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 d_inode(dentry)->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(d_inode(path.dentry));
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(d_inode(dentry), 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 this_cpu_sub(sockets_in_use, 1);
581 iput(SOCK_INODE(sock));
586 EXPORT_SYMBOL(sock_release);
588 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
590 u8 flags = *tx_flags;
592 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
593 flags |= SKBTX_HW_TSTAMP;
595 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
596 flags |= SKBTX_SW_TSTAMP;
598 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
599 flags |= SKBTX_SCHED_TSTAMP;
601 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
602 flags |= SKBTX_ACK_TSTAMP;
606 EXPORT_SYMBOL(__sock_tx_timestamp);
608 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
610 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
611 BUG_ON(ret == -EIOCBQUEUED);
615 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
617 int err = security_socket_sendmsg(sock, msg,
620 return err ?: sock_sendmsg_nosec(sock, msg);
622 EXPORT_SYMBOL(sock_sendmsg);
624 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
625 struct kvec *vec, size_t num, size_t size)
627 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
628 return sock_sendmsg(sock, msg);
630 EXPORT_SYMBOL(kernel_sendmsg);
633 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
635 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
638 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
639 struct scm_timestamping tss;
641 struct skb_shared_hwtstamps *shhwtstamps =
644 /* Race occurred between timestamp enabling and packet
645 receiving. Fill in the current time for now. */
646 if (need_software_tstamp && skb->tstamp.tv64 == 0)
647 __net_timestamp(skb);
649 if (need_software_tstamp) {
650 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
652 skb_get_timestamp(skb, &tv);
653 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
657 skb_get_timestampns(skb, &ts);
658 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
663 memset(&tss, 0, sizeof(tss));
664 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
665 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
668 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
669 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
672 put_cmsg(msg, SOL_SOCKET,
673 SCM_TIMESTAMPING, sizeof(tss), &tss);
675 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
677 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
682 if (!sock_flag(sk, SOCK_WIFI_STATUS))
684 if (!skb->wifi_acked_valid)
687 ack = skb->wifi_acked;
689 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
691 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
693 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
696 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
697 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
698 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
701 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
704 sock_recv_timestamp(msg, sk, skb);
705 sock_recv_drops(msg, sk, skb);
707 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
709 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
710 size_t size, int flags)
712 return sock->ops->recvmsg(sock, msg, size, flags);
715 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
718 int err = security_socket_recvmsg(sock, msg, size, flags);
720 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
722 EXPORT_SYMBOL(sock_recvmsg);
725 * kernel_recvmsg - Receive a message from a socket (kernel space)
726 * @sock: The socket to receive the message from
727 * @msg: Received message
728 * @vec: Input s/g array for message data
729 * @num: Size of input s/g array
730 * @size: Number of bytes to read
731 * @flags: Message flags (MSG_DONTWAIT, etc...)
733 * On return the msg structure contains the scatter/gather array passed in the
734 * vec argument. The array is modified so that it consists of the unfilled
735 * portion of the original array.
737 * The returned value is the total number of bytes received, or an error.
739 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
740 struct kvec *vec, size_t num, size_t size, int flags)
742 mm_segment_t oldfs = get_fs();
745 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
747 result = sock_recvmsg(sock, msg, size, flags);
751 EXPORT_SYMBOL(kernel_recvmsg);
753 static ssize_t sock_sendpage(struct file *file, struct page *page,
754 int offset, size_t size, loff_t *ppos, int more)
759 sock = file->private_data;
761 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
762 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
765 return kernel_sendpage(sock, page, offset, size, flags);
768 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
769 struct pipe_inode_info *pipe, size_t len,
772 struct socket *sock = file->private_data;
774 if (unlikely(!sock->ops->splice_read))
777 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
780 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
782 struct file *file = iocb->ki_filp;
783 struct socket *sock = file->private_data;
784 struct msghdr msg = {.msg_iter = *to,
788 if (file->f_flags & O_NONBLOCK)
789 msg.msg_flags = MSG_DONTWAIT;
791 if (iocb->ki_pos != 0)
794 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
797 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
802 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
804 struct file *file = iocb->ki_filp;
805 struct socket *sock = file->private_data;
806 struct msghdr msg = {.msg_iter = *from,
810 if (iocb->ki_pos != 0)
813 if (file->f_flags & O_NONBLOCK)
814 msg.msg_flags = MSG_DONTWAIT;
816 if (sock->type == SOCK_SEQPACKET)
817 msg.msg_flags |= MSG_EOR;
819 res = sock_sendmsg(sock, &msg);
820 *from = msg.msg_iter;
825 * Atomic setting of ioctl hooks to avoid race
826 * with module unload.
829 static DEFINE_MUTEX(br_ioctl_mutex);
830 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
832 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
834 mutex_lock(&br_ioctl_mutex);
835 br_ioctl_hook = hook;
836 mutex_unlock(&br_ioctl_mutex);
838 EXPORT_SYMBOL(brioctl_set);
840 static DEFINE_MUTEX(vlan_ioctl_mutex);
841 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
843 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
845 mutex_lock(&vlan_ioctl_mutex);
846 vlan_ioctl_hook = hook;
847 mutex_unlock(&vlan_ioctl_mutex);
849 EXPORT_SYMBOL(vlan_ioctl_set);
851 static DEFINE_MUTEX(dlci_ioctl_mutex);
852 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
854 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
856 mutex_lock(&dlci_ioctl_mutex);
857 dlci_ioctl_hook = hook;
858 mutex_unlock(&dlci_ioctl_mutex);
860 EXPORT_SYMBOL(dlci_ioctl_set);
862 static long sock_do_ioctl(struct net *net, struct socket *sock,
863 unsigned int cmd, unsigned long arg)
866 void __user *argp = (void __user *)arg;
868 err = sock->ops->ioctl(sock, cmd, arg);
871 * If this ioctl is unknown try to hand it down
874 if (err == -ENOIOCTLCMD)
875 err = dev_ioctl(net, cmd, argp);
881 * With an ioctl, arg may well be a user mode pointer, but we don't know
882 * what to do with it - that's up to the protocol still.
