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_storage *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_storage *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
218 err = get_user(len, ulen);
223 if (len < 0 || len > sizeof(struct sockaddr_storage))
226 if (audit_sockaddr(klen, kaddr))
228 if (copy_to_user(uaddr, kaddr, len))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen, ulen);
238 static struct kmem_cache *sock_inode_cachep __read_mostly;
240 static struct inode *sock_alloc_inode(struct super_block *sb)
242 struct socket_alloc *ei;
243 struct socket_wq *wq;
245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
250 kmem_cache_free(sock_inode_cachep, ei);
253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 struct socket_alloc *ei;
269 struct socket_wq *wq;
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
272 wq = rcu_dereference_protected(ei->socket.wq, 1);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static int init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
293 if (sock_inode_cachep == NULL)
298 static const struct super_operations sockfs_ops = {
299 .alloc_inode = sock_alloc_inode,
300 .destroy_inode = sock_destroy_inode,
301 .statfs = simple_statfs,
305 * sockfs_dname() is called from d_path().
307 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
309 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310 dentry->d_inode->i_ino);
313 static const struct dentry_operations sockfs_dentry_operations = {
314 .d_dname = sockfs_dname,
317 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318 int flags, const char *dev_name, void *data)
320 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321 &sockfs_dentry_operations, SOCKFS_MAGIC);
324 static struct vfsmount *sock_mnt __read_mostly;
326 static struct file_system_type sock_fs_type = {
328 .mount = sockfs_mount,
329 .kill_sb = kill_anon_super,
333 * Obtains the first available file descriptor and sets it up for use.
335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
349 struct file *sock_alloc_file(struct socket *sock, int flags)
351 struct qstr name = { .name = "" };
355 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
356 if (unlikely(!path.dentry))
357 return ERR_PTR(-ENOMEM);
358 path.mnt = mntget(sock_mnt);
360 d_instantiate(path.dentry, SOCK_INODE(sock));
361 SOCK_INODE(sock)->i_fop = &socket_file_ops;
363 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
365 if (unlikely(!file)) {
366 /* drop dentry, keep inode */
367 ihold(path.dentry->d_inode);
369 return ERR_PTR(-ENFILE);
373 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
375 file->private_data = sock;
378 EXPORT_SYMBOL(sock_alloc_file);
380 static int sock_map_fd(struct socket *sock, int flags)
382 struct file *newfile;
383 int fd = get_unused_fd_flags(flags);
384 if (unlikely(fd < 0))
387 newfile = sock_alloc_file(sock, flags);
388 if (likely(!IS_ERR(newfile))) {
389 fd_install(fd, newfile);
394 return PTR_ERR(newfile);
397 struct socket *sock_from_file(struct file *file, int *err)
399 if (file->f_op == &socket_file_ops)
400 return file->private_data; /* set in sock_map_fd */
405 EXPORT_SYMBOL(sock_from_file);
408 * sockfd_lookup - Go from a file number to its socket slot
410 * @err: pointer to an error code return
412 * The file handle passed in is locked and the socket it is bound
413 * too is returned. If an error occurs the err pointer is overwritten
414 * with a negative errno code and NULL is returned. The function checks
415 * for both invalid handles and passing a handle which is not a socket.
417 * On a success the socket object pointer is returned.
420 struct socket *sockfd_lookup(int fd, int *err)
431 sock = sock_from_file(file, err);
436 EXPORT_SYMBOL(sockfd_lookup);
438 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
444 file = fget_light(fd, fput_needed);
446 sock = sock_from_file(file, err);
449 fput_light(file, *fput_needed);
455 * sock_alloc - allocate a socket
457 * Allocate a new inode and socket object. The two are bound together
458 * and initialised. The socket is then returned. If we are out of inodes
462 static struct socket *sock_alloc(void)
467 inode = new_inode_pseudo(sock_mnt->mnt_sb);
471 sock = SOCKET_I(inode);
473 kmemcheck_annotate_bitfield(sock, type);
474 inode->i_ino = get_next_ino();
475 inode->i_mode = S_IFSOCK | S_IRWXUGO;
476 inode->i_uid = current_fsuid();
477 inode->i_gid = current_fsgid();
479 this_cpu_add(sockets_in_use, 1);
484 * In theory you can't get an open on this inode, but /proc provides
485 * a back door. Remember to keep it shut otherwise you'll let the
486 * creepy crawlies in.
489 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
494 const struct file_operations bad_sock_fops = {
495 .owner = THIS_MODULE,
496 .open = sock_no_open,
497 .llseek = noop_llseek,
501 * sock_release - close a socket
502 * @sock: socket to close
504 * The socket is released from the protocol stack if it has a release
505 * callback, and the inode is then released if the socket is bound to
506 * an inode not a file.
509 void sock_release(struct socket *sock)
512 struct module *owner = sock->ops->owner;
514 sock->ops->release(sock);
519 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
520 printk(KERN_ERR "sock_release: fasync list not empty!\n");
522 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
525 this_cpu_sub(sockets_in_use, 1);
527 iput(SOCK_INODE(sock));
532 EXPORT_SYMBOL(sock_release);
534 int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
537 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
538 *tx_flags |= SKBTX_HW_TSTAMP;
539 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
540 *tx_flags |= SKBTX_SW_TSTAMP;
541 if (sock_flag(sk, SOCK_WIFI_STATUS))
542 *tx_flags |= SKBTX_WIFI_STATUS;
545 EXPORT_SYMBOL(sock_tx_timestamp);
547 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
548 struct msghdr *msg, size_t size)
550 struct sock_iocb *si = kiocb_to_siocb(iocb);
552 sock_update_classid(sock->sk);
559 return sock->ops->sendmsg(iocb, sock, msg, size);
562 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
563 struct msghdr *msg, size_t size)
565 int err = security_socket_sendmsg(sock, msg, size);
567 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
570 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
573 struct sock_iocb siocb;
576 init_sync_kiocb(&iocb, NULL);
577 iocb.private = &siocb;
578 ret = __sock_sendmsg(&iocb, sock, msg, size);
579 if (-EIOCBQUEUED == ret)
580 ret = wait_on_sync_kiocb(&iocb);
583 EXPORT_SYMBOL(sock_sendmsg);
585 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
588 struct sock_iocb siocb;
591 init_sync_kiocb(&iocb, NULL);
592 iocb.private = &siocb;
593 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
594 if (-EIOCBQUEUED == ret)
595 ret = wait_on_sync_kiocb(&iocb);
599 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
600 struct kvec *vec, size_t num, size_t size)
602 mm_segment_t oldfs = get_fs();
607 * the following is safe, since for compiler definitions of kvec and
608 * iovec are identical, yielding the same in-core layout and alignment
610 msg->msg_iov = (struct iovec *)vec;
611 msg->msg_iovlen = num;
612 result = sock_sendmsg(sock, msg, size);
616 EXPORT_SYMBOL(kernel_sendmsg);
618 static int ktime2ts(ktime_t kt, struct timespec *ts)
621 *ts = ktime_to_timespec(kt);
629 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
631 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
634 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
635 struct timespec ts[3];
637 struct skb_shared_hwtstamps *shhwtstamps =
640 /* Race occurred between timestamp enabling and packet
641 receiving. Fill in the current time for now. */
