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,
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 rcu_lock */
1061 int sock_wake_async(struct socket_wq *wq, int how, int band)
1063 if (!wq || !wq->fasync_list)
1067 case SOCK_WAKE_WAITD:
1068 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1071 case SOCK_WAKE_SPACE:
1072 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1077 kill_fasync(&wq->fasync_list, SIGIO, band);
1080 kill_fasync(&wq->fasync_list, SIGURG, band);
1085 EXPORT_SYMBOL(sock_wake_async);
1087 int __sock_create(struct net *net, int family, int type, int protocol,
1088 struct socket **res, int kern)
1091 struct socket *sock;
1092 const struct net_proto_family *pf;
1095 * Check protocol is in range
1097 if (family < 0 || family >= NPROTO)
1098 return -EAFNOSUPPORT;
1099 if (type < 0 || type >= SOCK_MAX)
1104 This uglymoron is moved from INET layer to here to avoid
1105 deadlock in module load.
1107 if (family == PF_INET && type == SOCK_PACKET) {
1111 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1117 err = security_socket_create(family, type, protocol, kern);
1122 * Allocate the socket and allow the family to set things up. if
1123 * the protocol is 0, the family is instructed to select an appropriate
1126 sock = sock_alloc();
1128 net_warn_ratelimited("socket: no more sockets\n");
1129 return -ENFILE; /* Not exactly a match, but its the
1130 closest posix thing */
1135 #ifdef CONFIG_MODULES
1136 /* Attempt to load a protocol module if the find failed.
1138 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1139 * requested real, full-featured networking support upon configuration.
1140 * Otherwise module support will break!
1142 if (rcu_access_pointer(net_families[family]) == NULL)
1143 request_module("net-pf-%d", family);
1147 pf = rcu_dereference(net_families[family]);
1148 err = -EAFNOSUPPORT;
1153 * We will call the ->create function, that possibly is in a loadable
1154 * module, so we have to bump that loadable module refcnt first.
1156 if (!try_module_get(pf->owner))
1159 /* Now protected by module ref count */
1162 err = pf->create(net, sock, protocol, kern);
1164 goto out_module_put;
1167 * Now to bump the refcnt of the [loadable] module that owns this
1168 * socket at sock_release time we decrement its refcnt.
1170 if (!try_module_get(sock->ops->owner))
1171 goto out_module_busy;
1174 * Now that we're done with the ->create function, the [loadable]
1175 * module can have its refcnt decremented
1177 module_put(pf->owner);
1178 err = security_socket_post_create(sock, family, type, protocol, kern);
1180 goto out_sock_release;
1186 err = -EAFNOSUPPORT;
1189 module_put(pf->owner);
1196 goto out_sock_release;
1198 EXPORT_SYMBOL(__sock_create);
1200 int sock_create(int family, int type, int protocol, struct socket **res)
1202 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1204 EXPORT_SYMBOL(sock_create);
1206 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1208 return __sock_create(net, family, type, protocol, res, 1);
1210 EXPORT_SYMBOL(sock_create_kern);
1212 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1215 struct socket *sock;
1218 /* Check the SOCK_* constants for consistency. */
1219 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1220 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1221 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1222 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1224 flags = type & ~SOCK_TYPE_MASK;
1225 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1227 type &= SOCK_TYPE_MASK;
1229 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1230 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1232 retval = sock_create(family, type, protocol, &sock);
1236 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1241 /* It may be already another descriptor 8) Not kernel problem. */
1250 * Create a pair of connected sockets.
1253 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1254 int __user *, usockvec)
1256 struct socket *sock1, *sock2;
1258 struct file *newfile1, *newfile2;
1261 flags = type & ~SOCK_TYPE_MASK;
1262 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1264 type &= SOCK_TYPE_MASK;
1266 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1267 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1270 * Obtain the first socket and check if the underlying protocol
1271 * supports the socketpair call.
1274 err = sock_create(family, type, protocol, &sock1);
1278 err = sock_create(family, type, protocol, &sock2);
1282 err = sock1->ops->socketpair(sock1, sock2);
1284 goto out_release_both;
1286 fd1 = get_unused_fd_flags(flags);
1287 if (unlikely(fd1 < 0)) {
1289 goto out_release_both;
1292 fd2 = get_unused_fd_flags(flags);
1293 if (unlikely(fd2 < 0)) {
1295 goto out_put_unused_1;
1298 newfile1 = sock_alloc_file(sock1, flags, NULL);
1299 if (IS_ERR(newfile1)) {
1300 err = PTR_ERR(newfile1);
1301 goto out_put_unused_both;
1304 newfile2 = sock_alloc_file(sock2, flags, NULL);
1305 if (IS_ERR(newfile2)) {
1306 err = PTR_ERR(newfile2);
1310 err = put_user(fd1, &usockvec[0]);
1314 err = put_user(fd2, &usockvec[1]);
1318 audit_fd_pair(fd1, fd2);
1320 fd_install(fd1, newfile1);
1321 fd_install(fd2, newfile2);
1322 /* fd1 and fd2 may be already another descriptors.
