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 kiocb *iocb, struct socket *sock,
614 struct msghdr *msg, size_t size)
616 return sock->ops->sendmsg(iocb, sock, msg, size);
619 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
620 struct msghdr *msg, size_t size)
622 int err = security_socket_sendmsg(sock, msg, size);
624 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
627 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
628 size_t size, bool nosec)
633 init_sync_kiocb(&iocb, NULL);
634 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
635 __sock_sendmsg(&iocb, sock, msg, size);
636 BUG_ON(ret == -EIOCBQUEUED);
640 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
642 return do_sock_sendmsg(sock, msg, size, false);
644 EXPORT_SYMBOL(sock_sendmsg);
646 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
648 return do_sock_sendmsg(sock, msg, size, true);
651 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
652 struct kvec *vec, size_t num, size_t size)
654 mm_segment_t oldfs = get_fs();
659 * the following is safe, since for compiler definitions of kvec and
660 * iovec are identical, yielding the same in-core layout and alignment
662 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
663 result = sock_sendmsg(sock, msg, size);
667 EXPORT_SYMBOL(kernel_sendmsg);
670 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
672 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
675 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
676 struct scm_timestamping tss;
678 struct skb_shared_hwtstamps *shhwtstamps =
681 /* Race occurred between timestamp enabling and packet
682 receiving. Fill in the current time for now. */
683 if (need_software_tstamp && skb->tstamp.tv64 == 0)
684 __net_timestamp(skb);
686 if (need_software_tstamp) {
687 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
689 skb_get_timestamp(skb, &tv);
690 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
694 skb_get_timestampns(skb, &ts);
695 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
700 memset(&tss, 0, sizeof(tss));
701 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
702 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
705 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
706 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
709 put_cmsg(msg, SOL_SOCKET,
710 SCM_TIMESTAMPING, sizeof(tss), &tss);
712 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
714 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
719 if (!sock_flag(sk, SOCK_WIFI_STATUS))
721 if (!skb->wifi_acked_valid)
724 ack = skb->wifi_acked;
726 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
728 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
730 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
733 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
734 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
735 sizeof(__u32), &skb->dropcount);
738 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
741 sock_recv_timestamp(msg, sk, skb);
742 sock_recv_drops(msg, sk, skb);
744 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
746 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
747 struct msghdr *msg, size_t size, int flags)
749 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
752 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
753 struct msghdr *msg, size_t size, int flags)
755 int err = security_socket_recvmsg(sock, msg, size, flags);
757 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
760 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
761 size_t size, int flags)
766 init_sync_kiocb(&iocb, NULL);
767 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
768 BUG_ON(ret == -EIOCBQUEUED);
771 EXPORT_SYMBOL(sock_recvmsg);
773 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
774 size_t size, int flags)
779 init_sync_kiocb(&iocb, NULL);
780 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
781 BUG_ON(ret == -EIOCBQUEUED);
786 * kernel_recvmsg - Receive a message from a socket (kernel space)
787 * @sock: The socket to receive the message from
788 * @msg: Received message
789 * @vec: Input s/g array for message data
790 * @num: Size of input s/g array
791 * @size: Number of bytes to read
792 * @flags: Message flags (MSG_DONTWAIT, etc...)
794 * On return the msg structure contains the scatter/gather array passed in the
795 * vec argument. The array is modified so that it consists of the unfilled
796 * portion of the original array.
798 * The returned value is the total number of bytes received, or an error.
800 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
801 struct kvec *vec, size_t num, size_t size, int flags)
803 mm_segment_t oldfs = get_fs();
808 * the following is safe, since for compiler definitions of kvec and
809 * iovec are identical, yielding the same in-core layout and alignment
811 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
812 result = sock_recvmsg(sock, msg, size, flags);
816 EXPORT_SYMBOL(kernel_recvmsg);
818 static ssize_t sock_sendpage(struct file *file, struct page *page,
819 int offset, size_t size, loff_t *ppos, int more)
824 sock = file->private_data;
826 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
827 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
830 return kernel_sendpage(sock, page, offset, size, flags);
833 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
834 struct pipe_inode_info *pipe, size_t len,
837 struct socket *sock = file->private_data;
839 if (unlikely(!sock->ops->splice_read))
842 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
845 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
847 struct file *file = iocb->ki_filp;
848 struct socket *sock = file->private_data;
849 struct msghdr msg = {.msg_iter = *to};
852 if (file->f_flags & O_NONBLOCK)
853 msg.msg_flags = MSG_DONTWAIT;
855 if (iocb->ki_pos != 0)
858 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
861 res = __sock_recvmsg(iocb, sock, &msg,
862 iov_iter_count(to), msg.msg_flags);
867 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
869 struct file *file = iocb->ki_filp;
870 struct socket *sock = file->private_data;
871 struct msghdr msg = {.msg_iter = *from};
874 if (iocb->ki_pos != 0)
877 if (file->f_flags & O_NONBLOCK)
878 msg.msg_flags = MSG_DONTWAIT;
880 if (sock->type == SOCK_SEQPACKET)
881 msg.msg_flags |= MSG_EOR;
883 res = __sock_sendmsg(iocb, sock, &msg, iov_iter_count(from));
884 *from = msg.msg_iter;
889 * Atomic setting of ioctl hooks to avoid race
890 * with module unload.
