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;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
269 return &ei->vfs_inode;
272 static void sock_destroy_inode(struct inode *inode)
274 struct socket_alloc *ei;
275 struct socket_wq *wq;
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
278 wq = rcu_dereference_protected(ei->socket.wq, 1);
280 kmem_cache_free(sock_inode_cachep, ei);
283 static void init_once(void *foo)
285 struct socket_alloc *ei = (struct socket_alloc *)foo;
287 inode_init_once(&ei->vfs_inode);
290 static int init_inodecache(void)
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293 sizeof(struct socket_alloc),
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
299 if (sock_inode_cachep == NULL)
304 static const struct super_operations sockfs_ops = {
305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
311 * sockfs_dname() is called from d_path().
313 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316 d_inode(dentry)->i_ino);
319 static const struct dentry_operations sockfs_dentry_operations = {
320 .d_dname = sockfs_dname,
323 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
330 static struct vfsmount *sock_mnt __read_mostly;
332 static struct file_system_type sock_fs_type = {
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
339 * Obtains the first available file descriptor and sets it up for use.
341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
355 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
357 struct qstr name = { .name = "" };
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
371 path.mnt = mntget(sock_mnt);
373 d_instantiate(path.dentry, SOCK_INODE(sock));
375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 /* drop dentry, keep inode */
379 ihold(d_inode(path.dentry));
385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
386 file->private_data = sock;
389 EXPORT_SYMBOL(sock_alloc_file);
391 static int sock_map_fd(struct socket *sock, int flags)
393 struct file *newfile;
394 int fd = get_unused_fd_flags(flags);
395 if (unlikely(fd < 0))
398 newfile = sock_alloc_file(sock, flags, NULL);
399 if (likely(!IS_ERR(newfile))) {
400 fd_install(fd, newfile);
405 return PTR_ERR(newfile);
408 struct socket *sock_from_file(struct file *file, int *err)
410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
416 EXPORT_SYMBOL(sock_from_file);
419 * sockfd_lookup - Go from a file number to its socket slot
421 * @err: pointer to an error code return
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
428 * On a success the socket object pointer is returned.
431 struct socket *sockfd_lookup(int fd, int *err)
442 sock = sock_from_file(file, err);
447 EXPORT_SYMBOL(sockfd_lookup);
449 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451 struct fd f = fdget(fd);
456 sock = sock_from_file(f.file, err);
458 *fput_needed = f.flags;
466 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469 static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
472 const char *proto_name;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
483 if (proto_size + 1 > size)
486 strncpy(value, proto_name, proto_size + 1);
488 error = proto_size + 1;
495 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
501 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
523 int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
525 int err = simple_setattr(dentry, iattr);
527 if (!err && (iattr->ia_valid & ATTR_UID)) {
528 struct socket *sock = SOCKET_I(d_inode(dentry));
530 sock->sk->sk_uid = iattr->ia_uid;
536 static const struct inode_operations sockfs_inode_ops = {
537 .getxattr = sockfs_getxattr,
538 .listxattr = sockfs_listxattr,
539 .setattr = sockfs_setattr,
543 * sock_alloc - allocate a socket
545 * Allocate a new inode and socket object. The two are bound together
546 * and initialised. The socket is then returned. If we are out of inodes
550 static struct socket *sock_alloc(void)
555 inode = new_inode_pseudo(sock_mnt->mnt_sb);
559 sock = SOCKET_I(inode);
561 kmemcheck_annotate_bitfield(sock, type);
562 inode->i_ino = get_next_ino();
563 inode->i_mode = S_IFSOCK | S_IRWXUGO;
564 inode->i_uid = current_fsuid();
565 inode->i_gid = current_fsgid();
566 inode->i_op = &sockfs_inode_ops;
568 this_cpu_add(sockets_in_use, 1);
573 * sock_release - close a socket
574 * @sock: socket to close
576 * The socket is released from the protocol stack if it has a release
577 * callback, and the inode is then released if the socket is bound to
578 * an inode not a file.
581 void sock_release(struct socket *sock)
584 struct module *owner = sock->ops->owner;
586 sock->ops->release(sock);
591 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
592 pr_err("%s: fasync list not empty!\n", __func__);
594 this_cpu_sub(sockets_in_use, 1);
596 iput(SOCK_INODE(sock));
601 EXPORT_SYMBOL(sock_release);
603 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
605 u8 flags = *tx_flags;
607 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
608 flags |= SKBTX_HW_TSTAMP;
610 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
611 flags |= SKBTX_SW_TSTAMP;
613 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
614 flags |= SKBTX_SCHED_TSTAMP;
616 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
617 flags |= SKBTX_ACK_TSTAMP;
621 EXPORT_SYMBOL(__sock_tx_timestamp);
623 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
625 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
626 BUG_ON(ret == -EIOCBQUEUED);
630 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
632 int err = security_socket_sendmsg(sock, msg,
635 return err ?: sock_sendmsg_nosec(sock, msg);
637 EXPORT_SYMBOL(sock_sendmsg);
639 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
640 struct kvec *vec, size_t num, size_t size)
642 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
643 return sock_sendmsg(sock, msg);
645 EXPORT_SYMBOL(kernel_sendmsg);
648 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
650 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
653 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
654 struct scm_timestamping tss;
656 struct skb_shared_hwtstamps *shhwtstamps =
659 /* Race occurred between timestamp enabling and packet
660 receiving. Fill in the current time for now. */
661 if (need_software_tstamp && skb->tstamp.tv64 == 0)
662 __net_timestamp(skb);
664 if (need_software_tstamp) {
665 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
667 skb_get_timestamp(skb, &tv);
668 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
672 skb_get_timestampns(skb, &ts);
673 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
678 memset(&tss, 0, sizeof(tss));
679 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
680 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
683 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
684 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
687 put_cmsg(msg, SOL_SOCKET,
688 SCM_TIMESTAMPING, sizeof(tss), &tss);
690 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
692 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
697 if (!sock_flag(sk, SOCK_WIFI_STATUS))
699 if (!skb->wifi_acked_valid)
702 ack = skb->wifi_acked;
704 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
706 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
708 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
711 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
712 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
713 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
716 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
719 sock_recv_timestamp(msg, sk, skb);
720 sock_recv_drops(msg, sk, skb);
722 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
724 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
725 size_t size, int flags)
727 return sock->ops->recvmsg(sock, msg, size, flags);
730 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
733 int err = security_socket_recvmsg(sock, msg, size, flags);
735 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
737 EXPORT_SYMBOL(sock_recvmsg);
740 * kernel_recvmsg - Receive a message from a socket (kernel space)
741 * @sock: The socket to receive the message from
742 * @msg: Received message
743 * @vec: Input s/g array for message data
744 * @num: Size of input s/g array
745 * @size: Number of bytes to read
746 * @flags: Message flags (MSG_DONTWAIT, etc...)