1323 * Not kernel problem.
1339 sock_release(sock2);
1342 out_put_unused_both:
1347 sock_release(sock2);
1349 sock_release(sock1);
1355 * Bind a name to a socket. Nothing much to do here since it's
1356 * the protocol's responsibility to handle the local address.
1358 * We move the socket address to kernel space before we call
1359 * the protocol layer (having also checked the address is ok).
1362 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1364 struct socket *sock;
1365 struct sockaddr_storage address;
1366 int err, fput_needed;
1368 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1370 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1372 err = security_socket_bind(sock,
1373 (struct sockaddr *)&address,
1376 err = sock->ops->bind(sock,
1380 fput_light(sock->file, fput_needed);
1386 * Perform a listen. Basically, we allow the protocol to do anything
1387 * necessary for a listen, and if that works, we mark the socket as
1388 * ready for listening.
1391 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1393 struct socket *sock;
1394 int err, fput_needed;
1397 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1399 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1400 if ((unsigned int)backlog > somaxconn)
1401 backlog = somaxconn;
1403 err = security_socket_listen(sock, backlog);
1405 err = sock->ops->listen(sock, backlog);
1407 fput_light(sock->file, fput_needed);
1413 * For accept, we attempt to create a new socket, set up the link
1414 * with the client, wake up the client, then return the new
1415 * connected fd. We collect the address of the connector in kernel
1416 * space and move it to user at the very end. This is unclean because
1417 * we open the socket then return an error.
1419 * 1003.1g adds the ability to recvmsg() to query connection pending
1420 * status to recvmsg. We need to add that support in a way thats
1421 * clean when we restucture accept also.
1424 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1425 int __user *, upeer_addrlen, int, flags)
1427 struct socket *sock, *newsock;
1428 struct file *newfile;
1429 int err, len, newfd, fput_needed;
1430 struct sockaddr_storage address;
1432 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1435 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1436 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1438 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1443 newsock = sock_alloc();
1447 newsock->type = sock->type;
1448 newsock->ops = sock->ops;
1451 * We don't need try_module_get here, as the listening socket (sock)
1452 * has the protocol module (sock->ops->owner) held.
1454 __module_get(newsock->ops->owner);
1456 newfd = get_unused_fd_flags(flags);
1457 if (unlikely(newfd < 0)) {
1459 sock_release(newsock);
1462 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1463 if (IS_ERR(newfile)) {
1464 err = PTR_ERR(newfile);
1465 put_unused_fd(newfd);
1466 sock_release(newsock);
1470 err = security_socket_accept(sock, newsock);
1474 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1478 if (upeer_sockaddr) {
1479 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1481 err = -ECONNABORTED;
1484 err = move_addr_to_user(&address,
1485 len, upeer_sockaddr, upeer_addrlen);
1490 /* File flags are not inherited via accept() unlike another OSes. */
1492 fd_install(newfd, newfile);
1496 fput_light(sock->file, fput_needed);
1501 put_unused_fd(newfd);
1505 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1506 int __user *, upeer_addrlen)
1508 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1512 * Attempt to connect to a socket with the server address. The address
1513 * is in user space so we verify it is OK and move it to kernel space.
1515 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1518 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1519 * other SEQPACKET protocols that take time to connect() as it doesn't
1520 * include the -EINPROGRESS status for such sockets.
1523 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1526 struct socket *sock;
1527 struct sockaddr_storage address;
1528 int err, fput_needed;
1530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1533 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1538 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1542 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1543 sock->file->f_flags);
1545 fput_light(sock->file, fput_needed);
1551 * Get the local address ('name') of a socket object. Move the obtained
1552 * name to user space.
1555 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1556 int __user *, usockaddr_len)
1558 struct socket *sock;
1559 struct sockaddr_storage address;
1560 int len, err, fput_needed;
1562 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1566 err = security_socket_getsockname(sock);
1570 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1573 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1576 fput_light(sock->file, fput_needed);
1582 * Get the remote address ('name') of a socket object. Move the obtained
1583 * name to user space.
1586 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1587 int __user *, usockaddr_len)
1589 struct socket *sock;
1590 struct sockaddr_storage address;
1591 int len, err, fput_needed;
1593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1595 err = security_socket_getpeername(sock);
1597 fput_light(sock->file, fput_needed);
1602 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1605 err = move_addr_to_user(&address, len, usockaddr,
1607 fput_light(sock->file, fput_needed);
1613 * Send a datagram to a given address. We move the address into kernel
1614 * space and check the user space data area is readable before invoking
1618 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1619 unsigned int, flags, struct sockaddr __user *, addr,
1622 struct socket *sock;
1623 struct sockaddr_storage address;
1629 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1632 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1636 msg.msg_name = NULL;
1637 msg.msg_control = NULL;
1638 msg.msg_controllen = 0;
1639 msg.msg_namelen = 0;
1641 err = move_addr_to_kernel(addr, addr_len, &address);
1644 msg.msg_name = (struct sockaddr *)&address;
1645 msg.msg_namelen = addr_len;
1647 if (sock->file->f_flags & O_NONBLOCK)
1648 flags |= MSG_DONTWAIT;
1649 msg.msg_flags = flags;
1650 err = sock_sendmsg(sock, &msg);
1653 fput_light(sock->file, fput_needed);
1659 * Send a datagram down a socket.
