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/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.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>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
140 static const struct file_operations socket_file_ops = {
141 .owner = THIS_MODULE,
143 .read_iter = sock_read_iter,
144 .write_iter = sock_write_iter,
146 .unlocked_ioctl = sock_ioctl,
148 .compat_ioctl = compat_sock_ioctl,
151 .release = sock_close,
152 .fasync = sock_fasync,
153 .sendpage = sock_sendpage,
154 .splice_write = generic_splice_sendpage,
155 .splice_read = sock_splice_read,
159 * The protocol list. Each protocol is registered in here.
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
166 * Statistics counters of the socket lists
169 static DEFINE_PER_CPU(int, sockets_in_use);
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
190 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
194 if (copy_from_user(kaddr, uaddr, ulen))
196 return audit_sockaddr(ulen, kaddr);
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217 void __user *uaddr, int __user *ulen)
222 BUG_ON(klen > sizeof(struct sockaddr_storage));
223 err = get_user(len, ulen);
231 if (audit_sockaddr(klen, kaddr))
233 if (copy_to_user(uaddr, kaddr, len))
237 * "fromlen shall refer to the value before truncation.."
240 return __put_user(klen, ulen);
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
245 static struct inode *sock_alloc_inode(struct super_block *sb)
247 struct socket_alloc *ei;
248 struct socket_wq *wq;
250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 kmem_cache_free(sock_inode_cachep, ei);
258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
260 RCU_INIT_POINTER(ei->socket.wq, wq);
262 ei->socket.state = SS_UNCONNECTED;
263 ei->socket.flags = 0;
264 ei->socket.ops = NULL;
265 ei->socket.sk = NULL;
266 ei->socket.file = NULL;
268 return &ei->vfs_inode;
271 static void sock_destroy_inode(struct inode *inode)
273 struct socket_alloc *ei;
274 struct socket_wq *wq;
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 wq = rcu_dereference_protected(ei->socket.wq, 1);
279 kmem_cache_free(sock_inode_cachep, ei);
282 static void init_once(void *foo)
284 struct socket_alloc *ei = (struct socket_alloc *)foo;
286 inode_init_once(&ei->vfs_inode);
289 static int init_inodecache(void)
291 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
292 sizeof(struct socket_alloc),
294 (SLAB_HWCACHE_ALIGN |
295 SLAB_RECLAIM_ACCOUNT |
298 if (sock_inode_cachep == NULL)
303 static const struct super_operations sockfs_ops = {
304 .alloc_inode = sock_alloc_inode,
305 .destroy_inode = sock_destroy_inode,
306 .statfs = simple_statfs,
310 * sockfs_dname() is called from d_path().
312 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
315 dentry->d_inode->i_ino);
318 static const struct dentry_operations sockfs_dentry_operations = {
319 .d_dname = sockfs_dname,
322 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
323 int flags, const char *dev_name, void *data)
325 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
326 &sockfs_dentry_operations, SOCKFS_MAGIC);
329 static struct vfsmount *sock_mnt __read_mostly;
331 static struct file_system_type sock_fs_type = {
333 .mount = sockfs_mount,
334 .kill_sb = kill_anon_super,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356 struct qstr name = { .name = "" };
362 name.len = strlen(name.name);
363 } else if (sock->sk) {
364 name.name = sock->sk->sk_prot_creator->name;
365 name.len = strlen(name.name);
367 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
368 if (unlikely(!path.dentry))
369 return ERR_PTR(-ENOMEM);
370 path.mnt = mntget(sock_mnt);
372 d_instantiate(path.dentry, SOCK_INODE(sock));
374 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
376 if (unlikely(IS_ERR(file))) {
377 /* drop dentry, keep inode */
378 ihold(path.dentry->d_inode);
384 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
385 file->private_data = sock;
388 EXPORT_SYMBOL(sock_alloc_file);
390 static int sock_map_fd(struct socket *sock, int flags)
392 struct file *newfile;
393 int fd = get_unused_fd_flags(flags);
394 if (unlikely(fd < 0))
397 newfile = sock_alloc_file(sock, flags, NULL);
398 if (likely(!IS_ERR(newfile))) {
399 fd_install(fd, newfile);
404 return PTR_ERR(newfile);
407 struct socket *sock_from_file(struct file *file, int *err)
409 if (file->f_op == &socket_file_ops)
410 return file->private_data; /* set in sock_map_fd */
415 EXPORT_SYMBOL(sock_from_file);
418 * sockfd_lookup - Go from a file number to its socket slot
420 * @err: pointer to an error code return
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
427 * On a success the socket object pointer is returned.
430 struct socket *sockfd_lookup(int fd, int *err)
441 sock = sock_from_file(file, err);
446 EXPORT_SYMBOL(sockfd_lookup);
448 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450 struct fd f = fdget(fd);
455 sock = sock_from_file(f.file, err);
457 *fput_needed = f.flags;
465 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
466 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
467 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
468 static ssize_t sockfs_getxattr(struct dentry *dentry,
469 const char *name, void *value, size_t size)
471 const char *proto_name;
476 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
477 proto_name = dentry->d_name.name;
478 proto_size = strlen(proto_name);
482 if (proto_size + 1 > size)
485 strncpy(value, proto_name, proto_size + 1);
487 error = proto_size + 1;
494 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
500 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
510 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
515 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
522 static const struct inode_operations sockfs_inode_ops = {
523 .getxattr = sockfs_getxattr,
524 .listxattr = sockfs_listxattr,
528 * sock_alloc - allocate a socket
530 * Allocate a new inode and socket object. The two are bound together
531 * and initialised. The socket is then returned. If we are out of inodes
535 static struct socket *sock_alloc(void)
540 inode = new_inode_pseudo(sock_mnt->mnt_sb);
544 sock = SOCKET_I(inode);
546 kmemcheck_annotate_bitfield(sock, type);
547 inode->i_ino = get_next_ino();
548 inode->i_mode = S_IFSOCK | S_IRWXUGO;
549 inode->i_uid = current_fsuid();
550 inode->i_gid = current_fsgid();
551 inode->i_op = &sockfs_inode_ops;
553 this_cpu_add(sockets_in_use, 1);
558 * sock_release - close a socket
559 * @sock: socket to close
561 * The socket is released from the protocol stack if it has a release
562 * callback, and the inode is then released if the socket is bound to
563 * an inode not a file.