885 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
889 void __user *argp = (void __user *)arg;
893 sock = file->private_data;
896 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
897 err = dev_ioctl(net, cmd, argp);
899 #ifdef CONFIG_WEXT_CORE
900 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
901 err = dev_ioctl(net, cmd, argp);
908 if (get_user(pid, (int __user *)argp))
910 f_setown(sock->file, pid, 1);
915 err = put_user(f_getown(sock->file),
924 request_module("bridge");
926 mutex_lock(&br_ioctl_mutex);
928 err = br_ioctl_hook(net, cmd, argp);
929 mutex_unlock(&br_ioctl_mutex);
934 if (!vlan_ioctl_hook)
935 request_module("8021q");
937 mutex_lock(&vlan_ioctl_mutex);
939 err = vlan_ioctl_hook(net, argp);
940 mutex_unlock(&vlan_ioctl_mutex);
945 if (!dlci_ioctl_hook)
946 request_module("dlci");
948 mutex_lock(&dlci_ioctl_mutex);
950 err = dlci_ioctl_hook(cmd, argp);
951 mutex_unlock(&dlci_ioctl_mutex);
954 err = sock_do_ioctl(net, sock, cmd, arg);
960 int sock_create_lite(int family, int type, int protocol, struct socket **res)
963 struct socket *sock = NULL;
965 err = security_socket_create(family, type, protocol, 1);
976 err = security_socket_post_create(sock, family, type, protocol, 1);
988 EXPORT_SYMBOL(sock_create_lite);
990 /* No kernel lock held - perfect */
991 static unsigned int sock_poll(struct file *file, poll_table *wait)
993 unsigned int busy_flag = 0;
997 * We can't return errors to poll, so it's either yes or no.
999 sock = file->private_data;
1001 if (sk_can_busy_loop(sock->sk)) {
1002 /* this socket can poll_ll so tell the system call */
1003 busy_flag = POLL_BUSY_LOOP;
1005 /* once, only if requested by syscall */
1006 if (wait && (wait->_key & POLL_BUSY_LOOP))
1007 sk_busy_loop(sock->sk, 1);
1010 return busy_flag | sock->ops->poll(file, sock, wait);
1013 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1015 struct socket *sock = file->private_data;
1017 return sock->ops->mmap(file, sock, vma);
1020 static int sock_close(struct inode *inode, struct file *filp)
1022 sock_release(SOCKET_I(inode));
1027 * Update the socket async list
1029 * Fasync_list locking strategy.
1031 * 1. fasync_list is modified only under process context socket lock
1032 * i.e. under semaphore.
1033 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1034 * or under socket lock
1037 static int sock_fasync(int fd, struct file *filp, int on)
1039 struct socket *sock = filp->private_data;
1040 struct sock *sk = sock->sk;
1041 struct socket_wq *wq;
1047 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1048 fasync_helper(fd, filp, on, &wq->fasync_list);
1050 if (!wq->fasync_list)
1051 sock_reset_flag(sk, SOCK_FASYNC);
1053 sock_set_flag(sk, SOCK_FASYNC);
1059 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1061 int sock_wake_async(struct socket *sock, int how, int band)
1063 struct socket_wq *wq;
1068 wq = rcu_dereference(sock->wq);
1069 if (!wq || !wq->fasync_list) {
1074 case SOCK_WAKE_WAITD:
1075 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1078 case SOCK_WAKE_SPACE:
1079 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1084 kill_fasync(&wq->fasync_list, SIGIO, band);
1087 kill_fasync(&wq->fasync_list, SIGURG, band);
1092 EXPORT_SYMBOL(sock_wake_async);
1094 int __sock_create(struct net *net, int family, int type, int protocol,
1095 struct socket **res, int kern)
1098 struct socket *sock;
1099 const struct net_proto_family *pf;
1102 * Check protocol is in range
1104 if (family < 0 || family >= NPROTO)
1105 return -EAFNOSUPPORT;
1106 if (type < 0 || type >= SOCK_MAX)
1111 This uglymoron is moved from INET layer to here to avoid
1112 deadlock in module load.
1114 if (family == PF_INET && type == SOCK_PACKET) {
1118 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1124 err = security_socket_create(family, type, protocol, kern);
1129 * Allocate the socket and allow the family to set things up. if
1130 * the protocol is 0, the family is instructed to select an appropriate
1133 sock = sock_alloc();
1135 net_warn_ratelimited("socket: no more sockets\n");
1136 return -ENFILE; /* Not exactly a match, but its the
1137 closest posix thing */
1142 #ifdef CONFIG_MODULES
1143 /* Attempt to load a protocol module if the find failed.
1145 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1146 * requested real, full-featured networking support upon configuration.
1147 * Otherwise module support will break!
1149 if (rcu_access_pointer(net_families[family]) == NULL)
1150 request_module("net-pf-%d", family);
1154 pf = rcu_dereference(net_families[family]);
1155 err = -EAFNOSUPPORT;
1160 * We will call the ->create function, that possibly is in a loadable
1161 * module, so we have to bump that loadable module refcnt first.
1163 if (!try_module_get(pf->owner))
1166 /* Now protected by module ref count */
1169 err = pf->create(net, sock, protocol, kern);
1171 goto out_module_put;
1174 * Now to bump the refcnt of the [loadable] module that owns this
1175 * socket at sock_release time we decrement its refcnt.
1177 if (!try_module_get(sock->ops->owner))
1178 goto out_module_busy;
1181 * Now that we're done with the ->create function, the [loadable]
1182 * module can have its refcnt decremented
1184 module_put(pf->owner);
1185 err = security_socket_post_create(sock, family, type, protocol, kern);
1187 goto out_sock_release;
1193 err = -EAFNOSUPPORT;
1196 module_put(pf->owner);
1203 goto out_sock_release;
1205 EXPORT_SYMBOL(__sock_create);
1207 int sock_create(int family, int type, int protocol, struct socket **res)
1209 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1211 EXPORT_SYMBOL(sock_create);
1213 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1215 return __sock_create(net, family, type, protocol, res, 1);
1217 EXPORT_SYMBOL(sock_create_kern);
1219 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1222 struct socket *sock;
1225 /* Check the SOCK_* constants for consistency. */
1226 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1227 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1228 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1229 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1231 flags = type & ~SOCK_TYPE_MASK;
1232 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1234 type &= SOCK_TYPE_MASK;
1236 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1237 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1239 retval = sock_create(family, type, protocol, &sock);
1243 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1248 /* It may be already another descriptor 8) Not kernel problem. */
1257 * Create a pair of connected sockets.