1662 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1663 unsigned int, flags)
1665 return sys_sendto(fd, buff, len, flags, NULL, 0);
1669 * Receive a frame from the socket and optionally record the address of the
1670 * sender. We verify the buffers are writable and if needed move the
1671 * sender address from kernel to user space.
1674 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1675 unsigned int, flags, struct sockaddr __user *, addr,
1676 int __user *, addr_len)
1678 struct socket *sock;
1681 struct sockaddr_storage address;
1685 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1688 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1692 msg.msg_control = NULL;
1693 msg.msg_controllen = 0;
1694 /* Save some cycles and don't copy the address if not needed */
1695 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1696 /* We assume all kernel code knows the size of sockaddr_storage */
1697 msg.msg_namelen = 0;
1698 msg.msg_iocb = NULL;
1699 if (sock->file->f_flags & O_NONBLOCK)
1700 flags |= MSG_DONTWAIT;
1701 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1703 if (err >= 0 && addr != NULL) {
1704 err2 = move_addr_to_user(&address,
1705 msg.msg_namelen, addr, addr_len);
1710 fput_light(sock->file, fput_needed);
1716 * Receive a datagram from a socket.
1719 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1720 unsigned int, flags)
1722 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1726 * Set a socket option. Because we don't know the option lengths we have
1727 * to pass the user mode parameter for the protocols to sort out.
1730 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1731 char __user *, optval, int, optlen)
1733 int err, fput_needed;
1734 struct socket *sock;
1739 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1741 err = security_socket_setsockopt(sock, level, optname);
1745 if (level == SOL_SOCKET)
1747 sock_setsockopt(sock, level, optname, optval,
1751 sock->ops->setsockopt(sock, level, optname, optval,
1754 fput_light(sock->file, fput_needed);
1760 * Get a socket option. Because we don't know the option lengths we have
1761 * to pass a user mode parameter for the protocols to sort out.
1764 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1765 char __user *, optval, int __user *, optlen)
1767 int err, fput_needed;
1768 struct socket *sock;
1770 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1772 err = security_socket_getsockopt(sock, level, optname);
1776 if (level == SOL_SOCKET)
1778 sock_getsockopt(sock, level, optname, optval,
1782 sock->ops->getsockopt(sock, level, optname, optval,
1785 fput_light(sock->file, fput_needed);
1791 * Shutdown a socket.
1794 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1796 int err, fput_needed;
1797 struct socket *sock;
1799 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1801 err = security_socket_shutdown(sock, how);
1803 err = sock->ops->shutdown(sock, how);
1804 fput_light(sock->file, fput_needed);
1809 /* A couple of helpful macros for getting the address of the 32/64 bit
1810 * fields which are the same type (int / unsigned) on our platforms.
1812 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1813 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1814 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1816 struct used_address {
1817 struct sockaddr_storage name;
1818 unsigned int name_len;
1821 static int copy_msghdr_from_user(struct msghdr *kmsg,
1822 struct user_msghdr __user *umsg,
1823 struct sockaddr __user **save_addr,
1826 struct sockaddr __user *uaddr;
1827 struct iovec __user *uiov;
1831 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1832 __get_user(uaddr, &umsg->msg_name) ||
1833 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1834 __get_user(uiov, &umsg->msg_iov) ||
1835 __get_user(nr_segs, &umsg->msg_iovlen) ||
1836 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1837 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1838 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1842 kmsg->msg_namelen = 0;
1844 if (kmsg->msg_namelen < 0)
1847 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1848 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1853 if (uaddr && kmsg->msg_namelen) {
1855 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1861 kmsg->msg_name = NULL;
1862 kmsg->msg_namelen = 0;
1865 if (nr_segs > UIO_MAXIOV)
1868 kmsg->msg_iocb = NULL;
1870 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1871 UIO_FASTIOV, iov, &kmsg->msg_iter);
1874 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1875 struct msghdr *msg_sys, unsigned int flags,
1876 struct used_address *used_address)
1878 struct compat_msghdr __user *msg_compat =
1879 (struct compat_msghdr __user *)msg;
1880 struct sockaddr_storage address;
1881 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1882 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1883 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1884 /* 20 is size of ipv6_pktinfo */
1885 unsigned char *ctl_buf = ctl;
1889 msg_sys->msg_name = &address;
1891 if (MSG_CMSG_COMPAT & flags)
1892 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1894 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1900 if (msg_sys->msg_controllen > INT_MAX)
1902 ctl_len = msg_sys->msg_controllen;
1903 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1905 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1909 ctl_buf = msg_sys->msg_control;
1910 ctl_len = msg_sys->msg_controllen;
1911 } else if (ctl_len) {
1912 if (ctl_len > sizeof(ctl)) {
1913 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1914 if (ctl_buf == NULL)
1919 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1920 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1921 * checking falls down on this.