748 * On return the msg structure contains the scatter/gather array passed in the
749 * vec argument. The array is modified so that it consists of the unfilled
750 * portion of the original array.
752 * The returned value is the total number of bytes received, or an error.
754 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
755 struct kvec *vec, size_t num, size_t size, int flags)
757 mm_segment_t oldfs = get_fs();
760 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
762 result = sock_recvmsg(sock, msg, size, flags);
766 EXPORT_SYMBOL(kernel_recvmsg);
768 static ssize_t sock_sendpage(struct file *file, struct page *page,
769 int offset, size_t size, loff_t *ppos, int more)
774 sock = file->private_data;
776 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
777 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
780 return kernel_sendpage(sock, page, offset, size, flags);
783 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
784 struct pipe_inode_info *pipe, size_t len,
787 struct socket *sock = file->private_data;
789 if (unlikely(!sock->ops->splice_read))
792 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
795 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
797 struct file *file = iocb->ki_filp;
798 struct socket *sock = file->private_data;
799 struct msghdr msg = {.msg_iter = *to,
803 if (file->f_flags & O_NONBLOCK)
804 msg.msg_flags = MSG_DONTWAIT;
806 if (iocb->ki_pos != 0)
809 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
812 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
817 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
819 struct file *file = iocb->ki_filp;
820 struct socket *sock = file->private_data;
821 struct msghdr msg = {.msg_iter = *from,
825 if (iocb->ki_pos != 0)
828 if (file->f_flags & O_NONBLOCK)
829 msg.msg_flags = MSG_DONTWAIT;
831 if (sock->type == SOCK_SEQPACKET)
832 msg.msg_flags |= MSG_EOR;
834 res = sock_sendmsg(sock, &msg);
835 *from = msg.msg_iter;
840 * Atomic setting of ioctl hooks to avoid race
841 * with module unload.
844 static DEFINE_MUTEX(br_ioctl_mutex);
845 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
847 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
849 mutex_lock(&br_ioctl_mutex);
850 br_ioctl_hook = hook;
851 mutex_unlock(&br_ioctl_mutex);
853 EXPORT_SYMBOL(brioctl_set);
855 static DEFINE_MUTEX(vlan_ioctl_mutex);
856 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
858 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
860 mutex_lock(&vlan_ioctl_mutex);
861 vlan_ioctl_hook = hook;
862 mutex_unlock(&vlan_ioctl_mutex);
864 EXPORT_SYMBOL(vlan_ioctl_set);
866 static DEFINE_MUTEX(dlci_ioctl_mutex);
867 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
869 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
871 mutex_lock(&dlci_ioctl_mutex);
872 dlci_ioctl_hook = hook;
873 mutex_unlock(&dlci_ioctl_mutex);
875 EXPORT_SYMBOL(dlci_ioctl_set);
877 static long sock_do_ioctl(struct net *net, struct socket *sock,
878 unsigned int cmd, unsigned long arg)
881 void __user *argp = (void __user *)arg;
883 err = sock->ops->ioctl(sock, cmd, arg);
886 * If this ioctl is unknown try to hand it down
889 if (err == -ENOIOCTLCMD)
890 err = dev_ioctl(net, cmd, argp);
896 * With an ioctl, arg may well be a user mode pointer, but we don't know
897 * what to do with it - that's up to the protocol still.
900 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
904 void __user *argp = (void __user *)arg;
908 sock = file->private_data;
911 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
912 err = dev_ioctl(net, cmd, argp);
914 #ifdef CONFIG_WEXT_CORE
915 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
916 err = dev_ioctl(net, cmd, argp);
923 if (get_user(pid, (int __user *)argp))
925 f_setown(sock->file, pid, 1);
930 err = put_user(f_getown(sock->file),
939 request_module("bridge");
941 mutex_lock(&br_ioctl_mutex);
943 err = br_ioctl_hook(net, cmd, argp);
944 mutex_unlock(&br_ioctl_mutex);
949 if (!vlan_ioctl_hook)
950 request_module("8021q");
952 mutex_lock(&vlan_ioctl_mutex);
954 err = vlan_ioctl_hook(net, argp);
955 mutex_unlock(&vlan_ioctl_mutex);
960 if (!dlci_ioctl_hook)
961 request_module("dlci");
963 mutex_lock(&dlci_ioctl_mutex);
965 err = dlci_ioctl_hook(cmd, argp);
966 mutex_unlock(&dlci_ioctl_mutex);
969 err = sock_do_ioctl(net, sock, cmd, arg);
975 int sock_create_lite(int family, int type, int protocol, struct socket **res)
978 struct socket *sock = NULL;
980 err = security_socket_create(family, type, protocol, 1);
991 err = security_socket_post_create(sock, family, type, protocol, 1);
1003 EXPORT_SYMBOL(sock_create_lite);
1005 /* No kernel lock held - perfect */
1006 static unsigned int sock_poll(struct file *file, poll_table *wait)
1008 unsigned int busy_flag = 0;
1009 struct socket *sock;
1012 * We can't return errors to poll, so it's either yes or no.
1014 sock = file->private_data;
1016 if (sk_can_busy_loop(sock->sk)) {
1017 /* this socket can poll_ll so tell the system call */
1018 busy_flag = POLL_BUSY_LOOP;
1020 /* once, only if requested by syscall */
1021 if (wait && (wait->_key & POLL_BUSY_LOOP))
1022 sk_busy_loop(sock->sk, 1);
1025 return busy_flag | sock->ops->poll(file, sock, wait);
1028 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1030 struct socket *sock = file->private_data;
1032 return sock->ops->mmap(file, sock, vma);
1035 static int sock_close(struct inode *inode, struct file *filp)
1037 sock_release(SOCKET_I(inode));
1042 * Update the socket async list
1044 * Fasync_list locking strategy.
1046 * 1. fasync_list is modified only under process context socket lock
1047 * i.e. under semaphore.
