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 = new_sync_read,
144 .write = new_sync_write,
145 .read_iter = sock_read_iter,
146 .write_iter = sock_write_iter,
148 .unlocked_ioctl = sock_ioctl,
150 .compat_ioctl = compat_sock_ioctl,
153 .release = sock_close,
154 .fasync = sock_fasync,
155 .sendpage = sock_sendpage,
156 .splice_write = generic_splice_sendpage,
157 .splice_read = sock_splice_read,
161 * The protocol list. Each protocol is registered in here.
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
168 * Statistics counters of the socket lists
171 static DEFINE_PER_CPU(int, sockets_in_use);
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
192 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
196 if (copy_from_user(kaddr, uaddr, ulen))
198 return audit_sockaddr(ulen, kaddr);
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219 void __user *uaddr, int __user *ulen)
224 BUG_ON(klen > sizeof(struct sockaddr_storage));
225 err = get_user(len, ulen);
233 if (audit_sockaddr(klen, kaddr))
235 if (copy_to_user(uaddr, kaddr, len))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen, ulen);
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
247 static struct inode *sock_alloc_inode(struct super_block *sb)
249 struct socket_alloc *ei;
250 struct socket_wq *wq;
252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
257 kmem_cache_free(sock_inode_cachep, ei);
260 init_waitqueue_head(&wq->wait);
261 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
270 return &ei->vfs_inode;
273 static void sock_destroy_inode(struct inode *inode)
275 struct socket_alloc *ei;
276 struct socket_wq *wq;
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
279 wq = rcu_dereference_protected(ei->socket.wq, 1);
281 kmem_cache_free(sock_inode_cachep, ei);
284 static void init_once(void *foo)
286 struct socket_alloc *ei = (struct socket_alloc *)foo;
288 inode_init_once(&ei->vfs_inode);
291 static int init_inodecache(void)
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294 sizeof(struct socket_alloc),
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
300 if (sock_inode_cachep == NULL)
305 static const struct super_operations sockfs_ops = {
306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
312 * sockfs_dname() is called from d_path().
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 dentry->d_inode->i_ino);
320 static const struct dentry_operations sockfs_dentry_operations = {
321 .d_dname = sockfs_dname,
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
331 static struct vfsmount *sock_mnt __read_mostly;
333 static struct file_system_type sock_fs_type = {
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
340 * Obtains the first available file descriptor and sets it up for use.
342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
358 struct qstr name = { .name = "" };
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
372 path.mnt = mntget(sock_mnt);
374 d_instantiate(path.dentry, SOCK_INODE(sock));
376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 if (unlikely(IS_ERR(file))) {
379 /* drop dentry, keep inode */
380 ihold(path.dentry->d_inode);
386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387 file->private_data = sock;
390 EXPORT_SYMBOL(sock_alloc_file);
392 static int sock_map_fd(struct socket *sock, int flags)
394 struct file *newfile;
395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
399 newfile = sock_alloc_file(sock, flags, NULL);
400 if (likely(!IS_ERR(newfile))) {
401 fd_install(fd, newfile);
406 return PTR_ERR(newfile);
409 struct socket *sock_from_file(struct file *file, int *err)
411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
417 EXPORT_SYMBOL(sock_from_file);
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket *sockfd_lookup(int fd, int *err)
443 sock = sock_from_file(file, err);
448 EXPORT_SYMBOL(sockfd_lookup);
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452 struct fd f = fdget(fd);
457 sock = sock_from_file(f.file, err);
459 *fput_needed = f.flags;
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
473 const char *proto_name;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
484 if (proto_size + 1 > size)
487 strncpy(value, proto_name, proto_size + 1);
489 error = proto_size + 1;
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
524 static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
530 * sock_alloc - allocate a socket
532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
537 static struct socket *sock_alloc(void)
542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
546 sock = SOCKET_I(inode);
548 kmemcheck_annotate_bitfield(sock, type);
549 inode->i_ino = get_next_ino();
550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
553 inode->i_op = &sockfs_inode_ops;
555 this_cpu_add(sockets_in_use, 1);
560 * sock_release - close a socket
561 * @sock: socket to close
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
568 void sock_release(struct socket *sock)
571 struct module *owner = sock->ops->owner;
573 sock->ops->release(sock);
578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579 pr_err("%s: fasync list not empty!\n", __func__);
581 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
584 this_cpu_sub(sockets_in_use, 1);
586 iput(SOCK_INODE(sock));
591 EXPORT_SYMBOL(sock_release);
593 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
595 u8 flags = *tx_flags;
597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
598 flags |= SKBTX_HW_TSTAMP;
600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
601 flags |= SKBTX_SW_TSTAMP;
603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
604 flags |= SKBTX_SCHED_TSTAMP;
606 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
607 flags |= SKBTX_ACK_TSTAMP;
611 EXPORT_SYMBOL(__sock_tx_timestamp);
613 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg,
616 return sock->ops->sendmsg(sock, msg, size);
619 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
621 int err = security_socket_sendmsg(sock, msg, size);
623 return err ?: sock_sendmsg_nosec(sock, msg, size);
625 EXPORT_SYMBOL(sock_sendmsg);
627 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
628 struct kvec *vec, size_t num, size_t size)
630 mm_segment_t oldfs = get_fs();
635 * the following is safe, since for compiler definitions of kvec and
636 * iovec are identical, yielding the same in-core layout and alignment
638 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
639 result = sock_sendmsg(sock, msg, size);
643 EXPORT_SYMBOL(kernel_sendmsg);
646 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
648 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
651 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
652 struct scm_timestamping tss;
654 struct skb_shared_hwtstamps *shhwtstamps =
657 /* Race occurred between timestamp enabling and packet
658 receiving. Fill in the current time for now. */
659 if (need_software_tstamp && skb->tstamp.tv64 == 0)
660 __net_timestamp(skb);
662 if (need_software_tstamp) {
663 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
665 skb_get_timestamp(skb, &tv);
666 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
670 skb_get_timestampns(skb, &ts);
671 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
676 memset(&tss, 0, sizeof(tss));
677 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
678 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
681 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
682 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
685 put_cmsg(msg, SOL_SOCKET,
686 SCM_TIMESTAMPING, sizeof(tss), &tss);
688 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
690 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
695 if (!sock_flag(sk, SOCK_WIFI_STATUS))
697 if (!skb->wifi_acked_valid)
700 ack = skb->wifi_acked;
702 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
704 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
706 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
709 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
710 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
711 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
714 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
717 sock_recv_timestamp(msg, sk, skb);
718 sock_recv_drops(msg, sk, skb);
720 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
722 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
723 size_t size, int flags)
725 return sock->ops->recvmsg(sock, msg, size, flags);
728 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
731 int err = security_socket_recvmsg(sock, msg, size, flags);
733 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
735 EXPORT_SYMBOL(sock_recvmsg);
738 * kernel_recvmsg - Receive a message from a socket (kernel space)
739 * @sock: The socket to receive the message from
740 * @msg: Received message
741 * @vec: Input s/g array for message data
742 * @num: Size of input s/g array
743 * @size: Number of bytes to read
744 * @flags: Message flags (MSG_DONTWAIT, etc...)