566 void sock_release(struct socket *sock)
569 struct module *owner = sock->ops->owner;
571 sock->ops->release(sock);
576 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
577 pr_err("%s: fasync list not empty!\n", __func__);
579 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
582 this_cpu_sub(sockets_in_use, 1);
584 iput(SOCK_INODE(sock));
589 EXPORT_SYMBOL(sock_release);
591 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
593 u8 flags = *tx_flags;
595 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
596 flags |= SKBTX_HW_TSTAMP;
598 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
599 flags |= SKBTX_SW_TSTAMP;
601 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
602 flags |= SKBTX_SCHED_TSTAMP;
604 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
605 flags |= SKBTX_ACK_TSTAMP;
609 EXPORT_SYMBOL(__sock_tx_timestamp);
611 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg,
614 return sock->ops->sendmsg(sock, msg, size);
617 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
619 int err = security_socket_sendmsg(sock, msg, size);
621 return err ?: sock_sendmsg_nosec(sock, msg, size);
623 EXPORT_SYMBOL(sock_sendmsg);
625 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
626 struct kvec *vec, size_t num, size_t size)
628 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
629 return sock_sendmsg(sock, msg, size);
631 EXPORT_SYMBOL(kernel_sendmsg);
634 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
636 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
639 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
640 struct scm_timestamping tss;
642 struct skb_shared_hwtstamps *shhwtstamps =
645 /* Race occurred between timestamp enabling and packet
646 receiving. Fill in the current time for now. */
647 if (need_software_tstamp && skb->tstamp.tv64 == 0)
648 __net_timestamp(skb);
650 if (need_software_tstamp) {
651 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
653 skb_get_timestamp(skb, &tv);
654 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
658 skb_get_timestampns(skb, &ts);
659 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
664 memset(&tss, 0, sizeof(tss));
665 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
666 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
669 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
670 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
673 put_cmsg(msg, SOL_SOCKET,
674 SCM_TIMESTAMPING, sizeof(tss), &tss);
676 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
678 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
683 if (!sock_flag(sk, SOCK_WIFI_STATUS))
685 if (!skb->wifi_acked_valid)
688 ack = skb->wifi_acked;
690 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
692 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
694 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
697 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
698 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
699 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
702 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
705 sock_recv_timestamp(msg, sk, skb);
706 sock_recv_drops(msg, sk, skb);
708 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
710 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
711 size_t size, int flags)
713 return sock->ops->recvmsg(sock, msg, size, flags);
716 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
719 int err = security_socket_recvmsg(sock, msg, size, flags);
721 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
723 EXPORT_SYMBOL(sock_recvmsg);
726 * kernel_recvmsg - Receive a message from a socket (kernel space)
727 * @sock: The socket to receive the message from
728 * @msg: Received message
729 * @vec: Input s/g array for message data
730 * @num: Size of input s/g array
731 * @size: Number of bytes to read
732 * @flags: Message flags (MSG_DONTWAIT, etc...)
734 * On return the msg structure contains the scatter/gather array passed in the
735 * vec argument. The array is modified so that it consists of the unfilled
736 * portion of the original array.
738 * The returned value is the total number of bytes received, or an error.
740 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
741 struct kvec *vec, size_t num, size_t size, int flags)
743 mm_segment_t oldfs = get_fs();
746 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
748 result = sock_recvmsg(sock, msg, size, flags);
752 EXPORT_SYMBOL(kernel_recvmsg);
754 static ssize_t sock_sendpage(struct file *file, struct page *page,
755 int offset, size_t size, loff_t *ppos, int more)
760 sock = file->private_data;
762 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
763 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
766 return kernel_sendpage(sock, page, offset, size, flags);
769 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
770 struct pipe_inode_info *pipe, size_t len,
773 struct socket *sock = file->private_data;
775 if (unlikely(!sock->ops->splice_read))
778 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
781 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
783 struct file *file = iocb->ki_filp;
784 struct socket *sock = file->private_data;
785 struct msghdr msg = {.msg_iter = *to,
789 if (file->f_flags & O_NONBLOCK)
790 msg.msg_flags = MSG_DONTWAIT;
792 if (iocb->ki_pos != 0)
795 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
798 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
803 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
805 struct file *file = iocb->ki_filp;
806 struct socket *sock = file->private_data;
807 struct msghdr msg = {.msg_iter = *from,
811 if (iocb->ki_pos != 0)
814 if (file->f_flags & O_NONBLOCK)
815 msg.msg_flags = MSG_DONTWAIT;
817 if (sock->type == SOCK_SEQPACKET)
818 msg.msg_flags |= MSG_EOR;
820 res = sock_sendmsg(sock, &msg, iov_iter_count(from));
821 *from = msg.msg_iter;
826 * Atomic setting of ioctl hooks to avoid race
827 * with module unload.
830 static DEFINE_MUTEX(br_ioctl_mutex);
831 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
833 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
835 mutex_lock(&br_ioctl_mutex);
836 br_ioctl_hook = hook;
837 mutex_unlock(&br_ioctl_mutex);
839 EXPORT_SYMBOL(brioctl_set);
841 static DEFINE_MUTEX(vlan_ioctl_mutex);
842 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
844 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
846 mutex_lock(&vlan_ioctl_mutex);
847 vlan_ioctl_hook = hook;
848 mutex_unlock(&vlan_ioctl_mutex);
850 EXPORT_SYMBOL(vlan_ioctl_set);
852 static DEFINE_MUTEX(dlci_ioctl_mutex);
853 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
855 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
857 mutex_lock(&dlci_ioctl_mutex);
858 dlci_ioctl_hook = hook;
859 mutex_unlock(&dlci_ioctl_mutex);
861 EXPORT_SYMBOL(dlci_ioctl_set);
863 static long sock_do_ioctl(struct net *net, struct socket *sock,
864 unsigned int cmd, unsigned long arg)
867 void __user *argp = (void __user *)arg;
869 err = sock->ops->ioctl(sock, cmd, arg);
872 * If this ioctl is unknown try to hand it down
875 if (err == -ENOIOCTLCMD)
876 err = dev_ioctl(net, cmd, argp);
882 * With an ioctl, arg may well be a user mode pointer, but we don't know
883 * what to do with it - that's up to the protocol still.
886 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
890 void __user *argp = (void __user *)arg;
894 sock = file->private_data;
897 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
898 err = dev_ioctl(net, cmd, argp);
900 #ifdef CONFIG_WEXT_CORE
901 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
902 err = dev_ioctl(net, cmd, argp);
909 if (get_user(pid, (int __user *)argp))
911 f_setown(sock->file, pid, 1);
916 err = put_user(f_getown(sock->file),
925 request_module("bridge");
927 mutex_lock(&br_ioctl_mutex);
929 err = br_ioctl_hook(net, cmd, argp);
930 mutex_unlock(&br_ioctl_mutex);
935 if (!vlan_ioctl_hook)
936 request_module("8021q");
938 mutex_lock(&vlan_ioctl_mutex);
940 err = vlan_ioctl_hook(net, argp);
941 mutex_unlock(&vlan_ioctl_mutex);
946 if (!dlci_ioctl_hook)
947 request_module("dlci");
949 mutex_lock(&dlci_ioctl_mutex);
951 err = dlci_ioctl_hook(cmd, argp);
952 mutex_unlock(&dlci_ioctl_mutex);
955 err = sock_do_ioctl(net, sock, cmd, arg);
961 int sock_create_lite(int family, int type, int protocol, struct socket **res)
964 struct socket *sock = NULL;
966 err = security_socket_create(family, type, protocol, 1);
977 err = security_socket_post_create(sock, family, type, protocol, 1);
989 EXPORT_SYMBOL(sock_create_lite);
991 /* No kernel lock held - perfect */
992 static unsigned int sock_poll(struct file *file, poll_table *wait)
994 unsigned int busy_flag = 0;
998 * We can't return errors to poll, so it's either yes or no.