1923 if (copy_from_user(ctl_buf,
1924 (void __user __force *)msg_sys->msg_control,
1927 msg_sys->msg_control = ctl_buf;
1929 msg_sys->msg_flags = flags;
1931 if (sock->file->f_flags & O_NONBLOCK)
1932 msg_sys->msg_flags |= MSG_DONTWAIT;
1934 * If this is sendmmsg() and current destination address is same as
1935 * previously succeeded address, omit asking LSM's decision.
1936 * used_address->name_len is initialized to UINT_MAX so that the first
1937 * destination address never matches.
1939 if (used_address && msg_sys->msg_name &&
1940 used_address->name_len == msg_sys->msg_namelen &&
1941 !memcmp(&used_address->name, msg_sys->msg_name,
1942 used_address->name_len)) {
1943 err = sock_sendmsg_nosec(sock, msg_sys);
1946 err = sock_sendmsg(sock, msg_sys);
1948 * If this is sendmmsg() and sending to current destination address was
1949 * successful, remember it.
1951 if (used_address && err >= 0) {
1952 used_address->name_len = msg_sys->msg_namelen;
1953 if (msg_sys->msg_name)
1954 memcpy(&used_address->name, msg_sys->msg_name,
1955 used_address->name_len);
1960 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1967 * BSD sendmsg interface
1970 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1972 int fput_needed, err;
1973 struct msghdr msg_sys;
1974 struct socket *sock;
1976 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1980 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1982 fput_light(sock->file, fput_needed);
1987 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1989 if (flags & MSG_CMSG_COMPAT)
1991 return __sys_sendmsg(fd, msg, flags);
1995 * Linux sendmmsg interface
1998 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2001 int fput_needed, err, datagrams;
2002 struct socket *sock;
2003 struct mmsghdr __user *entry;
2004 struct compat_mmsghdr __user *compat_entry;
2005 struct msghdr msg_sys;
2006 struct used_address used_address;
2008 if (vlen > UIO_MAXIOV)
2013 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2017 used_address.name_len = UINT_MAX;
2019 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2022 while (datagrams < vlen) {
2023 if (MSG_CMSG_COMPAT & flags) {
2024 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2025 &msg_sys, flags, &used_address);
2028 err = __put_user(err, &compat_entry->msg_len);
2031 err = ___sys_sendmsg(sock,
2032 (struct user_msghdr __user *)entry,
2033 &msg_sys, flags, &used_address);
2036 err = put_user(err, &entry->msg_len);
2045 fput_light(sock->file, fput_needed);
2047 /* We only return an error if no datagrams were able to be sent */
2054 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2055 unsigned int, vlen, unsigned int, flags)
2057 if (flags & MSG_CMSG_COMPAT)
2059 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2062 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2063 struct msghdr *msg_sys, unsigned int flags, int nosec)
2065 struct compat_msghdr __user *msg_compat =
2066 (struct compat_msghdr __user *)msg;
2067 struct iovec iovstack[UIO_FASTIOV];
2068 struct iovec *iov = iovstack;
2069 unsigned long cmsg_ptr;
2073 /* kernel mode address */
2074 struct sockaddr_storage addr;
2076 /* user mode address pointers */
2077 struct sockaddr __user *uaddr;
2078 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2080 msg_sys->msg_name = &addr;
2082 if (MSG_CMSG_COMPAT & flags)
2083 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2085 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2088 total_len = iov_iter_count(&msg_sys->msg_iter);
2090 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2091 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2093 /* We assume all kernel code knows the size of sockaddr_storage */
2094 msg_sys->msg_namelen = 0;
2096 if (sock->file->f_flags & O_NONBLOCK)
2097 flags |= MSG_DONTWAIT;
2098 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2104 if (uaddr != NULL) {
2105 err = move_addr_to_user(&addr,
2106 msg_sys->msg_namelen, uaddr,
2111 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2115 if (MSG_CMSG_COMPAT & flags)
2116 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2117 &msg_compat->msg_controllen);
2119 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2120 &msg->msg_controllen);
2131 * BSD recvmsg interface
2134 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2136 int fput_needed, err;
2137 struct msghdr msg_sys;
2138 struct socket *sock;
2140 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2144 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2146 fput_light(sock->file, fput_needed);
2151 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2152 unsigned int, flags)
2154 if (flags & MSG_CMSG_COMPAT)
2156 return __sys_recvmsg(fd, msg, flags);
2160 * Linux recvmmsg interface
2163 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2164 unsigned int flags, struct timespec *timeout)
2166 int fput_needed, err, datagrams;
2167 struct socket *sock;
2168 struct mmsghdr __user *entry;
2169 struct compat_mmsghdr __user *compat_entry;
2170 struct msghdr msg_sys;
2171 struct timespec end_time;
2174 poll_select_set_timeout(&end_time, timeout->tv_sec,
2180 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2184 err = sock_error(sock->sk);
2189 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2191 while (datagrams < vlen) {
2193 * No need to ask LSM for more than the first datagram.