1048 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1049 * or under socket lock
1052 static int sock_fasync(int fd, struct file *filp, int on)
1054 struct socket *sock = filp->private_data;
1055 struct sock *sk = sock->sk;
1056 struct socket_wq *wq;
1062 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1063 fasync_helper(fd, filp, on, &wq->fasync_list);
1065 if (!wq->fasync_list)
1066 sock_reset_flag(sk, SOCK_FASYNC);
1068 sock_set_flag(sk, SOCK_FASYNC);
1074 /* This function may be called only under rcu_lock */
1076 int sock_wake_async(struct socket_wq *wq, int how, int band)
1078 if (!wq || !wq->fasync_list)
1082 case SOCK_WAKE_WAITD:
1083 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1086 case SOCK_WAKE_SPACE:
1087 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1092 kill_fasync(&wq->fasync_list, SIGIO, band);
1095 kill_fasync(&wq->fasync_list, SIGURG, band);
1100 EXPORT_SYMBOL(sock_wake_async);
1102 int __sock_create(struct net *net, int family, int type, int protocol,
1103 struct socket **res, int kern)
1106 struct socket *sock;
1107 const struct net_proto_family *pf;
1110 * Check protocol is in range
1112 if (family < 0 || family >= NPROTO)
1113 return -EAFNOSUPPORT;
1114 if (type < 0 || type >= SOCK_MAX)
1119 This uglymoron is moved from INET layer to here to avoid
1120 deadlock in module load.
1122 if (family == PF_INET && type == SOCK_PACKET) {
1126 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1132 err = security_socket_create(family, type, protocol, kern);
1137 * Allocate the socket and allow the family to set things up. if
1138 * the protocol is 0, the family is instructed to select an appropriate
1141 sock = sock_alloc();
1143 net_warn_ratelimited("socket: no more sockets\n");
1144 return -ENFILE; /* Not exactly a match, but its the
1145 closest posix thing */
1150 #ifdef CONFIG_MODULES
1151 /* Attempt to load a protocol module if the find failed.
1153 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1154 * requested real, full-featured networking support upon configuration.
1155 * Otherwise module support will break!
1157 if (rcu_access_pointer(net_families[family]) == NULL)
1158 request_module("net-pf-%d", family);
1162 pf = rcu_dereference(net_families[family]);
1163 err = -EAFNOSUPPORT;
1168 * We will call the ->create function, that possibly is in a loadable
1169 * module, so we have to bump that loadable module refcnt first.
1171 if (!try_module_get(pf->owner))
1174 /* Now protected by module ref count */
1177 err = pf->create(net, sock, protocol, kern);
1179 goto out_module_put;
1182 * Now to bump the refcnt of the [loadable] module that owns this
1183 * socket at sock_release time we decrement its refcnt.
1185 if (!try_module_get(sock->ops->owner))
1186 goto out_module_busy;
1189 * Now that we're done with the ->create function, the [loadable]
1190 * module can have its refcnt decremented
1192 module_put(pf->owner);
1193 err = security_socket_post_create(sock, family, type, protocol, kern);
1195 goto out_sock_release;
1201 err = -EAFNOSUPPORT;
1204 module_put(pf->owner);
1211 goto out_sock_release;
1213 EXPORT_SYMBOL(__sock_create);
1215 int sock_create(int family, int type, int protocol, struct socket **res)
1217 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1219 EXPORT_SYMBOL(sock_create);
1221 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1223 return __sock_create(net, family, type, protocol, res, 1);
1225 EXPORT_SYMBOL(sock_create_kern);
1227 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1230 struct socket *sock;
1233 /* Check the SOCK_* constants for consistency. */
1234 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1235 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1236 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1237 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1239 flags = type & ~SOCK_TYPE_MASK;
1240 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1242 type &= SOCK_TYPE_MASK;
1244 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1245 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1247 retval = sock_create(family, type, protocol, &sock);
1251 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1256 /* It may be already another descriptor 8) Not kernel problem. */
1265 * Create a pair of connected sockets.
1268 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1269 int __user *, usockvec)
1271 struct socket *sock1, *sock2;
1273 struct file *newfile1, *newfile2;
1276 flags = type & ~SOCK_TYPE_MASK;
1277 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1279 type &= SOCK_TYPE_MASK;
1281 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1282 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1285 * Obtain the first socket and check if the underlying protocol
1286 * supports the socketpair call.
1289 err = sock_create(family, type, protocol, &sock1);
1293 err = sock_create(family, type, protocol, &sock2);
1297 err = sock1->ops->socketpair(sock1, sock2);
1299 goto out_release_both;
1301 fd1 = get_unused_fd_flags(flags);
1302 if (unlikely(fd1 < 0)) {
1304 goto out_release_both;
1307 fd2 = get_unused_fd_flags(flags);
1308 if (unlikely(fd2 < 0)) {
1310 goto out_put_unused_1;
1313 newfile1 = sock_alloc_file(sock1, flags, NULL);
1314 if (IS_ERR(newfile1)) {
1315 err = PTR_ERR(newfile1);
1316 goto out_put_unused_both;
1319 newfile2 = sock_alloc_file(sock2, flags, NULL);
1320 if (IS_ERR(newfile2)) {
1321 err = PTR_ERR(newfile2);
1325 err = put_user(fd1, &usockvec[0]);
1329 err = put_user(fd2, &usockvec[1]);
1333 audit_fd_pair(fd1, fd2);
1335 fd_install(fd1, newfile1);
1336 fd_install(fd2, newfile2);
1337 /* fd1 and fd2 may be already another descriptors.
1338 * Not kernel problem.
1354 sock_release(sock2);
1357 out_put_unused_both:
1362 sock_release(sock2);
1364 sock_release(sock1);
1370 * Bind a name to a socket. Nothing much to do here since it's
1371 * the protocol's responsibility to handle the local address.
1373 * We move the socket address to kernel space before we call
1374 * the protocol layer (having also checked the address is ok).
1377 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1379 struct socket *sock;
1380 struct sockaddr_storage address;
1381 int err, fput_needed;
1383 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1385 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1387 err = security_socket_bind(sock,
1388 (struct sockaddr *)&address,
1391 err = sock->ops->bind(sock,
1395 fput_light(sock->file, fput_needed);
1401 * Perform a listen. Basically, we allow the protocol to do anything
1402 * necessary for a listen, and if that works, we mark the socket as
1403 * ready for listening.
1406 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1408 struct socket *sock;
1409 int err, fput_needed;
1412 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1414 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1415 if ((unsigned int)backlog > somaxconn)
1416 backlog = somaxconn;
1418 err = security_socket_listen(sock, backlog);
1420 err = sock->ops->listen(sock, backlog);
1422 fput_light(sock->file, fput_needed);
1428 * For accept, we attempt to create a new socket, set up the link
1429 * with the client, wake up the client, then return the new
1430 * connected fd. We collect the address of the connector in kernel
1431 * space and move it to user at the very end. This is unclean because
1432 * we open the socket then return an error.