642 if (need_software_tstamp && skb->tstamp.tv64 == 0)
643 __net_timestamp(skb);
645 if (need_software_tstamp) {
646 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
648 skb_get_timestamp(skb, &tv);
649 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
652 skb_get_timestampns(skb, &ts[0]);
653 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
654 sizeof(ts[0]), &ts[0]);
659 memset(ts, 0, sizeof(ts));
660 if (skb->tstamp.tv64 &&
661 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
662 skb_get_timestampns(skb, ts + 0);
666 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
667 ktime2ts(shhwtstamps->syststamp, ts + 1))
669 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
670 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
674 put_cmsg(msg, SOL_SOCKET,
675 SCM_TIMESTAMPING, sizeof(ts), &ts);
677 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
679 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
684 if (!sock_flag(sk, SOCK_WIFI_STATUS))
686 if (!skb->wifi_acked_valid)
689 ack = skb->wifi_acked;
691 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
693 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
695 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
698 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
699 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
700 sizeof(__u32), &skb->dropcount);
703 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
706 sock_recv_timestamp(msg, sk, skb);
707 sock_recv_drops(msg, sk, skb);
709 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
711 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
712 struct msghdr *msg, size_t size, int flags)
714 struct sock_iocb *si = kiocb_to_siocb(iocb);
716 sock_update_classid(sock->sk);
724 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
727 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
728 struct msghdr *msg, size_t size, int flags)
730 int err = security_socket_recvmsg(sock, msg, size, flags);
732 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
735 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
736 size_t size, int flags)
739 struct sock_iocb siocb;
742 init_sync_kiocb(&iocb, NULL);
743 iocb.private = &siocb;
744 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
745 if (-EIOCBQUEUED == ret)
746 ret = wait_on_sync_kiocb(&iocb);
749 EXPORT_SYMBOL(sock_recvmsg);
751 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
752 size_t size, int flags)
755 struct sock_iocb siocb;
758 init_sync_kiocb(&iocb, NULL);
759 iocb.private = &siocb;
760 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
761 if (-EIOCBQUEUED == ret)
762 ret = wait_on_sync_kiocb(&iocb);
767 * kernel_recvmsg - Receive a message from a socket (kernel space)
768 * @sock: The socket to receive the message from
769 * @msg: Received message
770 * @vec: Input s/g array for message data
771 * @num: Size of input s/g array
772 * @size: Number of bytes to read
773 * @flags: Message flags (MSG_DONTWAIT, etc...)
775 * On return the msg structure contains the scatter/gather array passed in the
776 * vec argument. The array is modified so that it consists of the unfilled
777 * portion of the original array.
779 * The returned value is the total number of bytes received, or an error.
781 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
782 struct kvec *vec, size_t num, size_t size, int flags)
784 mm_segment_t oldfs = get_fs();
789 * the following is safe, since for compiler definitions of kvec and
790 * iovec are identical, yielding the same in-core layout and alignment
792 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
793 result = sock_recvmsg(sock, msg, size, flags);
797 EXPORT_SYMBOL(kernel_recvmsg);
799 static void sock_aio_dtor(struct kiocb *iocb)
801 kfree(iocb->private);
804 static ssize_t sock_sendpage(struct file *file, struct page *page,
805 int offset, size_t size, loff_t *ppos, int more)
810 sock = file->private_data;
812 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
813 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
816 return kernel_sendpage(sock, page, offset, size, flags);
819 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
820 struct pipe_inode_info *pipe, size_t len,
823 struct socket *sock = file->private_data;
825 if (unlikely(!sock->ops->splice_read))
828 sock_update_classid(sock->sk);
830 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
833 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
834 struct sock_iocb *siocb)
836 if (!is_sync_kiocb(iocb)) {
837 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
840 iocb->ki_dtor = sock_aio_dtor;
844 iocb->private = siocb;
848 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
849 struct file *file, const struct iovec *iov,
850 unsigned long nr_segs)
852 struct socket *sock = file->private_data;
856 for (i = 0; i < nr_segs; i++)
857 size += iov[i].iov_len;
859 msg->msg_name = NULL;
860 msg->msg_namelen = 0;
861 msg->msg_control = NULL;
862 msg->msg_controllen = 0;
863 msg->msg_iov = (struct iovec *)iov;
864 msg->msg_iovlen = nr_segs;
865 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
867 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
870 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
871 unsigned long nr_segs, loff_t pos)
873 struct sock_iocb siocb, *x;
878 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
882 x = alloc_sock_iocb(iocb, &siocb);
885 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
888 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
889 struct file *file, const struct iovec *iov,
890 unsigned long nr_segs)
892 struct socket *sock = file->private_data;
896 for (i = 0; i < nr_segs; i++)
897 size += iov[i].iov_len;
899 msg->msg_name = NULL;
900 msg->msg_namelen = 0;
901 msg->msg_control = NULL;
902 msg->msg_controllen = 0;
903 msg->msg_iov = (struct iovec *)iov;
904 msg->msg_iovlen = nr_segs;
905 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
906 if (sock->type == SOCK_SEQPACKET)
907 msg->msg_flags |= MSG_EOR;
909 return __sock_sendmsg(iocb, sock, msg, size);
912 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
913 unsigned long nr_segs, loff_t pos)
915 struct sock_iocb siocb, *x;
920 x = alloc_sock_iocb(iocb, &siocb);
924 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
928 * Atomic setting of ioctl hooks to avoid race
929 * with module unload.
932 static DEFINE_MUTEX(br_ioctl_mutex);
933 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
935 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
937 mutex_lock(&br_ioctl_mutex);
938 br_ioctl_hook = hook;
939 mutex_unlock(&br_ioctl_mutex);
941 EXPORT_SYMBOL(brioctl_set);
943 static DEFINE_MUTEX(vlan_ioctl_mutex);
944 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
946 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
948 mutex_lock(&vlan_ioctl_mutex);
949 vlan_ioctl_hook = hook;
950 mutex_unlock(&vlan_ioctl_mutex);
952 EXPORT_SYMBOL(vlan_ioctl_set);
954 static DEFINE_MUTEX(dlci_ioctl_mutex);
955 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
957 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
959 mutex_lock(&dlci_ioctl_mutex);
960 dlci_ioctl_hook = hook;
961 mutex_unlock(&dlci_ioctl_mutex);
963 EXPORT_SYMBOL(dlci_ioctl_set);
965 static long sock_do_ioctl(struct net *net, struct socket *sock,
966 unsigned int cmd, unsigned long arg)
969 void __user *argp = (void __user *)arg;
971 err = sock->ops->ioctl(sock, cmd, arg);
974 * If this ioctl is unknown try to hand it down
977 if (err == -ENOIOCTLCMD)
978 err = dev_ioctl(net, cmd, argp);
984 * With an ioctl, arg may well be a user mode pointer, but we don't know
985 * what to do with it - that's up to the protocol still.