746 * On return the msg structure contains the scatter/gather array passed in the
747 * vec argument. The array is modified so that it consists of the unfilled
748 * portion of the original array.
750 * The returned value is the total number of bytes received, or an error.
752 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
753 struct kvec *vec, size_t num, size_t size, int flags)
755 mm_segment_t oldfs = get_fs();
760 * the following is safe, since for compiler definitions of kvec and
761 * iovec are identical, yielding the same in-core layout and alignment
763 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
764 result = sock_recvmsg(sock, msg, size, flags);
768 EXPORT_SYMBOL(kernel_recvmsg);
770 static ssize_t sock_sendpage(struct file *file, struct page *page,
771 int offset, size_t size, loff_t *ppos, int more)
776 sock = file->private_data;
778 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
779 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
782 return kernel_sendpage(sock, page, offset, size, flags);
785 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
786 struct pipe_inode_info *pipe, size_t len,
789 struct socket *sock = file->private_data;
791 if (unlikely(!sock->ops->splice_read))
794 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
797 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
799 struct file *file = iocb->ki_filp;
800 struct socket *sock = file->private_data;
801 struct msghdr msg = {.msg_iter = *to,
805 if (file->f_flags & O_NONBLOCK)
806 msg.msg_flags = MSG_DONTWAIT;
808 if (iocb->ki_pos != 0)
811 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
814 res = sock_recvmsg(sock, &msg, iocb->ki_nbytes, msg.msg_flags);
819 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
821 struct file *file = iocb->ki_filp;
822 struct socket *sock = file->private_data;
823 struct msghdr msg = {.msg_iter = *from,
827 if (iocb->ki_pos != 0)
830 if (file->f_flags & O_NONBLOCK)
831 msg.msg_flags = MSG_DONTWAIT;
833 if (sock->type == SOCK_SEQPACKET)
834 msg.msg_flags |= MSG_EOR;
836 res = sock_sendmsg(sock, &msg, iocb->ki_nbytes);
837 *from = msg.msg_iter;
842 * Atomic setting of ioctl hooks to avoid race
843 * with module unload.
846 static DEFINE_MUTEX(br_ioctl_mutex);
847 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
849 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
851 mutex_lock(&br_ioctl_mutex);
852 br_ioctl_hook = hook;
853 mutex_unlock(&br_ioctl_mutex);
855 EXPORT_SYMBOL(brioctl_set);
857 static DEFINE_MUTEX(vlan_ioctl_mutex);
858 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
860 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
862 mutex_lock(&vlan_ioctl_mutex);
863 vlan_ioctl_hook = hook;
864 mutex_unlock(&vlan_ioctl_mutex);
866 EXPORT_SYMBOL(vlan_ioctl_set);
868 static DEFINE_MUTEX(dlci_ioctl_mutex);
869 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
871 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
873 mutex_lock(&dlci_ioctl_mutex);
874 dlci_ioctl_hook = hook;
875 mutex_unlock(&dlci_ioctl_mutex);
877 EXPORT_SYMBOL(dlci_ioctl_set);
879 static long sock_do_ioctl(struct net *net, struct socket *sock,
880 unsigned int cmd, unsigned long arg)
883 void __user *argp = (void __user *)arg;
885 err = sock->ops->ioctl(sock, cmd, arg);
888 * If this ioctl is unknown try to hand it down
891 if (err == -ENOIOCTLCMD)
892 err = dev_ioctl(net, cmd, argp);
898 * With an ioctl, arg may well be a user mode pointer, but we don't know
899 * what to do with it - that's up to the protocol still.
902 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
906 void __user *argp = (void __user *)arg;
910 sock = file->private_data;
913 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
914 err = dev_ioctl(net, cmd, argp);
916 #ifdef CONFIG_WEXT_CORE
917 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
918 err = dev_ioctl(net, cmd, argp);
925 if (get_user(pid, (int __user *)argp))
927 f_setown(sock->file, pid, 1);
932 err = put_user(f_getown(sock->file),
941 request_module("bridge");
943 mutex_lock(&br_ioctl_mutex);
945 err = br_ioctl_hook(net, cmd, argp);
946 mutex_unlock(&br_ioctl_mutex);
951 if (!vlan_ioctl_hook)
952 request_module("8021q");
954 mutex_lock(&vlan_ioctl_mutex);
956 err = vlan_ioctl_hook(net, argp);
957 mutex_unlock(&vlan_ioctl_mutex);
962 if (!dlci_ioctl_hook)
963 request_module("dlci");
965 mutex_lock(&dlci_ioctl_mutex);
967 err = dlci_ioctl_hook(cmd, argp);
968 mutex_unlock(&dlci_ioctl_mutex);
971 err = sock_do_ioctl(net, sock, cmd, arg);
977 int sock_create_lite(int family, int type, int protocol, struct socket **res)
980 struct socket *sock = NULL;
982 err = security_socket_create(family, type, protocol, 1);
993 err = security_socket_post_create(sock, family, type, protocol, 1);
1005 EXPORT_SYMBOL(sock_create_lite);
1007 /* No kernel lock held - perfect */
1008 static unsigned int sock_poll(struct file *file, poll_table *wait)
1010 unsigned int busy_flag = 0;
1011 struct socket *sock;
1014 * We can't return errors to poll, so it's either yes or no.
1016 sock = file->private_data;
1018 if (sk_can_busy_loop(sock->sk)) {
1019 /* this socket can poll_ll so tell the system call */
1020 busy_flag = POLL_BUSY_LOOP;
1022 /* once, only if requested by syscall */
1023 if (wait && (wait->_key & POLL_BUSY_LOOP))
1024 sk_busy_loop(sock->sk, 1);
1027 return busy_flag | sock->ops->poll(file, sock, wait);
1030 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1032 struct socket *sock = file->private_data;
1034 return sock->ops->mmap(file, sock, vma);
1037 static int sock_close(struct inode *inode, struct file *filp)
1039 sock_release(SOCKET_I(inode));
1044 * Update the socket async list
1046 * Fasync_list locking strategy.
1048 * 1. fasync_list is modified only under process context socket lock
1049 * i.e. under semaphore.