1260 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1261 int __user *, usockvec)
1263 struct socket *sock1, *sock2;
1265 struct file *newfile1, *newfile2;
1268 flags = type & ~SOCK_TYPE_MASK;
1269 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1271 type &= SOCK_TYPE_MASK;
1273 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1274 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1277 * Obtain the first socket and check if the underlying protocol
1278 * supports the socketpair call.
1281 err = sock_create(family, type, protocol, &sock1);
1285 err = sock_create(family, type, protocol, &sock2);
1289 err = sock1->ops->socketpair(sock1, sock2);
1291 goto out_release_both;
1293 fd1 = get_unused_fd_flags(flags);
1294 if (unlikely(fd1 < 0)) {
1296 goto out_release_both;
1299 fd2 = get_unused_fd_flags(flags);
1300 if (unlikely(fd2 < 0)) {
1302 goto out_put_unused_1;
1305 newfile1 = sock_alloc_file(sock1, flags, NULL);
1306 if (unlikely(IS_ERR(newfile1))) {
1307 err = PTR_ERR(newfile1);
1308 goto out_put_unused_both;
1311 newfile2 = sock_alloc_file(sock2, flags, NULL);
1312 if (IS_ERR(newfile2)) {
1313 err = PTR_ERR(newfile2);
1317 err = put_user(fd1, &usockvec[0]);
1321 err = put_user(fd2, &usockvec[1]);
1325 audit_fd_pair(fd1, fd2);
1327 fd_install(fd1, newfile1);
1328 fd_install(fd2, newfile2);
1329 /* fd1 and fd2 may be already another descriptors.
1330 * Not kernel problem.
1346 sock_release(sock2);
1349 out_put_unused_both:
1354 sock_release(sock2);
1356 sock_release(sock1);
1362 * Bind a name to a socket. Nothing much to do here since it's
1363 * the protocol's responsibility to handle the local address.
1365 * We move the socket address to kernel space before we call
1366 * the protocol layer (having also checked the address is ok).
1369 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1371 struct socket *sock;
1372 struct sockaddr_storage address;
1373 int err, fput_needed;
1375 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1377 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1379 err = security_socket_bind(sock,
1380 (struct sockaddr *)&address,
1383 err = sock->ops->bind(sock,
1387 fput_light(sock->file, fput_needed);
1393 * Perform a listen. Basically, we allow the protocol to do anything
1394 * necessary for a listen, and if that works, we mark the socket as
1395 * ready for listening.
1398 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1400 struct socket *sock;
1401 int err, fput_needed;
1404 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1406 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1407 if ((unsigned int)backlog > somaxconn)
1408 backlog = somaxconn;
1410 err = security_socket_listen(sock, backlog);
1412 err = sock->ops->listen(sock, backlog);
1414 fput_light(sock->file, fput_needed);
1420 * For accept, we attempt to create a new socket, set up the link
1421 * with the client, wake up the client, then return the new
1422 * connected fd. We collect the address of the connector in kernel
1423 * space and move it to user at the very end. This is unclean because
1424 * we open the socket then return an error.
1426 * 1003.1g adds the ability to recvmsg() to query connection pending
1427 * status to recvmsg. We need to add that support in a way thats
1428 * clean when we restucture accept also.
1431 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1432 int __user *, upeer_addrlen, int, flags)
1434 struct socket *sock, *newsock;
1435 struct file *newfile;
1436 int err, len, newfd, fput_needed;
1437 struct sockaddr_storage address;
1439 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1442 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1443 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1445 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1450 newsock = sock_alloc();
1454 newsock->type = sock->type;
1455 newsock->ops = sock->ops;
1458 * We don't need try_module_get here, as the listening socket (sock)
1459 * has the protocol module (sock->ops->owner) held.
1461 __module_get(newsock->ops->owner);
1463 newfd = get_unused_fd_flags(flags);
1464 if (unlikely(newfd < 0)) {
1466 sock_release(newsock);
1469 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1470 if (unlikely(IS_ERR(newfile))) {
1471 err = PTR_ERR(newfile);
1472 put_unused_fd(newfd);
1473 sock_release(newsock);
1477 err = security_socket_accept(sock, newsock);
1481 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1485 if (upeer_sockaddr) {
1486 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1488 err = -ECONNABORTED;
1491 err = move_addr_to_user(&address,
1492 len, upeer_sockaddr, upeer_addrlen);
1497 /* File flags are not inherited via accept() unlike another OSes. */
1499 fd_install(newfd, newfile);
1503 fput_light(sock->file, fput_needed);
1508 put_unused_fd(newfd);
1512 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1513 int __user *, upeer_addrlen)
1515 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1519 * Attempt to connect to a socket with the server address. The address
1520 * is in user space so we verify it is OK and move it to kernel space.
1522 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1525 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1526 * other SEQPACKET protocols that take time to connect() as it doesn't
1527 * include the -EINPROGRESS status for such sockets.
1530 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1533 struct socket *sock;
1534 struct sockaddr_storage address;
1535 int err, fput_needed;
1537 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1540 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1545 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1549 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1550 sock->file->f_flags);
1552 fput_light(sock->file, fput_needed);
1558 * Get the local address ('name') of a socket object. Move the obtained
1559 * name to user space.
1562 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1563 int __user *, usockaddr_len)
1565 struct socket *sock;
1566 struct sockaddr_storage address;
1567 int len, err, fput_needed;
1569 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1573 err = security_socket_getsockname(sock);
1577 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1580 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1583 fput_light(sock->file, fput_needed);
1589 * Get the remote address ('name') of a socket object. Move the obtained
1590 * name to user space.