893 static DEFINE_MUTEX(br_ioctl_mutex);
894 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
896 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
898 mutex_lock(&br_ioctl_mutex);
899 br_ioctl_hook = hook;
900 mutex_unlock(&br_ioctl_mutex);
902 EXPORT_SYMBOL(brioctl_set);
904 static DEFINE_MUTEX(vlan_ioctl_mutex);
905 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
907 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
909 mutex_lock(&vlan_ioctl_mutex);
910 vlan_ioctl_hook = hook;
911 mutex_unlock(&vlan_ioctl_mutex);
913 EXPORT_SYMBOL(vlan_ioctl_set);
915 static DEFINE_MUTEX(dlci_ioctl_mutex);
916 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
918 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
920 mutex_lock(&dlci_ioctl_mutex);
921 dlci_ioctl_hook = hook;
922 mutex_unlock(&dlci_ioctl_mutex);
924 EXPORT_SYMBOL(dlci_ioctl_set);
926 static long sock_do_ioctl(struct net *net, struct socket *sock,
927 unsigned int cmd, unsigned long arg)
930 void __user *argp = (void __user *)arg;
932 err = sock->ops->ioctl(sock, cmd, arg);
935 * If this ioctl is unknown try to hand it down
938 if (err == -ENOIOCTLCMD)
939 err = dev_ioctl(net, cmd, argp);
945 * With an ioctl, arg may well be a user mode pointer, but we don't know
946 * what to do with it - that's up to the protocol still.
949 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
953 void __user *argp = (void __user *)arg;
957 sock = file->private_data;
960 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
961 err = dev_ioctl(net, cmd, argp);
963 #ifdef CONFIG_WEXT_CORE
964 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
965 err = dev_ioctl(net, cmd, argp);
972 if (get_user(pid, (int __user *)argp))
974 f_setown(sock->file, pid, 1);
979 err = put_user(f_getown(sock->file),
988 request_module("bridge");
990 mutex_lock(&br_ioctl_mutex);
992 err = br_ioctl_hook(net, cmd, argp);
993 mutex_unlock(&br_ioctl_mutex);
998 if (!vlan_ioctl_hook)
999 request_module("8021q");
1001 mutex_lock(&vlan_ioctl_mutex);
1002 if (vlan_ioctl_hook)
1003 err = vlan_ioctl_hook(net, argp);
1004 mutex_unlock(&vlan_ioctl_mutex);
1009 if (!dlci_ioctl_hook)
1010 request_module("dlci");
1012 mutex_lock(&dlci_ioctl_mutex);
1013 if (dlci_ioctl_hook)
1014 err = dlci_ioctl_hook(cmd, argp);
1015 mutex_unlock(&dlci_ioctl_mutex);
1018 err = sock_do_ioctl(net, sock, cmd, arg);
1024 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1027 struct socket *sock = NULL;
1029 err = security_socket_create(family, type, protocol, 1);
1033 sock = sock_alloc();
1040 err = security_socket_post_create(sock, family, type, protocol, 1);
1052 EXPORT_SYMBOL(sock_create_lite);
1054 /* No kernel lock held - perfect */
1055 static unsigned int sock_poll(struct file *file, poll_table *wait)
1057 unsigned int busy_flag = 0;
1058 struct socket *sock;
1061 * We can't return errors to poll, so it's either yes or no.
1063 sock = file->private_data;
1065 if (sk_can_busy_loop(sock->sk)) {
1066 /* this socket can poll_ll so tell the system call */
1067 busy_flag = POLL_BUSY_LOOP;
1069 /* once, only if requested by syscall */
1070 if (wait && (wait->_key & POLL_BUSY_LOOP))
1071 sk_busy_loop(sock->sk, 1);
1074 return busy_flag | sock->ops->poll(file, sock, wait);
1077 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1079 struct socket *sock = file->private_data;
1081 return sock->ops->mmap(file, sock, vma);
1084 static int sock_close(struct inode *inode, struct file *filp)
1086 sock_release(SOCKET_I(inode));
1091 * Update the socket async list
1093 * Fasync_list locking strategy.
1095 * 1. fasync_list is modified only under process context socket lock
1096 * i.e. under semaphore.
1097 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1098 * or under socket lock
1101 static int sock_fasync(int fd, struct file *filp, int on)
1103 struct socket *sock = filp->private_data;
1104 struct sock *sk = sock->sk;
1105 struct socket_wq *wq;
1111 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1112 fasync_helper(fd, filp, on, &wq->fasync_list);
1114 if (!wq->fasync_list)
1115 sock_reset_flag(sk, SOCK_FASYNC);
1117 sock_set_flag(sk, SOCK_FASYNC);
1123 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1125 int sock_wake_async(struct socket *sock, int how, int band)
1127 struct socket_wq *wq;
1132 wq = rcu_dereference(sock->wq);
1133 if (!wq || !wq->fasync_list) {
1138 case SOCK_WAKE_WAITD:
1139 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1142 case SOCK_WAKE_SPACE:
1143 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1148 kill_fasync(&wq->fasync_list, SIGIO, band);
1151 kill_fasync(&wq->fasync_list, SIGURG, band);
1156 EXPORT_SYMBOL(sock_wake_async);
1158 int __sock_create(struct net *net, int family, int type, int protocol,
1159 struct socket **res, int kern)
1162 struct socket *sock;
1163 const struct net_proto_family *pf;
1166 * Check protocol is in range
1168 if (family < 0 || family >= NPROTO)
1169 return -EAFNOSUPPORT;
1170 if (type < 0 || type >= SOCK_MAX)
1175 This uglymoron is moved from INET layer to here to avoid
1176 deadlock in module load.