1050 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1051 * or under socket lock
1054 static int sock_fasync(int fd, struct file *filp, int on)
1056 struct socket *sock = filp->private_data;
1057 struct sock *sk = sock->sk;
1058 struct socket_wq *wq;
1064 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1065 fasync_helper(fd, filp, on, &wq->fasync_list);
1067 if (!wq->fasync_list)
1068 sock_reset_flag(sk, SOCK_FASYNC);
1070 sock_set_flag(sk, SOCK_FASYNC);
1076 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1078 int sock_wake_async(struct socket *sock, int how, int band)
1080 struct socket_wq *wq;
1085 wq = rcu_dereference(sock->wq);
1086 if (!wq || !wq->fasync_list) {
1091 case SOCK_WAKE_WAITD:
1092 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1095 case SOCK_WAKE_SPACE:
1096 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1101 kill_fasync(&wq->fasync_list, SIGIO, band);
1104 kill_fasync(&wq->fasync_list, SIGURG, band);
1109 EXPORT_SYMBOL(sock_wake_async);
1111 int __sock_create(struct net *net, int family, int type, int protocol,
1112 struct socket **res, int kern)
1115 struct socket *sock;
1116 const struct net_proto_family *pf;
1119 * Check protocol is in range
1121 if (family < 0 || family >= NPROTO)
1122 return -EAFNOSUPPORT;
1123 if (type < 0 || type >= SOCK_MAX)
1128 This uglymoron is moved from INET layer to here to avoid
1129 deadlock in module load.
1131 if (family == PF_INET && type == SOCK_PACKET) {
1135 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1141 err = security_socket_create(family, type, protocol, kern);
1146 * Allocate the socket and allow the family to set things up. if
1147 * the protocol is 0, the family is instructed to select an appropriate
1150 sock = sock_alloc();
1152 net_warn_ratelimited("socket: no more sockets\n");
1153 return -ENFILE; /* Not exactly a match, but its the
1154 closest posix thing */
1159 #ifdef CONFIG_MODULES
1160 /* Attempt to load a protocol module if the find failed.
1162 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1163 * requested real, full-featured networking support upon configuration.
1164 * Otherwise module support will break!
1166 if (rcu_access_pointer(net_families[family]) == NULL)
1167 request_module("net-pf-%d", family);
1171 pf = rcu_dereference(net_families[family]);
1172 err = -EAFNOSUPPORT;
1177 * We will call the ->create function, that possibly is in a loadable
1178 * module, so we have to bump that loadable module refcnt first.
1180 if (!try_module_get(pf->owner))
1183 /* Now protected by module ref count */
1186 err = pf->create(net, sock, protocol, kern);
1188 goto out_module_put;
1191 * Now to bump the refcnt of the [loadable] module that owns this
1192 * socket at sock_release time we decrement its refcnt.
1194 if (!try_module_get(sock->ops->owner))
1195 goto out_module_busy;
1198 * Now that we're done with the ->create function, the [loadable]
1199 * module can have its refcnt decremented
1201 module_put(pf->owner);
1202 err = security_socket_post_create(sock, family, type, protocol, kern);
1204 goto out_sock_release;
1210 err = -EAFNOSUPPORT;
1213 module_put(pf->owner);
1220 goto out_sock_release;
1222 EXPORT_SYMBOL(__sock_create);
1224 int sock_create(int family, int type, int protocol, struct socket **res)
1226 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1228 EXPORT_SYMBOL(sock_create);
1230 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1232 return __sock_create(&init_net, family, type, protocol, res, 1);
1234 EXPORT_SYMBOL(sock_create_kern);
1236 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1239 struct socket *sock;
1242 /* Check the SOCK_* constants for consistency. */
1243 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1244 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1245 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1246 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1248 flags = type & ~SOCK_TYPE_MASK;
1249 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1251 type &= SOCK_TYPE_MASK;
1253 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1254 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1256 retval = sock_create(family, type, protocol, &sock);
1260 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1265 /* It may be already another descriptor 8) Not kernel problem. */
1274 * Create a pair of connected sockets.
1277 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1278 int __user *, usockvec)
1280 struct socket *sock1, *sock2;
1282 struct file *newfile1, *newfile2;
1285 flags = type & ~SOCK_TYPE_MASK;
1286 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1288 type &= SOCK_TYPE_MASK;
1290 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1291 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1294 * Obtain the first socket and check if the underlying protocol
1295 * supports the socketpair call.
1298 err = sock_create(family, type, protocol, &sock1);
1302 err = sock_create(family, type, protocol, &sock2);
1306 err = sock1->ops->socketpair(sock1, sock2);
1308 goto out_release_both;
1310 fd1 = get_unused_fd_flags(flags);
1311 if (unlikely(fd1 < 0)) {
1313 goto out_release_both;
1316 fd2 = get_unused_fd_flags(flags);
1317 if (unlikely(fd2 < 0)) {
1319 goto out_put_unused_1;
1322 newfile1 = sock_alloc_file(sock1, flags, NULL);
1323 if (unlikely(IS_ERR(newfile1))) {
1324 err = PTR_ERR(newfile1);
1325 goto out_put_unused_both;
1328 newfile2 = sock_alloc_file(sock2, flags, NULL);
1329 if (IS_ERR(newfile2)) {
1330 err = PTR_ERR(newfile2);
1334 err = put_user(fd1, &usockvec[0]);
1338 err = put_user(fd2, &usockvec[1]);
1342 audit_fd_pair(fd1, fd2);
1344 fd_install(fd1, newfile1);
1345 fd_install(fd2, newfile2);
1346 /* fd1 and fd2 may be already another descriptors.
1347 * Not kernel problem.
1363 sock_release(sock2);
1366 out_put_unused_both:
1371 sock_release(sock2);
1373 sock_release(sock1);
1379 * Bind a name to a socket. Nothing much to do here since it's
1380 * the protocol's responsibility to handle the local address.
1382 * We move the socket address to kernel space before we call
1383 * the protocol layer (having also checked the address is ok).
1386 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1388 struct socket *sock;
1389 struct sockaddr_storage address;
1390 int err, fput_needed;
1392 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1394 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1396 err = security_socket_bind(sock,
1397 (struct sockaddr *)&address,
1400 err = sock->ops->bind(sock,
1404 fput_light(sock->file, fput_needed);
1410 * Perform a listen. Basically, we allow the protocol to do anything
1411 * necessary for a listen, and if that works, we mark the socket as
1412 * ready for listening.
