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};
804 if (file->f_flags & O_NONBLOCK)
805 msg.msg_flags = MSG_DONTWAIT;
807 if (iocb->ki_pos != 0)
810 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
813 res = sock_recvmsg(sock, &msg, iocb->ki_nbytes, msg.msg_flags);
818 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
820 struct file *file = iocb->ki_filp;
821 struct socket *sock = file->private_data;
822 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, iocb->ki_nbytes);
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 socket lock or callback_lock or rcu_lock */
1076 int sock_wake_async(struct socket *sock, int how, int band)
1078 struct socket_wq *wq;
1083 wq = rcu_dereference(sock->wq);
1084 if (!wq || !wq->fasync_list) {
1089 case SOCK_WAKE_WAITD:
1090 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1093 case SOCK_WAKE_SPACE:
1094 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1099 kill_fasync(&wq->fasync_list, SIGIO, band);
1102 kill_fasync(&wq->fasync_list, SIGURG, band);
1107 EXPORT_SYMBOL(sock_wake_async);
1109 int __sock_create(struct net *net, int family, int type, int protocol,
1110 struct socket **res, int kern)
1113 struct socket *sock;
1114 const struct net_proto_family *pf;
1117 * Check protocol is in range
1119 if (family < 0 || family >= NPROTO)
1120 return -EAFNOSUPPORT;
1121 if (type < 0 || type >= SOCK_MAX)
1126 This uglymoron is moved from INET layer to here to avoid
1127 deadlock in module load.
1129 if (family == PF_INET && type == SOCK_PACKET) {
1133 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1139 err = security_socket_create(family, type, protocol, kern);
1144 * Allocate the socket and allow the family to set things up. if
1145 * the protocol is 0, the family is instructed to select an appropriate
1148 sock = sock_alloc();
1150 net_warn_ratelimited("socket: no more sockets\n");
1151 return -ENFILE; /* Not exactly a match, but its the
1152 closest posix thing */
1157 #ifdef CONFIG_MODULES
1158 /* Attempt to load a protocol module if the find failed.
1160 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1161 * requested real, full-featured networking support upon configuration.
1162 * Otherwise module support will break!
1164 if (rcu_access_pointer(net_families[family]) == NULL)
1165 request_module("net-pf-%d", family);
1169 pf = rcu_dereference(net_families[family]);
1170 err = -EAFNOSUPPORT;
1175 * We will call the ->create function, that possibly is in a loadable
1176 * module, so we have to bump that loadable module refcnt first.
1178 if (!try_module_get(pf->owner))
1181 /* Now protected by module ref count */
1184 err = pf->create(net, sock, protocol, kern);
1186 goto out_module_put;
1189 * Now to bump the refcnt of the [loadable] module that owns this
1190 * socket at sock_release time we decrement its refcnt.
1192 if (!try_module_get(sock->ops->owner))
1193 goto out_module_busy;
1196 * Now that we're done with the ->create function, the [loadable]
1197 * module can have its refcnt decremented
1199 module_put(pf->owner);
1200 err = security_socket_post_create(sock, family, type, protocol, kern);
1202 goto out_sock_release;
1208 err = -EAFNOSUPPORT;
1211 module_put(pf->owner);
1218 goto out_sock_release;
1220 EXPORT_SYMBOL(__sock_create);
1222 int sock_create(int family, int type, int protocol, struct socket **res)
1224 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1226 EXPORT_SYMBOL(sock_create);
1228 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1230 return __sock_create(&init_net, family, type, protocol, res, 1);
1232 EXPORT_SYMBOL(sock_create_kern);
1234 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1237 struct socket *sock;
1240 /* Check the SOCK_* constants for consistency. */
1241 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1242 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1243 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1244 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1246 flags = type & ~SOCK_TYPE_MASK;
1247 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1249 type &= SOCK_TYPE_MASK;
1251 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1252 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1254 retval = sock_create(family, type, protocol, &sock);
1258 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1263 /* It may be already another descriptor 8) Not kernel problem. */
1272 * Create a pair of connected sockets.
1275 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1276 int __user *, usockvec)
1278 struct socket *sock1, *sock2;
1280 struct file *newfile1, *newfile2;
1283 flags = type & ~SOCK_TYPE_MASK;
1284 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1286 type &= SOCK_TYPE_MASK;
1288 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1289 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1292 * Obtain the first socket and check if the underlying protocol
1293 * supports the socketpair call.
1296 err = sock_create(family, type, protocol, &sock1);
1300 err = sock_create(family, type, protocol, &sock2);
1304 err = sock1->ops->socketpair(sock1, sock2);
1306 goto out_release_both;
1308 fd1 = get_unused_fd_flags(flags);
1309 if (unlikely(fd1 < 0)) {
1311 goto out_release_both;
1314 fd2 = get_unused_fd_flags(flags);
1315 if (unlikely(fd2 < 0)) {
1317 goto out_put_unused_1;
1320 newfile1 = sock_alloc_file(sock1, flags, NULL);
1321 if (unlikely(IS_ERR(newfile1))) {
1322 err = PTR_ERR(newfile1);
1323 goto out_put_unused_both;
1326 newfile2 = sock_alloc_file(sock2, flags, NULL);
1327 if (IS_ERR(newfile2)) {
1328 err = PTR_ERR(newfile2);
1332 err = put_user(fd1, &usockvec[0]);
1336 err = put_user(fd2, &usockvec[1]);
1340 audit_fd_pair(fd1, fd2);
1342 fd_install(fd1, newfile1);
1343 fd_install(fd2, newfile2);
1344 /* fd1 and fd2 may be already another descriptors.