1178 if (family == PF_INET && type == SOCK_PACKET) {
1182 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1188 err = security_socket_create(family, type, protocol, kern);
1193 * Allocate the socket and allow the family to set things up. if
1194 * the protocol is 0, the family is instructed to select an appropriate
1197 sock = sock_alloc();
1199 net_warn_ratelimited("socket: no more sockets\n");
1200 return -ENFILE; /* Not exactly a match, but its the
1201 closest posix thing */
1206 #ifdef CONFIG_MODULES
1207 /* Attempt to load a protocol module if the find failed.
1209 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1210 * requested real, full-featured networking support upon configuration.
1211 * Otherwise module support will break!
1213 if (rcu_access_pointer(net_families[family]) == NULL)
1214 request_module("net-pf-%d", family);
1218 pf = rcu_dereference(net_families[family]);
1219 err = -EAFNOSUPPORT;
1224 * We will call the ->create function, that possibly is in a loadable
1225 * module, so we have to bump that loadable module refcnt first.
1227 if (!try_module_get(pf->owner))
1230 /* Now protected by module ref count */
1233 err = pf->create(net, sock, protocol, kern);
1235 goto out_module_put;
1238 * Now to bump the refcnt of the [loadable] module that owns this
1239 * socket at sock_release time we decrement its refcnt.
1241 if (!try_module_get(sock->ops->owner))
1242 goto out_module_busy;
1245 * Now that we're done with the ->create function, the [loadable]
1246 * module can have its refcnt decremented
1248 module_put(pf->owner);
1249 err = security_socket_post_create(sock, family, type, protocol, kern);
1251 goto out_sock_release;
1257 err = -EAFNOSUPPORT;
1260 module_put(pf->owner);
1267 goto out_sock_release;
1269 EXPORT_SYMBOL(__sock_create);
1271 int sock_create(int family, int type, int protocol, struct socket **res)
1273 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1275 EXPORT_SYMBOL(sock_create);
1277 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1279 return __sock_create(&init_net, family, type, protocol, res, 1);
1281 EXPORT_SYMBOL(sock_create_kern);
1283 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1286 struct socket *sock;
1289 /* Check the SOCK_* constants for consistency. */
1290 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1291 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1292 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1293 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1295 flags = type & ~SOCK_TYPE_MASK;
1296 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1298 type &= SOCK_TYPE_MASK;
1300 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1301 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1303 retval = sock_create(family, type, protocol, &sock);
1307 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1312 /* It may be already another descriptor 8) Not kernel problem. */
1321 * Create a pair of connected sockets.
1324 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1325 int __user *, usockvec)
1327 struct socket *sock1, *sock2;
1329 struct file *newfile1, *newfile2;
1332 flags = type & ~SOCK_TYPE_MASK;
1333 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1335 type &= SOCK_TYPE_MASK;
1337 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1338 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1341 * Obtain the first socket and check if the underlying protocol
1342 * supports the socketpair call.
1345 err = sock_create(family, type, protocol, &sock1);
1349 err = sock_create(family, type, protocol, &sock2);
1353 err = sock1->ops->socketpair(sock1, sock2);
1355 goto out_release_both;
1357 fd1 = get_unused_fd_flags(flags);
1358 if (unlikely(fd1 < 0)) {
1360 goto out_release_both;
1363 fd2 = get_unused_fd_flags(flags);
1364 if (unlikely(fd2 < 0)) {
1366 goto out_put_unused_1;
1369 newfile1 = sock_alloc_file(sock1, flags, NULL);
1370 if (unlikely(IS_ERR(newfile1))) {
1371 err = PTR_ERR(newfile1);
1372 goto out_put_unused_both;
1375 newfile2 = sock_alloc_file(sock2, flags, NULL);
1376 if (IS_ERR(newfile2)) {
1377 err = PTR_ERR(newfile2);
1381 err = put_user(fd1, &usockvec[0]);
1385 err = put_user(fd2, &usockvec[1]);
1389 audit_fd_pair(fd1, fd2);
1391 fd_install(fd1, newfile1);
1392 fd_install(fd2, newfile2);
1393 /* fd1 and fd2 may be already another descriptors.
1394 * Not kernel problem.
1410 sock_release(sock2);
1413 out_put_unused_both:
1418 sock_release(sock2);
1420 sock_release(sock1);
1426 * Bind a name to a socket. Nothing much to do here since it's
1427 * the protocol's responsibility to handle the local address.
1429 * We move the socket address to kernel space before we call
1430 * the protocol layer (having also checked the address is ok).
1433 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1435 struct socket *sock;
1436 struct sockaddr_storage address;
1437 int err, fput_needed;
1439 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1441 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1443 err = security_socket_bind(sock,
1444 (struct sockaddr *)&address,
1447 err = sock->ops->bind(sock,
1451 fput_light(sock->file, fput_needed);
1457 * Perform a listen. Basically, we allow the protocol to do anything
1458 * necessary for a listen, and if that works, we mark the socket as
1459 * ready for listening.
1462 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1464 struct socket *sock;
1465 int err, fput_needed;
1468 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1470 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1471 if ((unsigned int)backlog > somaxconn)
1472 backlog = somaxconn;
1474 err = security_socket_listen(sock, backlog);
1476 err = sock->ops->listen(sock, backlog);
1478 fput_light(sock->file, fput_needed);
1484 * For accept, we attempt to create a new socket, set up the link
1485 * with the client, wake up the client, then return the new
1486 * connected fd. We collect the address of the connector in kernel
1487 * space and move it to user at the very end. This is unclean because
1488 * we open the socket then return an error.
1490 * 1003.1g adds the ability to recvmsg() to query connection pending
1491 * status to recvmsg. We need to add that support in a way thats
1492 * clean when we restucture accept also.