1415 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1417 struct socket *sock;
1418 int err, fput_needed;
1421 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1423 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1424 if ((unsigned int)backlog > somaxconn)
1425 backlog = somaxconn;
1427 err = security_socket_listen(sock, backlog);
1429 err = sock->ops->listen(sock, backlog);
1431 fput_light(sock->file, fput_needed);
1437 * For accept, we attempt to create a new socket, set up the link
1438 * with the client, wake up the client, then return the new
1439 * connected fd. We collect the address of the connector in kernel
1440 * space and move it to user at the very end. This is unclean because
1441 * we open the socket then return an error.
1443 * 1003.1g adds the ability to recvmsg() to query connection pending
1444 * status to recvmsg. We need to add that support in a way thats
1445 * clean when we restucture accept also.
1448 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1449 int __user *, upeer_addrlen, int, flags)
1451 struct socket *sock, *newsock;
1452 struct file *newfile;
1453 int err, len, newfd, fput_needed;
1454 struct sockaddr_storage address;
1456 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1459 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1460 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1462 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1467 newsock = sock_alloc();
1471 newsock->type = sock->type;
1472 newsock->ops = sock->ops;
1475 * We don't need try_module_get here, as the listening socket (sock)
1476 * has the protocol module (sock->ops->owner) held.
1478 __module_get(newsock->ops->owner);
1480 newfd = get_unused_fd_flags(flags);
1481 if (unlikely(newfd < 0)) {
1483 sock_release(newsock);
1486 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1487 if (unlikely(IS_ERR(newfile))) {
1488 err = PTR_ERR(newfile);
1489 put_unused_fd(newfd);
1490 sock_release(newsock);
1494 err = security_socket_accept(sock, newsock);
1498 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1502 if (upeer_sockaddr) {
1503 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1505 err = -ECONNABORTED;
1508 err = move_addr_to_user(&address,
1509 len, upeer_sockaddr, upeer_addrlen);
1514 /* File flags are not inherited via accept() unlike another OSes. */
1516 fd_install(newfd, newfile);
1520 fput_light(sock->file, fput_needed);
1525 put_unused_fd(newfd);
1529 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1530 int __user *, upeer_addrlen)
1532 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1536 * Attempt to connect to a socket with the server address. The address
1537 * is in user space so we verify it is OK and move it to kernel space.
1539 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1542 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1543 * other SEQPACKET protocols that take time to connect() as it doesn't
1544 * include the -EINPROGRESS status for such sockets.
1547 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1550 struct socket *sock;
1551 struct sockaddr_storage address;
1552 int err, fput_needed;
1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1562 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1566 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1567 sock->file->f_flags);
1569 fput_light(sock->file, fput_needed);
1575 * Get the local address ('name') of a socket object. Move the obtained
1576 * name to user space.
1579 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1580 int __user *, usockaddr_len)
1582 struct socket *sock;
1583 struct sockaddr_storage address;
1584 int len, err, fput_needed;
1586 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1590 err = security_socket_getsockname(sock);
1594 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1597 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1600 fput_light(sock->file, fput_needed);
1606 * Get the remote address ('name') of a socket object. Move the obtained
1607 * name to user space.
1610 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1611 int __user *, usockaddr_len)
1613 struct socket *sock;
1614 struct sockaddr_storage address;
1615 int len, err, fput_needed;
1617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1619 err = security_socket_getpeername(sock);
1621 fput_light(sock->file, fput_needed);
1626 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1629 err = move_addr_to_user(&address, len, usockaddr,
1631 fput_light(sock->file, fput_needed);
1637 * Send a datagram to a given address. We move the address into kernel
1638 * space and check the user space data area is readable before invoking
1642 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1643 unsigned int, flags, struct sockaddr __user *, addr,
1646 struct socket *sock;
1647 struct sockaddr_storage address;
1655 if (unlikely(!access_ok(VERIFY_READ, buff, len)))
1657 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1661 iov.iov_base = buff;
1663 msg.msg_name = NULL;
1664 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1665 msg.msg_control = NULL;
1666 msg.msg_controllen = 0;
1667 msg.msg_namelen = 0;
1669 err = move_addr_to_kernel(addr, addr_len, &address);
1672 msg.msg_name = (struct sockaddr *)&address;
1673 msg.msg_namelen = addr_len;
1675 if (sock->file->f_flags & O_NONBLOCK)
1676 flags |= MSG_DONTWAIT;
1677 msg.msg_flags = flags;
1678 err = sock_sendmsg(sock, &msg, len);
1681 fput_light(sock->file, fput_needed);
1687 * Send a datagram down a socket.
1690 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1691 unsigned int, flags)
1693 return sys_sendto(fd, buff, len, flags, NULL, 0);
1697 * Receive a frame from the socket and optionally record the address of the
1698 * sender. We verify the buffers are writable and if needed move the
1699 * sender address from kernel to user space.
1702 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1703 unsigned int, flags, struct sockaddr __user *, addr,
1704 int __user *, addr_len)
1706 struct socket *sock;
1709 struct sockaddr_storage address;
1715 if (unlikely(!access_ok(VERIFY_WRITE, ubuf, size)))
1717 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1721 msg.msg_control = NULL;
1722 msg.msg_controllen = 0;
1724 iov.iov_base = ubuf;
1725 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1726 /* Save some cycles and don't copy the address if not needed */
1727 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1728 /* We assume all kernel code knows the size of sockaddr_storage */
1729 msg.msg_namelen = 0;
1730 if (sock->file->f_flags & O_NONBLOCK)
1731 flags |= MSG_DONTWAIT;
1732 err = sock_recvmsg(sock, &msg, size, flags);
1734 if (err >= 0 && addr != NULL) {
1735 err2 = move_addr_to_user(&address,
1736 msg.msg_namelen, addr, addr_len);
1741 fput_light(sock->file, fput_needed);
1747 * Receive a datagram from a socket.
1750 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1751 unsigned int, flags)
1753 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1757 * Set a socket option. Because we don't know the option lengths we have
1758 * to pass the user mode parameter for the protocols to sort out.
1761 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1762 char __user *, optval, int, optlen)
1764 int err, fput_needed;
1765 struct socket *sock;
1770 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1772 err = security_socket_setsockopt(sock, level, optname);
1776 if (level == SOL_SOCKET)
1778 sock_setsockopt(sock, level, optname, optval,
1782 sock->ops->setsockopt(sock, level, optname, optval,
1785 fput_light(sock->file, fput_needed);
1791 * Get a socket option. Because we don't know the option lengths we have
1792 * to pass a user mode parameter for the protocols to sort out.
1795 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1796 char __user *, optval, int __user *, optlen)
1798 int err, fput_needed;
1799 struct socket *sock;
1801 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1803 err = security_socket_getsockopt(sock, level, optname);
1807 if (level == SOL_SOCKET)
1809 sock_getsockopt(sock, level, optname, optval,
1813 sock->ops->getsockopt(sock, level, optname, optval,
1816 fput_light(sock->file, fput_needed);
1822 * Shutdown a socket.