1345 * Not kernel problem.
1361 sock_release(sock2);
1364 out_put_unused_both:
1369 sock_release(sock2);
1371 sock_release(sock1);
1377 * Bind a name to a socket. Nothing much to do here since it's
1378 * the protocol's responsibility to handle the local address.
1380 * We move the socket address to kernel space before we call
1381 * the protocol layer (having also checked the address is ok).
1384 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1386 struct socket *sock;
1387 struct sockaddr_storage address;
1388 int err, fput_needed;
1390 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1392 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1394 err = security_socket_bind(sock,
1395 (struct sockaddr *)&address,
1398 err = sock->ops->bind(sock,
1402 fput_light(sock->file, fput_needed);
1408 * Perform a listen. Basically, we allow the protocol to do anything
1409 * necessary for a listen, and if that works, we mark the socket as
1410 * ready for listening.
1413 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1415 struct socket *sock;
1416 int err, fput_needed;
1419 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1421 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1422 if ((unsigned int)backlog > somaxconn)
1423 backlog = somaxconn;
1425 err = security_socket_listen(sock, backlog);
1427 err = sock->ops->listen(sock, backlog);
1429 fput_light(sock->file, fput_needed);
1435 * For accept, we attempt to create a new socket, set up the link
1436 * with the client, wake up the client, then return the new
1437 * connected fd. We collect the address of the connector in kernel
1438 * space and move it to user at the very end. This is unclean because
1439 * we open the socket then return an error.
1441 * 1003.1g adds the ability to recvmsg() to query connection pending
1442 * status to recvmsg. We need to add that support in a way thats
1443 * clean when we restucture accept also.
1446 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1447 int __user *, upeer_addrlen, int, flags)
1449 struct socket *sock, *newsock;
1450 struct file *newfile;
1451 int err, len, newfd, fput_needed;
1452 struct sockaddr_storage address;
1454 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1457 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1458 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1460 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1465 newsock = sock_alloc();
1469 newsock->type = sock->type;
1470 newsock->ops = sock->ops;
1473 * We don't need try_module_get here, as the listening socket (sock)
1474 * has the protocol module (sock->ops->owner) held.
1476 __module_get(newsock->ops->owner);
1478 newfd = get_unused_fd_flags(flags);
1479 if (unlikely(newfd < 0)) {
1481 sock_release(newsock);
1484 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1485 if (unlikely(IS_ERR(newfile))) {
1486 err = PTR_ERR(newfile);
1487 put_unused_fd(newfd);
1488 sock_release(newsock);
1492 err = security_socket_accept(sock, newsock);
1496 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1500 if (upeer_sockaddr) {
1501 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1503 err = -ECONNABORTED;
1506 err = move_addr_to_user(&address,
1507 len, upeer_sockaddr, upeer_addrlen);
1512 /* File flags are not inherited via accept() unlike another OSes. */
1514 fd_install(newfd, newfile);
1518 fput_light(sock->file, fput_needed);
1523 put_unused_fd(newfd);
1527 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1528 int __user *, upeer_addrlen)
1530 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1534 * Attempt to connect to a socket with the server address. The address
1535 * is in user space so we verify it is OK and move it to kernel space.
1537 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1540 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1541 * other SEQPACKET protocols that take time to connect() as it doesn't
1542 * include the -EINPROGRESS status for such sockets.
1545 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1548 struct socket *sock;
1549 struct sockaddr_storage address;
1550 int err, fput_needed;
1552 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1555 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1560 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1564 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1565 sock->file->f_flags);
1567 fput_light(sock->file, fput_needed);
1573 * Get the local address ('name') of a socket object. Move the obtained
1574 * name to user space.
1577 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1578 int __user *, usockaddr_len)
1580 struct socket *sock;
1581 struct sockaddr_storage address;
1582 int len, err, fput_needed;
1584 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1588 err = security_socket_getsockname(sock);
1592 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1595 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1598 fput_light(sock->file, fput_needed);
1604 * Get the remote address ('name') of a socket object. Move the obtained
1605 * name to user space.