988 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
992 void __user *argp = (void __user *)arg;
996 sock = file->private_data;
999 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1000 err = dev_ioctl(net, cmd, argp);
1002 #ifdef CONFIG_WEXT_CORE
1003 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1004 err = dev_ioctl(net, cmd, argp);
1011 if (get_user(pid, (int __user *)argp))
1013 err = f_setown(sock->file, pid, 1);
1017 err = put_user(f_getown(sock->file),
1018 (int __user *)argp);
1026 request_module("bridge");
1028 mutex_lock(&br_ioctl_mutex);
1030 err = br_ioctl_hook(net, cmd, argp);
1031 mutex_unlock(&br_ioctl_mutex);
1036 if (!vlan_ioctl_hook)
1037 request_module("8021q");
1039 mutex_lock(&vlan_ioctl_mutex);
1040 if (vlan_ioctl_hook)
1041 err = vlan_ioctl_hook(net, argp);
1042 mutex_unlock(&vlan_ioctl_mutex);
1047 if (!dlci_ioctl_hook)
1048 request_module("dlci");
1050 mutex_lock(&dlci_ioctl_mutex);
1051 if (dlci_ioctl_hook)
1052 err = dlci_ioctl_hook(cmd, argp);
1053 mutex_unlock(&dlci_ioctl_mutex);
1056 err = sock_do_ioctl(net, sock, cmd, arg);
1062 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1065 struct socket *sock = NULL;
1067 err = security_socket_create(family, type, protocol, 1);
1071 sock = sock_alloc();
1078 err = security_socket_post_create(sock, family, type, protocol, 1);
1090 EXPORT_SYMBOL(sock_create_lite);
1092 /* No kernel lock held - perfect */
1093 static unsigned int sock_poll(struct file *file, poll_table *wait)
1095 struct socket *sock;
1098 * We can't return errors to poll, so it's either yes or no.
1100 sock = file->private_data;
1101 return sock->ops->poll(file, sock, wait);
1104 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1106 struct socket *sock = file->private_data;
1108 return sock->ops->mmap(file, sock, vma);
1111 static int sock_close(struct inode *inode, struct file *filp)
1114 * It was possible the inode is NULL we were
1115 * closing an unfinished socket.
1119 printk(KERN_DEBUG "sock_close: NULL inode\n");
1122 sock_release(SOCKET_I(inode));
1127 * Update the socket async list
1129 * Fasync_list locking strategy.
1131 * 1. fasync_list is modified only under process context socket lock
1132 * i.e. under semaphore.
1133 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1134 * or under socket lock
1137 static int sock_fasync(int fd, struct file *filp, int on)
1139 struct socket *sock = filp->private_data;
1140 struct sock *sk = sock->sk;
1141 struct socket_wq *wq;
1147 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1148 fasync_helper(fd, filp, on, &wq->fasync_list);
1150 if (!wq->fasync_list)
1151 sock_reset_flag(sk, SOCK_FASYNC);
1153 sock_set_flag(sk, SOCK_FASYNC);
1159 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1161 int sock_wake_async(struct socket *sock, int how, int band)
1163 struct socket_wq *wq;
1168 wq = rcu_dereference(sock->wq);
1169 if (!wq || !wq->fasync_list) {
1174 case SOCK_WAKE_WAITD:
1175 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1178 case SOCK_WAKE_SPACE:
1179 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1184 kill_fasync(&wq->fasync_list, SIGIO, band);
1187 kill_fasync(&wq->fasync_list, SIGURG, band);
1192 EXPORT_SYMBOL(sock_wake_async);
1194 int __sock_create(struct net *net, int family, int type, int protocol,
1195 struct socket **res, int kern)
1198 struct socket *sock;
1199 const struct net_proto_family *pf;
1202 * Check protocol is in range
1204 if (family < 0 || family >= NPROTO)
1205 return -EAFNOSUPPORT;
1206 if (type < 0 || type >= SOCK_MAX)
1211 This uglymoron is moved from INET layer to here to avoid
1212 deadlock in module load.
1214 if (family == PF_INET && type == SOCK_PACKET) {
1218 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1224 err = security_socket_create(family, type, protocol, kern);
1229 * Allocate the socket and allow the family to set things up. if
1230 * the protocol is 0, the family is instructed to select an appropriate
1233 sock = sock_alloc();
1235 net_warn_ratelimited("socket: no more sockets\n");
1236 return -ENFILE; /* Not exactly a match, but its the
1237 closest posix thing */
1242 #ifdef CONFIG_MODULES
1243 /* Attempt to load a protocol module if the find failed.
1245 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1246 * requested real, full-featured networking support upon configuration.
1247 * Otherwise module support will break!
1249 if (rcu_access_pointer(net_families[family]) == NULL)
1250 request_module("net-pf-%d", family);
1254 pf = rcu_dereference(net_families[family]);
1255 err = -EAFNOSUPPORT;
1260 * We will call the ->create function, that possibly is in a loadable
1261 * module, so we have to bump that loadable module refcnt first.
1263 if (!try_module_get(pf->owner))
1266 /* Now protected by module ref count */
1269 err = pf->create(net, sock, protocol, kern);
1271 goto out_module_put;
1274 * Now to bump the refcnt of the [loadable] module that owns this
1275 * socket at sock_release time we decrement its refcnt.
1277 if (!try_module_get(sock->ops->owner))
1278 goto out_module_busy;
1281 * Now that we're done with the ->create function, the [loadable]
1282 * module can have its refcnt decremented
1284 module_put(pf->owner);
1285 err = security_socket_post_create(sock, family, type, protocol, kern);
1287 goto out_sock_release;
1293 err = -EAFNOSUPPORT;
1296 module_put(pf->owner);
1303 goto out_sock_release;
1305 EXPORT_SYMBOL(__sock_create);
1307 int sock_create(int family, int type, int protocol, struct socket **res)
1309 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1311 EXPORT_SYMBOL(sock_create);
1313 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1315 return __sock_create(&init_net, family, type, protocol, res, 1);
1317 EXPORT_SYMBOL(sock_create_kern);
1319 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1322 struct socket *sock;
1325 /* Check the SOCK_* constants for consistency. */
1326 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1327 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1328 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1329 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1331 flags = type & ~SOCK_TYPE_MASK;
1332 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1334 type &= SOCK_TYPE_MASK;
1336 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1337 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1339 retval = sock_create(family, type, protocol, &sock);
1343 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1348 /* It may be already another descriptor 8) Not kernel problem. */
1357 * Create a pair of connected sockets.