1825 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1827 int err, fput_needed;
1828 struct socket *sock;
1830 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1832 err = security_socket_shutdown(sock, how);
1834 err = sock->ops->shutdown(sock, how);
1835 fput_light(sock->file, fput_needed);
1840 /* A couple of helpful macros for getting the address of the 32/64 bit
1841 * fields which are the same type (int / unsigned) on our platforms.
1843 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1844 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1845 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1847 struct used_address {
1848 struct sockaddr_storage name;
1849 unsigned int name_len;
1852 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1853 struct user_msghdr __user *umsg,
1854 struct sockaddr __user **save_addr,
1857 struct sockaddr __user *uaddr;
1858 struct iovec __user *uiov;
1862 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1863 __get_user(uaddr, &umsg->msg_name) ||
1864 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1865 __get_user(uiov, &umsg->msg_iov) ||
1866 __get_user(nr_segs, &umsg->msg_iovlen) ||
1867 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1868 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1869 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1873 kmsg->msg_namelen = 0;
1875 if (kmsg->msg_namelen < 0)
1878 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1879 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1884 if (uaddr && kmsg->msg_namelen) {
1886 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1892 kmsg->msg_name = NULL;
1893 kmsg->msg_namelen = 0;
1896 if (nr_segs > UIO_MAXIOV)
1899 kmsg->msg_iocb = NULL;
1901 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
1903 UIO_FASTIOV, *iov, iov);
1905 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1906 *iov, nr_segs, err);
1910 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1911 struct msghdr *msg_sys, unsigned int flags,
1912 struct used_address *used_address)
1914 struct compat_msghdr __user *msg_compat =
1915 (struct compat_msghdr __user *)msg;
1916 struct sockaddr_storage address;
1917 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1918 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1919 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1920 /* 20 is size of ipv6_pktinfo */
1921 unsigned char *ctl_buf = ctl;
1922 int ctl_len, total_len;
1925 msg_sys->msg_name = &address;
1927 if (MSG_CMSG_COMPAT & flags)
1928 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1930 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1937 if (msg_sys->msg_controllen > INT_MAX)
1939 ctl_len = msg_sys->msg_controllen;
1940 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1942 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1946 ctl_buf = msg_sys->msg_control;
1947 ctl_len = msg_sys->msg_controllen;
1948 } else if (ctl_len) {
1949 if (ctl_len > sizeof(ctl)) {
1950 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1951 if (ctl_buf == NULL)
1956 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1957 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1958 * checking falls down on this.
1960 if (copy_from_user(ctl_buf,
1961 (void __user __force *)msg_sys->msg_control,
1964 msg_sys->msg_control = ctl_buf;
1966 msg_sys->msg_flags = flags;
1968 if (sock->file->f_flags & O_NONBLOCK)
1969 msg_sys->msg_flags |= MSG_DONTWAIT;
1971 * If this is sendmmsg() and current destination address is same as
1972 * previously succeeded address, omit asking LSM's decision.
1973 * used_address->name_len is initialized to UINT_MAX so that the first
1974 * destination address never matches.
1976 if (used_address && msg_sys->msg_name &&
1977 used_address->name_len == msg_sys->msg_namelen &&
1978 !memcmp(&used_address->name, msg_sys->msg_name,
1979 used_address->name_len)) {
1980 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1983 err = sock_sendmsg(sock, msg_sys, total_len);
1985 * If this is sendmmsg() and sending to current destination address was
1986 * successful, remember it.
1988 if (used_address && err >= 0) {
1989 used_address->name_len = msg_sys->msg_namelen;
1990 if (msg_sys->msg_name)
1991 memcpy(&used_address->name, msg_sys->msg_name,
1992 used_address->name_len);
1997 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1999 if (iov != iovstack)
2005 * BSD sendmsg interface
2008 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2010 int fput_needed, err;
2011 struct msghdr msg_sys;
2012 struct socket *sock;
2014 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2018 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2020 fput_light(sock->file, fput_needed);
2025 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2027 if (flags & MSG_CMSG_COMPAT)
2029 return __sys_sendmsg(fd, msg, flags);
2033 * Linux sendmmsg interface
2036 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2039 int fput_needed, err, datagrams;
2040 struct socket *sock;
2041 struct mmsghdr __user *entry;
2042 struct compat_mmsghdr __user *compat_entry;
2043 struct msghdr msg_sys;
2044 struct used_address used_address;
2046 if (vlen > UIO_MAXIOV)
2051 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2055 used_address.name_len = UINT_MAX;
2057 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2060 while (datagrams < vlen) {
2061 if (MSG_CMSG_COMPAT & flags) {
2062 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2063 &msg_sys, flags, &used_address);
2066 err = __put_user(err, &compat_entry->msg_len);
2069 err = ___sys_sendmsg(sock,
2070 (struct user_msghdr __user *)entry,
2071 &msg_sys, flags, &used_address);
2074 err = put_user(err, &entry->msg_len);
2083 fput_light(sock->file, fput_needed);
2085 /* We only return an error if no datagrams were able to be sent */
2092 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2093 unsigned int, vlen, unsigned int, flags)
2095 if (flags & MSG_CMSG_COMPAT)
2097 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2100 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2101 struct msghdr *msg_sys, unsigned int flags, int nosec)
2103 struct compat_msghdr __user *msg_compat =
2104 (struct compat_msghdr __user *)msg;
2105 struct iovec iovstack[UIO_FASTIOV];
2106 struct iovec *iov = iovstack;
2107 unsigned long cmsg_ptr;
2111 /* kernel mode address */
2112 struct sockaddr_storage addr;
2114 /* user mode address pointers */
2115 struct sockaddr __user *uaddr;
2116 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2118 msg_sys->msg_name = &addr;
2120 if (MSG_CMSG_COMPAT & flags)
2121 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2123 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2128 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2129 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2131 /* We assume all kernel code knows the size of sockaddr_storage */
2132 msg_sys->msg_namelen = 0;
2134 if (sock->file->f_flags & O_NONBLOCK)
2135 flags |= MSG_DONTWAIT;
2136 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2142 if (uaddr != NULL) {
2143 err = move_addr_to_user(&addr,
2144 msg_sys->msg_namelen, uaddr,
2149 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2153 if (MSG_CMSG_COMPAT & flags)
2154 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2155 &msg_compat->msg_controllen);
2157 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2158 &msg->msg_controllen);
2164 if (iov != iovstack)
2170 * BSD recvmsg interface
2173 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2175 int fput_needed, err;
2176 struct msghdr msg_sys;
2177 struct socket *sock;
2179 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2183 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2185 fput_light(sock->file, fput_needed);
2190 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2191 unsigned int, flags)
2193 if (flags & MSG_CMSG_COMPAT)
2195 return __sys_recvmsg(fd, msg, flags);
2199 * Linux recvmmsg interface
2202 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2203 unsigned int flags, struct timespec *timeout)
2205 int fput_needed, err, datagrams;
2206 struct socket *sock;
2207 struct mmsghdr __user *entry;
2208 struct compat_mmsghdr __user *compat_entry;
2209 struct msghdr msg_sys;
2210 struct timespec end_time;
2213 poll_select_set_timeout(&end_time, timeout->tv_sec,
2219 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2223 err = sock_error(sock->sk);
2228 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2230 while (datagrams < vlen) {
2232 * No need to ask LSM for more than the first datagram.