1593 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1594 int __user *, usockaddr_len)
1596 struct socket *sock;
1597 struct sockaddr_storage address;
1598 int len, err, fput_needed;
1600 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1602 err = security_socket_getpeername(sock);
1604 fput_light(sock->file, fput_needed);
1609 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1612 err = move_addr_to_user(&address, len, usockaddr,
1614 fput_light(sock->file, fput_needed);
1620 * Send a datagram to a given address. We move the address into kernel
1621 * space and check the user space data area is readable before invoking
1625 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1626 unsigned int, flags, struct sockaddr __user *, addr,
1629 struct socket *sock;
1630 struct sockaddr_storage address;
1636 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1639 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1643 msg.msg_name = NULL;
1644 msg.msg_control = NULL;
1645 msg.msg_controllen = 0;
1646 msg.msg_namelen = 0;
1648 err = move_addr_to_kernel(addr, addr_len, &address);
1651 msg.msg_name = (struct sockaddr *)&address;
1652 msg.msg_namelen = addr_len;
1654 if (sock->file->f_flags & O_NONBLOCK)
1655 flags |= MSG_DONTWAIT;
1656 msg.msg_flags = flags;
1657 err = sock_sendmsg(sock, &msg);
1660 fput_light(sock->file, fput_needed);
1666 * Send a datagram down a socket.
1669 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1670 unsigned int, flags)
1672 return sys_sendto(fd, buff, len, flags, NULL, 0);
1676 * Receive a frame from the socket and optionally record the address of the
1677 * sender. We verify the buffers are writable and if needed move the
1678 * sender address from kernel to user space.
1681 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1682 unsigned int, flags, struct sockaddr __user *, addr,
1683 int __user *, addr_len)
1685 struct socket *sock;
1688 struct sockaddr_storage address;
1692 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1695 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1699 msg.msg_control = NULL;
1700 msg.msg_controllen = 0;
1701 /* Save some cycles and don't copy the address if not needed */
1702 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1703 /* We assume all kernel code knows the size of sockaddr_storage */
1704 msg.msg_namelen = 0;
1705 if (sock->file->f_flags & O_NONBLOCK)
1706 flags |= MSG_DONTWAIT;
1707 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1709 if (err >= 0 && addr != NULL) {
1710 err2 = move_addr_to_user(&address,
1711 msg.msg_namelen, addr, addr_len);
1716 fput_light(sock->file, fput_needed);
1722 * Receive a datagram from a socket.
1725 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1726 unsigned int, flags)
1728 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1732 * Set a socket option. Because we don't know the option lengths we have
1733 * to pass the user mode parameter for the protocols to sort out.
1736 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1737 char __user *, optval, int, optlen)
1739 int err, fput_needed;
1740 struct socket *sock;
1745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1747 err = security_socket_setsockopt(sock, level, optname);
1751 if (level == SOL_SOCKET)
1753 sock_setsockopt(sock, level, optname, optval,
1757 sock->ops->setsockopt(sock, level, optname, optval,
1760 fput_light(sock->file, fput_needed);
1766 * Get a socket option. Because we don't know the option lengths we have
1767 * to pass a user mode parameter for the protocols to sort out.
1770 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1771 char __user *, optval, int __user *, optlen)
1773 int err, fput_needed;
1774 struct socket *sock;
1776 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1778 err = security_socket_getsockopt(sock, level, optname);
1782 if (level == SOL_SOCKET)
1784 sock_getsockopt(sock, level, optname, optval,
1788 sock->ops->getsockopt(sock, level, optname, optval,
1791 fput_light(sock->file, fput_needed);
1797 * Shutdown a socket.
1800 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1802 int err, fput_needed;
1803 struct socket *sock;
1805 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1807 err = security_socket_shutdown(sock, how);
1809 err = sock->ops->shutdown(sock, how);
1810 fput_light(sock->file, fput_needed);
1815 /* A couple of helpful macros for getting the address of the 32/64 bit
1816 * fields which are the same type (int / unsigned) on our platforms.
1818 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1819 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1820 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1822 struct used_address {
1823 struct sockaddr_storage name;
1824 unsigned int name_len;
1827 static int copy_msghdr_from_user(struct msghdr *kmsg,
1828 struct user_msghdr __user *umsg,
1829 struct sockaddr __user **save_addr,
1832 struct sockaddr __user *uaddr;
1833 struct iovec __user *uiov;
1837 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1838 __get_user(uaddr, &umsg->msg_name) ||
1839 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1840 __get_user(uiov, &umsg->msg_iov) ||
1841 __get_user(nr_segs, &umsg->msg_iovlen) ||
1842 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1843 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1844 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1848 kmsg->msg_namelen = 0;
1850 if (kmsg->msg_namelen < 0)
1853 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1854 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1859 if (uaddr && kmsg->msg_namelen) {
1861 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1867 kmsg->msg_name = NULL;
1868 kmsg->msg_namelen = 0;
1871 if (nr_segs > UIO_MAXIOV)
1874 kmsg->msg_iocb = NULL;
1876 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1877 UIO_FASTIOV, iov, &kmsg->msg_iter);
1880 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1881 struct msghdr *msg_sys, unsigned int flags,
1882 struct used_address *used_address)
1884 struct compat_msghdr __user *msg_compat =
1885 (struct compat_msghdr __user *)msg;
1886 struct sockaddr_storage address;
1887 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1888 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1889 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1890 /* 20 is size of ipv6_pktinfo */
1891 unsigned char *ctl_buf = ctl;
1895 msg_sys->msg_name = &address;
1897 if (MSG_CMSG_COMPAT & flags)
1898 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1900 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1906 if (msg_sys->msg_controllen > INT_MAX)
1908 ctl_len = msg_sys->msg_controllen;
1909 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1911 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1915 ctl_buf = msg_sys->msg_control;
1916 ctl_len = msg_sys->msg_controllen;
1917 } else if (ctl_len) {
1918 if (ctl_len > sizeof(ctl)) {
1919 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1920 if (ctl_buf == NULL)
1925 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1926 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1927 * checking falls down on this.
