dev_ioctl(net, cmd,
2914 (struct ifreq __user __force *) &kifr);
2918 case SIOCBONDSLAVEINFOQUERY:
2919 case SIOCBONDINFOQUERY:
2920 uifr = compat_alloc_user_space(sizeof(*uifr));
2921 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2924 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2927 datap = compat_ptr(data);
2928 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2931 return dev_ioctl(net, cmd, uifr);
2937 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2938 struct compat_ifreq __user *u_ifreq32)
2940 struct ifreq __user *u_ifreq64;
2941 char tmp_buf[IFNAMSIZ];
2942 void __user *data64;
2945 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2948 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2950 data64 = compat_ptr(data32);
2952 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2954 /* Don't check these user accesses, just let that get trapped
2955 * in the ioctl handler instead.
2957 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2960 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2963 return dev_ioctl(net, cmd, u_ifreq64);
2966 static int dev_ifsioc(struct net *net, struct socket *sock,
2967 unsigned int cmd, struct compat_ifreq __user *uifr32)
2969 struct ifreq __user *uifr;
2972 uifr = compat_alloc_user_space(sizeof(*uifr));
2973 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2976 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2987 case SIOCGIFBRDADDR:
2988 case SIOCGIFDSTADDR:
2989 case SIOCGIFNETMASK:
2994 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3002 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3003 struct compat_ifreq __user *uifr32)
3006 struct compat_ifmap __user *uifmap32;
3007 mm_segment_t old_fs;
3010 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3011 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3012 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3013 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3014 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3015 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3016 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3017 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3023 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3026 if (cmd == SIOCGIFMAP && !err) {
3027 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3028 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3029 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3030 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3031 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3032 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3033 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3040 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3043 compat_uptr_t uptr32;
3044 struct ifreq __user *uifr;
3046 uifr = compat_alloc_user_space(sizeof(*uifr));
3047 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3050 if (get_user(uptr32, &uifr32->ifr_data))
3053 uptr = compat_ptr(uptr32);
3055 if (put_user(uptr, &uifr->ifr_data))
3058 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3063 struct sockaddr rt_dst; /* target address */
3064 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3065 struct sockaddr rt_genmask; /* target network mask (IP) */
3066 unsigned short rt_flags;
3069 unsigned char rt_tos;
3070 unsigned char rt_class;
3072 short rt_metric; /* +1 for binary compatibility! */
3073 /* char * */ u32 rt_dev; /* forcing the device at add */
3074 u32 rt_mtu; /* per route MTU/Window */
3075 u32 rt_window; /* Window clamping */
3076 unsigned short rt_irtt; /* Initial RTT */
3079 struct in6_rtmsg32 {
3080 struct in6_addr rtmsg_dst;
3081 struct in6_addr rtmsg_src;
3082 struct in6_addr rtmsg_gateway;
3092 static int routing_ioctl(struct net *net, struct socket *sock,
3093 unsigned int cmd, void __user *argp)
3097 struct in6_rtmsg r6;
3101 mm_segment_t old_fs = get_fs();
3103 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3104 struct in6_rtmsg32 __user *ur6 = argp;
3105 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3106 3 * sizeof(struct in6_addr));
3107 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3108 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3109 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3110 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3111 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3112 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3113 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3117 struct rtentry32 __user *ur4 = argp;
3118 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3119 3 * sizeof(struct sockaddr));
3120 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3121 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3122 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3123 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3124 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3125 ret |= __get_user(rtdev, &(ur4->rt_dev));