1000 sock = file->private_data;
1002 if (sk_can_busy_loop(sock->sk)) {
1003 /* this socket can poll_ll so tell the system call */
1004 busy_flag = POLL_BUSY_LOOP;
1006 /* once, only if requested by syscall */
1007 if (wait && (wait->_key & POLL_BUSY_LOOP))
1008 sk_busy_loop(sock->sk, 1);
1011 return busy_flag | sock->ops->poll(file, sock, wait);
1014 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1016 struct socket *sock = file->private_data;
1018 return sock->ops->mmap(file, sock, vma);
1021 static int sock_close(struct inode *inode, struct file *filp)
1023 sock_release(SOCKET_I(inode));
1028 * Update the socket async list
1030 * Fasync_list locking strategy.
1032 * 1. fasync_list is modified only under process context socket lock
1033 * i.e. under semaphore.
1034 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1035 * or under socket lock
1038 static int sock_fasync(int fd, struct file *filp, int on)
1040 struct socket *sock = filp->private_data;
1041 struct sock *sk = sock->sk;
1042 struct socket_wq *wq;
1048 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1049 fasync_helper(fd, filp, on, &wq->fasync_list);
1051 if (!wq->fasync_list)
1052 sock_reset_flag(sk, SOCK_FASYNC);
1054 sock_set_flag(sk, SOCK_FASYNC);
1060 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1062 int sock_wake_async(struct socket *sock, int how, int band)
1064 struct socket_wq *wq;
1069 wq = rcu_dereference(sock->wq);
1070 if (!wq || !wq->fasync_list) {
1075 case SOCK_WAKE_WAITD:
1076 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1079 case SOCK_WAKE_SPACE:
1080 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1085 kill_fasync(&wq->fasync_list, SIGIO, band);
1088 kill_fasync(&wq->fasync_list, SIGURG, band);
1093 EXPORT_SYMBOL(sock_wake_async);
1095 int __sock_create(struct net *net, int family, int type, int protocol,
1096 struct socket **res, int kern)
1099 struct socket *sock;
1100 const struct net_proto_family *pf;
1103 * Check protocol is in range
1105 if (family < 0 || family >= NPROTO)
1106 return -EAFNOSUPPORT;
1107 if (type < 0 || type >= SOCK_MAX)
1112 This uglymoron is moved from INET layer to here to avoid
1113 deadlock in module load.
1115 if (family == PF_INET && type == SOCK_PACKET) {
1119 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1125 err = security_socket_create(family, type, protocol, kern);
1130 * Allocate the socket and allow the family to set things up. if
1131 * the protocol is 0, the family is instructed to select an appropriate
1134 sock = sock_alloc();
1136 net_warn_ratelimited("socket: no more sockets\n");
1137 return -ENFILE; /* Not exactly a match, but its the
1138 closest posix thing */
1143 #ifdef CONFIG_MODULES
1144 /* Attempt to load a protocol module if the find failed.
1146 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1147 * requested real, full-featured networking support upon configuration.
1148 * Otherwise module support will break!
1150 if (rcu_access_pointer(net_families[family]) == NULL)
1151 request_module("net-pf-%d", family);
1155 pf = rcu_dereference(net_families[family]);
1156 err = -EAFNOSUPPORT;
1161 * We will call the ->create function, that possibly is in a loadable
1162 * module, so we have to bump that loadable module refcnt first.
1164 if (!try_module_get(pf->owner))
1167 /* Now protected by module ref count */
1170 err = pf->create(net, sock, protocol, kern);
1172 goto out_module_put;
1175 * Now to bump the refcnt of the [loadable] module that owns this
1176 * socket at sock_release time we decrement its refcnt.
1178 if (!try_module_get(sock->ops->owner))
1179 goto out_module_busy;
1182 * Now that we're done with the ->create function, the [loadable]
1183 * module can have its refcnt decremented
1185 module_put(pf->owner);
1186 err = security_socket_post_create(sock, family, type, protocol, kern);
1188 goto out_sock_release;
1194 err = -EAFNOSUPPORT;
1197 module_put(pf->owner);
1204 goto out_sock_release;
1206 EXPORT_SYMBOL(__sock_create);
1208 int sock_create(int family, int type, int protocol, struct socket **res)
1210 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1212 EXPORT_SYMBOL(sock_create);
1214 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1216 return __sock_create(&init_net, family, type, protocol, res, 1);
1218 EXPORT_SYMBOL(sock_create_kern);
1220 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1223 struct socket *sock;
1226 /* Check the SOCK_* constants for consistency. */
1227 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1228 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1229 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1230 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1232 flags = type & ~SOCK_TYPE_MASK;
1233 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1235 type &= SOCK_TYPE_MASK;
1237 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1238 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1240 retval = sock_create(family, type, protocol, &sock);
1244 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1249 /* It may be already another descriptor 8) Not kernel problem. */
1258 * Create a pair of connected sockets.
1261 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1262 int __user *, usockvec)
1264 struct socket *sock1, *sock2;
1266 struct file *newfile1, *newfile2;
1269 flags = type & ~SOCK_TYPE_MASK;
1270 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1272 type &= SOCK_TYPE_MASK;
1274 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1275 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1278 * Obtain the first socket and check if the underlying protocol
1279 * supports the socketpair call.
1282 err = sock_create(family, type, protocol, &sock1);
1286 err = sock_create(family, type, protocol, &sock2);
1290 err = sock1->ops->socketpair(sock1, sock2);
1292 goto out_release_both;
1294 fd1 = get_unused_fd_flags(flags);
1295 if (unlikely(fd1 < 0)) {
1297 goto out_release_both;
1300 fd2 = get_unused_fd_flags(flags);
1301 if (unlikely(fd2 < 0)) {
1303 goto out_put_unused_1;
1306 newfile1 = sock_alloc_file(sock1, flags, NULL);
1307 if (unlikely(IS_ERR(newfile1))) {
1308 err = PTR_ERR(newfile1);
1309 goto out_put_unused_both;
1312 newfile2 = sock_alloc_file(sock2, flags, NULL);
1313 if (IS_ERR(newfile2)) {
1314 err = PTR_ERR(newfile2);
1318 err = put_user(fd1, &usockvec[0]);
1322 err = put_user(fd2, &usockvec[1]);
1326 audit_fd_pair(fd1, fd2);
1328 fd_install(fd1, newfile1);
1329 fd_install(fd2, newfile2);
1330 /* fd1 and fd2 may be already another descriptors.