1929 if (copy_from_user(ctl_buf,
1930 (void __user __force *)msg_sys->msg_control,
1933 msg_sys->msg_control = ctl_buf;
1935 msg_sys->msg_flags = flags;
1937 if (sock->file->f_flags & O_NONBLOCK)
1938 msg_sys->msg_flags |= MSG_DONTWAIT;
1940 * If this is sendmmsg() and current destination address is same as
1941 * previously succeeded address, omit asking LSM's decision.
1942 * used_address->name_len is initialized to UINT_MAX so that the first
1943 * destination address never matches.
1945 if (used_address && msg_sys->msg_name &&
1946 used_address->name_len == msg_sys->msg_namelen &&
1947 !memcmp(&used_address->name, msg_sys->msg_name,
1948 used_address->name_len)) {
1949 err = sock_sendmsg_nosec(sock, msg_sys);
1952 err = sock_sendmsg(sock, msg_sys);
1954 * If this is sendmmsg() and sending to current destination address was
1955 * successful, remember it.
1957 if (used_address && err >= 0) {
1958 used_address->name_len = msg_sys->msg_namelen;
1959 if (msg_sys->msg_name)
1960 memcpy(&used_address->name, msg_sys->msg_name,
1961 used_address->name_len);
1966 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1973 * BSD sendmsg interface
1976 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1978 int fput_needed, err;
1979 struct msghdr msg_sys;
1980 struct socket *sock;
1982 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1986 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1988 fput_light(sock->file, fput_needed);
1993 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1995 if (flags & MSG_CMSG_COMPAT)
1997 return __sys_sendmsg(fd, msg, flags);
2001 * Linux sendmmsg interface
2004 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2007 int fput_needed, err, datagrams;
2008 struct socket *sock;
2009 struct mmsghdr __user *entry;
2010 struct compat_mmsghdr __user *compat_entry;
2011 struct msghdr msg_sys;
2012 struct used_address used_address;
2014 if (vlen > UIO_MAXIOV)
2019 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2023 used_address.name_len = UINT_MAX;
2025 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2028 while (datagrams < vlen) {
2029 if (MSG_CMSG_COMPAT & flags) {
2030 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2031 &msg_sys, flags, &used_address);
2034 err = __put_user(err, &compat_entry->msg_len);
2037 err = ___sys_sendmsg(sock,
2038 (struct user_msghdr __user *)entry,
2039 &msg_sys, flags, &used_address);
2042 err = put_user(err, &entry->msg_len);
2051 fput_light(sock->file, fput_needed);
2053 /* We only return an error if no datagrams were able to be sent */
2060 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2061 unsigned int, vlen, unsigned int, flags)
2063 if (flags & MSG_CMSG_COMPAT)
2065 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2068 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2069 struct msghdr *msg_sys, unsigned int flags, int nosec)
2071 struct compat_msghdr __user *msg_compat =
2072 (struct compat_msghdr __user *)msg;
2073 struct iovec iovstack[UIO_FASTIOV];
2074 struct iovec *iov = iovstack;
2075 unsigned long cmsg_ptr;
2079 /* kernel mode address */
2080 struct sockaddr_storage addr;
2082 /* user mode address pointers */
2083 struct sockaddr __user *uaddr;
2084 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2086 msg_sys->msg_name = &addr;
2088 if (MSG_CMSG_COMPAT & flags)
2089 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2091 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2094 total_len = iov_iter_count(&msg_sys->msg_iter);
2096 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2097 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2099 /* We assume all kernel code knows the size of sockaddr_storage */
2100 msg_sys->msg_namelen = 0;
2102 if (sock->file->f_flags & O_NONBLOCK)
2103 flags |= MSG_DONTWAIT;
2104 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2110 if (uaddr != NULL) {
2111 err = move_addr_to_user(&addr,
2112 msg_sys->msg_namelen, uaddr,
2117 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2121 if (MSG_CMSG_COMPAT & flags)
2122 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2123 &msg_compat->msg_controllen);
2125 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2126 &msg->msg_controllen);
2137 * BSD recvmsg interface
2140 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2142 int fput_needed, err;
2143 struct msghdr msg_sys;
2144 struct socket *sock;
2146 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2150 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2152 fput_light(sock->file, fput_needed);
2157 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2158 unsigned int, flags)
2160 if (flags & MSG_CMSG_COMPAT)
2162 return __sys_recvmsg(fd, msg, flags);
2166 * Linux recvmmsg interface
2169 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2170 unsigned int flags, struct timespec *timeout)
2172 int fput_needed, err, datagrams;
2173 struct socket *sock;
2174 struct mmsghdr __user *entry;
2175 struct compat_mmsghdr __user *compat_entry;
2176 struct msghdr msg_sys;
2177 struct timespec end_time;
2180 poll_select_set_timeout(&end_time, timeout->tv_sec,
2186 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2190 err = sock_error(sock->sk);
2195 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2197 while (datagrams < vlen) {
2199 * No need to ask LSM for more than the first datagram.
2201 if (MSG_CMSG_COMPAT & flags) {
2202 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2203 &msg_sys, flags & ~MSG_WAITFORONE,
2207 err = __put_user(err, &compat_entry->msg_len);
2210 err = ___sys_recvmsg(sock,
2211 (struct user_msghdr __user *)entry,
2212 &msg_sys, flags & ~MSG_WAITFORONE,
2216 err = put_user(err, &entry->msg_len);
2224 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2225 if (flags & MSG_WAITFORONE)
2226 flags |= MSG_DONTWAIT;
2229 ktime_get_ts(timeout);
2230 *timeout = timespec_sub(end_time, *timeout);
2231 if (timeout->tv_sec < 0) {
2232 timeout->tv_sec = timeout->tv_nsec = 0;
2236 /* Timeout, return less than vlen datagrams */
2237 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2241 /* Out of band data, return right away */
2242 if (msg_sys.msg_flags & MSG_OOB)
2247 fput_light(sock->file, fput_needed);
2252 if (datagrams != 0) {
2254 * We may return less entries than requested (vlen) if the
2255 * sock is non block and there aren't enough datagrams...