3127 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3128 r4.rt_dev = (char __user __force *)devname;
3142 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3149 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3150 * for some operations; this forces use of the newer bridge-utils that
3151 * use compatible ioctls
3153 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3157 if (get_user(tmp, argp))
3159 if (tmp == BRCTL_GET_VERSION)
3160 return BRCTL_VERSION + 1;
3164 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3165 unsigned int cmd, unsigned long arg)
3167 void __user *argp = compat_ptr(arg);
3168 struct sock *sk = sock->sk;
3169 struct net *net = sock_net(sk);
3171 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3172 return siocdevprivate_ioctl(net, cmd, argp);
3177 return old_bridge_ioctl(argp);
3179 return dev_ifname32(net, argp);
3181 return dev_ifconf(net, argp);
3183 return ethtool_ioctl(net, argp);
3185 return compat_siocwandev(net, argp);
3188 return compat_sioc_ifmap(net, cmd, argp);
3189 case SIOCBONDENSLAVE:
3190 case SIOCBONDRELEASE:
3191 case SIOCBONDSETHWADDR:
3192 case SIOCBONDSLAVEINFOQUERY:
3193 case SIOCBONDINFOQUERY:
3194 case SIOCBONDCHANGEACTIVE:
3195 return bond_ioctl(net, cmd, argp);
3198 return routing_ioctl(net, sock, cmd, argp);
3200 return do_siocgstamp(net, sock, cmd, argp);
3202 return do_siocgstampns(net, sock, cmd, argp);
3204 return compat_siocshwtstamp(net, argp);
3216 return sock_ioctl(file, cmd, arg);
3233 case SIOCSIFHWBROADCAST:
3235 case SIOCGIFBRDADDR:
3236 case SIOCSIFBRDADDR:
3237 case SIOCGIFDSTADDR:
3238 case SIOCSIFDSTADDR:
3239 case SIOCGIFNETMASK:
3240 case SIOCSIFNETMASK:
3251 return dev_ifsioc(net, sock, cmd, argp);
3257 return sock_do_ioctl(net, sock, cmd, arg);
3260 /* Prevent warning from compat_sys_ioctl, these always
3261 * result in -EINVAL in the native case anyway. */
3274 return -ENOIOCTLCMD;
3277 static long compat_sock_ioctl(struct file *file, unsigned cmd,
3280 struct socket *sock = file->private_data;
3281 int ret = -ENOIOCTLCMD;
3288 if (sock->ops->compat_ioctl)
3289 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3291 if (ret == -ENOIOCTLCMD &&
3292 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3293 ret = compat_wext_handle_ioctl(net, cmd, arg);
3295 if (ret == -ENOIOCTLCMD)
3296 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3302 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3304 return sock->ops->bind(sock, addr, addrlen);
3306 EXPORT_SYMBOL(kernel_bind);
3308 int kernel_listen(struct socket *sock, int backlog)
3310 return sock->ops->listen(sock, backlog);
3312 EXPORT_SYMBOL(kernel_listen);
3314 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3316 struct sock *sk = sock->sk;
3319 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3324 err = sock->ops->accept(sock, *newsock, flags);
3326 sock_release(*newsock);
3331 (*newsock)->ops = sock->ops;
3332 __module_get((*newsock)->ops->owner);
3337 EXPORT_SYMBOL(kernel_accept);
3339 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3342 return sock->ops->connect(sock, addr, addrlen, flags);
3344 EXPORT_SYMBOL(kernel_connect);
3346 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3349 return sock->ops->getname(sock, addr, addrlen, 0);
3351 EXPORT_SYMBOL(kernel_getsockname);
3353 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3356 return sock->ops->getname(sock, addr, addrlen, 1);
3358 EXPORT_SYMBOL(kernel_getpeername);
3360 int kernel_getsockopt(struct socket *sock, int level, int optname,
3361 char *optval, int *optlen)
3363 mm_segment_t oldfs = get_fs();
3364 char __user *uoptval;
3365 int __user *uoptlen;
3368 uoptval = (char __user __force *) optval;
3369 uoptlen = (int __user __force *) optlen;
3372 if (level == SOL_SOCKET)
3373 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3375 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3380 EXPORT_SYMBOL(kernel_getsockopt);
3382 int kernel_setsockopt(struct socket *sock, int level, int optname,
3383 char *optval, unsigned int optlen)
3385 mm_segment_t oldfs = get_fs();
3386 char __user *uoptval;
3389 uoptval = (char __user __force *) optval;
3392 if (level == SOL_SOCKET)
3393 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3395 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3400 EXPORT_SYMBOL(kernel_setsockopt);
3402 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3403 size_t size, int flags)
3405 sock_update_classid(sock->sk);
3407 if (sock->ops->sendpage)
3408 return sock->ops->sendpage(sock, page, offset, size, flags);
3410 return sock_no_sendpage(sock, page, offset, size, flags);
3412 EXPORT_SYMBOL(kernel_sendpage);
3414 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3416 mm_segment_t oldfs = get_fs();
3420 err = sock->ops->ioctl(sock, cmd, arg);
3425 EXPORT_SYMBOL(kernel_sock_ioctl);
3427 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3429 return sock->ops->shutdown(sock, how);
3431 EXPORT_SYMBOL(kernel_sock_shutdown);