1331 * Not kernel problem.
1347 sock_release(sock2);
1350 out_put_unused_both:
1355 sock_release(sock2);
1357 sock_release(sock1);
1363 * Bind a name to a socket. Nothing much to do here since it's
1364 * the protocol's responsibility to handle the local address.
1366 * We move the socket address to kernel space before we call
1367 * the protocol layer (having also checked the address is ok).
1370 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1372 struct socket *sock;
1373 struct sockaddr_storage address;
1374 int err, fput_needed;
1376 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1378 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1380 err = security_socket_bind(sock,
1381 (struct sockaddr *)&address,
1384 err = sock->ops->bind(sock,
1388 fput_light(sock->file, fput_needed);
1394 * Perform a listen. Basically, we allow the protocol to do anything
1395 * necessary for a listen, and if that works, we mark the socket as
1396 * ready for listening.
1399 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1401 struct socket *sock;
1402 int err, fput_needed;
1405 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1407 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1408 if ((unsigned int)backlog > somaxconn)
1409 backlog = somaxconn;
1411 err = security_socket_listen(sock, backlog);
1413 err = sock->ops->listen(sock, backlog);
1415 fput_light(sock->file, fput_needed);
1421 * For accept, we attempt to create a new socket, set up the link
1422 * with the client, wake up the client, then return the new
1423 * connected fd. We collect the address of the connector in kernel
1424 * space and move it to user at the very end. This is unclean because
1425 * we open the socket then return an error.
1427 * 1003.1g adds the ability to recvmsg() to query connection pending
1428 * status to recvmsg. We need to add that support in a way thats
1429 * clean when we restucture accept also.
1432 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1433 int __user *, upeer_addrlen, int, flags)
1435 struct socket *sock, *newsock;
1436 struct file *newfile;
1437 int err, len, newfd, fput_needed;
1438 struct sockaddr_storage address;
1440 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1443 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1444 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1446 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1451 newsock = sock_alloc();
1455 newsock->type = sock->type;
1456 newsock->ops = sock->ops;
1459 * We don't need try_module_get here, as the listening socket (sock)
1460 * has the protocol module (sock->ops->owner) held.
1462 __module_get(newsock->ops->owner);
1464 newfd = get_unused_fd_flags(flags);
1465 if (unlikely(newfd < 0)) {
1467 sock_release(newsock);
1470 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1471 if (unlikely(IS_ERR(newfile))) {
1472 err = PTR_ERR(newfile);
1473 put_unused_fd(newfd);
1474 sock_release(newsock);
1478 err = security_socket_accept(sock, newsock);
1482 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1486 if (upeer_sockaddr) {
1487 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1489 err = -ECONNABORTED;
1492 err = move_addr_to_user(&address,
1493 len, upeer_sockaddr, upeer_addrlen);
1498 /* File flags are not inherited via accept() unlike another OSes. */
1500 fd_install(newfd, newfile);
1504 fput_light(sock->file, fput_needed);
1509 put_unused_fd(newfd);
1513 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1514 int __user *, upeer_addrlen)
1516 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1520 * Attempt to connect to a socket with the server address. The address
1521 * is in user space so we verify it is OK and move it to kernel space.
1523 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1526 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1527 * other SEQPACKET protocols that take time to connect() as it doesn't
1528 * include the -EINPROGRESS status for such sockets.
1531 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1534 struct socket *sock;
1535 struct sockaddr_storage address;
1536 int err, fput_needed;
1538 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1541 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1546 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1550 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1551 sock->file->f_flags);
1553 fput_light(sock->file, fput_needed);
1559 * Get the local address ('name') of a socket object. Move the obtained
1560 * name to user space.
1563 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1564 int __user *, usockaddr_len)
1566 struct socket *sock;
1567 struct sockaddr_storage address;
1568 int len, err, fput_needed;
1570 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1574 err = security_socket_getsockname(sock);
1578 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1581 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1584 fput_light(sock->file, fput_needed);
1590 * Get the remote address ('name') of a socket object. Move the obtained
1591 * name to user space.
1594 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1595 int __user *, usockaddr_len)
1597 struct socket *sock;
1598 struct sockaddr_storage address;
1599 int len, err, fput_needed;
1601 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1603 err = security_socket_getpeername(sock);
1605 fput_light(sock->file, fput_needed);
1610 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1613 err = move_addr_to_user(&address, len, usockaddr,
1615 fput_light(sock->file, fput_needed);
1621 * Send a datagram to a given address. We move the address into kernel
1622 * space and check the user space data area is readable before invoking
1626 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1627 unsigned int, flags, struct sockaddr __user *, addr,
1630 struct socket *sock;
1631 struct sockaddr_storage address;
1637 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1640 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1644 msg.msg_name = NULL;
1645 msg.msg_control = NULL;
1646 msg.msg_controllen = 0;
1647 msg.msg_namelen = 0;
1649 err = move_addr_to_kernel(addr, addr_len, &address);
1652 msg.msg_name = (struct sockaddr *)&address;
1653 msg.msg_namelen = addr_len;
1655 if (sock->file->f_flags & O_NONBLOCK)
1656 flags |= MSG_DONTWAIT;
1657 msg.msg_flags = flags;
1658 err = sock_sendmsg(sock, &msg, iov_iter_count(&msg.msg_iter));
1661 fput_light(sock->file, fput_needed);
1667 * Send a datagram down a socket.
1670 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1671 unsigned int, flags)
1673 return sys_sendto(fd, buff, len, flags, NULL, 0);
1677 * Receive a frame from the socket and optionally record the address of the
1678 * sender. We verify the buffers are writable and if needed move the
1679 * sender address from kernel to user space.
1682 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1683 unsigned int, flags, struct sockaddr __user *, addr,
1684 int __user *, addr_len)
1686 struct socket *sock;
1689 struct sockaddr_storage address;
1693 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1696 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1700 msg.msg_control = NULL;
1701 msg.msg_controllen = 0;
1702 /* Save some cycles and don't copy the address if not needed */
1703 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1704 /* We assume all kernel code knows the size of sockaddr_storage */
1705 msg.msg_namelen = 0;
1706 if (sock->file->f_flags & O_NONBLOCK)
1707 flags |= MSG_DONTWAIT;
1708 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1710 if (err >= 0 && addr != NULL) {
1711 err2 = move_addr_to_user(&address,
1712 msg.msg_namelen, addr, addr_len);
1717 fput_light(sock->file, fput_needed);
1723 * Receive a datagram from a socket.
1726 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1727 unsigned int, flags)
1729 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1733 * Set a socket option. Because we don't know the option lengths we have
1734 * to pass the user mode parameter for the protocols to sort out.