1495 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1496 int __user *, upeer_addrlen, int, flags)
1498 struct socket *sock, *newsock;
1499 struct file *newfile;
1500 int err, len, newfd, fput_needed;
1501 struct sockaddr_storage address;
1503 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1506 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1507 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1509 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1514 newsock = sock_alloc();
1518 newsock->type = sock->type;
1519 newsock->ops = sock->ops;
1522 * We don't need try_module_get here, as the listening socket (sock)
1523 * has the protocol module (sock->ops->owner) held.
1525 __module_get(newsock->ops->owner);
1527 newfd = get_unused_fd_flags(flags);
1528 if (unlikely(newfd < 0)) {
1530 sock_release(newsock);
1533 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1534 if (unlikely(IS_ERR(newfile))) {
1535 err = PTR_ERR(newfile);
1536 put_unused_fd(newfd);
1537 sock_release(newsock);
1541 err = security_socket_accept(sock, newsock);
1545 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1549 if (upeer_sockaddr) {
1550 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1552 err = -ECONNABORTED;
1555 err = move_addr_to_user(&address,
1556 len, upeer_sockaddr, upeer_addrlen);
1561 /* File flags are not inherited via accept() unlike another OSes. */
1563 fd_install(newfd, newfile);
1567 fput_light(sock->file, fput_needed);
1572 put_unused_fd(newfd);
1576 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1577 int __user *, upeer_addrlen)
1579 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1583 * Attempt to connect to a socket with the server address. The address
1584 * is in user space so we verify it is OK and move it to kernel space.
1586 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1589 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1590 * other SEQPACKET protocols that take time to connect() as it doesn't
1591 * include the -EINPROGRESS status for such sockets.
1594 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1597 struct socket *sock;
1598 struct sockaddr_storage address;
1599 int err, fput_needed;
1601 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1604 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1609 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1613 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1614 sock->file->f_flags);
1616 fput_light(sock->file, fput_needed);
1622 * Get the local address ('name') of a socket object. Move the obtained
1623 * name to user space.
1626 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1627 int __user *, usockaddr_len)
1629 struct socket *sock;
1630 struct sockaddr_storage address;
1631 int len, err, fput_needed;
1633 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1637 err = security_socket_getsockname(sock);
1641 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1644 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1647 fput_light(sock->file, fput_needed);
1653 * Get the remote address ('name') of a socket object. Move the obtained
1654 * name to user space.
1657 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1658 int __user *, usockaddr_len)
1660 struct socket *sock;
1661 struct sockaddr_storage address;
1662 int len, err, fput_needed;
1664 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1666 err = security_socket_getpeername(sock);
1668 fput_light(sock->file, fput_needed);
1673 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1676 err = move_addr_to_user(&address, len, usockaddr,
1678 fput_light(sock->file, fput_needed);
1684 * Send a datagram to a given address. We move the address into kernel
1685 * space and check the user space data area is readable before invoking
1689 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1690 unsigned int, flags, struct sockaddr __user *, addr,
1693 struct socket *sock;
1694 struct sockaddr_storage address;
1702 if (unlikely(!access_ok(VERIFY_READ, buff, len)))
1704 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1708 iov.iov_base = buff;
1710 msg.msg_name = NULL;
1711 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1712 msg.msg_control = NULL;
1713 msg.msg_controllen = 0;
1714 msg.msg_namelen = 0;
1716 err = move_addr_to_kernel(addr, addr_len, &address);
1719 msg.msg_name = (struct sockaddr *)&address;
1720 msg.msg_namelen = addr_len;
1722 if (sock->file->f_flags & O_NONBLOCK)
1723 flags |= MSG_DONTWAIT;
1724 msg.msg_flags = flags;
1725 err = sock_sendmsg(sock, &msg, len);
1728 fput_light(sock->file, fput_needed);
1734 * Send a datagram down a socket.
1737 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1738 unsigned int, flags)
1740 return sys_sendto(fd, buff, len, flags, NULL, 0);
1744 * Receive a frame from the socket and optionally record the address of the
1745 * sender. We verify the buffers are writable and if needed move the
1746 * sender address from kernel to user space.
1749 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1750 unsigned int, flags, struct sockaddr __user *, addr,
1751 int __user *, addr_len)
1753 struct socket *sock;
1756 struct sockaddr_storage address;
1762 if (unlikely(!access_ok(VERIFY_WRITE, ubuf, size)))
1764 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1768 msg.msg_control = NULL;
1769 msg.msg_controllen = 0;
1771 iov.iov_base = ubuf;
1772 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1773 /* Save some cycles and don't copy the address if not needed */
1774 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1775 /* We assume all kernel code knows the size of sockaddr_storage */
1776 msg.msg_namelen = 0;
1777 if (sock->file->f_flags & O_NONBLOCK)
1778 flags |= MSG_DONTWAIT;
1779 err = sock_recvmsg(sock, &msg, size, flags);
1781 if (err >= 0 && addr != NULL) {
1782 err2 = move_addr_to_user(&address,
1783 msg.msg_namelen, addr, addr_len);
1788 fput_light(sock->file, fput_needed);
1794 * Receive a datagram from a socket.
1797 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1798 unsigned int, flags)
1800 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1804 * Set a socket option. Because we don't know the option lengths we have
1805 * to pass the user mode parameter for the protocols to sort out.
1808 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1809 char __user *, optval, int, optlen)
1811 int err, fput_needed;
1812 struct socket *sock;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1819 err = security_socket_setsockopt(sock, level, optname);
1823 if (level == SOL_SOCKET)
1825 sock_setsockopt(sock, level, optname, optval,
1829 sock->ops->setsockopt(sock, level, optname, optval,
1832 fput_light(sock->file, fput_needed);
1838 * Get a socket option. Because we don't know the option lengths we have
1839 * to pass a user mode parameter for the protocols to sort out.