1360 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1361 int __user *, usockvec)
1363 struct socket *sock1, *sock2;
1365 struct file *newfile1, *newfile2;
1368 flags = type & ~SOCK_TYPE_MASK;
1369 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1371 type &= SOCK_TYPE_MASK;
1373 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1374 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1377 * Obtain the first socket and check if the underlying protocol
1378 * supports the socketpair call.
1381 err = sock_create(family, type, protocol, &sock1);
1385 err = sock_create(family, type, protocol, &sock2);
1389 err = sock1->ops->socketpair(sock1, sock2);
1391 goto out_release_both;
1393 fd1 = get_unused_fd_flags(flags);
1394 if (unlikely(fd1 < 0)) {
1396 goto out_release_both;
1398 fd2 = get_unused_fd_flags(flags);
1399 if (unlikely(fd2 < 0)) {
1402 goto out_release_both;
1405 newfile1 = sock_alloc_file(sock1, flags);
1406 if (unlikely(IS_ERR(newfile1))) {
1407 err = PTR_ERR(newfile1);
1410 goto out_release_both;
1413 newfile2 = sock_alloc_file(sock2, flags);
1414 if (IS_ERR(newfile2)) {
1415 err = PTR_ERR(newfile2);
1419 sock_release(sock2);
1423 audit_fd_pair(fd1, fd2);
1424 fd_install(fd1, newfile1);
1425 fd_install(fd2, newfile2);
1426 /* fd1 and fd2 may be already another descriptors.
1427 * Not kernel problem.
1430 err = put_user(fd1, &usockvec[0]);
1432 err = put_user(fd2, &usockvec[1]);
1441 sock_release(sock2);
1443 sock_release(sock1);
1449 * Bind a name to a socket. Nothing much to do here since it's
1450 * the protocol's responsibility to handle the local address.
1452 * We move the socket address to kernel space before we call
1453 * the protocol layer (having also checked the address is ok).
1456 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1458 struct socket *sock;
1459 struct sockaddr_storage address;
1460 int err, fput_needed;
1462 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1464 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1466 err = security_socket_bind(sock,
1467 (struct sockaddr *)&address,
1470 err = sock->ops->bind(sock,
1474 fput_light(sock->file, fput_needed);
1480 * Perform a listen. Basically, we allow the protocol to do anything
1481 * necessary for a listen, and if that works, we mark the socket as
1482 * ready for listening.
1485 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1487 struct socket *sock;
1488 int err, fput_needed;
1491 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1493 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1494 if ((unsigned int)backlog > somaxconn)
1495 backlog = somaxconn;
1497 err = security_socket_listen(sock, backlog);
1499 err = sock->ops->listen(sock, backlog);
1501 fput_light(sock->file, fput_needed);
1507 * For accept, we attempt to create a new socket, set up the link
1508 * with the client, wake up the client, then return the new
1509 * connected fd. We collect the address of the connector in kernel
1510 * space and move it to user at the very end. This is unclean because
1511 * we open the socket then return an error.
1513 * 1003.1g adds the ability to recvmsg() to query connection pending
1514 * status to recvmsg. We need to add that support in a way thats
1515 * clean when we restucture accept also.
1518 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1519 int __user *, upeer_addrlen, int, flags)
1521 struct socket *sock, *newsock;
1522 struct file *newfile;
1523 int err, len, newfd, fput_needed;
1524 struct sockaddr_storage address;
1526 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1529 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1530 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1532 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1537 newsock = sock_alloc();
1541 newsock->type = sock->type;
1542 newsock->ops = sock->ops;
1545 * We don't need try_module_get here, as the listening socket (sock)
1546 * has the protocol module (sock->ops->owner) held.
1548 __module_get(newsock->ops->owner);
1550 newfd = get_unused_fd_flags(flags);
1551 if (unlikely(newfd < 0)) {
1553 sock_release(newsock);
1556 newfile = sock_alloc_file(newsock, flags);
1557 if (unlikely(IS_ERR(newfile))) {
1558 err = PTR_ERR(newfile);
1559 put_unused_fd(newfd);
1560 sock_release(newsock);
1564 err = security_socket_accept(sock, newsock);
1568 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1572 if (upeer_sockaddr) {
1573 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1575 err = -ECONNABORTED;
1578 err = move_addr_to_user(&address,
1579 len, upeer_sockaddr, upeer_addrlen);
1584 /* File flags are not inherited via accept() unlike another OSes. */
1586 fd_install(newfd, newfile);
1590 fput_light(sock->file, fput_needed);
1595 put_unused_fd(newfd);
1599 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1600 int __user *, upeer_addrlen)
1602 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1606 * Attempt to connect to a socket with the server address. The address
1607 * is in user space so we verify it is OK and move it to kernel space.
1609 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1612 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1613 * other SEQPACKET protocols that take time to connect() as it doesn't
1614 * include the -EINPROGRESS status for such sockets.
1617 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1620 struct socket *sock;
1621 struct sockaddr_storage address;
1622 int err, fput_needed;
1624 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1627 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1632 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1636 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1637 sock->file->f_flags);
1639 fput_light(sock->file, fput_needed);
1645 * Get the local address ('name') of a socket object. Move the obtained
1646 * name to user space.
1649 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1650 int __user *, usockaddr_len)
1652 struct socket *sock;
1653 struct sockaddr_storage address;
1654 int len, err, fput_needed;
1656 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1660 err = security_socket_getsockname(sock);
1664 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1667 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1670 fput_light(sock->file, fput_needed);
1676 * Get the remote address ('name') of a socket object. Move the obtained
1677 * name to user space.
1680 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1681 int __user *, usockaddr_len)
1683 struct socket *sock;
1684 struct sockaddr_storage address;
1685 int len, err, fput_needed;
1687 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1689 err = security_socket_getpeername(sock);
1691 fput_light(sock->file, fput_needed);
1696 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1699 err = move_addr_to_user(&address, len, usockaddr,
1701 fput_light(sock->file, fput_needed);
1707 * Send a datagram to a given address. We move the address into kernel
1708 * space and check the user space data area is readable before invoking
1712 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1713 unsigned int, flags, struct sockaddr __user *, addr,
1716 struct socket *sock;
1717 struct sockaddr_storage address;
1725 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1729 iov.iov_base = buff;
1731 msg.msg_name = NULL;
1734 msg.msg_control = NULL;
1735 msg.msg_controllen = 0;
1736 msg.msg_namelen = 0;
1738 err = move_addr_to_kernel(addr, addr_len, &address);
1741 msg.msg_name = (struct sockaddr *)&address;
1742 msg.msg_namelen = addr_len;
1744 if (sock->file->f_flags & O_NONBLOCK)
1745 flags |= MSG_DONTWAIT;
1746 msg.msg_flags = flags;
1747 err = sock_sendmsg(sock, &msg, len);
1750 fput_light(sock->file, fput_needed);
1756 * Send a datagram down a socket.