1608 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1609 int __user *, usockaddr_len)
1611 struct socket *sock;
1612 struct sockaddr_storage address;
1613 int len, err, fput_needed;
1615 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1617 err = security_socket_getpeername(sock);
1619 fput_light(sock->file, fput_needed);
1624 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1627 err = move_addr_to_user(&address, len, usockaddr,
1629 fput_light(sock->file, fput_needed);
1635 * Send a datagram to a given address. We move the address into kernel
1636 * space and check the user space data area is readable before invoking
1640 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1641 unsigned int, flags, struct sockaddr __user *, addr,
1644 struct socket *sock;
1645 struct sockaddr_storage address;
1653 if (unlikely(!access_ok(VERIFY_READ, buff, len)))
1655 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1659 iov.iov_base = buff;
1661 msg.msg_name = NULL;
1662 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1663 msg.msg_control = NULL;
1664 msg.msg_controllen = 0;
1665 msg.msg_namelen = 0;
1667 err = move_addr_to_kernel(addr, addr_len, &address);
1670 msg.msg_name = (struct sockaddr *)&address;
1671 msg.msg_namelen = addr_len;
1673 if (sock->file->f_flags & O_NONBLOCK)
1674 flags |= MSG_DONTWAIT;
1675 msg.msg_flags = flags;
1676 err = sock_sendmsg(sock, &msg, len);
1679 fput_light(sock->file, fput_needed);
1685 * Send a datagram down a socket.
1688 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1689 unsigned int, flags)
1691 return sys_sendto(fd, buff, len, flags, NULL, 0);
1695 * Receive a frame from the socket and optionally record the address of the
1696 * sender. We verify the buffers are writable and if needed move the
1697 * sender address from kernel to user space.
1700 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1701 unsigned int, flags, struct sockaddr __user *, addr,
1702 int __user *, addr_len)
1704 struct socket *sock;
1707 struct sockaddr_storage address;
1713 if (unlikely(!access_ok(VERIFY_WRITE, ubuf, size)))
1715 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1719 msg.msg_control = NULL;
1720 msg.msg_controllen = 0;
1722 iov.iov_base = ubuf;
1723 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1724 /* Save some cycles and don't copy the address if not needed */
1725 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1726 /* We assume all kernel code knows the size of sockaddr_storage */
1727 msg.msg_namelen = 0;
1728 if (sock->file->f_flags & O_NONBLOCK)
1729 flags |= MSG_DONTWAIT;
1730 err = sock_recvmsg(sock, &msg, size, flags);
1732 if (err >= 0 && addr != NULL) {
1733 err2 = move_addr_to_user(&address,
1734 msg.msg_namelen, addr, addr_len);
1739 fput_light(sock->file, fput_needed);
1745 * Receive a datagram from a socket.
1748 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1749 unsigned int, flags)
1751 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1755 * Set a socket option. Because we don't know the option lengths we have
1756 * to pass the user mode parameter for the protocols to sort out.
1759 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1760 char __user *, optval, int, optlen)
1762 int err, fput_needed;
1763 struct socket *sock;
1768 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1770 err = security_socket_setsockopt(sock, level, optname);
1774 if (level == SOL_SOCKET)
1776 sock_setsockopt(sock, level, optname, optval,
1780 sock->ops->setsockopt(sock, level, optname, optval,
1783 fput_light(sock->file, fput_needed);
1789 * Get a socket option. Because we don't know the option lengths we have
1790 * to pass a user mode parameter for the protocols to sort out.
1793 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1794 char __user *, optval, int __user *, optlen)
1796 int err, fput_needed;
1797 struct socket *sock;
1799 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1801 err = security_socket_getsockopt(sock, level, optname);
1805 if (level == SOL_SOCKET)
1807 sock_getsockopt(sock, level, optname, optval,
1811 sock->ops->getsockopt(sock, level, optname, optval,
1814 fput_light(sock->file, fput_needed);
1820 * Shutdown a socket.
1823 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1825 int err, fput_needed;
1826 struct socket *sock;
1828 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1830 err = security_socket_shutdown(sock, how);
1832 err = sock->ops->shutdown(sock, how);
1833 fput_light(sock->file, fput_needed);
1838 /* A couple of helpful macros for getting the address of the 32/64 bit
1839 * fields which are the same type (int / unsigned) on our platforms.
1841 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1842 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1843 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1845 struct used_address {
1846 struct sockaddr_storage name;
1847 unsigned int name_len;
1850 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1851 struct user_msghdr __user *umsg,
1852 struct sockaddr __user **save_addr,
1855 struct sockaddr __user *uaddr;
1856 struct iovec __user *uiov;
1860 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1861 __get_user(uaddr, &umsg->msg_name) ||
1862 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1863 __get_user(uiov, &umsg->msg_iov) ||
1864 __get_user(nr_segs, &umsg->msg_iovlen) ||
1865 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1866 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1867 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1871 kmsg->msg_namelen = 0;
1873 if (kmsg->msg_namelen < 0)
1876 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1877 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1882 if (uaddr && kmsg->msg_namelen) {
1884 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1890 kmsg->msg_name = NULL;
1891 kmsg->msg_namelen = 0;
1894 if (nr_segs > UIO_MAXIOV)
1897 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
1899 UIO_FASTIOV, *iov, iov);
1901 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1902 *iov, nr_segs, err);
1906 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1907 struct msghdr *msg_sys, unsigned int flags,
1908 struct used_address *used_address)
1910 struct compat_msghdr __user *msg_compat =
1911 (struct compat_msghdr __user *)msg;
1912 struct sockaddr_storage address;
1913 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1914 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1915 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1916 /* 20 is size of ipv6_pktinfo */
1917 unsigned char *ctl_buf = ctl;
1918 int ctl_len, total_len;
1921 msg_sys->msg_name = &address;
1923 if (MSG_CMSG_COMPAT & flags)
1924 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1926 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1933 if (msg_sys->msg_controllen > INT_MAX)
1935 ctl_len = msg_sys->msg_controllen;
1936 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1938 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1942 ctl_buf = msg_sys->msg_control;
1943 ctl_len = msg_sys->msg_controllen;
1944 } else if (ctl_len) {
1945 if (ctl_len > sizeof(ctl)) {
1946 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1947 if (ctl_buf == NULL)
1952 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1953 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1954 * checking falls down on this.