2234 if (MSG_CMSG_COMPAT & flags) {
2235 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2236 &msg_sys, flags & ~MSG_WAITFORONE,
2240 err = __put_user(err, &compat_entry->msg_len);
2243 err = ___sys_recvmsg(sock,
2244 (struct user_msghdr __user *)entry,
2245 &msg_sys, flags & ~MSG_WAITFORONE,
2249 err = put_user(err, &entry->msg_len);
2257 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2258 if (flags & MSG_WAITFORONE)
2259 flags |= MSG_DONTWAIT;
2262 ktime_get_ts(timeout);
2263 *timeout = timespec_sub(end_time, *timeout);
2264 if (timeout->tv_sec < 0) {
2265 timeout->tv_sec = timeout->tv_nsec = 0;
2269 /* Timeout, return less than vlen datagrams */
2270 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2274 /* Out of band data, return right away */
2275 if (msg_sys.msg_flags & MSG_OOB)
2280 fput_light(sock->file, fput_needed);
2285 if (datagrams != 0) {
2287 * We may return less entries than requested (vlen) if the
2288 * sock is non block and there aren't enough datagrams...
2290 if (err != -EAGAIN) {
2292 * ... or if recvmsg returns an error after we
2293 * received some datagrams, where we record the
2294 * error to return on the next call or if the
2295 * app asks about it using getsockopt(SO_ERROR).
2297 sock->sk->sk_err = -err;
2306 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2307 unsigned int, vlen, unsigned int, flags,
2308 struct timespec __user *, timeout)
2311 struct timespec timeout_sys;
2313 if (flags & MSG_CMSG_COMPAT)
2317 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2319 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2322 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2324 if (datagrams > 0 &&
2325 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2326 datagrams = -EFAULT;
2331 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2332 /* Argument list sizes for sys_socketcall */
2333 #define AL(x) ((x) * sizeof(unsigned long))
2334 static const unsigned char nargs[21] = {
2335 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2336 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2337 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2344 * System call vectors.
2346 * Argument checking cleaned up. Saved 20% in size.
2347 * This function doesn't need to set the kernel lock because
2348 * it is set by the callees.
2351 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2353 unsigned long a[AUDITSC_ARGS];
2354 unsigned long a0, a1;
2358 if (call < 1 || call > SYS_SENDMMSG)
2362 if (len > sizeof(a))
2365 /* copy_from_user should be SMP safe. */
2366 if (copy_from_user(a, args, len))
2369 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2378 err = sys_socket(a0, a1, a[2]);
2381 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2384 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2387 err = sys_listen(a0, a1);
2390 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2391 (int __user *)a[2], 0);
2393 case SYS_GETSOCKNAME:
2395 sys_getsockname(a0, (struct sockaddr __user *)a1,
2396 (int __user *)a[2]);
2398 case SYS_GETPEERNAME:
2400 sys_getpeername(a0, (struct sockaddr __user *)a1,
2401 (int __user *)a[2]);
2403 case SYS_SOCKETPAIR:
2404 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2407 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2410 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2411 (struct sockaddr __user *)a[4], a[5]);
2414 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2417 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2418 (struct sockaddr __user *)a[4],
2419 (int __user *)a[5]);
2422 err = sys_shutdown(a0, a1);
2424 case SYS_SETSOCKOPT:
2425 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2427 case SYS_GETSOCKOPT:
2429 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2430 (int __user *)a[4]);
2433 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2436 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2439 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2442 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2443 (struct timespec __user *)a[4]);
2446 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2447 (int __user *)a[2], a[3]);
2456 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2459 * sock_register - add a socket protocol handler
2460 * @ops: description of protocol
2462 * This function is called by a protocol handler that wants to
2463 * advertise its address family, and have it linked into the
2464 * socket interface. The value ops->family corresponds to the
2465 * socket system call protocol family.
2467 int sock_register(const struct net_proto_family *ops)
2471 if (ops->family >= NPROTO) {
2472 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2476 spin_lock(&net_family_lock);
2477 if (rcu_dereference_protected(net_families[ops->family],
2478 lockdep_is_held(&net_family_lock)))
2481 rcu_assign_pointer(net_families[ops->family], ops);
2484 spin_unlock(&net_family_lock);
2486 pr_info("NET: Registered protocol family %d\n", ops->family);
2489 EXPORT_SYMBOL(sock_register);
2492 * sock_unregister - remove a protocol handler
2493 * @family: protocol family to remove
2495 * This function is called by a protocol handler that wants to
2496 * remove its address family, and have it unlinked from the
2497 * new socket creation.
2499 * If protocol handler is a module, then it can use module reference
2500 * counts to protect against new references. If protocol handler is not
2501 * a module then it needs to provide its own protection in
2502 * the ops->create routine.
2504 void sock_unregister(int family)
2506 BUG_ON(family < 0 || family >= NPROTO);
2508 spin_lock(&net_family_lock);
2509 RCU_INIT_POINTER(net_families[family], NULL);
2510 spin_unlock(&net_family_lock);
2514 pr_info("NET: Unregistered protocol family %d\n", family);
2516 EXPORT_SYMBOL(sock_unregister);
2518 static int __init sock_init(void)
2522 * Initialize the network sysctl infrastructure.
2524 err = net_sysctl_init();
2529 * Initialize skbuff SLAB cache
2534 * Initialize the protocols module.
2539 err = register_filesystem(&sock_fs_type);
2542 sock_mnt = kern_mount(&sock_fs_type);
2543 if (IS_ERR(sock_mnt)) {
2544 err = PTR_ERR(sock_mnt);
2548 /* The real protocol initialization is performed in later initcalls.