1842 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1843 char __user *, optval, int __user *, optlen)
1845 int err, fput_needed;
1846 struct socket *sock;
1848 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1850 err = security_socket_getsockopt(sock, level, optname);
1854 if (level == SOL_SOCKET)
1856 sock_getsockopt(sock, level, optname, optval,
1860 sock->ops->getsockopt(sock, level, optname, optval,
1863 fput_light(sock->file, fput_needed);
1869 * Shutdown a socket.
1872 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1874 int err, fput_needed;
1875 struct socket *sock;
1877 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1879 err = security_socket_shutdown(sock, how);
1881 err = sock->ops->shutdown(sock, how);
1882 fput_light(sock->file, fput_needed);
1887 /* A couple of helpful macros for getting the address of the 32/64 bit
1888 * fields which are the same type (int / unsigned) on our platforms.
1890 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1891 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1892 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1894 struct used_address {
1895 struct sockaddr_storage name;
1896 unsigned int name_len;
1899 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1900 struct user_msghdr __user *umsg,
1901 struct sockaddr __user **save_addr,
1904 struct sockaddr __user *uaddr;
1905 struct iovec __user *uiov;
1909 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1910 __get_user(uaddr, &umsg->msg_name) ||
1911 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1912 __get_user(uiov, &umsg->msg_iov) ||
1913 __get_user(nr_segs, &umsg->msg_iovlen) ||
1914 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1915 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1916 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1920 kmsg->msg_namelen = 0;
1922 if (kmsg->msg_namelen < 0)
1925 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1926 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1931 if (uaddr && kmsg->msg_namelen) {
1933 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1939 kmsg->msg_name = NULL;
1940 kmsg->msg_namelen = 0;
1943 if (nr_segs > UIO_MAXIOV)
1946 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
1948 UIO_FASTIOV, *iov, iov);
1950 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1951 *iov, nr_segs, err);
1955 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1956 struct msghdr *msg_sys, unsigned int flags,
1957 struct used_address *used_address)
1959 struct compat_msghdr __user *msg_compat =
1960 (struct compat_msghdr __user *)msg;
1961 struct sockaddr_storage address;
1962 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1963 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1964 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1965 /* 20 is size of ipv6_pktinfo */
1966 unsigned char *ctl_buf = ctl;
1967 int ctl_len, total_len;
1970 msg_sys->msg_name = &address;
1972 if (MSG_CMSG_COMPAT & flags)
1973 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1975 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1982 if (msg_sys->msg_controllen > INT_MAX)
1984 ctl_len = msg_sys->msg_controllen;
1985 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1987 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1991 ctl_buf = msg_sys->msg_control;
1992 ctl_len = msg_sys->msg_controllen;
1993 } else if (ctl_len) {
1994 if (ctl_len > sizeof(ctl)) {
1995 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1996 if (ctl_buf == NULL)
2001 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2002 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2003 * checking falls down on this.
2005 if (copy_from_user(ctl_buf,
2006 (void __user __force *)msg_sys->msg_control,
2009 msg_sys->msg_control = ctl_buf;
2011 msg_sys->msg_flags = flags;
2013 if (sock->file->f_flags & O_NONBLOCK)
2014 msg_sys->msg_flags |= MSG_DONTWAIT;
2016 * If this is sendmmsg() and current destination address is same as
2017 * previously succeeded address, omit asking LSM's decision.
2018 * used_address->name_len is initialized to UINT_MAX so that the first
2019 * destination address never matches.
2021 if (used_address && msg_sys->msg_name &&
2022 used_address->name_len == msg_sys->msg_namelen &&
2023 !memcmp(&used_address->name, msg_sys->msg_name,
2024 used_address->name_len)) {
2025 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2028 err = sock_sendmsg(sock, msg_sys, total_len);
2030 * If this is sendmmsg() and sending to current destination address was
2031 * successful, remember it.