1956 if (copy_from_user(ctl_buf,
1957 (void __user __force *)msg_sys->msg_control,
1960 msg_sys->msg_control = ctl_buf;
1962 msg_sys->msg_flags = flags;
1964 if (sock->file->f_flags & O_NONBLOCK)
1965 msg_sys->msg_flags |= MSG_DONTWAIT;
1967 * If this is sendmmsg() and current destination address is same as
1968 * previously succeeded address, omit asking LSM's decision.
1969 * used_address->name_len is initialized to UINT_MAX so that the first
1970 * destination address never matches.
1972 if (used_address && msg_sys->msg_name &&
1973 used_address->name_len == msg_sys->msg_namelen &&
1974 !memcmp(&used_address->name, msg_sys->msg_name,
1975 used_address->name_len)) {
1976 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1979 err = sock_sendmsg(sock, msg_sys, total_len);
1981 * If this is sendmmsg() and sending to current destination address was
1982 * successful, remember it.
1984 if (used_address && err >= 0) {
1985 used_address->name_len = msg_sys->msg_namelen;
1986 if (msg_sys->msg_name)
1987 memcpy(&used_address->name, msg_sys->msg_name,
1988 used_address->name_len);
1993 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1995 if (iov != iovstack)
2001 * BSD sendmsg interface
2004 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2006 int fput_needed, err;
2007 struct msghdr msg_sys;
2008 struct socket *sock;
2010 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2014 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2016 fput_light(sock->file, fput_needed);
2021 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2023 if (flags & MSG_CMSG_COMPAT)
2025 return __sys_sendmsg(fd, msg, flags);
2029 * Linux sendmmsg interface
2032 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2035 int fput_needed, err, datagrams;
2036 struct socket *sock;
2037 struct mmsghdr __user *entry;
2038 struct compat_mmsghdr __user *compat_entry;
2039 struct msghdr msg_sys;
2040 struct used_address used_address;
2042 if (vlen > UIO_MAXIOV)
2047 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2051 used_address.name_len = UINT_MAX;
2053 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2056 while (datagrams < vlen) {
2057 if (MSG_CMSG_COMPAT & flags) {
2058 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2059 &msg_sys, flags, &used_address);
2062 err = __put_user(err, &compat_entry->msg_len);
2065 err = ___sys_sendmsg(sock,
2066 (struct user_msghdr __user *)entry,
2067 &msg_sys, flags, &used_address);
2070 err = put_user(err, &entry->msg_len);
2079 fput_light(sock->file, fput_needed);
2081 /* We only return an error if no datagrams were able to be sent */
2088 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2089 unsigned int, vlen, unsigned int, flags)
2091 if (flags & MSG_CMSG_COMPAT)
2093 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2096 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2097 struct msghdr *msg_sys, unsigned int flags, int nosec)
2099 struct compat_msghdr __user *msg_compat =
2100 (struct compat_msghdr __user *)msg;
2101 struct iovec iovstack[UIO_FASTIOV];
2102 struct iovec *iov = iovstack;
2103 unsigned long cmsg_ptr;
2107 /* kernel mode address */
2108 struct sockaddr_storage addr;
2110 /* user mode address pointers */
2111 struct sockaddr __user *uaddr;
2112 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2114 msg_sys->msg_name = &addr;
2116 if (MSG_CMSG_COMPAT & flags)
2117 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2119 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2124 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2125 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2127 /* We assume all kernel code knows the size of sockaddr_storage */
2128 msg_sys->msg_namelen = 0;
2130 if (sock->file->f_flags & O_NONBLOCK)
2131 flags |= MSG_DONTWAIT;
2132 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2138 if (uaddr != NULL) {
2139 err = move_addr_to_user(&addr,
2140 msg_sys->msg_namelen, uaddr,
2145 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2149 if (MSG_CMSG_COMPAT & flags)
2150 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2151 &msg_compat->msg_controllen);
2153 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2154 &msg->msg_controllen);
2160 if (iov != iovstack)
2166 * BSD recvmsg interface
2169 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2171 int fput_needed, err;
2172 struct msghdr msg_sys;
2173 struct socket *sock;
2175 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2179 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2181 fput_light(sock->file, fput_needed);
2186 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2187 unsigned int, flags)
2189 if (flags & MSG_CMSG_COMPAT)
2191 return __sys_recvmsg(fd, msg, flags);
2195 * Linux recvmmsg interface
2198 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2199 unsigned int flags, struct timespec *timeout)
2201 int fput_needed, err, datagrams;
2202 struct socket *sock;
2203 struct mmsghdr __user *entry;
2204 struct compat_mmsghdr __user *compat_entry;
2205 struct msghdr msg_sys;
2206 struct timespec end_time;
2209 poll_select_set_timeout(&end_time, timeout->tv_sec,
2215 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2219 err = sock_error(sock->sk);
2224 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2226 while (datagrams < vlen) {
2228 * No need to ask LSM for more than the first datagram.