1434 * 1003.1g adds the ability to recvmsg() to query connection pending
1435 * status to recvmsg. We need to add that support in a way thats
1436 * clean when we restucture accept also.
1439 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1440 int __user *, upeer_addrlen, int, flags)
1442 struct socket *sock, *newsock;
1443 struct file *newfile;
1444 int err, len, newfd, fput_needed;
1445 struct sockaddr_storage address;
1447 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1450 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1451 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1453 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1458 newsock = sock_alloc();
1462 newsock->type = sock->type;
1463 newsock->ops = sock->ops;
1466 * We don't need try_module_get here, as the listening socket (sock)
1467 * has the protocol module (sock->ops->owner) held.
1469 __module_get(newsock->ops->owner);
1471 newfd = get_unused_fd_flags(flags);
1472 if (unlikely(newfd < 0)) {
1474 sock_release(newsock);
1477 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1478 if (IS_ERR(newfile)) {
1479 err = PTR_ERR(newfile);
1480 put_unused_fd(newfd);
1481 sock_release(newsock);
1485 err = security_socket_accept(sock, newsock);
1489 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1493 if (upeer_sockaddr) {
1494 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1496 err = -ECONNABORTED;
1499 err = move_addr_to_user(&address,
1500 len, upeer_sockaddr, upeer_addrlen);
1505 /* File flags are not inherited via accept() unlike another OSes. */
1507 fd_install(newfd, newfile);
1511 fput_light(sock->file, fput_needed);
1516 put_unused_fd(newfd);
1520 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1521 int __user *, upeer_addrlen)
1523 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1527 * Attempt to connect to a socket with the server address. The address
1528 * is in user space so we verify it is OK and move it to kernel space.
1530 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1533 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1534 * other SEQPACKET protocols that take time to connect() as it doesn't
1535 * include the -EINPROGRESS status for such sockets.
1538 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1541 struct socket *sock;
1542 struct sockaddr_storage address;
1543 int err, fput_needed;
1545 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1548 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1553 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1557 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1558 sock->file->f_flags);
1560 fput_light(sock->file, fput_needed);
1566 * Get the local address ('name') of a socket object. Move the obtained
1567 * name to user space.
1570 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1571 int __user *, usockaddr_len)
1573 struct socket *sock;
1574 struct sockaddr_storage address;
1575 int len, err, fput_needed;
1577 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1581 err = security_socket_getsockname(sock);
1585 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1588 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1591 fput_light(sock->file, fput_needed);
1597 * Get the remote address ('name') of a socket object. Move the obtained
1598 * name to user space.
1601 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1602 int __user *, usockaddr_len)
1604 struct socket *sock;
1605 struct sockaddr_storage address;
1606 int len, err, fput_needed;
1608 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1610 err = security_socket_getpeername(sock);
1612 fput_light(sock->file, fput_needed);
1617 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1620 err = move_addr_to_user(&address, len, usockaddr,
1622 fput_light(sock->file, fput_needed);
1628 * Send a datagram to a given address. We move the address into kernel
1629 * space and check the user space data area is readable before invoking
1633 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1634 unsigned int, flags, struct sockaddr __user *, addr,
1637 struct socket *sock;
1638 struct sockaddr_storage address;
1644 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1647 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1651 msg.msg_name = NULL;
1652 msg.msg_control = NULL;
1653 msg.msg_controllen = 0;
1654 msg.msg_namelen = 0;
1656 err = move_addr_to_kernel(addr, addr_len, &address);
1659 msg.msg_name = (struct sockaddr *)&address;
1660 msg.msg_namelen = addr_len;
1662 if (sock->file->f_flags & O_NONBLOCK)
1663 flags |= MSG_DONTWAIT;
1664 msg.msg_flags = flags;
1665 err = sock_sendmsg(sock, &msg);
1668 fput_light(sock->file, fput_needed);
1674 * Send a datagram down a socket.
1677 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1678 unsigned int, flags)
1680 return sys_sendto(fd, buff, len, flags, NULL, 0);
1684 * Receive a frame from the socket and optionally record the address of the
1685 * sender. We verify the buffers are writable and if needed move the
1686 * sender address from kernel to user space.
1689 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1690 unsigned int, flags, struct sockaddr __user *, addr,
1691 int __user *, addr_len)
1693 struct socket *sock;
1696 struct sockaddr_storage address;
1700 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1703 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707 msg.msg_control = NULL;
1708 msg.msg_controllen = 0;
1709 /* Save some cycles and don't copy the address if not needed */
1710 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1711 /* We assume all kernel code knows the size of sockaddr_storage */
1712 msg.msg_namelen = 0;
1713 msg.msg_iocb = NULL;
1714 if (sock->file->f_flags & O_NONBLOCK)
1715 flags |= MSG_DONTWAIT;
1716 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1718 if (err >= 0 && addr != NULL) {
1719 err2 = move_addr_to_user(&address,
1720 msg.msg_namelen, addr, addr_len);
1725 fput_light(sock->file, fput_needed);
1731 * Receive a datagram from a socket.
1734 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1735 unsigned int, flags)
1737 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1741 * Set a socket option. Because we don't know the option lengths we have
1742 * to pass the user mode parameter for the protocols to sort out.
1745 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1746 char __user *, optval, int, optlen)
1748 int err, fput_needed;
1749 struct socket *sock;
1754 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1756 err = security_socket_setsockopt(sock, level, optname);
1760 if (level == SOL_SOCKET)
1762 sock_setsockopt(sock, level, optname, optval,
1766 sock->ops->setsockopt(sock, level, optname, optval,
1769 fput_light(sock->file, fput_needed);
1775 * Get a socket option. Because we don't know the option lengths we have
1776 * to pass a user mode parameter for the protocols to sort out.
1779 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1780 char __user *, optval, int __user *, optlen)
1782 int err, fput_needed;
1783 struct socket *sock;
1785 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1787 err = security_socket_getsockopt(sock, level, optname);
1791 if (level == SOL_SOCKET)
1793 sock_getsockopt(sock, level, optname, optval,
1797 sock->ops->getsockopt(sock, level, optname, optval,
1800 fput_light(sock->file, fput_needed);
1806 * Shutdown a socket.
1809 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1811 int err, fput_needed;
1812 struct socket *sock;
1814 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1816 err = security_socket_shutdown(sock, how);
1818 err = sock->ops->shutdown(sock, how);
1819 fput_light(sock->file, fput_needed);
1824 /* A couple of helpful macros for getting the address of the 32/64 bit
1825 * fields which are the same type (int / unsigned) on our platforms.