2033 if (used_address && err >= 0) {
2034 used_address->name_len = msg_sys->msg_namelen;
2035 if (msg_sys->msg_name)
2036 memcpy(&used_address->name, msg_sys->msg_name,
2037 used_address->name_len);
2042 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2044 if (iov != iovstack)
2050 * BSD sendmsg interface
2053 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2055 int fput_needed, err;
2056 struct msghdr msg_sys;
2057 struct socket *sock;
2059 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2063 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2065 fput_light(sock->file, fput_needed);
2070 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2072 if (flags & MSG_CMSG_COMPAT)
2074 return __sys_sendmsg(fd, msg, flags);
2078 * Linux sendmmsg interface
2081 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2084 int fput_needed, err, datagrams;
2085 struct socket *sock;
2086 struct mmsghdr __user *entry;
2087 struct compat_mmsghdr __user *compat_entry;
2088 struct msghdr msg_sys;
2089 struct used_address used_address;
2091 if (vlen > UIO_MAXIOV)
2096 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2100 used_address.name_len = UINT_MAX;
2102 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2105 while (datagrams < vlen) {
2106 if (MSG_CMSG_COMPAT & flags) {
2107 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2108 &msg_sys, flags, &used_address);
2111 err = __put_user(err, &compat_entry->msg_len);
2114 err = ___sys_sendmsg(sock,
2115 (struct user_msghdr __user *)entry,
2116 &msg_sys, flags, &used_address);
2119 err = put_user(err, &entry->msg_len);
2128 fput_light(sock->file, fput_needed);
2130 /* We only return an error if no datagrams were able to be sent */
2137 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2138 unsigned int, vlen, unsigned int, flags)
2140 if (flags & MSG_CMSG_COMPAT)
2142 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2145 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2146 struct msghdr *msg_sys, unsigned int flags, int nosec)
2148 struct compat_msghdr __user *msg_compat =
2149 (struct compat_msghdr __user *)msg;
2150 struct iovec iovstack[UIO_FASTIOV];
2151 struct iovec *iov = iovstack;
2152 unsigned long cmsg_ptr;
2156 /* kernel mode address */
2157 struct sockaddr_storage addr;
2159 /* user mode address pointers */
2160 struct sockaddr __user *uaddr;
2161 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2163 msg_sys->msg_name = &addr;
2165 if (MSG_CMSG_COMPAT & flags)
2166 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2168 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2173 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2174 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2176 /* We assume all kernel code knows the size of sockaddr_storage */
2177 msg_sys->msg_namelen = 0;
2179 if (sock->file->f_flags & O_NONBLOCK)
2180 flags |= MSG_DONTWAIT;
2181 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2187 if (uaddr != NULL) {
2188 err = move_addr_to_user(&addr,
2189 msg_sys->msg_namelen, uaddr,
2194 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2198 if (MSG_CMSG_COMPAT & flags)
2199 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2200 &msg_compat->msg_controllen);
2202 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2203 &msg->msg_controllen);
2209 if (iov != iovstack)
2215 * BSD recvmsg interface
2218 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2220 int fput_needed, err;
2221 struct msghdr msg_sys;
2222 struct socket *sock;
2224 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2228 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2230 fput_light(sock->file, fput_needed);
2235 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2236 unsigned int, flags)
2238 if (flags & MSG_CMSG_COMPAT)
2240 return __sys_recvmsg(fd, msg, flags);
2244 * Linux recvmmsg interface
2247 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2248 unsigned int flags, struct timespec *timeout)
2250 int fput_needed, err, datagrams;
2251 struct socket *sock;
2252 struct mmsghdr __user *entry;
2253 struct compat_mmsghdr __user *compat_entry;
2254 struct msghdr msg_sys;
2255 struct timespec end_time;
2258 poll_select_set_timeout(&end_time, timeout->tv_sec,
2264 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2268 err = sock_error(sock->sk);
2273 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2275 while (datagrams < vlen) {
2277 * No need to ask LSM for more than the first datagram.
2279 if (MSG_CMSG_COMPAT & flags) {
2280 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2281 &msg_sys, flags & ~MSG_WAITFORONE,
2285 err = __put_user(err, &compat_entry->msg_len);
2288 err = ___sys_recvmsg(sock,
2289 (struct user_msghdr __user *)entry,
2290 &msg_sys, flags & ~MSG_WAITFORONE,
2294 err = put_user(err, &entry->msg_len);
2302 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2303 if (flags & MSG_WAITFORONE)
2304 flags |= MSG_DONTWAIT;
2307 ktime_get_ts(timeout);
2308 *timeout = timespec_sub(end_time, *timeout);
2309 if (timeout->tv_sec < 0) {
2310 timeout->tv_sec = timeout->tv_nsec = 0;
2314 /* Timeout, return less than vlen datagrams */
2315 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2319 /* Out of band data, return right away */
2320 if (msg_sys.msg_flags & MSG_OOB)
2325 fput_light(sock->file, fput_needed);
2330 if (datagrams != 0) {
2332 * We may return less entries than requested (vlen) if the
2333 * sock is non block and there aren't enough datagrams...
2335 if (err != -EAGAIN) {
2337 * ... or if recvmsg returns an error after we
2338 * received some datagrams, where we record the
2339 * error to return on the next call or if the
2340 * app asks about it using getsockopt(SO_ERROR).
2342 sock->sk->sk_err = -err;
2351 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2352 unsigned int, vlen, unsigned int, flags,
2353 struct timespec __user *, timeout)
2356 struct timespec timeout_sys;
2358 if (flags & MSG_CMSG_COMPAT)
2362 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2364 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2367 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2369 if (datagrams > 0 &&
2370 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2371 datagrams = -EFAULT;
2376 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2377 /* Argument list sizes for sys_socketcall */
2378 #define AL(x) ((x) * sizeof(unsigned long))
2379 static const unsigned char nargs[21] = {
2380 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2381 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2382 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2389 * System call vectors.
2391 * Argument checking cleaned up. Saved 20% in size.
2392 * This function doesn't need to set the kernel lock because
2393 * it is set by the callees.
2396 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2398 unsigned long a[AUDITSC_ARGS];
2399 unsigned long a0, a1;
2403 if (call < 1 || call > SYS_SENDMMSG)
2407 if (len > sizeof(a))
2410 /* copy_from_user should be SMP safe. */
2411 if (copy_from_user(a, args, len))
2414 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2423 err = sys_socket(a0, a1, a[2]);
2426 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2429 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2432 err = sys_listen(a0, a1);
2435 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2436 (int __user *)a[2], 0);
2438 case SYS_GETSOCKNAME:
2440 sys_getsockname(a0, (struct sockaddr __user *)a1,
2441 (int __user *)a[2]);
2443 case SYS_GETPEERNAME:
2445 sys_getpeername(a0, (struct sockaddr __user *)a1,
2446 (int __user *)a[2]);
2448 case SYS_SOCKETPAIR:
2449 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2452 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2455 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2456 (struct sockaddr __user *)a[4], a[5]);
2459 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2462 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2463 (struct sockaddr __user *)a[4],
2464 (int __user *)a[5]);
2467 err = sys_shutdown(a0, a1);
2469 case SYS_SETSOCKOPT:
2470 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2472 case SYS_GETSOCKOPT:
2474 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2475 (int __user *)a[4]);
2478 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2481 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2484 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2487 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2488 (struct timespec __user *)a[4]);
2491 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2492 (int __user *)a[2], a[3]);
2501 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2504 * sock_register - add a socket protocol handler
2505 * @ops: description of protocol
2507 * This function is called by a protocol handler that wants to
2508 * advertise its address family, and have it linked into the
2509 * socket interface. The value ops->family corresponds to the
2510 * socket system call protocol family.