2230 if (MSG_CMSG_COMPAT & flags) {
2231 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2232 &msg_sys, flags & ~MSG_WAITFORONE,
2236 err = __put_user(err, &compat_entry->msg_len);
2239 err = ___sys_recvmsg(sock,
2240 (struct user_msghdr __user *)entry,
2241 &msg_sys, flags & ~MSG_WAITFORONE,
2245 err = put_user(err, &entry->msg_len);
2253 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2254 if (flags & MSG_WAITFORONE)
2255 flags |= MSG_DONTWAIT;
2258 ktime_get_ts(timeout);
2259 *timeout = timespec_sub(end_time, *timeout);
2260 if (timeout->tv_sec < 0) {
2261 timeout->tv_sec = timeout->tv_nsec = 0;
2265 /* Timeout, return less than vlen datagrams */
2266 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2270 /* Out of band data, return right away */
2271 if (msg_sys.msg_flags & MSG_OOB)
2276 fput_light(sock->file, fput_needed);
2281 if (datagrams != 0) {
2283 * We may return less entries than requested (vlen) if the
2284 * sock is non block and there aren't enough datagrams...
2286 if (err != -EAGAIN) {
2288 * ... or if recvmsg returns an error after we
2289 * received some datagrams, where we record the
2290 * error to return on the next call or if the
2291 * app asks about it using getsockopt(SO_ERROR).
2293 sock->sk->sk_err = -err;
2302 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2303 unsigned int, vlen, unsigned int, flags,
2304 struct timespec __user *, timeout)
2307 struct timespec timeout_sys;
2309 if (flags & MSG_CMSG_COMPAT)
2313 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2315 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2318 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2320 if (datagrams > 0 &&
2321 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2322 datagrams = -EFAULT;
2327 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2328 /* Argument list sizes for sys_socketcall */
2329 #define AL(x) ((x) * sizeof(unsigned long))
2330 static const unsigned char nargs[21] = {
2331 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2332 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2333 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2340 * System call vectors.
2342 * Argument checking cleaned up. Saved 20% in size.
2343 * This function doesn't need to set the kernel lock because
2344 * it is set by the callees.
2347 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2349 unsigned long a[AUDITSC_ARGS];
2350 unsigned long a0, a1;
2354 if (call < 1 || call > SYS_SENDMMSG)
2358 if (len > sizeof(a))
2361 /* copy_from_user should be SMP safe. */
2362 if (copy_from_user(a, args, len))
2365 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2374 err = sys_socket(a0, a1, a[2]);
2377 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2380 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2383 err = sys_listen(a0, a1);
2386 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2387 (int __user *)a[2], 0);
2389 case SYS_GETSOCKNAME:
2391 sys_getsockname(a0, (struct sockaddr __user *)a1,
2392 (int __user *)a[2]);
2394 case SYS_GETPEERNAME:
2396 sys_getpeername(a0, (struct sockaddr __user *)a1,
2397 (int __user *)a[2]);
2399 case SYS_SOCKETPAIR:
2400 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2403 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2406 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2407 (struct sockaddr __user *)a[4], a[5]);
2410 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2413 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2414 (struct sockaddr __user *)a[4],
2415 (int __user *)a[5]);
2418 err = sys_shutdown(a0, a1);
2420 case SYS_SETSOCKOPT:
2421 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2423 case SYS_GETSOCKOPT:
2425 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2426 (int __user *)a[4]);
2429 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2432 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2435 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2438 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2439 (struct timespec __user *)a[4]);
2442 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2443 (int __user *)a[2], a[3]);
2452 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2455 * sock_register - add a socket protocol handler
2456 * @ops: description of protocol
2458 * This function is called by a protocol handler that wants to
2459 * advertise its address family, and have it linked into the
2460 * socket interface. The value ops->family corresponds to the
2461 * socket system call protocol family.