1827 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1828 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1829 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1831 struct used_address {
1832 struct sockaddr_storage name;
1833 unsigned int name_len;
1836 static int copy_msghdr_from_user(struct msghdr *kmsg,
1837 struct user_msghdr __user *umsg,
1838 struct sockaddr __user **save_addr,
1841 struct sockaddr __user *uaddr;
1842 struct iovec __user *uiov;
1846 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1847 __get_user(uaddr, &umsg->msg_name) ||
1848 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1849 __get_user(uiov, &umsg->msg_iov) ||
1850 __get_user(nr_segs, &umsg->msg_iovlen) ||
1851 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1852 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1853 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1857 kmsg->msg_namelen = 0;
1859 if (kmsg->msg_namelen < 0)
1862 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1863 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1868 if (uaddr && kmsg->msg_namelen) {
1870 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1876 kmsg->msg_name = NULL;
1877 kmsg->msg_namelen = 0;
1880 if (nr_segs > UIO_MAXIOV)
1883 kmsg->msg_iocb = NULL;
1885 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1886 UIO_FASTIOV, iov, &kmsg->msg_iter);
1889 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1890 struct msghdr *msg_sys, unsigned int flags,
1891 struct used_address *used_address)
1893 struct compat_msghdr __user *msg_compat =
1894 (struct compat_msghdr __user *)msg;
1895 struct sockaddr_storage address;
1896 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1897 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1898 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1899 /* 20 is size of ipv6_pktinfo */
1900 unsigned char *ctl_buf = ctl;
1904 msg_sys->msg_name = &address;
1906 if (MSG_CMSG_COMPAT & flags)
1907 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1909 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1915 if (msg_sys->msg_controllen > INT_MAX)
1917 ctl_len = msg_sys->msg_controllen;
1918 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1920 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1924 ctl_buf = msg_sys->msg_control;
1925 ctl_len = msg_sys->msg_controllen;
1926 } else if (ctl_len) {
1927 if (ctl_len > sizeof(ctl)) {
1928 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1929 if (ctl_buf == NULL)
1934 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1935 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1936 * checking falls down on this.
1938 if (copy_from_user(ctl_buf,
1939 (void __user __force *)msg_sys->msg_control,
1942 msg_sys->msg_control = ctl_buf;
1944 msg_sys->msg_flags = flags;
1946 if (sock->file->f_flags & O_NONBLOCK)
1947 msg_sys->msg_flags |= MSG_DONTWAIT;
1949 * If this is sendmmsg() and current destination address is same as
1950 * previously succeeded address, omit asking LSM's decision.
1951 * used_address->name_len is initialized to UINT_MAX so that the first
1952 * destination address never matches.
1954 if (used_address && msg_sys->msg_name &&
1955 used_address->name_len == msg_sys->msg_namelen &&
1956 !memcmp(&used_address->name, msg_sys->msg_name,
1957 used_address->name_len)) {
1958 err = sock_sendmsg_nosec(sock, msg_sys);
1961 err = sock_sendmsg(sock, msg_sys);
1963 * If this is sendmmsg() and sending to current destination address was
1964 * successful, remember it.
1966 if (used_address && err >= 0) {
1967 used_address->name_len = msg_sys->msg_namelen;
1968 if (msg_sys->msg_name)
1969 memcpy(&used_address->name, msg_sys->msg_name,
1970 used_address->name_len);
1975 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1982 * BSD sendmsg interface
1985 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1987 int fput_needed, err;
1988 struct msghdr msg_sys;
1989 struct socket *sock;
1991 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1995 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1997 fput_light(sock->file, fput_needed);
2002 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2004 if (flags & MSG_CMSG_COMPAT)
2006 return __sys_sendmsg(fd, msg, flags);
2010 * Linux sendmmsg interface
2013 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2016 int fput_needed, err, datagrams;
2017 struct socket *sock;
2018 struct mmsghdr __user *entry;
2019 struct compat_mmsghdr __user *compat_entry;
2020 struct msghdr msg_sys;
2021 struct used_address used_address;
2023 if (vlen > UIO_MAXIOV)
2028 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2032 used_address.name_len = UINT_MAX;
2034 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2037 while (datagrams < vlen) {
2038 if (MSG_CMSG_COMPAT & flags) {
2039 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2040 &msg_sys, flags, &used_address);
2043 err = __put_user(err, &compat_entry->msg_len);
2046 err = ___sys_sendmsg(sock,
2047 (struct user_msghdr __user *)entry,
2048 &msg_sys, flags, &used_address);
2051 err = put_user(err, &entry->msg_len);
2058 if (msg_data_left(&msg_sys))
2062 fput_light(sock->file, fput_needed);
2064 /* We only return an error if no datagrams were able to be sent */
2071 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2072 unsigned int, vlen, unsigned int, flags)
2074 if (flags & MSG_CMSG_COMPAT)
2076 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2079 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2080 struct msghdr *msg_sys, unsigned int flags, int nosec)
2082 struct compat_msghdr __user *msg_compat =
2083 (struct compat_msghdr __user *)msg;
2084 struct iovec iovstack[UIO_FASTIOV];
2085 struct iovec *iov = iovstack;
2086 unsigned long cmsg_ptr;
2090 /* kernel mode address */
2091 struct sockaddr_storage addr;
2093 /* user mode address pointers */
2094 struct sockaddr __user *uaddr;
2095 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2097 msg_sys->msg_name = &addr;
2099 if (MSG_CMSG_COMPAT & flags)
2100 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2102 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2105 total_len = iov_iter_count(&msg_sys->msg_iter);
2107 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2108 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2110 /* We assume all kernel code knows the size of sockaddr_storage */
2111 msg_sys->msg_namelen = 0;
2113 if (sock->file->f_flags & O_NONBLOCK)
2114 flags |= MSG_DONTWAIT;
2115 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2121 if (uaddr != NULL) {
2122 err = move_addr_to_user(&addr,
2123 msg_sys->msg_namelen, uaddr,
2128 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2132 if (MSG_CMSG_COMPAT & flags)
2133 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2134 &msg_compat->msg_controllen);
2136 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2137 &msg->msg_controllen);
2148 * BSD recvmsg interface
2151 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2153 int fput_needed, err;
2154 struct msghdr msg_sys;
2155 struct socket *sock;
2157 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2161 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2163 fput_light(sock->file, fput_needed);
2168 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2169 unsigned int, flags)
2171 if (flags & MSG_CMSG_COMPAT)
2173 return __sys_recvmsg(fd, msg, flags);
2177 * Linux recvmmsg interface
2180 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2181 unsigned int flags, struct timespec *timeout)
2183 int fput_needed, err, datagrams;
2184 struct socket *sock;
2185 struct mmsghdr __user *entry;
2186 struct compat_mmsghdr __user *compat_entry;
2187 struct msghdr msg_sys;
2188 struct timespec end_time;
2191 poll_select_set_timeout(&end_time, timeout->tv_sec,
2197 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2201 err = sock_error(sock->sk);
2206 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2208 while (datagrams < vlen) {
2210 * No need to ask LSM for more than the first datagram.