2195 if (MSG_CMSG_COMPAT & flags) {
2196 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2197 &msg_sys, flags & ~MSG_WAITFORONE,
2201 err = __put_user(err, &compat_entry->msg_len);
2204 err = ___sys_recvmsg(sock,
2205 (struct user_msghdr __user *)entry,
2206 &msg_sys, flags & ~MSG_WAITFORONE,
2210 err = put_user(err, &entry->msg_len);
2218 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2219 if (flags & MSG_WAITFORONE)
2220 flags |= MSG_DONTWAIT;
2223 ktime_get_ts(timeout);
2224 *timeout = timespec_sub(end_time, *timeout);
2225 if (timeout->tv_sec < 0) {
2226 timeout->tv_sec = timeout->tv_nsec = 0;
2230 /* Timeout, return less than vlen datagrams */
2231 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2235 /* Out of band data, return right away */
2236 if (msg_sys.msg_flags & MSG_OOB)
2241 fput_light(sock->file, fput_needed);
2246 if (datagrams != 0) {
2248 * We may return less entries than requested (vlen) if the
2249 * sock is non block and there aren't enough datagrams...
2251 if (err != -EAGAIN) {
2253 * ... or if recvmsg returns an error after we
2254 * received some datagrams, where we record the
2255 * error to return on the next call or if the
2256 * app asks about it using getsockopt(SO_ERROR).
2258 sock->sk->sk_err = -err;
2267 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2268 unsigned int, vlen, unsigned int, flags,
2269 struct timespec __user *, timeout)
2272 struct timespec timeout_sys;
2274 if (flags & MSG_CMSG_COMPAT)
2278 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2280 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2283 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2285 if (datagrams > 0 &&
2286 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2287 datagrams = -EFAULT;
2292 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2293 /* Argument list sizes for sys_socketcall */
2294 #define AL(x) ((x) * sizeof(unsigned long))
2295 static const unsigned char nargs[21] = {
2296 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2297 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2298 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2305 * System call vectors.
2307 * Argument checking cleaned up. Saved 20% in size.
2308 * This function doesn't need to set the kernel lock because
2309 * it is set by the callees.
2312 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2314 unsigned long a[AUDITSC_ARGS];
2315 unsigned long a0, a1;
2319 if (call < 1 || call > SYS_SENDMMSG)
2323 if (len > sizeof(a))
2326 /* copy_from_user should be SMP safe. */
2327 if (copy_from_user(a, args, len))
2330 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2339 err = sys_socket(a0, a1, a[2]);
2342 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2345 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2348 err = sys_listen(a0, a1);
2351 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2352 (int __user *)a[2], 0);
2354 case SYS_GETSOCKNAME:
2356 sys_getsockname(a0, (struct sockaddr __user *)a1,
2357 (int __user *)a[2]);
2359 case SYS_GETPEERNAME:
2361 sys_getpeername(a0, (struct sockaddr __user *)a1,
2362 (int __user *)a[2]);
2364 case SYS_SOCKETPAIR:
2365 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2368 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2371 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2372 (struct sockaddr __user *)a[4], a[5]);
2375 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2378 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2379 (struct sockaddr __user *)a[4],
2380 (int __user *)a[5]);
2383 err = sys_shutdown(a0, a1);
2385 case SYS_SETSOCKOPT:
2386 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2388 case SYS_GETSOCKOPT:
2390 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2391 (int __user *)a[4]);
2394 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2397 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2400 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2403 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2404 (struct timespec __user *)a[4]);
2407 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2408 (int __user *)a[2], a[3]);
2417 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2420 * sock_register - add a socket protocol handler
2421 * @ops: description of protocol
2423 * This function is called by a protocol handler that wants to
2424 * advertise its address family, and have it linked into the
2425 * socket interface. The value ops->family corresponds to the
2426 * socket system call protocol family.
2428 int sock_register(const struct net_proto_family *ops)
2432 if (ops->family >= NPROTO) {
2433 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2437 spin_lock(&net_family_lock);
2438 if (rcu_dereference_protected(net_families[ops->family],
2439 lockdep_is_held(&net_family_lock)))
2442 rcu_assign_pointer(net_families[ops->family], ops);
2445 spin_unlock(&net_family_lock);
2447 pr_info("NET: Registered protocol family %d\n", ops->family);
2450 EXPORT_SYMBOL(sock_register);
2453 * sock_unregister - remove a protocol handler
2454 * @family: protocol family to remove
2456 * This function is called by a protocol handler that wants to
2457 * remove its address family, and have it unlinked from the
2458 * new socket creation.