2257 if (err != -EAGAIN) {
2259 * ... or if recvmsg returns an error after we
2260 * received some datagrams, where we record the
2261 * error to return on the next call or if the
2262 * app asks about it using getsockopt(SO_ERROR).
2264 sock->sk->sk_err = -err;
2273 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2274 unsigned int, vlen, unsigned int, flags,
2275 struct timespec __user *, timeout)
2278 struct timespec timeout_sys;
2280 if (flags & MSG_CMSG_COMPAT)
2284 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2286 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2289 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2291 if (datagrams > 0 &&
2292 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2293 datagrams = -EFAULT;
2298 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2299 /* Argument list sizes for sys_socketcall */
2300 #define AL(x) ((x) * sizeof(unsigned long))
2301 static const unsigned char nargs[21] = {
2302 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2303 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2304 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2311 * System call vectors.
2313 * Argument checking cleaned up. Saved 20% in size.
2314 * This function doesn't need to set the kernel lock because
2315 * it is set by the callees.
2318 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2320 unsigned long a[AUDITSC_ARGS];
2321 unsigned long a0, a1;
2325 if (call < 1 || call > SYS_SENDMMSG)
2329 if (len > sizeof(a))
2332 /* copy_from_user should be SMP safe. */
2333 if (copy_from_user(a, args, len))
2336 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2345 err = sys_socket(a0, a1, a[2]);
2348 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2351 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2354 err = sys_listen(a0, a1);
2357 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2358 (int __user *)a[2], 0);
2360 case SYS_GETSOCKNAME:
2362 sys_getsockname(a0, (struct sockaddr __user *)a1,
2363 (int __user *)a[2]);
2365 case SYS_GETPEERNAME:
2367 sys_getpeername(a0, (struct sockaddr __user *)a1,
2368 (int __user *)a[2]);
2370 case SYS_SOCKETPAIR:
2371 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2374 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2377 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2378 (struct sockaddr __user *)a[4], a[5]);
2381 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2384 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2385 (struct sockaddr __user *)a[4],
2386 (int __user *)a[5]);
2389 err = sys_shutdown(a0, a1);
2391 case SYS_SETSOCKOPT:
2392 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2394 case SYS_GETSOCKOPT:
2396 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2397 (int __user *)a[4]);
2400 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2403 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2406 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2409 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2410 (struct timespec __user *)a[4]);
2413 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2414 (int __user *)a[2], a[3]);
2423 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2426 * sock_register - add a socket protocol handler
2427 * @ops: description of protocol
2429 * This function is called by a protocol handler that wants to
2430 * advertise its address family, and have it linked into the
2431 * socket interface. The value ops->family corresponds to the
2432 * socket system call protocol family.
2434 int sock_register(const struct net_proto_family *ops)
2438 if (ops->family >= NPROTO) {
2439 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2443 spin_lock(&net_family_lock);
2444 if (rcu_dereference_protected(net_families[ops->family],
2445 lockdep_is_held(&net_family_lock)))
2448 rcu_assign_pointer(net_families[ops->family], ops);
2451 spin_unlock(&net_family_lock);
2453 pr_info("NET: Registered protocol family %d\n", ops->family);
2456 EXPORT_SYMBOL(sock_register);
2459 * sock_unregister - remove a protocol handler
2460 * @family: protocol family to remove
2462 * This function is called by a protocol handler that wants to
2463 * remove its address family, and have it unlinked from the
2464 * new socket creation.
2466 * If protocol handler is a module, then it can use module reference
2467 * counts to protect against new references. If protocol handler is not
2468 * a module then it needs to provide its own protection in
2469 * the ops->create routine.
2471 void sock_unregister(int family)
2473 BUG_ON(family < 0 || family >= NPROTO);
2475 spin_lock(&net_family_lock);
2476 RCU_INIT_POINTER(net_families[family], NULL);
2477 spin_unlock(&net_family_lock);
2481 pr_info("NET: Unregistered protocol family %d\n", family);
2483 EXPORT_SYMBOL(sock_unregister);
2485 static int __init sock_init(void)
2489 * Initialize the network sysctl infrastructure.
2491 err = net_sysctl_init();
2496 * Initialize skbuff SLAB cache
2501 * Initialize the protocols module.
2506 err = register_filesystem(&sock_fs_type);
2509 sock_mnt = kern_mount(&sock_fs_type);
2510 if (IS_ERR(sock_mnt)) {
2511 err = PTR_ERR(sock_mnt);
2515 /* The real protocol initialization is performed in later initcalls.
2518 #ifdef CONFIG_NETFILTER
2519 err = netfilter_init();
2524 ptp_classifier_init();
2530 unregister_filesystem(&sock_fs_type);
2535 core_initcall(sock_init); /* early initcall */
2537 #ifdef CONFIG_PROC_FS
2538 void socket_seq_show(struct seq_file *seq)
2543 for_each_possible_cpu(cpu)
2544 counter += per_cpu(sockets_in_use, cpu);
2546 /* It can be negative, by the way. 8) */
2550 seq_printf(seq, "sockets: used %d\n", counter);
2552 #endif /* CONFIG_PROC_FS */
2554 #ifdef CONFIG_COMPAT
2555 static int do_siocgstamp(struct net *net, struct socket *sock,
2556 unsigned int cmd, void __user *up)
2558 mm_segment_t old_fs = get_fs();
2563 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2566 err = compat_put_timeval(&ktv, up);
2571 static int do_siocgstampns(struct net *net, struct socket *sock,
2572 unsigned int cmd, void __user *up)
2574 mm_segment_t old_fs = get_fs();
2575 struct timespec kts;
2579 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2582 err = compat_put_timespec(&kts, up);
2587 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2589 struct ifreq __user *uifr;
2592 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2593 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2596 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2600 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2606 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2608 struct compat_ifconf ifc32;
2610 struct ifconf __user *uifc;
2611 struct compat_ifreq __user *ifr32;
2612 struct ifreq __user *ifr;
2616 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2619 memset(&ifc, 0, sizeof(ifc));
2620 if (ifc32.ifcbuf == 0) {
2624 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2626 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2627 sizeof(struct ifreq);
2628 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2630 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2631 ifr32 = compat_ptr(ifc32.ifcbuf);
2632 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2633 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2639 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2642 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2646 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2650 ifr32 = compat_ptr(ifc32.ifcbuf);
2652 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2653 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2654 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2660 if (ifc32.ifcbuf == 0) {
2661 /* Translate from 64-bit structure multiple to
2665 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2670 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2676 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2678 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2679 bool convert_in = false, convert_out = false;
2680 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2681 struct ethtool_rxnfc __user *rxnfc;
2682 struct ifreq __user *ifr;
2683 u32 rule_cnt = 0, actual_rule_cnt;
2688 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2691 compat_rxnfc = compat_ptr(data);
2693 if (get_user(ethcmd, &compat_rxnfc->cmd))
2696 /* Most ethtool structures are defined without padding.