2551 #ifdef CONFIG_NETFILTER
2552 err = netfilter_init();
2557 ptp_classifier_init();
2563 unregister_filesystem(&sock_fs_type);
2568 core_initcall(sock_init); /* early initcall */
2570 #ifdef CONFIG_PROC_FS
2571 void socket_seq_show(struct seq_file *seq)
2576 for_each_possible_cpu(cpu)
2577 counter += per_cpu(sockets_in_use, cpu);
2579 /* It can be negative, by the way. 8) */
2583 seq_printf(seq, "sockets: used %d\n", counter);
2585 #endif /* CONFIG_PROC_FS */
2587 #ifdef CONFIG_COMPAT
2588 static int do_siocgstamp(struct net *net, struct socket *sock,
2589 unsigned int cmd, void __user *up)
2591 mm_segment_t old_fs = get_fs();
2596 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2599 err = compat_put_timeval(&ktv, up);
2604 static int do_siocgstampns(struct net *net, struct socket *sock,
2605 unsigned int cmd, void __user *up)
2607 mm_segment_t old_fs = get_fs();
2608 struct timespec kts;
2612 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2615 err = compat_put_timespec(&kts, up);
2620 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2622 struct ifreq __user *uifr;
2625 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2626 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2629 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2633 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2639 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2641 struct compat_ifconf ifc32;
2643 struct ifconf __user *uifc;
2644 struct compat_ifreq __user *ifr32;
2645 struct ifreq __user *ifr;
2649 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2652 memset(&ifc, 0, sizeof(ifc));
2653 if (ifc32.ifcbuf == 0) {
2657 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2659 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2660 sizeof(struct ifreq);
2661 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2663 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2664 ifr32 = compat_ptr(ifc32.ifcbuf);
2665 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2666 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2672 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2675 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2679 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2683 ifr32 = compat_ptr(ifc32.ifcbuf);
2685 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2686 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2687 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2693 if (ifc32.ifcbuf == 0) {
2694 /* Translate from 64-bit structure multiple to
2698 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2703 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2709 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2711 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2712 bool convert_in = false, convert_out = false;
2713 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2714 struct ethtool_rxnfc __user *rxnfc;
2715 struct ifreq __user *ifr;
2716 u32 rule_cnt = 0, actual_rule_cnt;
2721 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2724 compat_rxnfc = compat_ptr(data);
2726 if (get_user(ethcmd, &compat_rxnfc->cmd))
2729 /* Most ethtool structures are defined without padding.
2730 * Unfortunately struct ethtool_rxnfc is an exception.
2735 case ETHTOOL_GRXCLSRLALL:
2736 /* Buffer size is variable */
2737 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2739 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2741 buf_size += rule_cnt * sizeof(u32);
2743 case ETHTOOL_GRXRINGS:
2744 case ETHTOOL_GRXCLSRLCNT:
2745 case ETHTOOL_GRXCLSRULE:
2746 case ETHTOOL_SRXCLSRLINS:
2749 case ETHTOOL_SRXCLSRLDEL:
2750 buf_size += sizeof(struct ethtool_rxnfc);
2755 ifr = compat_alloc_user_space(buf_size);
2756 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2758 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2761 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2762 &ifr->ifr_ifru.ifru_data))
2766 /* We expect there to be holes between fs.m_ext and
2767 * fs.ring_cookie and at the end of fs, but nowhere else.
2769 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2770 sizeof(compat_rxnfc->fs.m_ext) !=
2771 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2772 sizeof(rxnfc->fs.m_ext));
2774 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2775 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2776 offsetof(struct ethtool_rxnfc, fs.location) -
2777 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2779 if (copy_in_user(rxnfc, compat_rxnfc,
2780 (void __user *)(&rxnfc->fs.m_ext + 1) -
2781 (void __user *)rxnfc) ||
2782 copy_in_user(&rxnfc->fs.ring_cookie,
2783 &compat_rxnfc->fs.ring_cookie,
2784 (void __user *)(&rxnfc->fs.location + 1) -
2785 (void __user *)&rxnfc->fs.ring_cookie) ||
2786 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2787 sizeof(rxnfc->rule_cnt)))
2791 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2796 if (copy_in_user(compat_rxnfc, rxnfc,
2797 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2798 (const void __user *)rxnfc) ||
2799 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2800 &rxnfc->fs.ring_cookie,
2801 (const void __user *)(&rxnfc->fs.location + 1) -
2802 (const void __user *)&rxnfc->fs.ring_cookie) ||
2803 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2804 sizeof(rxnfc->rule_cnt)))
2807 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2808 /* As an optimisation, we only copy the actual
2809 * number of rules that the underlying
2810 * function returned. Since Mallory might
2811 * change the rule count in user memory, we
2812 * check that it is less than the rule count
2813 * originally given (as the user buffer size),
2814 * which has been range-checked.
2816 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2818 if (actual_rule_cnt < rule_cnt)
2819 rule_cnt = actual_rule_cnt;
2820 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2821 &rxnfc->rule_locs[0],
2822 rule_cnt * sizeof(u32)))
2830 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2833 compat_uptr_t uptr32;
2834 struct ifreq __user *uifr;
2836 uifr = compat_alloc_user_space(sizeof(*uifr));
2837 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2840 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2843 uptr = compat_ptr(uptr32);
2845 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2848 return dev_ioctl(net, SIOCWANDEV, uifr);
2851 static int bond_ioctl(struct net *net, unsigned int cmd,
2852 struct compat_ifreq __user *ifr32)
2855 mm_segment_t old_fs;
2859 case SIOCBONDENSLAVE:
2860 case SIOCBONDRELEASE:
2861 case SIOCBONDSETHWADDR:
2862 case SIOCBONDCHANGEACTIVE:
2863 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2868 err = dev_ioctl(net, cmd,
2869 (struct ifreq __user __force *) &kifr);
2874 return -ENOIOCTLCMD;
2878 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2879 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2880 struct compat_ifreq __user *u_ifreq32)
2882 struct ifreq __user *u_ifreq64;
2883 char tmp_buf[IFNAMSIZ];
2884 void __user *data64;
2887 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2890 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2892 data64 = compat_ptr(data32);
2894 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2896 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2899 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2902 return dev_ioctl(net, cmd, u_ifreq64);
2905 static int dev_ifsioc(struct net *net, struct socket *sock,
2906 unsigned int cmd, struct compat_ifreq __user *uifr32)
2908 struct ifreq __user *uifr;
2911 uifr = compat_alloc_user_space(sizeof(*uifr));
2912 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2915 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2926 case SIOCGIFBRDADDR:
2927 case SIOCGIFDSTADDR:
2928 case SIOCGIFNETMASK:
2933 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2941 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2942 struct compat_ifreq __user *uifr32)
2945 struct compat_ifmap __user *uifmap32;
2946 mm_segment_t old_fs;
2949 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2950 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2951 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2952 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2953 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2954 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2955 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2956 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2962 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2965 if (cmd == SIOCGIFMAP && !err) {
2966 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2967 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2968 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2969 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2970 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2971 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2972 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2981 struct sockaddr rt_dst; /* target address */
2982 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2983 struct sockaddr rt_genmask; /* target network mask (IP) */
2984 unsigned short rt_flags;
2987 unsigned char rt_tos;
2988 unsigned char rt_class;
2990 short rt_metric; /* +1 for binary compatibility! */
2991 /* char * */ u32 rt_dev; /* forcing the device at add */
2992 u32 rt_mtu; /* per route MTU/Window */
2993 u32 rt_window; /* Window clamping */
2994 unsigned short rt_irtt; /* Initial RTT */
2997 struct in6_rtmsg32 {
2998 struct in6_addr rtmsg_dst;
2999 struct in6_addr rtmsg_src;
3000 struct in6_addr rtmsg_gateway;
3010 static int routing_ioctl(struct net *net, struct socket *sock,
3011 unsigned int cmd, void __user *argp)
3015 struct in6_rtmsg r6;
3019 mm_segment_t old_fs = get_fs();
3021 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3022 struct in6_rtmsg32 __user *ur6 = argp;
3023 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3024 3 * sizeof(struct in6_addr));
3025 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3026 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3027 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3028 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3029 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3030 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3031 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3035 struct rtentry32 __user *ur4 = argp;
3036 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3037 3 * sizeof(struct sockaddr));
3038 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3039 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3040 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3041 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3042 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3043 ret |= get_user(rtdev, &(ur4->rt_dev));
3045 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3046 r4.rt_dev = (char __user __force *)devname;
3060 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3067 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3068 * for some operations; this forces use of the newer bridge-utils that
3069 * use compatible ioctls
3071 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3075 if (get_user(tmp, argp))
3077 if (tmp == BRCTL_GET_VERSION)
3078 return BRCTL_VERSION + 1;
3082 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3083 unsigned int cmd, unsigned long arg)
3085 void __user *argp = compat_ptr(arg);
3086 struct sock *sk = sock->sk;
3087 struct net *net = sock_net(sk);
3089 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3090 return compat_ifr_data_ioctl(net, cmd, argp);
3095 return old_bridge_ioctl(argp);
3097 return dev_ifname32(net, argp);
3099 return dev_ifconf(net, argp);
3101 return ethtool_ioctl(net, argp);
3103 return compat_siocwandev(net, argp);
3106 return compat_sioc_ifmap(net, cmd, argp);
3107 case SIOCBONDENSLAVE:
3108 case SIOCBONDRELEASE:
3109 case SIOCBONDSETHWADDR:
3110 case SIOCBONDCHANGEACTIVE:
3111 return bond_ioctl(net, cmd, argp);
3114 return routing_ioctl(net, sock, cmd, argp);
3116 return do_siocgstamp(net, sock, cmd, argp);
3118 return do_siocgstampns(net, sock, cmd, argp);
3119 case SIOCBONDSLAVEINFOQUERY:
3120 case SIOCBONDINFOQUERY:
3123 return compat_ifr_data_ioctl(net, cmd, argp);
3135 return sock_ioctl(file, cmd, arg);
3152 case SIOCSIFHWBROADCAST:
3154 case SIOCGIFBRDADDR:
3155 case SIOCSIFBRDADDR:
3156 case SIOCGIFDSTADDR:
3157 case SIOCSIFDSTADDR:
3158 case SIOCGIFNETMASK:
3159 case SIOCSIFNETMASK:
3170 return dev_ifsioc(net, sock, cmd, argp);
3176 return sock_do_ioctl(net, sock, cmd, arg);
3179 return -ENOIOCTLCMD;
3182 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3185 struct socket *sock = file->private_data;
3186 int ret = -ENOIOCTLCMD;
3193 if (sock->ops->compat_ioctl)
3194 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3196 if (ret == -ENOIOCTLCMD &&
3197 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3198 ret = compat_wext_handle_ioctl(net, cmd, arg);
3200 if (ret == -ENOIOCTLCMD)
3201 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3207 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3209 return sock->ops->bind(sock, addr, addrlen);
3211 EXPORT_SYMBOL(kernel_bind);
3213 int kernel_listen(struct socket *sock, int backlog)
3215 return sock->ops->listen(sock, backlog);
3217 EXPORT_SYMBOL(kernel_listen);
3219 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3221 struct sock *sk = sock->sk;
3224 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3229 err = sock->ops->accept(sock, *newsock, flags);
3231 sock_release(*newsock);
3236 (*newsock)->ops = sock->ops;
3237 __module_get((*newsock)->ops->owner);
3242 EXPORT_SYMBOL(kernel_accept);
3244 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3247 return sock->ops->connect(sock, addr, addrlen, flags);
3249 EXPORT_SYMBOL(kernel_connect);
3251 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3254 return sock->ops->getname(sock, addr, addrlen, 0);
3256 EXPORT_SYMBOL(kernel_getsockname);
3258 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3261 return sock->ops->getname(sock, addr, addrlen, 1);
3263 EXPORT_SYMBOL(kernel_getpeername);
3265 int kernel_getsockopt(struct socket *sock, int level, int optname,
3266 char *optval, int *optlen)
3268 mm_segment_t oldfs = get_fs();
3269 char __user *uoptval;
3270 int __user *uoptlen;
3273 uoptval = (char __user __force *) optval;
3274 uoptlen = (int __user __force *) optlen;
3277 if (level == SOL_SOCKET)
3278 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3280 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3285 EXPORT_SYMBOL(kernel_getsockopt);
3287 int kernel_setsockopt(struct socket *sock, int level, int optname,
3288 char *optval, unsigned int optlen)
3290 mm_segment_t oldfs = get_fs();
3291 char __user *uoptval;
3294 uoptval = (char __user __force *) optval;
3297 if (level == SOL_SOCKET)
3298 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3300 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3305 EXPORT_SYMBOL(kernel_setsockopt);
3307 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3308 size_t size, int flags)
3310 if (sock->ops->sendpage)
3311 return sock->ops->sendpage(sock, page, offset, size, flags);
3313 return sock_no_sendpage(sock, page, offset, size, flags);
3315 EXPORT_SYMBOL(kernel_sendpage);
3317 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3319 mm_segment_t oldfs = get_fs();
3323 err = sock->ops->ioctl(sock, cmd, arg);
3328 EXPORT_SYMBOL(kernel_sock_ioctl);
3330 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3332 return sock->ops->shutdown(sock, how);
3334 EXPORT_SYMBOL(kernel_sock_shutdown);