1737 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1738 char __user *, optval, int, optlen)
1740 int err, fput_needed;
1741 struct socket *sock;
1746 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1748 err = security_socket_setsockopt(sock, level, optname);
1752 if (level == SOL_SOCKET)
1754 sock_setsockopt(sock, level, optname, optval,
1758 sock->ops->setsockopt(sock, level, optname, optval,
1761 fput_light(sock->file, fput_needed);
1767 * Get a socket option. Because we don't know the option lengths we have
1768 * to pass a user mode parameter for the protocols to sort out.
1771 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1772 char __user *, optval, int __user *, optlen)
1774 int err, fput_needed;
1775 struct socket *sock;
1777 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1779 err = security_socket_getsockopt(sock, level, optname);
1783 if (level == SOL_SOCKET)
1785 sock_getsockopt(sock, level, optname, optval,
1789 sock->ops->getsockopt(sock, level, optname, optval,
1792 fput_light(sock->file, fput_needed);
1798 * Shutdown a socket.
1801 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1803 int err, fput_needed;
1804 struct socket *sock;
1806 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1808 err = security_socket_shutdown(sock, how);
1810 err = sock->ops->shutdown(sock, how);
1811 fput_light(sock->file, fput_needed);
1816 /* A couple of helpful macros for getting the address of the 32/64 bit
1817 * fields which are the same type (int / unsigned) on our platforms.
1819 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1820 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1821 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1823 struct used_address {
1824 struct sockaddr_storage name;
1825 unsigned int name_len;
1828 static int copy_msghdr_from_user(struct msghdr *kmsg,
1829 struct user_msghdr __user *umsg,
1830 struct sockaddr __user **save_addr,
1833 struct sockaddr __user *uaddr;
1834 struct iovec __user *uiov;
1838 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1839 __get_user(uaddr, &umsg->msg_name) ||
1840 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1841 __get_user(uiov, &umsg->msg_iov) ||
1842 __get_user(nr_segs, &umsg->msg_iovlen) ||
1843 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1844 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1845 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1849 kmsg->msg_namelen = 0;
1851 if (kmsg->msg_namelen < 0)
1854 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1855 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1860 if (uaddr && kmsg->msg_namelen) {
1862 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1868 kmsg->msg_name = NULL;
1869 kmsg->msg_namelen = 0;
1872 if (nr_segs > UIO_MAXIOV)
1875 kmsg->msg_iocb = NULL;
1877 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1878 UIO_FASTIOV, iov, &kmsg->msg_iter);
1881 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1882 struct msghdr *msg_sys, unsigned int flags,
1883 struct used_address *used_address)
1885 struct compat_msghdr __user *msg_compat =
1886 (struct compat_msghdr __user *)msg;
1887 struct sockaddr_storage address;
1888 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1889 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1890 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1891 /* 20 is size of ipv6_pktinfo */
1892 unsigned char *ctl_buf = ctl;
1893 int ctl_len, total_len;
1896 msg_sys->msg_name = &address;
1898 if (MSG_CMSG_COMPAT & flags)
1899 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1901 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1904 total_len = iov_iter_count(&msg_sys->msg_iter);
1908 if (msg_sys->msg_controllen > INT_MAX)
1910 ctl_len = msg_sys->msg_controllen;
1911 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1913 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1917 ctl_buf = msg_sys->msg_control;
1918 ctl_len = msg_sys->msg_controllen;
1919 } else if (ctl_len) {
1920 if (ctl_len > sizeof(ctl)) {
1921 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1922 if (ctl_buf == NULL)
1927 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1928 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1929 * checking falls down on this.
1931 if (copy_from_user(ctl_buf,
1932 (void __user __force *)msg_sys->msg_control,
1935 msg_sys->msg_control = ctl_buf;
1937 msg_sys->msg_flags = flags;
1939 if (sock->file->f_flags & O_NONBLOCK)
1940 msg_sys->msg_flags |= MSG_DONTWAIT;
1942 * If this is sendmmsg() and current destination address is same as
1943 * previously succeeded address, omit asking LSM's decision.
1944 * used_address->name_len is initialized to UINT_MAX so that the first
1945 * destination address never matches.
1947 if (used_address && msg_sys->msg_name &&
1948 used_address->name_len == msg_sys->msg_namelen &&
1949 !memcmp(&used_address->name, msg_sys->msg_name,
1950 used_address->name_len)) {
1951 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1954 err = sock_sendmsg(sock, msg_sys, total_len);
1956 * If this is sendmmsg() and sending to current destination address was
1957 * successful, remember it.
1959 if (used_address && err >= 0) {
1960 used_address->name_len = msg_sys->msg_namelen;
1961 if (msg_sys->msg_name)
1962 memcpy(&used_address->name, msg_sys->msg_name,
1963 used_address->name_len);
1968 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1975 * BSD sendmsg interface
1978 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1980 int fput_needed, err;
1981 struct msghdr msg_sys;
1982 struct socket *sock;
1984 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1988 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1990 fput_light(sock->file, fput_needed);
1995 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1997 if (flags & MSG_CMSG_COMPAT)
1999 return __sys_sendmsg(fd, msg, flags);
2003 * Linux sendmmsg interface
2006 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2009 int fput_needed, err, datagrams;
2010 struct socket *sock;
2011 struct mmsghdr __user *entry;
2012 struct compat_mmsghdr __user *compat_entry;
2013 struct msghdr msg_sys;
2014 struct used_address used_address;
2016 if (vlen > UIO_MAXIOV)
2021 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2025 used_address.name_len = UINT_MAX;
2027 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2030 while (datagrams < vlen) {
2031 if (MSG_CMSG_COMPAT & flags) {
2032 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2033 &msg_sys, flags, &used_address);
2036 err = __put_user(err, &compat_entry->msg_len);
2039 err = ___sys_sendmsg(sock,
2040 (struct user_msghdr __user *)entry,
2041 &msg_sys, flags, &used_address);
2044 err = put_user(err, &entry->msg_len);
2053 fput_light(sock->file, fput_needed);
2055 /* We only return an error if no datagrams were able to be sent */
2062 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2063 unsigned int, vlen, unsigned int, flags)
2065 if (flags & MSG_CMSG_COMPAT)
2067 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2070 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2071 struct msghdr *msg_sys, unsigned int flags, int nosec)
2073 struct compat_msghdr __user *msg_compat =
2074 (struct compat_msghdr __user *)msg;
2075 struct iovec iovstack[UIO_FASTIOV];
2076 struct iovec *iov = iovstack;
2077 unsigned long cmsg_ptr;
2081 /* kernel mode address */
2082 struct sockaddr_storage addr;
2084 /* user mode address pointers */
2085 struct sockaddr __user *uaddr;
2086 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2088 msg_sys->msg_name = &addr;
2090 if (MSG_CMSG_COMPAT & flags)
2091 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2093 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2096 total_len = iov_iter_count(&msg_sys->msg_iter);
2098 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2099 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2101 /* We assume all kernel code knows the size of sockaddr_storage */
2102 msg_sys->msg_namelen = 0;
2104 if (sock->file->f_flags & O_NONBLOCK)
2105 flags |= MSG_DONTWAIT;
2106 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2112 if (uaddr != NULL) {
2113 err = move_addr_to_user(&addr,
2114 msg_sys->msg_namelen, uaddr,
2119 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2123 if (MSG_CMSG_COMPAT & flags)
2124 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2125 &msg_compat->msg_controllen);
2127 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2128 &msg->msg_controllen);
2139 * BSD recvmsg interface
2142 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2144 int fput_needed, err;
2145 struct msghdr msg_sys;
2146 struct socket *sock;
2148 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2152 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2154 fput_light(sock->file, fput_needed);
2159 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2160 unsigned int, flags)
2162 if (flags & MSG_CMSG_COMPAT)
2164 return __sys_recvmsg(fd, msg, flags);
2168 * Linux recvmmsg interface
2171 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2172 unsigned int flags, struct timespec *timeout)
2174 int fput_needed, err, datagrams;
2175 struct socket *sock;
2176 struct mmsghdr __user *entry;
2177 struct compat_mmsghdr __user *compat_entry;
2178 struct msghdr msg_sys;
2179 struct timespec end_time;
2182 poll_select_set_timeout(&end_time, timeout->tv_sec,
2188 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2192 err = sock_error(sock->sk);
2197 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2199 while (datagrams < vlen) {
2201 * No need to ask LSM for more than the first datagram.