2512 int sock_register(const struct net_proto_family *ops)
2516 if (ops->family >= NPROTO) {
2517 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2521 spin_lock(&net_family_lock);
2522 if (rcu_dereference_protected(net_families[ops->family],
2523 lockdep_is_held(&net_family_lock)))
2526 rcu_assign_pointer(net_families[ops->family], ops);
2529 spin_unlock(&net_family_lock);
2531 pr_info("NET: Registered protocol family %d\n", ops->family);
2534 EXPORT_SYMBOL(sock_register);
2537 * sock_unregister - remove a protocol handler
2538 * @family: protocol family to remove
2540 * This function is called by a protocol handler that wants to
2541 * remove its address family, and have it unlinked from the
2542 * new socket creation.
2544 * If protocol handler is a module, then it can use module reference
2545 * counts to protect against new references. If protocol handler is not
2546 * a module then it needs to provide its own protection in
2547 * the ops->create routine.
2549 void sock_unregister(int family)
2551 BUG_ON(family < 0 || family >= NPROTO);
2553 spin_lock(&net_family_lock);
2554 RCU_INIT_POINTER(net_families[family], NULL);
2555 spin_unlock(&net_family_lock);
2559 pr_info("NET: Unregistered protocol family %d\n", family);
2561 EXPORT_SYMBOL(sock_unregister);
2563 static int __init sock_init(void)
2567 * Initialize the network sysctl infrastructure.
2569 err = net_sysctl_init();
2574 * Initialize skbuff SLAB cache
2579 * Initialize the protocols module.
2584 err = register_filesystem(&sock_fs_type);
2587 sock_mnt = kern_mount(&sock_fs_type);
2588 if (IS_ERR(sock_mnt)) {
2589 err = PTR_ERR(sock_mnt);
2593 /* The real protocol initialization is performed in later initcalls.
2596 #ifdef CONFIG_NETFILTER
2597 err = netfilter_init();
2602 ptp_classifier_init();
2608 unregister_filesystem(&sock_fs_type);
2613 core_initcall(sock_init); /* early initcall */
2615 #ifdef CONFIG_PROC_FS
2616 void socket_seq_show(struct seq_file *seq)
2621 for_each_possible_cpu(cpu)
2622 counter += per_cpu(sockets_in_use, cpu);
2624 /* It can be negative, by the way. 8) */
2628 seq_printf(seq, "sockets: used %d\n", counter);
2630 #endif /* CONFIG_PROC_FS */
2632 #ifdef CONFIG_COMPAT
2633 static int do_siocgstamp(struct net *net, struct socket *sock,
2634 unsigned int cmd, void __user *up)
2636 mm_segment_t old_fs = get_fs();
2641 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2644 err = compat_put_timeval(&ktv, up);
2649 static int do_siocgstampns(struct net *net, struct socket *sock,
2650 unsigned int cmd, void __user *up)
2652 mm_segment_t old_fs = get_fs();
2653 struct timespec kts;
2657 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2660 err = compat_put_timespec(&kts, up);
2665 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2667 struct ifreq __user *uifr;
2670 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2671 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2674 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2678 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2684 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2686 struct compat_ifconf ifc32;
2688 struct ifconf __user *uifc;
2689 struct compat_ifreq __user *ifr32;
2690 struct ifreq __user *ifr;
2694 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2697 memset(&ifc, 0, sizeof(ifc));
2698 if (ifc32.ifcbuf == 0) {
2702 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2704 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2705 sizeof(struct ifreq);
2706 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2708 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2709 ifr32 = compat_ptr(ifc32.ifcbuf);
2710 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2711 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2717 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2720 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2724 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2728 ifr32 = compat_ptr(ifc32.ifcbuf);
2730 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2731 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2732 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2738 if (ifc32.ifcbuf == 0) {
2739 /* Translate from 64-bit structure multiple to
2743 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2748 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2754 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2756 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2757 bool convert_in = false, convert_out = false;
2758 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2759 struct ethtool_rxnfc __user *rxnfc;
2760 struct ifreq __user *ifr;
2761 u32 rule_cnt = 0, actual_rule_cnt;
2766 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2769 compat_rxnfc = compat_ptr(data);
2771 if (get_user(ethcmd, &compat_rxnfc->cmd))
2774 /* Most ethtool structures are defined without padding.
2775 * Unfortunately struct ethtool_rxnfc is an exception.
2780 case ETHTOOL_GRXCLSRLALL:
2781 /* Buffer size is variable */
2782 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2784 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2786 buf_size += rule_cnt * sizeof(u32);
2788 case ETHTOOL_GRXRINGS:
2789 case ETHTOOL_GRXCLSRLCNT:
2790 case ETHTOOL_GRXCLSRULE:
2791 case ETHTOOL_SRXCLSRLINS:
2794 case ETHTOOL_SRXCLSRLDEL:
2795 buf_size += sizeof(struct ethtool_rxnfc);
2800 ifr = compat_alloc_user_space(buf_size);
2801 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2803 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2806 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2807 &ifr->ifr_ifru.ifru_data))
2811 /* We expect there to be holes between fs.m_ext and
2812 * fs.ring_cookie and at the end of fs, but nowhere else.