2460 * If protocol handler is a module, then it can use module reference
2461 * counts to protect against new references. If protocol handler is not
2462 * a module then it needs to provide its own protection in
2463 * the ops->create routine.
2465 void sock_unregister(int family)
2467 BUG_ON(family < 0 || family >= NPROTO);
2469 spin_lock(&net_family_lock);
2470 RCU_INIT_POINTER(net_families[family], NULL);
2471 spin_unlock(&net_family_lock);
2475 pr_info("NET: Unregistered protocol family %d\n", family);
2477 EXPORT_SYMBOL(sock_unregister);
2479 static int __init sock_init(void)
2483 * Initialize the network sysctl infrastructure.
2485 err = net_sysctl_init();
2490 * Initialize skbuff SLAB cache
2495 * Initialize the protocols module.
2500 err = register_filesystem(&sock_fs_type);
2503 sock_mnt = kern_mount(&sock_fs_type);
2504 if (IS_ERR(sock_mnt)) {
2505 err = PTR_ERR(sock_mnt);
2509 /* The real protocol initialization is performed in later initcalls.
2512 #ifdef CONFIG_NETFILTER
2513 err = netfilter_init();
2518 ptp_classifier_init();
2524 unregister_filesystem(&sock_fs_type);
2529 core_initcall(sock_init); /* early initcall */
2531 #ifdef CONFIG_PROC_FS
2532 void socket_seq_show(struct seq_file *seq)
2537 for_each_possible_cpu(cpu)
2538 counter += per_cpu(sockets_in_use, cpu);
2540 /* It can be negative, by the way. 8) */
2544 seq_printf(seq, "sockets: used %d\n", counter);
2546 #endif /* CONFIG_PROC_FS */
2548 #ifdef CONFIG_COMPAT
2549 static int do_siocgstamp(struct net *net, struct socket *sock,
2550 unsigned int cmd, void __user *up)
2552 mm_segment_t old_fs = get_fs();
2557 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2560 err = compat_put_timeval(&ktv, up);
2565 static int do_siocgstampns(struct net *net, struct socket *sock,
2566 unsigned int cmd, void __user *up)
2568 mm_segment_t old_fs = get_fs();
2569 struct timespec kts;
2573 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2576 err = compat_put_timespec(&kts, up);
2581 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2583 struct ifreq __user *uifr;
2586 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2587 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2590 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2594 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2600 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2602 struct compat_ifconf ifc32;
2604 struct ifconf __user *uifc;
2605 struct compat_ifreq __user *ifr32;
2606 struct ifreq __user *ifr;
2610 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2613 memset(&ifc, 0, sizeof(ifc));
2614 if (ifc32.ifcbuf == 0) {
2618 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2620 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2621 sizeof(struct ifreq);
2622 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2624 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2625 ifr32 = compat_ptr(ifc32.ifcbuf);
2626 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2627 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2633 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2636 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2640 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2644 ifr32 = compat_ptr(ifc32.ifcbuf);
2646 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2647 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2648 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2654 if (ifc32.ifcbuf == 0) {
2655 /* Translate from 64-bit structure multiple to
2659 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2664 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2670 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2672 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2673 bool convert_in = false, convert_out = false;
2674 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2675 struct ethtool_rxnfc __user *rxnfc;
2676 struct ifreq __user *ifr;
2677 u32 rule_cnt = 0, actual_rule_cnt;
2682 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2685 compat_rxnfc = compat_ptr(data);
2687 if (get_user(ethcmd, &compat_rxnfc->cmd))
2690 /* Most ethtool structures are defined without padding.
2691 * Unfortunately struct ethtool_rxnfc is an exception.
2696 case ETHTOOL_GRXCLSRLALL:
2697 /* Buffer size is variable */
2698 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2700 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2702 buf_size += rule_cnt * sizeof(u32);
2704 case ETHTOOL_GRXRINGS:
2705 case ETHTOOL_GRXCLSRLCNT:
2706 case ETHTOOL_GRXCLSRULE:
2707 case ETHTOOL_SRXCLSRLINS:
2710 case ETHTOOL_SRXCLSRLDEL:
2711 buf_size += sizeof(struct ethtool_rxnfc);
2716 ifr = compat_alloc_user_space(buf_size);
2717 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2719 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2722 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2723 &ifr->ifr_ifru.ifru_data))
2727 /* We expect there to be holes between fs.m_ext and
2728 * fs.ring_cookie and at the end of fs, but nowhere else.