2212 if (MSG_CMSG_COMPAT & flags) {
2213 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2214 &msg_sys, flags & ~MSG_WAITFORONE,
2218 err = __put_user(err, &compat_entry->msg_len);
2221 err = ___sys_recvmsg(sock,
2222 (struct user_msghdr __user *)entry,
2223 &msg_sys, flags & ~MSG_WAITFORONE,
2227 err = put_user(err, &entry->msg_len);
2235 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2236 if (flags & MSG_WAITFORONE)
2237 flags |= MSG_DONTWAIT;
2240 ktime_get_ts(timeout);
2241 *timeout = timespec_sub(end_time, *timeout);
2242 if (timeout->tv_sec < 0) {
2243 timeout->tv_sec = timeout->tv_nsec = 0;
2247 /* Timeout, return less than vlen datagrams */
2248 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2252 /* Out of band data, return right away */
2253 if (msg_sys.msg_flags & MSG_OOB)
2260 if (datagrams == 0) {
2266 * We may return less entries than requested (vlen) if the
2267 * sock is non block and there aren't enough datagrams...
2269 if (err != -EAGAIN) {
2271 * ... or if recvmsg returns an error after we
2272 * received some datagrams, where we record the
2273 * error to return on the next call or if the
2274 * app asks about it using getsockopt(SO_ERROR).
2276 sock->sk->sk_err = -err;
2279 fput_light(sock->file, fput_needed);
2284 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2285 unsigned int, vlen, unsigned int, flags,
2286 struct timespec __user *, timeout)
2289 struct timespec timeout_sys;
2291 if (flags & MSG_CMSG_COMPAT)
2295 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2297 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2300 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2302 if (datagrams > 0 &&
2303 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2304 datagrams = -EFAULT;
2309 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2310 /* Argument list sizes for sys_socketcall */
2311 #define AL(x) ((x) * sizeof(unsigned long))
2312 static const unsigned char nargs[21] = {
2313 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2314 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2315 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2322 * System call vectors.
2324 * Argument checking cleaned up. Saved 20% in size.
2325 * This function doesn't need to set the kernel lock because
2326 * it is set by the callees.
2329 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2331 unsigned long a[AUDITSC_ARGS];
2332 unsigned long a0, a1;
2336 if (call < 1 || call > SYS_SENDMMSG)
2340 if (len > sizeof(a))
2343 /* copy_from_user should be SMP safe. */
2344 if (copy_from_user(a, args, len))
2347 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2356 err = sys_socket(a0, a1, a[2]);
2359 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2362 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2365 err = sys_listen(a0, a1);
2368 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2369 (int __user *)a[2], 0);
2371 case SYS_GETSOCKNAME:
2373 sys_getsockname(a0, (struct sockaddr __user *)a1,
2374 (int __user *)a[2]);
2376 case SYS_GETPEERNAME:
2378 sys_getpeername(a0, (struct sockaddr __user *)a1,
2379 (int __user *)a[2]);
2381 case SYS_SOCKETPAIR:
2382 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2385 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2388 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2389 (struct sockaddr __user *)a[4], a[5]);
2392 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2395 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2396 (struct sockaddr __user *)a[4],
2397 (int __user *)a[5]);
2400 err = sys_shutdown(a0, a1);
2402 case SYS_SETSOCKOPT:
2403 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2405 case SYS_GETSOCKOPT:
2407 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2408 (int __user *)a[4]);
2411 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2414 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2417 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2420 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2421 (struct timespec __user *)a[4]);
2424 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2425 (int __user *)a[2], a[3]);
2434 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2437 * sock_register - add a socket protocol handler
2438 * @ops: description of protocol
2440 * This function is called by a protocol handler that wants to
2441 * advertise its address family, and have it linked into the
2442 * socket interface. The value ops->family corresponds to the
2443 * socket system call protocol family.
2445 int sock_register(const struct net_proto_family *ops)
2449 if (ops->family >= NPROTO) {
2450 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2454 spin_lock(&net_family_lock);
2455 if (rcu_dereference_protected(net_families[ops->family],
2456 lockdep_is_held(&net_family_lock)))
2459 rcu_assign_pointer(net_families[ops->family], ops);
2462 spin_unlock(&net_family_lock);
2464 pr_info("NET: Registered protocol family %d\n", ops->family);
2467 EXPORT_SYMBOL(sock_register);
2470 * sock_unregister - remove a protocol handler
2471 * @family: protocol family to remove
2473 * This function is called by a protocol handler that wants to
2474 * remove its address family, and have it unlinked from the
2475 * new socket creation.
2477 * If protocol handler is a module, then it can use module reference
2478 * counts to protect against new references. If protocol handler is not
2479 * a module then it needs to provide its own protection in
2480 * the ops->create routine.
2482 void sock_unregister(int family)
2484 BUG_ON(family < 0 || family >= NPROTO);
2486 spin_lock(&net_family_lock);
2487 RCU_INIT_POINTER(net_families[family], NULL);
2488 spin_unlock(&net_family_lock);
2492 pr_info("NET: Unregistered protocol family %d\n", family);
2494 EXPORT_SYMBOL(sock_unregister);
2496 static int __init sock_init(void)
2500 * Initialize the network sysctl infrastructure.
2502 err = net_sysctl_init();
2507 * Initialize skbuff SLAB cache
2512 * Initialize the protocols module.
2517 err = register_filesystem(&sock_fs_type);
2520 sock_mnt = kern_mount(&sock_fs_type);
2521 if (IS_ERR(sock_mnt)) {
2522 err = PTR_ERR(sock_mnt);
2526 /* The real protocol initialization is performed in later initcalls.