2697 * Unfortunately struct ethtool_rxnfc is an exception.
2702 case ETHTOOL_GRXCLSRLALL:
2703 /* Buffer size is variable */
2704 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2706 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2708 buf_size += rule_cnt * sizeof(u32);
2710 case ETHTOOL_GRXRINGS:
2711 case ETHTOOL_GRXCLSRLCNT:
2712 case ETHTOOL_GRXCLSRULE:
2713 case ETHTOOL_SRXCLSRLINS:
2716 case ETHTOOL_SRXCLSRLDEL:
2717 buf_size += sizeof(struct ethtool_rxnfc);
2722 ifr = compat_alloc_user_space(buf_size);
2723 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2725 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2728 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2729 &ifr->ifr_ifru.ifru_data))
2733 /* We expect there to be holes between fs.m_ext and
2734 * fs.ring_cookie and at the end of fs, but nowhere else.
2736 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2737 sizeof(compat_rxnfc->fs.m_ext) !=
2738 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2739 sizeof(rxnfc->fs.m_ext));
2741 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2742 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2743 offsetof(struct ethtool_rxnfc, fs.location) -
2744 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2746 if (copy_in_user(rxnfc, compat_rxnfc,
2747 (void __user *)(&rxnfc->fs.m_ext + 1) -
2748 (void __user *)rxnfc) ||
2749 copy_in_user(&rxnfc->fs.ring_cookie,
2750 &compat_rxnfc->fs.ring_cookie,
2751 (void __user *)(&rxnfc->fs.location + 1) -
2752 (void __user *)&rxnfc->fs.ring_cookie) ||
2753 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2754 sizeof(rxnfc->rule_cnt)))
2758 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2763 if (copy_in_user(compat_rxnfc, rxnfc,
2764 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2765 (const void __user *)rxnfc) ||
2766 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2767 &rxnfc->fs.ring_cookie,
2768 (const void __user *)(&rxnfc->fs.location + 1) -
2769 (const void __user *)&rxnfc->fs.ring_cookie) ||
2770 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2771 sizeof(rxnfc->rule_cnt)))
2774 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2775 /* As an optimisation, we only copy the actual
2776 * number of rules that the underlying
2777 * function returned. Since Mallory might
2778 * change the rule count in user memory, we
2779 * check that it is less than the rule count
2780 * originally given (as the user buffer size),
2781 * which has been range-checked.
2783 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2785 if (actual_rule_cnt < rule_cnt)
2786 rule_cnt = actual_rule_cnt;
2787 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2788 &rxnfc->rule_locs[0],
2789 rule_cnt * sizeof(u32)))
2797 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2800 compat_uptr_t uptr32;
2801 struct ifreq __user *uifr;
2803 uifr = compat_alloc_user_space(sizeof(*uifr));
2804 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2807 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2810 uptr = compat_ptr(uptr32);
2812 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2815 return dev_ioctl(net, SIOCWANDEV, uifr);
2818 static int bond_ioctl(struct net *net, unsigned int cmd,
2819 struct compat_ifreq __user *ifr32)
2822 mm_segment_t old_fs;
2826 case SIOCBONDENSLAVE:
2827 case SIOCBONDRELEASE:
2828 case SIOCBONDSETHWADDR:
2829 case SIOCBONDCHANGEACTIVE:
2830 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2835 err = dev_ioctl(net, cmd,
2836 (struct ifreq __user __force *) &kifr);
2841 return -ENOIOCTLCMD;
2845 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2846 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2847 struct compat_ifreq __user *u_ifreq32)
2849 struct ifreq __user *u_ifreq64;
2850 char tmp_buf[IFNAMSIZ];
2851 void __user *data64;
2854 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2857 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2859 data64 = compat_ptr(data32);
2861 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2863 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2866 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2869 return dev_ioctl(net, cmd, u_ifreq64);
2872 static int dev_ifsioc(struct net *net, struct socket *sock,
2873 unsigned int cmd, struct compat_ifreq __user *uifr32)
2875 struct ifreq __user *uifr;
2878 uifr = compat_alloc_user_space(sizeof(*uifr));
2879 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2882 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2893 case SIOCGIFBRDADDR:
2894 case SIOCGIFDSTADDR:
2895 case SIOCGIFNETMASK:
2900 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2908 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2909 struct compat_ifreq __user *uifr32)
2912 struct compat_ifmap __user *uifmap32;
2913 mm_segment_t old_fs;
2916 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2917 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2918 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2919 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2920 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2921 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2922 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2923 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2929 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2932 if (cmd == SIOCGIFMAP && !err) {
2933 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2934 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2935 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2936 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2937 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2938 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2939 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2948 struct sockaddr rt_dst; /* target address */
2949 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2950 struct sockaddr rt_genmask; /* target network mask (IP) */
2951 unsigned short rt_flags;
2954 unsigned char rt_tos;
2955 unsigned char rt_class;
2957 short rt_metric; /* +1 for binary compatibility! */
2958 /* char * */ u32 rt_dev; /* forcing the device at add */
2959 u32 rt_mtu; /* per route MTU/Window */
2960 u32 rt_window; /* Window clamping */