2730 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2731 sizeof(compat_rxnfc->fs.m_ext) !=
2732 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2733 sizeof(rxnfc->fs.m_ext));
2735 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2736 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2737 offsetof(struct ethtool_rxnfc, fs.location) -
2738 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2740 if (copy_in_user(rxnfc, compat_rxnfc,
2741 (void __user *)(&rxnfc->fs.m_ext + 1) -
2742 (void __user *)rxnfc) ||
2743 copy_in_user(&rxnfc->fs.ring_cookie,
2744 &compat_rxnfc->fs.ring_cookie,
2745 (void __user *)(&rxnfc->fs.location + 1) -
2746 (void __user *)&rxnfc->fs.ring_cookie) ||
2747 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2748 sizeof(rxnfc->rule_cnt)))
2752 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2757 if (copy_in_user(compat_rxnfc, rxnfc,
2758 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2759 (const void __user *)rxnfc) ||
2760 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2761 &rxnfc->fs.ring_cookie,
2762 (const void __user *)(&rxnfc->fs.location + 1) -
2763 (const void __user *)&rxnfc->fs.ring_cookie) ||
2764 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2765 sizeof(rxnfc->rule_cnt)))
2768 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2769 /* As an optimisation, we only copy the actual
2770 * number of rules that the underlying
2771 * function returned. Since Mallory might
2772 * change the rule count in user memory, we
2773 * check that it is less than the rule count
2774 * originally given (as the user buffer size),
2775 * which has been range-checked.
2777 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2779 if (actual_rule_cnt < rule_cnt)
2780 rule_cnt = actual_rule_cnt;
2781 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2782 &rxnfc->rule_locs[0],
2783 rule_cnt * sizeof(u32)))
2791 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2794 compat_uptr_t uptr32;
2795 struct ifreq __user *uifr;
2797 uifr = compat_alloc_user_space(sizeof(*uifr));
2798 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2801 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2804 uptr = compat_ptr(uptr32);
2806 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2809 return dev_ioctl(net, SIOCWANDEV, uifr);
2812 static int bond_ioctl(struct net *net, unsigned int cmd,
2813 struct compat_ifreq __user *ifr32)
2816 mm_segment_t old_fs;
2820 case SIOCBONDENSLAVE:
2821 case SIOCBONDRELEASE:
2822 case SIOCBONDSETHWADDR:
2823 case SIOCBONDCHANGEACTIVE:
2824 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2829 err = dev_ioctl(net, cmd,
2830 (struct ifreq __user __force *) &kifr);
2835 return -ENOIOCTLCMD;
2839 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2840 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2841 struct compat_ifreq __user *u_ifreq32)
2843 struct ifreq __user *u_ifreq64;
2844 char tmp_buf[IFNAMSIZ];
2845 void __user *data64;
2848 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2851 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2853 data64 = compat_ptr(data32);
2855 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2857 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2860 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2863 return dev_ioctl(net, cmd, u_ifreq64);
2866 static int dev_ifsioc(struct net *net, struct socket *sock,
2867 unsigned int cmd, struct compat_ifreq __user *uifr32)
2869 struct ifreq __user *uifr;
2872 uifr = compat_alloc_user_space(sizeof(*uifr));
2873 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2876 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2887 case SIOCGIFBRDADDR:
2888 case SIOCGIFDSTADDR:
2889 case SIOCGIFNETMASK:
2894 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2902 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2903 struct compat_ifreq __user *uifr32)
2906 struct compat_ifmap __user *uifmap32;
2907 mm_segment_t old_fs;
2910 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2911 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2912 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2913 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2914 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2915 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2916 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2917 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2923 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2926 if (cmd == SIOCGIFMAP && !err) {
2927 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2928 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2929 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2930 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2931 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2932 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2933 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2942 struct sockaddr rt_dst; /* target address */
2943 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2944 struct sockaddr rt_genmask; /* target network mask (IP) */
2945 unsigned short rt_flags;
2948 unsigned char rt_tos;
2949 unsigned char rt_class;
2951 short rt_metric; /* +1 for binary compatibility! */
2952 /* char * */ u32 rt_dev; /* forcing the device at add */
2953 u32 rt_mtu; /* per route MTU/Window */
2954 u32 rt_window; /* Window clamping */
2955 unsigned short rt_irtt; /* Initial RTT */
2958 struct in6_rtmsg32 {
2959 struct in6_addr rtmsg_dst;
2960 struct in6_addr rtmsg_src;
2961 struct in6_addr rtmsg_gateway;
2971 static int routing_ioctl(struct net *net, struct socket *sock,
2972 unsigned int cmd, void __user *argp)
2976 struct in6_rtmsg r6;
2980 mm_segment_t old_fs = get_fs();
2982 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2983 struct in6_rtmsg32 __user *ur6 = argp;