1759 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1760 unsigned int, flags)
1762 return sys_sendto(fd, buff, len, flags, NULL, 0);
1766 * Receive a frame from the socket and optionally record the address of the
1767 * sender. We verify the buffers are writable and if needed move the
1768 * sender address from kernel to user space.
1771 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1772 unsigned int, flags, struct sockaddr __user *, addr,
1773 int __user *, addr_len)
1775 struct socket *sock;
1778 struct sockaddr_storage address;
1784 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1788 msg.msg_control = NULL;
1789 msg.msg_controllen = 0;
1793 iov.iov_base = ubuf;
1794 msg.msg_name = (struct sockaddr *)&address;
1795 msg.msg_namelen = sizeof(address);
1796 if (sock->file->f_flags & O_NONBLOCK)
1797 flags |= MSG_DONTWAIT;
1798 err = sock_recvmsg(sock, &msg, size, flags);
1800 if (err >= 0 && addr != NULL) {
1801 err2 = move_addr_to_user(&address,
1802 msg.msg_namelen, addr, addr_len);
1807 fput_light(sock->file, fput_needed);
1813 * Receive a datagram from a socket.
1816 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1819 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1823 * Set a socket option. Because we don't know the option lengths we have
1824 * to pass the user mode parameter for the protocols to sort out.
1827 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1828 char __user *, optval, int, optlen)
1830 int err, fput_needed;
1831 struct socket *sock;
1836 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1838 err = security_socket_setsockopt(sock, level, optname);
1842 if (level == SOL_SOCKET)
1844 sock_setsockopt(sock, level, optname, optval,
1848 sock->ops->setsockopt(sock, level, optname, optval,
1851 fput_light(sock->file, fput_needed);
1857 * Get a socket option. Because we don't know the option lengths we have
1858 * to pass a user mode parameter for the protocols to sort out.
1861 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1862 char __user *, optval, int __user *, optlen)
1864 int err, fput_needed;
1865 struct socket *sock;
1867 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1869 err = security_socket_getsockopt(sock, level, optname);
1873 if (level == SOL_SOCKET)
1875 sock_getsockopt(sock, level, optname, optval,
1879 sock->ops->getsockopt(sock, level, optname, optval,
1882 fput_light(sock->file, fput_needed);
1888 * Shutdown a socket.
1891 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1893 int err, fput_needed;
1894 struct socket *sock;
1896 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1898 err = security_socket_shutdown(sock, how);
1900 err = sock->ops->shutdown(sock, how);
1901 fput_light(sock->file, fput_needed);
1906 /* A couple of helpful macros for getting the address of the 32/64 bit
1907 * fields which are the same type (int / unsigned) on our platforms.
1909 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1910 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1911 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1913 struct used_address {
1914 struct sockaddr_storage name;
1915 unsigned int name_len;
1918 static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1919 struct msghdr *msg_sys, unsigned int flags,
1920 struct used_address *used_address)
1922 struct compat_msghdr __user *msg_compat =
1923 (struct compat_msghdr __user *)msg;
1924 struct sockaddr_storage address;
1925 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1926 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1927 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1928 /* 20 is size of ipv6_pktinfo */
1929 unsigned char *ctl_buf = ctl;
1930 int err, ctl_len, total_len;
1933 if (MSG_CMSG_COMPAT & flags) {
1934 if (get_compat_msghdr(msg_sys, msg_compat))
1936 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1939 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1941 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1944 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
1950 /* This will also move the address data into kernel space */
1951 if (MSG_CMSG_COMPAT & flags) {
1952 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1954 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1961 if (msg_sys->msg_controllen > INT_MAX)
1963 ctl_len = msg_sys->msg_controllen;
1964 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1966 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1970 ctl_buf = msg_sys->msg_control;
1971 ctl_len = msg_sys->msg_controllen;
1972 } else if (ctl_len) {
1973 if (ctl_len > sizeof(ctl)) {
1974 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1975 if (ctl_buf == NULL)
1980 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1981 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1982 * checking falls down on this.
1984 if (copy_from_user(ctl_buf,
1985 (void __user __force *)msg_sys->msg_control,
1988 msg_sys->msg_control = ctl_buf;
1990 msg_sys->msg_flags = flags;
1992 if (sock->file->f_flags & O_NONBLOCK)
1993 msg_sys->msg_flags |= MSG_DONTWAIT;
1995 * If this is sendmmsg() and current destination address is same as
1996 * previously succeeded address, omit asking LSM's decision.
1997 * used_address->name_len is initialized to UINT_MAX so that the first
1998 * destination address never matches.
2000 if (used_address && msg_sys->msg_name &&
2001 used_address->name_len == msg_sys->msg_namelen &&
2002 !memcmp(&used_address->name, msg_sys->msg_name,
2003 used_address->name_len)) {
2004 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2007 err = sock_sendmsg(sock, msg_sys, total_len);
2009 * If this is sendmmsg() and sending to current destination address was
2010 * successful, remember it.
2012 if (used_address && err >= 0) {
2013 used_address->name_len = msg_sys->msg_namelen;
2014 if (msg_sys->msg_name)
2015 memcpy(&used_address->name, msg_sys->msg_name,
2016 used_address->name_len);
2021 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2023 if (iov != iovstack)
2030 * BSD sendmsg interface
2033 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2035 int fput_needed, err;
2036 struct msghdr msg_sys;
2037 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2042 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2044 fput_light(sock->file, fput_needed);
2050 * Linux sendmmsg interface
2053 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2056 int fput_needed, err, datagrams;
2057 struct socket *sock;
2058 struct mmsghdr __user *entry;
2059 struct compat_mmsghdr __user *compat_entry;
2060 struct msghdr msg_sys;
2061 struct used_address used_address;
2063 if (vlen > UIO_MAXIOV)
2068 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2072 used_address.name_len = UINT_MAX;
2074 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2077 while (datagrams < vlen) {
2078 if (MSG_CMSG_COMPAT & flags) {
2079 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2080 &msg_sys, flags, &used_address);
2083 err = __put_user(err, &compat_entry->msg_len);
2086 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
2087 &msg_sys, flags, &used_address);
2090 err = put_user(err, &entry->msg_len);
2099 fput_light(sock->file, fput_needed);
2101 /* We only return an error if no datagrams were able to be sent */
2108 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2109 unsigned int, vlen, unsigned int, flags)
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 int 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, 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 if (get_compat_msghdr(msg_sys, msg_compat))
2134 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2137 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2139 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2142 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2149 * Save the user-mode address (verify_iovec will change the
2150 * kernel msghdr to use the kernel address space)
2153 uaddr = (__force void __user *)msg_sys->msg_name;
2154 uaddr_len = COMPAT_NAMELEN(msg);
2155 if (MSG_CMSG_COMPAT & flags) {
2156 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2158 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2163 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2164 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2166 if (sock->file->f_flags & O_NONBLOCK)
2167 flags |= MSG_DONTWAIT;
2168 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2174 if (uaddr != NULL) {
2175 err = move_addr_to_user(&addr,
2176 msg_sys->msg_namelen, uaddr,
2181 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2185 if (MSG_CMSG_COMPAT & flags)
2186 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2187 &msg_compat->msg_controllen);
2189 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2190 &msg->msg_controllen);
2196 if (iov != iovstack)
2203 * BSD recvmsg interface
2206 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2207 unsigned int, flags)
2209 int fput_needed, err;
2210 struct msghdr msg_sys;
2211 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2216 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2218 fput_light(sock->file, fput_needed);
2224 * Linux recvmmsg interface
2227 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2228 unsigned int flags, struct timespec *timeout)
2230 int fput_needed, err, datagrams;
2231 struct socket *sock;
2232 struct mmsghdr __user *entry;
2233 struct compat_mmsghdr __user *compat_entry;
2234 struct msghdr msg_sys;
2235 struct timespec end_time;
2238 poll_select_set_timeout(&end_time, timeout->tv_sec,
2244 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2248 err = sock_error(sock->sk);
2253 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2255 while (datagrams < vlen) {
2257 * No need to ask LSM for more than the first datagram.