2463 int sock_register(const struct net_proto_family *ops)
2467 if (ops->family >= NPROTO) {
2468 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2472 spin_lock(&net_family_lock);
2473 if (rcu_dereference_protected(net_families[ops->family],
2474 lockdep_is_held(&net_family_lock)))
2477 rcu_assign_pointer(net_families[ops->family], ops);
2480 spin_unlock(&net_family_lock);
2482 pr_info("NET: Registered protocol family %d\n", ops->family);
2485 EXPORT_SYMBOL(sock_register);
2488 * sock_unregister - remove a protocol handler
2489 * @family: protocol family to remove
2491 * This function is called by a protocol handler that wants to
2492 * remove its address family, and have it unlinked from the
2493 * new socket creation.
2495 * If protocol handler is a module, then it can use module reference
2496 * counts to protect against new references. If protocol handler is not
2497 * a module then it needs to provide its own protection in
2498 * the ops->create routine.
2500 void sock_unregister(int family)
2502 BUG_ON(family < 0 || family >= NPROTO);
2504 spin_lock(&net_family_lock);
2505 RCU_INIT_POINTER(net_families[family], NULL);
2506 spin_unlock(&net_family_lock);
2510 pr_info("NET: Unregistered protocol family %d\n", family);
2512 EXPORT_SYMBOL(sock_unregister);
2514 static int __init sock_init(void)
2518 * Initialize the network sysctl infrastructure.
2520 err = net_sysctl_init();
2525 * Initialize skbuff SLAB cache
2530 * Initialize the protocols module.
2535 err = register_filesystem(&sock_fs_type);
2538 sock_mnt = kern_mount(&sock_fs_type);
2539 if (IS_ERR(sock_mnt)) {
2540 err = PTR_ERR(sock_mnt);
2544 /* The real protocol initialization is performed in later initcalls.
2547 #ifdef CONFIG_NETFILTER
2548 err = netfilter_init();
2553 ptp_classifier_init();
2559 unregister_filesystem(&sock_fs_type);
2564 core_initcall(sock_init); /* early initcall */
2566 #ifdef CONFIG_PROC_FS
2567 void socket_seq_show(struct seq_file *seq)
2572 for_each_possible_cpu(cpu)
2573 counter += per_cpu(sockets_in_use, cpu);
2575 /* It can be negative, by the way. 8) */
2579 seq_printf(seq, "sockets: used %d\n", counter);
2581 #endif /* CONFIG_PROC_FS */
2583 #ifdef CONFIG_COMPAT
2584 static int do_siocgstamp(struct net *net, struct socket *sock,
2585 unsigned int cmd, void __user *up)
2587 mm_segment_t old_fs = get_fs();
2592 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2595 err = compat_put_timeval(&ktv, up);
2600 static int do_siocgstampns(struct net *net, struct socket *sock,
2601 unsigned int cmd, void __user *up)
2603 mm_segment_t old_fs = get_fs();
2604 struct timespec kts;
2608 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2611 err = compat_put_timespec(&kts, up);
2616 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2618 struct ifreq __user *uifr;
2621 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2622 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2625 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2629 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2635 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2637 struct compat_ifconf ifc32;
2639 struct ifconf __user *uifc;
2640 struct compat_ifreq __user *ifr32;
2641 struct ifreq __user *ifr;
2645 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2648 memset(&ifc, 0, sizeof(ifc));
2649 if (ifc32.ifcbuf == 0) {
2653 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2655 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2656 sizeof(struct ifreq);
2657 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2659 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2660 ifr32 = compat_ptr(ifc32.ifcbuf);
2661 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2662 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2668 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2671 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2675 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2679 ifr32 = compat_ptr(ifc32.ifcbuf);
2681 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2682 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2683 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2689 if (ifc32.ifcbuf == 0) {
2690 /* Translate from 64-bit structure multiple to
2694 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2699 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2705 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2707 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2708 bool convert_in = false, convert_out = false;
2709 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2710 struct ethtool_rxnfc __user *rxnfc;
2711 struct ifreq __user *ifr;
2712 u32 rule_cnt = 0, actual_rule_cnt;
2717 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2720 compat_rxnfc = compat_ptr(data);
2722 if (get_user(ethcmd, &compat_rxnfc->cmd))
2725 /* Most ethtool structures are defined without padding.
2726 * Unfortunately struct ethtool_rxnfc is an exception.
2731 case ETHTOOL_GRXCLSRLALL:
2732 /* Buffer size is variable */
2733 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2735 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2737 buf_size += rule_cnt * sizeof(u32);
2739 case ETHTOOL_GRXRINGS:
2740 case ETHTOOL_GRXCLSRLCNT:
2741 case ETHTOOL_GRXCLSRULE:
2742 case ETHTOOL_SRXCLSRLINS:
2745 case ETHTOOL_SRXCLSRLDEL:
2746 buf_size += sizeof(struct ethtool_rxnfc);
2751 ifr = compat_alloc_user_space(buf_size);
2752 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2754 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2757 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2758 &ifr->ifr_ifru.ifru_data))
2762 /* We expect there to be holes between fs.m_ext and
2763 * fs.ring_cookie and at the end of fs, but nowhere else.