2984 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2985 3 * sizeof(struct in6_addr));
2986 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2987 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2988 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2989 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2990 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2991 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2992 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2996 struct rtentry32 __user *ur4 = argp;
2997 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2998 3 * sizeof(struct sockaddr));
2999 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3000 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3001 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3002 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3003 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3004 ret |= get_user(rtdev, &(ur4->rt_dev));
3006 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3007 r4.rt_dev = (char __user __force *)devname;
3021 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3028 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3029 * for some operations; this forces use of the newer bridge-utils that
3030 * use compatible ioctls
3032 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3036 if (get_user(tmp, argp))
3038 if (tmp == BRCTL_GET_VERSION)
3039 return BRCTL_VERSION + 1;
3043 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3044 unsigned int cmd, unsigned long arg)
3046 void __user *argp = compat_ptr(arg);
3047 struct sock *sk = sock->sk;
3048 struct net *net = sock_net(sk);
3050 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3051 return compat_ifr_data_ioctl(net, cmd, argp);
3056 return old_bridge_ioctl(argp);
3058 return dev_ifname32(net, argp);
3060 return dev_ifconf(net, argp);
3062 return ethtool_ioctl(net, argp);
3064 return compat_siocwandev(net, argp);
3067 return compat_sioc_ifmap(net, cmd, argp);
3068 case SIOCBONDENSLAVE:
3069 case SIOCBONDRELEASE:
3070 case SIOCBONDSETHWADDR:
3071 case SIOCBONDCHANGEACTIVE:
3072 return bond_ioctl(net, cmd, argp);
3075 return routing_ioctl(net, sock, cmd, argp);
3077 return do_siocgstamp(net, sock, cmd, argp);
3079 return do_siocgstampns(net, sock, cmd, argp);
3080 case SIOCBONDSLAVEINFOQUERY:
3081 case SIOCBONDINFOQUERY:
3084 return compat_ifr_data_ioctl(net, cmd, argp);
3096 return sock_ioctl(file, cmd, arg);
3113 case SIOCSIFHWBROADCAST:
3115 case SIOCGIFBRDADDR:
3116 case SIOCSIFBRDADDR:
3117 case SIOCGIFDSTADDR:
3118 case SIOCSIFDSTADDR:
3119 case SIOCGIFNETMASK:
3120 case SIOCSIFNETMASK:
3131 return dev_ifsioc(net, sock, cmd, argp);
3137 return sock_do_ioctl(net, sock, cmd, arg);
3140 return -ENOIOCTLCMD;
3143 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3146 struct socket *sock = file->private_data;
3147 int ret = -ENOIOCTLCMD;
3154 if (sock->ops->compat_ioctl)
3155 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3157 if (ret == -ENOIOCTLCMD &&
3158 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3159 ret = compat_wext_handle_ioctl(net, cmd, arg);
3161 if (ret == -ENOIOCTLCMD)
3162 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3168 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3170 return sock->ops->bind(sock, addr, addrlen);
3172 EXPORT_SYMBOL(kernel_bind);
3174 int kernel_listen(struct socket *sock, int backlog)
3176 return sock->ops->listen(sock, backlog);
3178 EXPORT_SYMBOL(kernel_listen);
3180 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3182 struct sock *sk = sock->sk;
3185 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3190 err = sock->ops->accept(sock, *newsock, flags);
3192 sock_release(*newsock);
3197 (*newsock)->ops = sock->ops;
3198 __module_get((*newsock)->ops->owner);
3203 EXPORT_SYMBOL(kernel_accept);
3205 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3208 return sock->ops->connect(sock, addr, addrlen, flags);
3210 EXPORT_SYMBOL(kernel_connect);
3212 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3215 return sock->ops->getname(sock, addr, addrlen, 0);
3217 EXPORT_SYMBOL(kernel_getsockname);
3219 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3222 return sock->ops->getname(sock, addr, addrlen, 1);
3224 EXPORT_SYMBOL(kernel_getpeername);
3226 int kernel_getsockopt(struct socket *sock, int level, int optname,
3227 char *optval, int *optlen)
3229 mm_segment_t oldfs = get_fs();
3230 char __user *uoptval;
3231 int __user *uoptlen;
3234 uoptval = (char __user __force *) optval;
3235 uoptlen = (int __user __force *) optlen;
3238 if (level == SOL_SOCKET)
3239 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3241 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3246 EXPORT_SYMBOL(kernel_getsockopt);
3248 int kernel_setsockopt(struct socket *sock, int level, int optname,
3249 char *optval, unsigned int optlen)
3251 mm_segment_t oldfs = get_fs();
3252 char __user *uoptval;
3255 uoptval = (char __user __force *) optval;
3258 if (level == SOL_SOCKET)
3259 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3261 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3266 EXPORT_SYMBOL(kernel_setsockopt);
3268 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3269 size_t size, int flags)
3271 if (sock->ops->sendpage)
3272 return sock->ops->sendpage(sock, page, offset, size, flags);
3274 return sock_no_sendpage(sock, page, offset, size, flags);
3276 EXPORT_SYMBOL(kernel_sendpage);
3278 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3280 mm_segment_t oldfs = get_fs();
3284 err = sock->ops->ioctl(sock, cmd, arg);
3289 EXPORT_SYMBOL(kernel_sock_ioctl);
3291 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3293 return sock->ops->shutdown(sock, how);
3295 EXPORT_SYMBOL(kernel_sock_shutdown);