2203 if (MSG_CMSG_COMPAT & flags) {
2204 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2205 &msg_sys, flags & ~MSG_WAITFORONE,
2209 err = __put_user(err, &compat_entry->msg_len);
2212 err = ___sys_recvmsg(sock,
2213 (struct user_msghdr __user *)entry,
2214 &msg_sys, flags & ~MSG_WAITFORONE,
2218 err = put_user(err, &entry->msg_len);
2226 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2227 if (flags & MSG_WAITFORONE)
2228 flags |= MSG_DONTWAIT;
2231 ktime_get_ts(timeout);
2232 *timeout = timespec_sub(end_time, *timeout);
2233 if (timeout->tv_sec < 0) {
2234 timeout->tv_sec = timeout->tv_nsec = 0;
2238 /* Timeout, return less than vlen datagrams */
2239 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2243 /* Out of band data, return right away */
2244 if (msg_sys.msg_flags & MSG_OOB)
2249 fput_light(sock->file, fput_needed);
2254 if (datagrams != 0) {
2256 * We may return less entries than requested (vlen) if the
2257 * sock is non block and there aren't enough datagrams...
2259 if (err != -EAGAIN) {
2261 * ... or if recvmsg returns an error after we
2262 * received some datagrams, where we record the
2263 * error to return on the next call or if the
2264 * app asks about it using getsockopt(SO_ERROR).
2266 sock->sk->sk_err = -err;
2275 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2276 unsigned int, vlen, unsigned int, flags,
2277 struct timespec __user *, timeout)
2280 struct timespec timeout_sys;
2282 if (flags & MSG_CMSG_COMPAT)
2286 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2288 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2291 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2293 if (datagrams > 0 &&
2294 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2295 datagrams = -EFAULT;
2300 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2301 /* Argument list sizes for sys_socketcall */
2302 #define AL(x) ((x) * sizeof(unsigned long))
2303 static const unsigned char nargs[21] = {
2304 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2305 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2306 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2313 * System call vectors.
2315 * Argument checking cleaned up. Saved 20% in size.
2316 * This function doesn't need to set the kernel lock because
2317 * it is set by the callees.
2320 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2322 unsigned long a[AUDITSC_ARGS];
2323 unsigned long a0, a1;
2327 if (call < 1 || call > SYS_SENDMMSG)
2331 if (len > sizeof(a))
2334 /* copy_from_user should be SMP safe. */
2335 if (copy_from_user(a, args, len))
2338 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2347 err = sys_socket(a0, a1, a[2]);
2350 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2353 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2356 err = sys_listen(a0, a1);
2359 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2360 (int __user *)a[2], 0);
2362 case SYS_GETSOCKNAME:
2364 sys_getsockname(a0, (struct sockaddr __user *)a1,
2365 (int __user *)a[2]);
2367 case SYS_GETPEERNAME:
2369 sys_getpeername(a0, (struct sockaddr __user *)a1,
2370 (int __user *)a[2]);
2372 case SYS_SOCKETPAIR:
2373 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2376 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2379 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2380 (struct sockaddr __user *)a[4], a[5]);
2383 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2386 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2387 (struct sockaddr __user *)a[4],
2388 (int __user *)a[5]);
2391 err = sys_shutdown(a0, a1);
2393 case SYS_SETSOCKOPT:
2394 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2396 case SYS_GETSOCKOPT:
2398 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2399 (int __user *)a[4]);
2402 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2405 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2408 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2411 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2412 (struct timespec __user *)a[4]);
2415 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2416 (int __user *)a[2], a[3]);
2425 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2428 * sock_register - add a socket protocol handler
2429 * @ops: description of protocol
2431 * This function is called by a protocol handler that wants to
2432 * advertise its address family, and have it linked into the
2433 * socket interface. The value ops->family corresponds to the
2434 * socket system call protocol family.
2436 int sock_register(const struct net_proto_family *ops)
2440 if (ops->family >= NPROTO) {
2441 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2445 spin_lock(&net_family_lock);
2446 if (rcu_dereference_protected(net_families[ops->family],
2447 lockdep_is_held(&net_family_lock)))
2450 rcu_assign_pointer(net_families[ops->family], ops);
2453 spin_unlock(&net_family_lock);
2455 pr_info("NET: Registered protocol family %d\n", ops->family);
2458 EXPORT_SYMBOL(sock_register);
2461 * sock_unregister - remove a protocol handler
2462 * @family: protocol family to remove
2464 * This function is called by a protocol handler that wants to
2465 * remove its address family, and have it unlinked from the
2466 * new socket creation.
2468 * If protocol handler is a module, then it can use module reference
2469 * counts to protect against new references. If protocol handler is not
2470 * a module then it needs to provide its own protection in
2471 * the ops->create routine.
2473 void sock_unregister(int family)
2475 BUG_ON(family < 0 || family >= NPROTO);
2477 spin_lock(&net_family_lock);
2478 RCU_INIT_POINTER(net_families[family], NULL);
2479 spin_unlock(&net_family_lock);
2483 pr_info("NET: Unregistered protocol family %d\n", family);
2485 EXPORT_SYMBOL(sock_unregister);
2487 static int __init sock_init(void)
2491 * Initialize the network sysctl infrastructure.
2493 err = net_sysctl_init();
2498 * Initialize skbuff SLAB cache
2503 * Initialize the protocols module.