2765 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2766 sizeof(compat_rxnfc->fs.m_ext) !=
2767 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2768 sizeof(rxnfc->fs.m_ext));
2770 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2771 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2772 offsetof(struct ethtool_rxnfc, fs.location) -
2773 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2775 if (copy_in_user(rxnfc, compat_rxnfc,
2776 (void __user *)(&rxnfc->fs.m_ext + 1) -
2777 (void __user *)rxnfc) ||
2778 copy_in_user(&rxnfc->fs.ring_cookie,
2779 &compat_rxnfc->fs.ring_cookie,
2780 (void __user *)(&rxnfc->fs.location + 1) -
2781 (void __user *)&rxnfc->fs.ring_cookie) ||
2782 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2783 sizeof(rxnfc->rule_cnt)))
2787 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2792 if (copy_in_user(compat_rxnfc, rxnfc,
2793 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2794 (const void __user *)rxnfc) ||
2795 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2796 &rxnfc->fs.ring_cookie,
2797 (const void __user *)(&rxnfc->fs.location + 1) -
2798 (const void __user *)&rxnfc->fs.ring_cookie) ||
2799 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2800 sizeof(rxnfc->rule_cnt)))
2803 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2804 /* As an optimisation, we only copy the actual
2805 * number of rules that the underlying
2806 * function returned. Since Mallory might
2807 * change the rule count in user memory, we
2808 * check that it is less than the rule count
2809 * originally given (as the user buffer size),
2810 * which has been range-checked.
2812 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2814 if (actual_rule_cnt < rule_cnt)
2815 rule_cnt = actual_rule_cnt;
2816 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2817 &rxnfc->rule_locs[0],
2818 rule_cnt * sizeof(u32)))
2826 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2829 compat_uptr_t uptr32;
2830 struct ifreq __user *uifr;
2832 uifr = compat_alloc_user_space(sizeof(*uifr));
2833 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2836 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2839 uptr = compat_ptr(uptr32);
2841 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2844 return dev_ioctl(net, SIOCWANDEV, uifr);
2847 static int bond_ioctl(struct net *net, unsigned int cmd,
2848 struct compat_ifreq __user *ifr32)
2851 mm_segment_t old_fs;
2855 case SIOCBONDENSLAVE:
2856 case SIOCBONDRELEASE:
2857 case SIOCBONDSETHWADDR:
2858 case SIOCBONDCHANGEACTIVE:
2859 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2864 err = dev_ioctl(net, cmd,
2865 (struct ifreq __user __force *) &kifr);
2870 return -ENOIOCTLCMD;
2874 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2875 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2876 struct compat_ifreq __user *u_ifreq32)
2878 struct ifreq __user *u_ifreq64;
2879 char tmp_buf[IFNAMSIZ];
2880 void __user *data64;
2883 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2886 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2888 data64 = compat_ptr(data32);
2890 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2892 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2895 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2898 return dev_ioctl(net, cmd, u_ifreq64);
2901 static int dev_ifsioc(struct net *net, struct socket *sock,
2902 unsigned int cmd, struct compat_ifreq __user *uifr32)
2904 struct ifreq __user *uifr;
2907 uifr = compat_alloc_user_space(sizeof(*uifr));
2908 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2911 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2922 case SIOCGIFBRDADDR:
2923 case SIOCGIFDSTADDR:
2924 case SIOCGIFNETMASK:
2929 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2937 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2938 struct compat_ifreq __user *uifr32)
2941 struct compat_ifmap __user *uifmap32;
2942 mm_segment_t old_fs;
2945 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2946 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2947 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2948 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2949 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2950 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2951 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2952 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2958 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2961 if (cmd == SIOCGIFMAP && !err) {
2962 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2963 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2964 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2965 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2966 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2967 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2968 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2977 struct sockaddr rt_dst; /* target address */
2978 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2979 struct sockaddr rt_genmask; /* target network mask (IP) */
2980 unsigned short rt_flags;
2983 unsigned char rt_tos;
2984 unsigned char rt_class;
2986 short rt_metric; /* +1 for binary compatibility! */
2987 /* char * */ u32 rt_dev; /* forcing the device at add */
2988 u32 rt_mtu; /* per route MTU/Window */
2989 u32 rt_window; /* Window clamping */
2990 unsigned short rt_irtt; /* Initial RTT */
2993 struct in6_rtmsg32 {
2994 struct in6_addr rtmsg_dst;
2995 struct in6_addr rtmsg_src;
2996 struct in6_addr rtmsg_gateway;
3006 static int routing_ioctl(struct net *net, struct socket *sock,
3007 unsigned int cmd, void __user *argp)
3011 struct in6_rtmsg r6;
3015 mm_segment_t old_fs = get_fs();
3017 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3018 struct in6_rtmsg32 __user *ur6 = argp;