2961 unsigned short rt_irtt; /* Initial RTT */
2964 struct in6_rtmsg32 {
2965 struct in6_addr rtmsg_dst;
2966 struct in6_addr rtmsg_src;
2967 struct in6_addr rtmsg_gateway;
2977 static int routing_ioctl(struct net *net, struct socket *sock,
2978 unsigned int cmd, void __user *argp)
2982 struct in6_rtmsg r6;
2986 mm_segment_t old_fs = get_fs();
2988 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2989 struct in6_rtmsg32 __user *ur6 = argp;
2990 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2991 3 * sizeof(struct in6_addr));
2992 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2993 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2994 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2995 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2996 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2997 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2998 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3002 struct rtentry32 __user *ur4 = argp;
3003 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3004 3 * sizeof(struct sockaddr));
3005 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3006 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3007 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3008 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3009 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3010 ret |= get_user(rtdev, &(ur4->rt_dev));
3012 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3013 r4.rt_dev = (char __user __force *)devname;
3027 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3034 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3035 * for some operations; this forces use of the newer bridge-utils that
3036 * use compatible ioctls
3038 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3042 if (get_user(tmp, argp))
3044 if (tmp == BRCTL_GET_VERSION)
3045 return BRCTL_VERSION + 1;
3049 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3050 unsigned int cmd, unsigned long arg)
3052 void __user *argp = compat_ptr(arg);
3053 struct sock *sk = sock->sk;
3054 struct net *net = sock_net(sk);
3056 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3057 return compat_ifr_data_ioctl(net, cmd, argp);
3062 return old_bridge_ioctl(argp);
3064 return dev_ifname32(net, argp);
3066 return dev_ifconf(net, argp);
3068 return ethtool_ioctl(net, argp);
3070 return compat_siocwandev(net, argp);
3073 return compat_sioc_ifmap(net, cmd, argp);
3074 case SIOCBONDENSLAVE:
3075 case SIOCBONDRELEASE:
3076 case SIOCBONDSETHWADDR:
3077 case SIOCBONDCHANGEACTIVE:
3078 return bond_ioctl(net, cmd, argp);
3081 return routing_ioctl(net, sock, cmd, argp);
3083 return do_siocgstamp(net, sock, cmd, argp);
3085 return do_siocgstampns(net, sock, cmd, argp);
3086 case SIOCBONDSLAVEINFOQUERY:
3087 case SIOCBONDINFOQUERY:
3090 return compat_ifr_data_ioctl(net, cmd, argp);
3102 return sock_ioctl(file, cmd, arg);
3119 case SIOCSIFHWBROADCAST:
3121 case SIOCGIFBRDADDR:
3122 case SIOCSIFBRDADDR:
3123 case SIOCGIFDSTADDR:
3124 case SIOCSIFDSTADDR:
3125 case SIOCGIFNETMASK:
3126 case SIOCSIFNETMASK:
3137 return dev_ifsioc(net, sock, cmd, argp);
3143 return sock_do_ioctl(net, sock, cmd, arg);
3146 return -ENOIOCTLCMD;
3149 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3152 struct socket *sock = file->private_data;
3153 int ret = -ENOIOCTLCMD;
3160 if (sock->ops->compat_ioctl)
3161 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3163 if (ret == -ENOIOCTLCMD &&
3164 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3165 ret = compat_wext_handle_ioctl(net, cmd, arg);
3167 if (ret == -ENOIOCTLCMD)
3168 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3174 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3176 return sock->ops->bind(sock, addr, addrlen);
3178 EXPORT_SYMBOL(kernel_bind);
3180 int kernel_listen(struct socket *sock, int backlog)
3182 return sock->ops->listen(sock, backlog);
3184 EXPORT_SYMBOL(kernel_listen);
3186 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3188 struct sock *sk = sock->sk;
3191 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3196 err = sock->ops->accept(sock, *newsock, flags);
3198 sock_release(*newsock);
3203 (*newsock)->ops = sock->ops;
3204 __module_get((*newsock)->ops->owner);
3209 EXPORT_SYMBOL(kernel_accept);
3211 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3214 return sock->ops->connect(sock, addr, addrlen, flags);
3216 EXPORT_SYMBOL(kernel_connect);
3218 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3221 return sock->ops->getname(sock, addr, addrlen, 0);
3223 EXPORT_SYMBOL(kernel_getsockname);
3225 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3228 return sock->ops->getname(sock, addr, addrlen, 1);
3230 EXPORT_SYMBOL(kernel_getpeername);
3232 int kernel_getsockopt(struct socket *sock, int level, int optname,
3233 char *optval, int *optlen)
3235 mm_segment_t oldfs = get_fs();
3236 char __user *uoptval;
3237 int __user *uoptlen;
3240 uoptval = (char __user __force *) optval;
3241 uoptlen = (int __user __force *) optlen;
3244 if (level == SOL_SOCKET)
3245 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3247 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3252 EXPORT_SYMBOL(kernel_getsockopt);
3254 int kernel_setsockopt(struct socket *sock, int level, int optname,
3255 char *optval, unsigned int optlen)
3257 mm_segment_t oldfs = get_fs();
3258 char __user *uoptval;
3261 uoptval = (char __user __force *) optval;
3264 if (level == SOL_SOCKET)
3265 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3267 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3272 EXPORT_SYMBOL(kernel_setsockopt);
3274 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3275 size_t size, int flags)
3277 if (sock->ops->sendpage)
3278 return sock->ops->sendpage(sock, page, offset, size, flags);
3280 return sock_no_sendpage(sock, page, offset, size, flags);
3282 EXPORT_SYMBOL(kernel_sendpage);
3284 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3286 mm_segment_t oldfs = get_fs();
3290 err = sock->ops->ioctl(sock, cmd, arg);
3295 EXPORT_SYMBOL(kernel_sock_ioctl);
3297 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3299 return sock->ops->shutdown(sock, how);
3301 EXPORT_SYMBOL(kernel_sock_shutdown);