2259 if (MSG_CMSG_COMPAT & flags) {
2260 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2261 &msg_sys, flags & ~MSG_WAITFORONE,
2265 err = __put_user(err, &compat_entry->msg_len);
2268 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2269 &msg_sys, flags & ~MSG_WAITFORONE,
2273 err = put_user(err, &entry->msg_len);
2281 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2282 if (flags & MSG_WAITFORONE)
2283 flags |= MSG_DONTWAIT;
2286 ktime_get_ts(timeout);
2287 *timeout = timespec_sub(end_time, *timeout);
2288 if (timeout->tv_sec < 0) {
2289 timeout->tv_sec = timeout->tv_nsec = 0;
2293 /* Timeout, return less than vlen datagrams */
2294 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2298 /* Out of band data, return right away */
2299 if (msg_sys.msg_flags & MSG_OOB)
2304 fput_light(sock->file, fput_needed);
2309 if (datagrams != 0) {
2311 * We may return less entries than requested (vlen) if the
2312 * sock is non block and there aren't enough datagrams...
2314 if (err != -EAGAIN) {
2316 * ... or if recvmsg returns an error after we
2317 * received some datagrams, where we record the
2318 * error to return on the next call or if the
2319 * app asks about it using getsockopt(SO_ERROR).
2321 sock->sk->sk_err = -err;
2330 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2331 unsigned int, vlen, unsigned int, flags,
2332 struct timespec __user *, timeout)
2335 struct timespec timeout_sys;
2338 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2340 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2343 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2345 if (datagrams > 0 &&
2346 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2347 datagrams = -EFAULT;
2352 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2353 /* Argument list sizes for sys_socketcall */
2354 #define AL(x) ((x) * sizeof(unsigned long))
2355 static const unsigned char nargs[21] = {
2356 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2357 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2358 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2365 * System call vectors.
2367 * Argument checking cleaned up. Saved 20% in size.
2368 * This function doesn't need to set the kernel lock because
2369 * it is set by the callees.
2372 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2375 unsigned long a0, a1;
2379 if (call < 1 || call > SYS_SENDMMSG)
2383 if (len > sizeof(a))
2386 /* copy_from_user should be SMP safe. */
2387 if (copy_from_user(a, args, len))
2390 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2397 err = sys_socket(a0, a1, a[2]);
2400 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2403 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2406 err = sys_listen(a0, a1);
2409 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2410 (int __user *)a[2], 0);
2412 case SYS_GETSOCKNAME:
2414 sys_getsockname(a0, (struct sockaddr __user *)a1,
2415 (int __user *)a[2]);
2417 case SYS_GETPEERNAME:
2419 sys_getpeername(a0, (struct sockaddr __user *)a1,
2420 (int __user *)a[2]);
2422 case SYS_SOCKETPAIR:
2423 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2426 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2429 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2430 (struct sockaddr __user *)a[4], a[5]);
2433 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2436 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2437 (struct sockaddr __user *)a[4],
2438 (int __user *)a[5]);
2441 err = sys_shutdown(a0, a1);
2443 case SYS_SETSOCKOPT:
2444 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2446 case SYS_GETSOCKOPT:
2448 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2449 (int __user *)a[4]);
2452 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2455 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2458 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2461 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2462 (struct timespec __user *)a[4]);
2465 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2466 (int __user *)a[2], a[3]);
2475 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2478 * sock_register - add a socket protocol handler
2479 * @ops: description of protocol
2481 * This function is called by a protocol handler that wants to
2482 * advertise its address family, and have it linked into the
2483 * socket interface. The value ops->family coresponds to the
2484 * socket system call protocol family.
2486 int sock_register(const struct net_proto_family *ops)
2490 if (ops->family >= NPROTO) {
2491 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2496 spin_lock(&net_family_lock);
2497 if (rcu_dereference_protected(net_families[ops->family],
2498 lockdep_is_held(&net_family_lock)))
2501 rcu_assign_pointer(net_families[ops->family], ops);
2504 spin_unlock(&net_family_lock);
2506 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2509 EXPORT_SYMBOL(sock_register);
2512 * sock_unregister - remove a protocol handler
2513 * @family: protocol family to remove
2515 * This function is called by a protocol handler that wants to
2516 * remove its address family, and have it unlinked from the
2517 * new socket creation.
2519 * If protocol handler is a module, then it can use module reference
2520 * counts to protect against new references. If protocol handler is not
2521 * a module then it needs to provide its own protection in
2522 * the ops->create routine.
2524 void sock_unregister(int family)
2526 BUG_ON(family < 0 || family >= NPROTO);
2528 spin_lock(&net_family_lock);
2529 RCU_INIT_POINTER(net_families[family], NULL);
2530 spin_unlock(&net_family_lock);
2534 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2536 EXPORT_SYMBOL(sock_unregister);
2538 static int __init sock_init(void)
2542 * Initialize the network sysctl infrastructure.
2544 err = net_sysctl_init();
2549 * Initialize sock SLAB cache.
2555 * Initialize skbuff SLAB cache
2560 * Initialize the protocols module.
2565 err = register_filesystem(&sock_fs_type);
2568 sock_mnt = kern_mount(&sock_fs_type);
2569 if (IS_ERR(sock_mnt)) {
2570 err = PTR_ERR(sock_mnt);
2574 /* The real protocol initialization is performed in later initcalls.