3019 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3020 3 * sizeof(struct in6_addr));
3021 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3022 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3023 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3024 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3025 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3026 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3027 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3031 struct rtentry32 __user *ur4 = argp;
3032 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3033 3 * sizeof(struct sockaddr));
3034 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3035 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3036 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3037 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3038 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3039 ret |= get_user(rtdev, &(ur4->rt_dev));
3041 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3042 r4.rt_dev = (char __user __force *)devname;
3056 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3063 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3064 * for some operations; this forces use of the newer bridge-utils that
3065 * use compatible ioctls
3067 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3071 if (get_user(tmp, argp))
3073 if (tmp == BRCTL_GET_VERSION)
3074 return BRCTL_VERSION + 1;
3078 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3079 unsigned int cmd, unsigned long arg)
3081 void __user *argp = compat_ptr(arg);
3082 struct sock *sk = sock->sk;
3083 struct net *net = sock_net(sk);
3085 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3086 return compat_ifr_data_ioctl(net, cmd, argp);
3091 return old_bridge_ioctl(argp);
3093 return dev_ifname32(net, argp);
3095 return dev_ifconf(net, argp);
3097 return ethtool_ioctl(net, argp);
3099 return compat_siocwandev(net, argp);
3102 return compat_sioc_ifmap(net, cmd, argp);
3103 case SIOCBONDENSLAVE:
3104 case SIOCBONDRELEASE:
3105 case SIOCBONDSETHWADDR:
3106 case SIOCBONDCHANGEACTIVE:
3107 return bond_ioctl(net, cmd, argp);
3110 return routing_ioctl(net, sock, cmd, argp);
3112 return do_siocgstamp(net, sock, cmd, argp);
3114 return do_siocgstampns(net, sock, cmd, argp);
3115 case SIOCBONDSLAVEINFOQUERY:
3116 case SIOCBONDINFOQUERY:
3119 return compat_ifr_data_ioctl(net, cmd, argp);
3131 return sock_ioctl(file, cmd, arg);
3148 case SIOCSIFHWBROADCAST:
3150 case SIOCGIFBRDADDR:
3151 case SIOCSIFBRDADDR:
3152 case SIOCGIFDSTADDR:
3153 case SIOCSIFDSTADDR:
3154 case SIOCGIFNETMASK:
3155 case SIOCSIFNETMASK:
3166 return dev_ifsioc(net, sock, cmd, argp);
3172 return sock_do_ioctl(net, sock, cmd, arg);
3175 return -ENOIOCTLCMD;
3178 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3181 struct socket *sock = file->private_data;
3182 int ret = -ENOIOCTLCMD;
3189 if (sock->ops->compat_ioctl)
3190 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3192 if (ret == -ENOIOCTLCMD &&
3193 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3194 ret = compat_wext_handle_ioctl(net, cmd, arg);
3196 if (ret == -ENOIOCTLCMD)
3197 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3203 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3205 return sock->ops->bind(sock, addr, addrlen);
3207 EXPORT_SYMBOL(kernel_bind);
3209 int kernel_listen(struct socket *sock, int backlog)
3211 return sock->ops->listen(sock, backlog);
3213 EXPORT_SYMBOL(kernel_listen);
3215 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3217 struct sock *sk = sock->sk;
3220 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3225 err = sock->ops->accept(sock, *newsock, flags);
3227 sock_release(*newsock);
3232 (*newsock)->ops = sock->ops;
3233 __module_get((*newsock)->ops->owner);
3238 EXPORT_SYMBOL(kernel_accept);
3240 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3243 return sock->ops->connect(sock, addr, addrlen, flags);
3245 EXPORT_SYMBOL(kernel_connect);
3247 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3250 return sock->ops->getname(sock, addr, addrlen, 0);
3252 EXPORT_SYMBOL(kernel_getsockname);
3254 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3257 return sock->ops->getname(sock, addr, addrlen, 1);
3259 EXPORT_SYMBOL(kernel_getpeername);
3261 int kernel_getsockopt(struct socket *sock, int level, int optname,
3262 char *optval, int *optlen)
3264 mm_segment_t oldfs = get_fs();
3265 char __user *uoptval;
3266 int __user *uoptlen;
3269 uoptval = (char __user __force *) optval;
3270 uoptlen = (int __user __force *) optlen;
3273 if (level == SOL_SOCKET)
3274 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3276 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3281 EXPORT_SYMBOL(kernel_getsockopt);
3283 int kernel_setsockopt(struct socket *sock, int level, int optname,
3284 char *optval, unsigned int optlen)
3286 mm_segment_t oldfs = get_fs();
3287 char __user *uoptval;
3290 uoptval = (char __user __force *) optval;
3293 if (level == SOL_SOCKET)
3294 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3296 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3301 EXPORT_SYMBOL(kernel_setsockopt);
3303 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3304 size_t size, int flags)
3306 if (sock->ops->sendpage)
3307 return sock->ops->sendpage(sock, page, offset, size, flags);
3309 return sock_no_sendpage(sock, page, offset, size, flags);
3311 EXPORT_SYMBOL(kernel_sendpage);
3313 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3315 mm_segment_t oldfs = get_fs();
3319 err = sock->ops->ioctl(sock, cmd, arg);
3324 EXPORT_SYMBOL(kernel_sock_ioctl);
3326 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3328 return sock->ops->shutdown(sock, how);
3330 EXPORT_SYMBOL(kernel_sock_shutdown);