2529 #ifdef CONFIG_NETFILTER
2530 err = netfilter_init();
2535 ptp_classifier_init();
2541 unregister_filesystem(&sock_fs_type);
2546 core_initcall(sock_init); /* early initcall */
2548 #ifdef CONFIG_PROC_FS
2549 void socket_seq_show(struct seq_file *seq)
2554 for_each_possible_cpu(cpu)
2555 counter += per_cpu(sockets_in_use, cpu);
2557 /* It can be negative, by the way. 8) */
2561 seq_printf(seq, "sockets: used %d\n", counter);
2563 #endif /* CONFIG_PROC_FS */
2565 #ifdef CONFIG_COMPAT
2566 static int do_siocgstamp(struct net *net, struct socket *sock,
2567 unsigned int cmd, void __user *up)
2569 mm_segment_t old_fs = get_fs();
2574 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2577 err = compat_put_timeval(&ktv, up);
2582 static int do_siocgstampns(struct net *net, struct socket *sock,
2583 unsigned int cmd, void __user *up)
2585 mm_segment_t old_fs = get_fs();
2586 struct timespec kts;
2590 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2593 err = compat_put_timespec(&kts, up);
2598 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2600 struct ifreq __user *uifr;
2603 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2604 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2607 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2611 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2617 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2619 struct compat_ifconf ifc32;
2621 struct ifconf __user *uifc;
2622 struct compat_ifreq __user *ifr32;
2623 struct ifreq __user *ifr;
2627 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2630 memset(&ifc, 0, sizeof(ifc));
2631 if (ifc32.ifcbuf == 0) {
2635 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2637 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2638 sizeof(struct ifreq);
2639 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2641 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2642 ifr32 = compat_ptr(ifc32.ifcbuf);
2643 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2644 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2650 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2653 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2657 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2661 ifr32 = compat_ptr(ifc32.ifcbuf);
2663 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2664 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2665 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2671 if (ifc32.ifcbuf == 0) {
2672 /* Translate from 64-bit structure multiple to
2676 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2681 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2687 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2689 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2690 bool convert_in = false, convert_out = false;
2691 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2692 struct ethtool_rxnfc __user *rxnfc;
2693 struct ifreq __user *ifr;
2694 u32 rule_cnt = 0, actual_rule_cnt;
2699 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2702 compat_rxnfc = compat_ptr(data);
2704 if (get_user(ethcmd, &compat_rxnfc->cmd))
2707 /* Most ethtool structures are defined without padding.
2708 * Unfortunately struct ethtool_rxnfc is an exception.
2713 case ETHTOOL_GRXCLSRLALL:
2714 /* Buffer size is variable */
2715 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2717 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2719 buf_size += rule_cnt * sizeof(u32);
2721 case ETHTOOL_GRXRINGS:
2722 case ETHTOOL_GRXCLSRLCNT:
2723 case ETHTOOL_GRXCLSRULE:
2724 case ETHTOOL_SRXCLSRLINS:
2727 case ETHTOOL_SRXCLSRLDEL:
2728 buf_size += sizeof(struct ethtool_rxnfc);
2733 ifr = compat_alloc_user_space(buf_size);
2734 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2736 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2739 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2740 &ifr->ifr_ifru.ifru_data))
2744 /* We expect there to be holes between fs.m_ext and
2745 * fs.ring_cookie and at the end of fs, but nowhere else.
2747 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2748 sizeof(compat_rxnfc->fs.m_ext) !=
2749 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2750 sizeof(rxnfc->fs.m_ext));
2752 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2753 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2754 offsetof(struct ethtool_rxnfc, fs.location) -
2755 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2757 if (copy_in_user(rxnfc, compat_rxnfc,
2758 (void __user *)(&rxnfc->fs.m_ext + 1) -
2759 (void __user *)rxnfc) ||
2760 copy_in_user(&rxnfc->fs.ring_cookie,
2761 &compat_rxnfc->fs.ring_cookie,
2762 (void __user *)(&rxnfc->fs.location + 1) -
2763 (void __user *)&rxnfc->fs.ring_cookie) ||
2764 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2765 sizeof(rxnfc->rule_cnt)))
2769 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2774 if (copy_in_user(compat_rxnfc, rxnfc,
2775 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2776 (const void __user *)rxnfc) ||
2777 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2778 &rxnfc->fs.ring_cookie,
2779 (const void __user *)(&rxnfc->fs.location + 1) -
2780 (const void __user *)&rxnfc->fs.ring_cookie) ||
2781 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2782 sizeof(rxnfc->rule_cnt)))
2785 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2786 /* As an optimisation, we only copy the actual
2787 * number of rules that the underlying
2788 * function returned. Since Mallory might
2789 * change the rule count in user memory, we
2790 * check that it is less than the rule count
2791 * originally given (as the user buffer size),
2792 * which has been range-checked.
2794 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2796 if (actual_rule_cnt < rule_cnt)
2797 rule_cnt = actual_rule_cnt;
2798 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2799 &rxnfc->rule_locs[0],
2800 rule_cnt * sizeof(u32)))
2808 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2811 compat_uptr_t uptr32;
2812 struct ifreq __user *uifr;
2814 uifr = compat_alloc_user_space(sizeof(*uifr));
2815 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2818 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2821 uptr = compat_ptr(uptr32);
2823 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2826 return dev_ioctl(net, SIOCWANDEV, uifr);
2829 static int bond_ioctl(struct net *net, unsigned int cmd,
2830 struct compat_ifreq __user *ifr32)
2833 mm_segment_t old_fs;
2837 case SIOCBONDENSLAVE:
2838 case SIOCBONDRELEASE:
2839 case SIOCBONDSETHWADDR:
2840 case SIOCBONDCHANGEACTIVE:
2841 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2846 err = dev_ioctl(net, cmd,
2847 (struct ifreq __user __force *) &kifr);
2852 return -ENOIOCTLCMD;
2856 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2857 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2858 struct compat_ifreq __user *u_ifreq32)
2860 struct ifreq __user *u_ifreq64;
2861 char tmp_buf[IFNAMSIZ];
2862 void __user *data64;
2865 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2868 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2870 data64 = compat_ptr(data32);
2872 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2874 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2877 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2880 return dev_ioctl(net, cmd, u_ifreq64);
2883 static int dev_ifsioc(struct net *net, struct socket *sock,
2884 unsigned int cmd, struct compat_ifreq __user *uifr32)
2886 struct ifreq __user *uifr;
2889 uifr = compat_alloc_user_space(sizeof(*uifr));
2890 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2893 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2904 case SIOCGIFBRDADDR:
2905 case SIOCGIFDSTADDR:
2906 case SIOCGIFNETMASK:
2911 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2919 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2920 struct compat_ifreq __user *uifr32)
2923 struct compat_ifmap __user *uifmap32;
2924 mm_segment_t old_fs;
2927 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2928 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2929 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2930 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2931 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2932 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2933 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2934 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2940 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2943 if (cmd == SIOCGIFMAP && !err) {
2944 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2945 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2946 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2947 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2948 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2949 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2950 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2959 struct sockaddr rt_dst; /* target address */
2960 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2961 struct sockaddr rt_genmask; /* target network mask (IP) */
2962 unsigned short rt_flags;
2965 unsigned char rt_tos;
2966 unsigned char rt_class;
2968 short rt_metric; /* +1 for binary compatibility! */
2969 /* char * */ u32 rt_dev; /* forcing the device at add */
2970 u32 rt_mtu; /* per route MTU/Window */
2971 u32 rt_window; /* Window clamping */
2972 unsigned short rt_irtt; /* Initial RTT */
2975 struct in6_rtmsg32 {
2976 struct in6_addr rtmsg_dst;
2977 struct in6_addr rtmsg_src;
2978 struct in6_addr rtmsg_gateway;
2988 static int routing_ioctl(struct net *net, struct socket *sock,
2989 unsigned int cmd, void __user *argp)
2993 struct in6_rtmsg r6;
2997 mm_segment_t old_fs = get_fs();
2999 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3000 struct in6_rtmsg32 __user *ur6 = argp;
3001 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3002 3 * sizeof(struct in6_addr));
3003 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3004 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3005 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3006 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3007 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3008 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3009 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3013 struct rtentry32 __user *ur4 = argp;
3014 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3015 3 * sizeof(struct sockaddr));
3016 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3017 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3018 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3019 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3020 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3021 ret |= get_user(rtdev, &(ur4->rt_dev));
3023 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3024 r4.rt_dev = (char __user __force *)devname;
3038 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3045 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3046 * for some operations; this forces use of the newer bridge-utils that
3047 * use compatible ioctls
3049 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3053 if (get_user(tmp, argp))
3055 if (tmp == BRCTL_GET_VERSION)
3056 return BRCTL_VERSION + 1;
3060 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3061 unsigned int cmd, unsigned long arg)
3063 void __user *argp = compat_ptr(arg);
3064 struct sock *sk = sock->sk;
3065 struct net *net = sock_net(sk);
3067 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3068 return compat_ifr_data_ioctl(net, cmd, argp);
3073 return old_bridge_ioctl(argp);
3075 return dev_ifname32(net, argp);
3077 return dev_ifconf(net, argp);
3079 return ethtool_ioctl(net, argp);
3081 return compat_siocwandev(net, argp);
3084 return compat_sioc_ifmap(net, cmd, argp);
3085 case SIOCBONDENSLAVE:
3086 case SIOCBONDRELEASE:
3087 case SIOCBONDSETHWADDR:
3088 case SIOCBONDCHANGEACTIVE:
3089 return bond_ioctl(net, cmd, argp);
3092 return routing_ioctl(net, sock, cmd, argp);
3094 return do_siocgstamp(net, sock, cmd, argp);
3096 return do_siocgstampns(net, sock, cmd, argp);
3097 case SIOCBONDSLAVEINFOQUERY:
3098 case SIOCBONDINFOQUERY:
3101 return compat_ifr_data_ioctl(net, cmd, argp);
3113 return sock_ioctl(file, cmd, arg);
3130 case SIOCSIFHWBROADCAST:
3132 case SIOCGIFBRDADDR:
3133 case SIOCSIFBRDADDR:
3134 case SIOCGIFDSTADDR:
3135 case SIOCSIFDSTADDR:
3136 case SIOCGIFNETMASK:
3137 case SIOCSIFNETMASK:
3148 return dev_ifsioc(net, sock, cmd, argp);
3154 return sock_do_ioctl(net, sock, cmd, arg);
3157 return -ENOIOCTLCMD;
3160 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3163 struct socket *sock = file->private_data;
3164 int ret = -ENOIOCTLCMD;
3171 if (sock->ops->compat_ioctl)
3172 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3174 if (ret == -ENOIOCTLCMD &&
3175 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3176 ret = compat_wext_handle_ioctl(net, cmd, arg);
3178 if (ret == -ENOIOCTLCMD)
3179 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3185 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3187 return sock->ops->bind(sock, addr, addrlen);
3189 EXPORT_SYMBOL(kernel_bind);
3191 int kernel_listen(struct socket *sock, int backlog)
3193 return sock->ops->listen(sock, backlog);
3195 EXPORT_SYMBOL(kernel_listen);
3197 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3199 struct sock *sk = sock->sk;
3202 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3207 err = sock->ops->accept(sock, *newsock, flags);
3209 sock_release(*newsock);
3214 (*newsock)->ops = sock->ops;
3215 __module_get((*newsock)->ops->owner);
3220 EXPORT_SYMBOL(kernel_accept);
3222 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3225 return sock->ops->connect(sock, addr, addrlen, flags);
3227 EXPORT_SYMBOL(kernel_connect);
3229 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3232 return sock->ops->getname(sock, addr, addrlen, 0);
3234 EXPORT_SYMBOL(kernel_getsockname);
3236 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3239 return sock->ops->getname(sock, addr, addrlen, 1);
3241 EXPORT_SYMBOL(kernel_getpeername);
3243 int kernel_getsockopt(struct socket *sock, int level, int optname,
3244 char *optval, int *optlen)
3246 mm_segment_t oldfs = get_fs();
3247 char __user *uoptval;
3248 int __user *uoptlen;
3251 uoptval = (char __user __force *) optval;
3252 uoptlen = (int __user __force *) optlen;
3255 if (level == SOL_SOCKET)
3256 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3258 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3263 EXPORT_SYMBOL(kernel_getsockopt);
3265 int kernel_setsockopt(struct socket *sock, int level, int optname,
3266 char *optval, unsigned int optlen)
3268 mm_segment_t oldfs = get_fs();
3269 char __user *uoptval;
3272 uoptval = (char __user __force *) optval;
3275 if (level == SOL_SOCKET)
3276 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3278 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3283 EXPORT_SYMBOL(kernel_setsockopt);
3285 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3286 size_t size, int flags)
3288 if (sock->ops->sendpage)
3289 return sock->ops->sendpage(sock, page, offset, size, flags);
3291 return sock_no_sendpage(sock, page, offset, size, flags);
3293 EXPORT_SYMBOL(kernel_sendpage);
3295 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3297 mm_segment_t oldfs = get_fs();
3301 err = sock->ops->ioctl(sock, cmd, arg);
3306 EXPORT_SYMBOL(kernel_sock_ioctl);
3308 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3310 return sock->ops->shutdown(sock, how);
3312 EXPORT_SYMBOL(kernel_sock_shutdown);