2577 #ifdef CONFIG_NETFILTER
2581 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2582 skb_timestamping_init();
2589 unregister_filesystem(&sock_fs_type);
2594 core_initcall(sock_init); /* early initcall */
2596 #ifdef CONFIG_PROC_FS
2597 void socket_seq_show(struct seq_file *seq)
2602 for_each_possible_cpu(cpu)
2603 counter += per_cpu(sockets_in_use, cpu);
2605 /* It can be negative, by the way. 8) */
2609 seq_printf(seq, "sockets: used %d\n", counter);
2611 #endif /* CONFIG_PROC_FS */
2613 #ifdef CONFIG_COMPAT
2614 static int do_siocgstamp(struct net *net, struct socket *sock,
2615 unsigned int cmd, void __user *up)
2617 mm_segment_t old_fs = get_fs();
2622 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2625 err = compat_put_timeval(up, &ktv);
2630 static int do_siocgstampns(struct net *net, struct socket *sock,
2631 unsigned int cmd, void __user *up)
2633 mm_segment_t old_fs = get_fs();
2634 struct timespec kts;
2638 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2641 err = compat_put_timespec(up, &kts);
2646 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2648 struct ifreq __user *uifr;
2651 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2652 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2655 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2659 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2665 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2667 struct compat_ifconf ifc32;
2669 struct ifconf __user *uifc;
2670 struct compat_ifreq __user *ifr32;
2671 struct ifreq __user *ifr;
2675 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2678 memset(&ifc, 0, sizeof(ifc));
2679 if (ifc32.ifcbuf == 0) {
2683 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2685 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2686 sizeof(struct ifreq);
2687 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2689 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2690 ifr32 = compat_ptr(ifc32.ifcbuf);
2691 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2692 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2698 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2701 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2705 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2709 ifr32 = compat_ptr(ifc32.ifcbuf);
2711 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2712 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2713 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2719 if (ifc32.ifcbuf == 0) {
2720 /* Translate from 64-bit structure multiple to
2724 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2729 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2735 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2737 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2738 bool convert_in = false, convert_out = false;
2739 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2740 struct ethtool_rxnfc __user *rxnfc;
2741 struct ifreq __user *ifr;
2742 u32 rule_cnt = 0, actual_rule_cnt;
2747 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2750 compat_rxnfc = compat_ptr(data);
2752 if (get_user(ethcmd, &compat_rxnfc->cmd))
2755 /* Most ethtool structures are defined without padding.
2756 * Unfortunately struct ethtool_rxnfc is an exception.
2761 case ETHTOOL_GRXCLSRLALL:
2762 /* Buffer size is variable */
2763 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2765 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2767 buf_size += rule_cnt * sizeof(u32);
2769 case ETHTOOL_GRXRINGS:
2770 case ETHTOOL_GRXCLSRLCNT:
2771 case ETHTOOL_GRXCLSRULE:
2772 case ETHTOOL_SRXCLSRLINS:
2775 case ETHTOOL_SRXCLSRLDEL:
2776 buf_size += sizeof(struct ethtool_rxnfc);
2781 ifr = compat_alloc_user_space(buf_size);
2782 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2784 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2787 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2788 &ifr->ifr_ifru.ifru_data))
2792 /* We expect there to be holes between fs.m_ext and
2793 * fs.ring_cookie and at the end of fs, but nowhere else.
2795 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2796 sizeof(compat_rxnfc->fs.m_ext) !=
2797 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2798 sizeof(rxnfc->fs.m_ext));
2800 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2801 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2802 offsetof(struct ethtool_rxnfc, fs.location) -
2803 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2805 if (copy_in_user(rxnfc, compat_rxnfc,
2806 (void *)(&rxnfc->fs.m_ext + 1) -
2808 copy_in_user(&rxnfc->fs.ring_cookie,
2809 &compat_rxnfc->fs.ring_cookie,
2810 (void *)(&rxnfc->fs.location + 1) -
2811 (void *)&rxnfc->fs.ring_cookie) ||
2812 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2813 sizeof(rxnfc->rule_cnt)))
2817 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2822 if (copy_in_user(compat_rxnfc, rxnfc,
2823 (const void *)(&rxnfc->fs.m_ext + 1) -
2824 (const void *)rxnfc) ||
2825 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2826 &rxnfc->fs.ring_cookie,
2827 (const void *)(&rxnfc->fs.location + 1) -
2828 (const void *)&rxnfc->fs.ring_cookie) ||
2829 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2830 sizeof(rxnfc->rule_cnt)))
2833 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2834 /* As an optimisation, we only copy the actual
2835 * number of rules that the underlying
2836 * function returned. Since Mallory might
2837 * change the rule count in user memory, we
2838 * check that it is less than the rule count
2839 * originally given (as the user buffer size),
2840 * which has been range-checked.
2842 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2844 if (actual_rule_cnt < rule_cnt)
2845 rule_cnt = actual_rule_cnt;
2846 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2847 &rxnfc->rule_locs[0],
2848 rule_cnt * sizeof(u32)))
2856 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2859 compat_uptr_t uptr32;
2860 struct ifreq __user *uifr;
2862 uifr = compat_alloc_user_space(sizeof(*uifr));
2863 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2866 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2869 uptr = compat_ptr(uptr32);
2871 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2874 return dev_ioctl(net, SIOCWANDEV, uifr);
2877 static int bond_ioctl(struct net *net, unsigned int cmd,
2878 struct compat_ifreq __user *ifr32)
2881 struct ifreq __user *uifr;
2882 mm_segment_t old_fs;
2888 case SIOCBONDENSLAVE:
2889 case SIOCBONDRELEASE:
2890 case SIOCBONDSETHWADDR:
2891 case SIOCBONDCHANGEACTIVE:
2892 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2897 err = dev_ioctl(net, cmd,
2898 (struct ifreq __user __force *) &kifr);
2902 case SIOCBONDSLAVEINFOQUERY:
2903 case SIOCBONDINFOQUERY:
2904 uifr = compat_alloc_user_space(sizeof(*uifr));
2905 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2908 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2911 datap = compat_ptr(data);
2912 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2915 return dev_ioctl(net, cmd, uifr);
2917 return -ENOIOCTLCMD;
2921 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2922 struct compat_ifreq __user *u_ifreq32)
2924 struct ifreq __user *u_ifreq64;
2925 char tmp_buf[IFNAMSIZ];
2926 void __user *data64;
2929 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2932 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2934 data64 = compat_ptr(data32);
2936 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2938 /* Don't check these user accesses, just let that get trapped
2939 * in the ioctl handler instead.