2508 err = register_filesystem(&sock_fs_type);
2511 sock_mnt = kern_mount(&sock_fs_type);
2512 if (IS_ERR(sock_mnt)) {
2513 err = PTR_ERR(sock_mnt);
2517 /* The real protocol initialization is performed in later initcalls.
2520 #ifdef CONFIG_NETFILTER
2521 err = netfilter_init();
2526 ptp_classifier_init();
2532 unregister_filesystem(&sock_fs_type);
2537 core_initcall(sock_init); /* early initcall */
2539 #ifdef CONFIG_PROC_FS
2540 void socket_seq_show(struct seq_file *seq)
2545 for_each_possible_cpu(cpu)
2546 counter += per_cpu(sockets_in_use, cpu);
2548 /* It can be negative, by the way. 8) */
2552 seq_printf(seq, "sockets: used %d\n", counter);
2554 #endif /* CONFIG_PROC_FS */
2556 #ifdef CONFIG_COMPAT
2557 static int do_siocgstamp(struct net *net, struct socket *sock,
2558 unsigned int cmd, void __user *up)
2560 mm_segment_t old_fs = get_fs();
2565 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2568 err = compat_put_timeval(&ktv, up);
2573 static int do_siocgstampns(struct net *net, struct socket *sock,
2574 unsigned int cmd, void __user *up)
2576 mm_segment_t old_fs = get_fs();
2577 struct timespec kts;
2581 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2584 err = compat_put_timespec(&kts, up);
2589 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2591 struct ifreq __user *uifr;
2594 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2595 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2598 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2602 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2608 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2610 struct compat_ifconf ifc32;
2612 struct ifconf __user *uifc;
2613 struct compat_ifreq __user *ifr32;
2614 struct ifreq __user *ifr;
2618 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2621 memset(&ifc, 0, sizeof(ifc));
2622 if (ifc32.ifcbuf == 0) {
2626 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2628 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2629 sizeof(struct ifreq);
2630 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2632 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2633 ifr32 = compat_ptr(ifc32.ifcbuf);
2634 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2635 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2641 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2644 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2648 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2652 ifr32 = compat_ptr(ifc32.ifcbuf);
2654 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2655 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2656 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2662 if (ifc32.ifcbuf == 0) {
2663 /* Translate from 64-bit structure multiple to
2667 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2672 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2678 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2680 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2681 bool convert_in = false, convert_out = false;
2682 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2683 struct ethtool_rxnfc __user *rxnfc;
2684 struct ifreq __user *ifr;
2685 u32 rule_cnt = 0, actual_rule_cnt;
2690 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2693 compat_rxnfc = compat_ptr(data);
2695 if (get_user(ethcmd, &compat_rxnfc->cmd))
2698 /* Most ethtool structures are defined without padding.
2699 * Unfortunately struct ethtool_rxnfc is an exception.
2704 case ETHTOOL_GRXCLSRLALL:
2705 /* Buffer size is variable */
2706 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2708 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2710 buf_size += rule_cnt * sizeof(u32);
2712 case ETHTOOL_GRXRINGS:
2713 case ETHTOOL_GRXCLSRLCNT:
2714 case ETHTOOL_GRXCLSRULE:
2715 case ETHTOOL_SRXCLSRLINS:
2718 case ETHTOOL_SRXCLSRLDEL:
2719 buf_size += sizeof(struct ethtool_rxnfc);
2724 ifr = compat_alloc_user_space(buf_size);
2725 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2727 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2730 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2731 &ifr->ifr_ifru.ifru_data))
2735 /* We expect there to be holes between fs.m_ext and
2736 * fs.ring_cookie and at the end of fs, but nowhere else.
2738 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2739 sizeof(compat_rxnfc->fs.m_ext) !=
2740 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2741 sizeof(rxnfc->fs.m_ext));
2743 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2744 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2745 offsetof(struct ethtool_rxnfc, fs.location) -
2746 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2748 if (copy_in_user(rxnfc, compat_rxnfc,
2749 (void __user *)(&rxnfc->fs.m_ext + 1) -
2750 (void __user *)rxnfc) ||
2751 copy_in_user(&rxnfc->fs.ring_cookie,
2752 &compat_rxnfc->fs.ring_cookie,
2753 (void __user *)(&rxnfc->fs.location + 1) -
2754 (void __user *)&rxnfc->fs.ring_cookie) ||
2755 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2756 sizeof(rxnfc->rule_cnt)))
2760 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2765 if (copy_in_user(compat_rxnfc, rxnfc,
2766 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2767 (const void __user *)rxnfc) ||
2768 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2769 &rxnfc->fs.ring_cookie,
2770 (const void __user *)(&rxnfc->fs.location + 1) -
2771 (const void __user *)&rxnfc->fs.ring_cookie) ||
2772 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2773 sizeof(rxnfc->rule_cnt)))
2776 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2777 /* As an optimisation, we only copy the actual
2778 * number of rules that the underlying
2779 * function returned. Since Mallory might
2780 * change the rule count in user memory, we
2781 * check that it is less than the rule count
2782 * originally given (as the user buffer size),
2783 * which has been range-checked.