2941 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2944 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2947 return dev_ioctl(net, cmd, u_ifreq64);
2950 static int dev_ifsioc(struct net *net, struct socket *sock,
2951 unsigned int cmd, struct compat_ifreq __user *uifr32)
2953 struct ifreq __user *uifr;
2956 uifr = compat_alloc_user_space(sizeof(*uifr));
2957 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2960 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2971 case SIOCGIFBRDADDR:
2972 case SIOCGIFDSTADDR:
2973 case SIOCGIFNETMASK:
2978 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2986 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2987 struct compat_ifreq __user *uifr32)
2990 struct compat_ifmap __user *uifmap32;
2991 mm_segment_t old_fs;
2994 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2995 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2996 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2997 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2998 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2999 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3000 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3001 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3007 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3010 if (cmd == SIOCGIFMAP && !err) {
3011 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3012 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3013 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3014 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3015 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3016 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3017 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3024 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3027 compat_uptr_t uptr32;
3028 struct ifreq __user *uifr;
3030 uifr = compat_alloc_user_space(sizeof(*uifr));
3031 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3034 if (get_user(uptr32, &uifr32->ifr_data))
3037 uptr = compat_ptr(uptr32);
3039 if (put_user(uptr, &uifr->ifr_data))
3042 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3047 struct sockaddr rt_dst; /* target address */
3048 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3049 struct sockaddr rt_genmask; /* target network mask (IP) */
3050 unsigned short rt_flags;
3053 unsigned char rt_tos;
3054 unsigned char rt_class;
3056 short rt_metric; /* +1 for binary compatibility! */
3057 /* char * */ u32 rt_dev; /* forcing the device at add */
3058 u32 rt_mtu; /* per route MTU/Window */
3059 u32 rt_window; /* Window clamping */
3060 unsigned short rt_irtt; /* Initial RTT */
3063 struct in6_rtmsg32 {
3064 struct in6_addr rtmsg_dst;
3065 struct in6_addr rtmsg_src;
3066 struct in6_addr rtmsg_gateway;
3076 static int routing_ioctl(struct net *net, struct socket *sock,
3077 unsigned int cmd, void __user *argp)
3081 struct in6_rtmsg r6;
3085 mm_segment_t old_fs = get_fs();
3087 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3088 struct in6_rtmsg32 __user *ur6 = argp;
3089 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3090 3 * sizeof(struct in6_addr));
3091 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3092 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3093 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3094 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3095 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3096 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3097 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3101 struct rtentry32 __user *ur4 = argp;
3102 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3103 3 * sizeof(struct sockaddr));
3104 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3105 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3106 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3107 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3108 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3109 ret |= __get_user(rtdev, &(ur4->rt_dev));
3111 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3112 r4.rt_dev = (char __user __force *)devname;
3126 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3133 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3134 * for some operations; this forces use of the newer bridge-utils that
3135 * use compatible ioctls
3137 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3141 if (get_user(tmp, argp))
3143 if (tmp == BRCTL_GET_VERSION)
3144 return BRCTL_VERSION + 1;
3148 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3149 unsigned int cmd, unsigned long arg)
3151 void __user *argp = compat_ptr(arg);
3152 struct sock *sk = sock->sk;
3153 struct net *net = sock_net(sk);
3155 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3156 return siocdevprivate_ioctl(net, cmd, argp);
3161 return old_bridge_ioctl(argp);
3163 return dev_ifname32(net, argp);
3165 return dev_ifconf(net, argp);
3167 return ethtool_ioctl(net, argp);
3169 return compat_siocwandev(net, argp);
3172 return compat_sioc_ifmap(net, cmd, argp);
3173 case SIOCBONDENSLAVE:
3174 case SIOCBONDRELEASE:
3175 case SIOCBONDSETHWADDR:
3176 case SIOCBONDSLAVEINFOQUERY:
3177 case SIOCBONDINFOQUERY:
3178 case SIOCBONDCHANGEACTIVE:
3179 return bond_ioctl(net, cmd, argp);
3182 return routing_ioctl(net, sock, cmd, argp);
3184 return do_siocgstamp(net, sock, cmd, argp);
3186 return do_siocgstampns(net, sock, cmd, argp);
3188 return compat_siocshwtstamp(net, argp);
3200 return sock_ioctl(file, cmd, arg);
3217 case SIOCSIFHWBROADCAST:
3219 case SIOCGIFBRDADDR:
3220 case SIOCSIFBRDADDR:
3221 case SIOCGIFDSTADDR:
3222 case SIOCSIFDSTADDR:
3223 case SIOCGIFNETMASK:
3224 case SIOCSIFNETMASK:
3235 return dev_ifsioc(net, sock, cmd, argp);
3241 return sock_do_ioctl(net, sock, cmd, arg);
3244 return -ENOIOCTLCMD;
3247 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3250 struct socket *sock = file->private_data;
3251 int ret = -ENOIOCTLCMD;
3258 if (sock->ops->compat_ioctl)
3259 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3261 if (ret == -ENOIOCTLCMD &&
3262 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3263 ret = compat_wext_handle_ioctl(net, cmd, arg);
3265 if (ret == -ENOIOCTLCMD)
3266 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3272 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3274 return sock->ops->bind(sock, addr, addrlen);
3276 EXPORT_SYMBOL(kernel_bind);
3278 int kernel_listen(struct socket *sock, int backlog)
3280 return sock->ops->listen(sock, backlog);
3282 EXPORT_SYMBOL(kernel_listen);
3284 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3286 struct sock *sk = sock->sk;
3289 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3294 err = sock->ops->accept(sock, *newsock, flags);
3296 sock_release(*newsock);
3301 (*newsock)->ops = sock->ops;
3302 __module_get((*newsock)->ops->owner);
3307 EXPORT_SYMBOL(kernel_accept);
3309 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3312 return sock->ops->connect(sock, addr, addrlen, flags);
3314 EXPORT_SYMBOL(kernel_connect);
3316 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3319 return sock->ops->getname(sock, addr, addrlen, 0);
3321 EXPORT_SYMBOL(kernel_getsockname);
3323 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3326 return sock->ops->getname(sock, addr, addrlen, 1);
3328 EXPORT_SYMBOL(kernel_getpeername);
3330 int kernel_getsockopt(struct socket *sock, int level, int optname,
3331 char *optval, int *optlen)
3333 mm_segment_t oldfs = get_fs();
3334 char __user *uoptval;
3335 int __user *uoptlen;
3338 uoptval = (char __user __force *) optval;
3339 uoptlen = (int __user __force *) optlen;
3342 if (level == SOL_SOCKET)
3343 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3345 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3350 EXPORT_SYMBOL(kernel_getsockopt);
3352 int kernel_setsockopt(struct socket *sock, int level, int optname,
3353 char *optval, unsigned int optlen)
3355 mm_segment_t oldfs = get_fs();
3356 char __user *uoptval;
3359 uoptval = (char __user __force *) optval;
3362 if (level == SOL_SOCKET)
3363 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3365 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3370 EXPORT_SYMBOL(kernel_setsockopt);
3372 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3373 size_t size, int flags)
3375 sock_update_classid(sock->sk);
3377 if (sock->ops->sendpage)
3378 return sock->ops->sendpage(sock, page, offset, size, flags);
3380 return sock_no_sendpage(sock, page, offset, size, flags);
3382 EXPORT_SYMBOL(kernel_sendpage);
3384 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3386 mm_segment_t oldfs = get_fs();
3390 err = sock->ops->ioctl(sock, cmd, arg);
3395 EXPORT_SYMBOL(kernel_sock_ioctl);
3397 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3399 return sock->ops->shutdown(sock, how);
3401 EXPORT_SYMBOL(kernel_sock_shutdown);