2785 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2787 if (actual_rule_cnt < rule_cnt)
2788 rule_cnt = actual_rule_cnt;
2789 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2790 &rxnfc->rule_locs[0],
2791 rule_cnt * sizeof(u32)))
2799 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2802 compat_uptr_t uptr32;
2803 struct ifreq __user *uifr;
2805 uifr = compat_alloc_user_space(sizeof(*uifr));
2806 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2809 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2812 uptr = compat_ptr(uptr32);
2814 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2817 return dev_ioctl(net, SIOCWANDEV, uifr);
2820 static int bond_ioctl(struct net *net, unsigned int cmd,
2821 struct compat_ifreq __user *ifr32)
2824 mm_segment_t old_fs;
2828 case SIOCBONDENSLAVE:
2829 case SIOCBONDRELEASE:
2830 case SIOCBONDSETHWADDR:
2831 case SIOCBONDCHANGEACTIVE:
2832 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2837 err = dev_ioctl(net, cmd,
2838 (struct ifreq __user __force *) &kifr);
2843 return -ENOIOCTLCMD;
2847 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2848 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2849 struct compat_ifreq __user *u_ifreq32)
2851 struct ifreq __user *u_ifreq64;
2852 char tmp_buf[IFNAMSIZ];
2853 void __user *data64;
2856 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2859 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2861 data64 = compat_ptr(data32);
2863 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2865 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2868 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2871 return dev_ioctl(net, cmd, u_ifreq64);
2874 static int dev_ifsioc(struct net *net, struct socket *sock,
2875 unsigned int cmd, struct compat_ifreq __user *uifr32)
2877 struct ifreq __user *uifr;
2880 uifr = compat_alloc_user_space(sizeof(*uifr));
2881 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2884 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2895 case SIOCGIFBRDADDR:
2896 case SIOCGIFDSTADDR:
2897 case SIOCGIFNETMASK:
2902 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2910 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2911 struct compat_ifreq __user *uifr32)
2914 struct compat_ifmap __user *uifmap32;
2915 mm_segment_t old_fs;
2918 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2919 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2920 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2921 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2922 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2923 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2924 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2925 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2931 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2934 if (cmd == SIOCGIFMAP && !err) {
2935 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2936 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2937 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2938 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2939 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2940 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2941 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2950 struct sockaddr rt_dst; /* target address */
2951 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2952 struct sockaddr rt_genmask; /* target network mask (IP) */
2953 unsigned short rt_flags;
2956 unsigned char rt_tos;
2957 unsigned char rt_class;
2959 short rt_metric; /* +1 for binary compatibility! */
2960 /* char * */ u32 rt_dev; /* forcing the device at add */
2961 u32 rt_mtu; /* per route MTU/Window */
2962 u32 rt_window; /* Window clamping */
2963 unsigned short rt_irtt; /* Initial RTT */
2966 struct in6_rtmsg32 {
2967 struct in6_addr rtmsg_dst;
2968 struct in6_addr rtmsg_src;
2969 struct in6_addr rtmsg_gateway;
2979 static int routing_ioctl(struct net *net, struct socket *sock,
2980 unsigned int cmd, void __user *argp)
2984 struct in6_rtmsg r6;
2988 mm_segment_t old_fs = get_fs();
2990 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2991 struct in6_rtmsg32 __user *ur6 = argp;
2992 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2993 3 * sizeof(struct in6_addr));
2994 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2995 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2996 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2997 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2998 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2999 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3000 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3004 struct rtentry32 __user *ur4 = argp;
3005 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3006 3 * sizeof(struct sockaddr));
3007 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3008 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3009 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3010 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3011 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3012 ret |= get_user(rtdev, &(ur4->rt_dev));
3014 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3015 r4.rt_dev = (char __user __force *)devname;
3029 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3036 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3037 * for some operations; this forces use of the newer bridge-utils that
3038 * use compatible ioctls
3040 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3044 if (get_user(tmp, argp))
3046 if (tmp == BRCTL_GET_VERSION)
3047 return BRCTL_VERSION + 1;
3051 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3052 unsigned int cmd, unsigned long arg)
3054 void __user *argp = compat_ptr(arg);
3055 struct sock *sk = sock->sk;
3056 struct net *net = sock_net(sk);
3058 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3059 return compat_ifr_data_ioctl(net, cmd, argp);
3064 return old_bridge_ioctl(argp);
3066 return dev_ifname32(net, argp);
3068 return dev_ifconf(net, argp);
3070 return ethtool_ioctl(net, argp);
3072 return compat_siocwandev(net, argp);
3075 return compat_sioc_ifmap(net, cmd, argp);
3076 case SIOCBONDENSLAVE:
3077 case SIOCBONDRELEASE:
3078 case SIOCBONDSETHWADDR:
3079 case SIOCBONDCHANGEACTIVE:
3080 return bond_ioctl(net, cmd, argp);
3083 return routing_ioctl(net, sock, cmd, argp);
3085 return do_siocgstamp(net, sock, cmd, argp);
3087 return do_siocgstampns(net, sock, cmd, argp);
3088 case SIOCBONDSLAVEINFOQUERY:
3089 case SIOCBONDINFOQUERY:
3092 return compat_ifr_data_ioctl(net, cmd, argp);
3104 return sock_ioctl(file, cmd, arg);
3121 case SIOCSIFHWBROADCAST:
3123 case SIOCGIFBRDADDR:
3124 case SIOCSIFBRDADDR:
3125 case SIOCGIFDSTADDR:
3126 case SIOCSIFDSTADDR:
3127 case SIOCGIFNETMASK:
3128 case SIOCSIFNETMASK:
3139 return dev_ifsioc(net, sock, cmd, argp);
3145 return sock_do_ioctl(net, sock, cmd, arg);
3148 return -ENOIOCTLCMD;
3151 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3154 struct socket *sock = file->private_data;
3155 int ret = -ENOIOCTLCMD;
3162 if (sock->ops->compat_ioctl)
3163 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3165 if (ret == -ENOIOCTLCMD &&
3166 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3167 ret = compat_wext_handle_ioctl(net, cmd, arg);
3169 if (ret == -ENOIOCTLCMD)
3170 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3176 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3178 return sock->ops->bind(sock, addr, addrlen);
3180 EXPORT_SYMBOL(kernel_bind);
3182 int kernel_listen(struct socket *sock, int backlog)
3184 return sock->ops->listen(sock, backlog);
3186 EXPORT_SYMBOL(kernel_listen);
3188 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3190 struct sock *sk = sock->sk;
3193 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3198 err = sock->ops->accept(sock, *newsock, flags);
3200 sock_release(*newsock);
3205 (*newsock)->ops = sock->ops;
3206 __module_get((*newsock)->ops->owner);
3211 EXPORT_SYMBOL(kernel_accept);
3213 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3216 return sock->ops->connect(sock, addr, addrlen, flags);
3218 EXPORT_SYMBOL(kernel_connect);
3220 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3223 return sock->ops->getname(sock, addr, addrlen, 0);
3225 EXPORT_SYMBOL(kernel_getsockname);
3227 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3230 return sock->ops->getname(sock, addr, addrlen, 1);
3232 EXPORT_SYMBOL(kernel_getpeername);
3234 int kernel_getsockopt(struct socket *sock, int level, int optname,
3235 char *optval, int *optlen)
3237 mm_segment_t oldfs = get_fs();
3238 char __user *uoptval;
3239 int __user *uoptlen;
3242 uoptval = (char __user __force *) optval;
3243 uoptlen = (int __user __force *) optlen;
3246 if (level == SOL_SOCKET)
3247 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3249 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3254 EXPORT_SYMBOL(kernel_getsockopt);
3256 int kernel_setsockopt(struct socket *sock, int level, int optname,
3257 char *optval, unsigned int optlen)
3259 mm_segment_t oldfs = get_fs();
3260 char __user *uoptval;
3263 uoptval = (char __user __force *) optval;
3266 if (level == SOL_SOCKET)
3267 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3269 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3274 EXPORT_SYMBOL(kernel_setsockopt);
3276 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3277 size_t size, int flags)
3279 if (sock->ops->sendpage)
3280 return sock->ops->sendpage(sock, page, offset, size, flags);
3282 return sock_no_sendpage(sock, page, offset, size, flags);
3284 EXPORT_SYMBOL(kernel_sendpage);
3286 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3288 mm_segment_t oldfs = get_fs();
3292 err = sock->ops->ioctl(sock, cmd, arg);
3297 EXPORT_SYMBOL(kernel_sock_ioctl);
3299 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3301 return sock->ops->shutdown(sock, how);
3303 EXPORT_SYMBOL(kernel_sock_shutdown);