2814 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2815 sizeof(compat_rxnfc->fs.m_ext) !=
2816 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2817 sizeof(rxnfc->fs.m_ext));
2819 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2820 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2821 offsetof(struct ethtool_rxnfc, fs.location) -
2822 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2824 if (copy_in_user(rxnfc, compat_rxnfc,
2825 (void __user *)(&rxnfc->fs.m_ext + 1) -
2826 (void __user *)rxnfc) ||
2827 copy_in_user(&rxnfc->fs.ring_cookie,
2828 &compat_rxnfc->fs.ring_cookie,
2829 (void __user *)(&rxnfc->fs.location + 1) -
2830 (void __user *)&rxnfc->fs.ring_cookie) ||
2831 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2832 sizeof(rxnfc->rule_cnt)))
2836 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2841 if (copy_in_user(compat_rxnfc, rxnfc,
2842 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2843 (const void __user *)rxnfc) ||
2844 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2845 &rxnfc->fs.ring_cookie,
2846 (const void __user *)(&rxnfc->fs.location + 1) -
2847 (const void __user *)&rxnfc->fs.ring_cookie) ||
2848 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2849 sizeof(rxnfc->rule_cnt)))
2852 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2853 /* As an optimisation, we only copy the actual
2854 * number of rules that the underlying
2855 * function returned. Since Mallory might
2856 * change the rule count in user memory, we
2857 * check that it is less than the rule count
2858 * originally given (as the user buffer size),
2859 * which has been range-checked.
2861 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2863 if (actual_rule_cnt < rule_cnt)
2864 rule_cnt = actual_rule_cnt;
2865 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2866 &rxnfc->rule_locs[0],
2867 rule_cnt * sizeof(u32)))
2875 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2878 compat_uptr_t uptr32;
2879 struct ifreq __user *uifr;
2881 uifr = compat_alloc_user_space(sizeof(*uifr));
2882 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2885 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2888 uptr = compat_ptr(uptr32);
2890 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2893 return dev_ioctl(net, SIOCWANDEV, uifr);
2896 static int bond_ioctl(struct net *net, unsigned int cmd,
2897 struct compat_ifreq __user *ifr32)
2900 mm_segment_t old_fs;
2904 case SIOCBONDENSLAVE:
2905 case SIOCBONDRELEASE:
2906 case SIOCBONDSETHWADDR:
2907 case SIOCBONDCHANGEACTIVE:
2908 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2913 err = dev_ioctl(net, cmd,
2914 (struct ifreq __user __force *) &kifr);
2919 return -ENOIOCTLCMD;
2923 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2924 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2925 struct compat_ifreq __user *u_ifreq32)
2927 struct ifreq __user *u_ifreq64;
2928 char tmp_buf[IFNAMSIZ];
2929 void __user *data64;
2932 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2935 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2937 data64 = compat_ptr(data32);
2939 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2941 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2944 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2947 return dev_ioctl(net, cmd, u_ifreq64);
2950 static int dev_ifsioc(struct net *net, struct socket *sock,
2951 unsigned int cmd, struct compat_ifreq __user *uifr32)
2953 struct ifreq __user *uifr;
2956 uifr = compat_alloc_user_space(sizeof(*uifr));
2957 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2960 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2971 case SIOCGIFBRDADDR:
2972 case SIOCGIFDSTADDR:
2973 case SIOCGIFNETMASK:
2978 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2986 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2987 struct compat_ifreq __user *uifr32)
2990 struct compat_ifmap __user *uifmap32;
2991 mm_segment_t old_fs;
2994 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2995 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2996 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2997 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2998 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2999 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3000 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3001 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3007 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3010 if (cmd == SIOCGIFMAP && !err) {
3011 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3012 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3013 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3014 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3015 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3016 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3017 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3026 struct sockaddr rt_dst; /* target address */
3027 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3028 struct sockaddr rt_genmask; /* target network mask (IP) */
3029 unsigned short rt_flags;
3032 unsigned char rt_tos;
3033 unsigned char rt_class;
3035 short rt_metric; /* +1 for binary compatibility! */
3036 /* char * */ u32 rt_dev; /* forcing the device at add */
3037 u32 rt_mtu; /* per route MTU/Window */
3038 u32 rt_window; /* Window clamping */
3039 unsigned short rt_irtt; /* Initial RTT */
3042 struct in6_rtmsg32 {
3043 struct in6_addr rtmsg_dst;
3044 struct in6_addr rtmsg_src;
3045 struct in6_addr rtmsg_gateway;
3055 static int routing_ioctl(struct net *net, struct socket *sock,
3056 unsigned int cmd, void __user *argp)
3060 struct in6_rtmsg r6;
3064 mm_segment_t old_fs = get_fs();
3066 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3067 struct in6_rtmsg32 __user *ur6 = argp;
3068 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3069 3 * sizeof(struct in6_addr));
3070 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3071 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3072 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3073 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3074 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3075 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3076 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3080 struct rtentry32 __user *ur4 = argp;
3081 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3082 3 * sizeof(struct sockaddr));
3083 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3084 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3085 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3086 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3087 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3088 ret |= get_user(rtdev, &(ur4->rt_dev));
3090 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3091 r4.rt_dev = (char __user __force *)devname;
3105 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3112 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3113 * for some operations; this forces use of the newer bridge-utils that
3114 * use compatible ioctls
3116 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3120 if (get_user(tmp, argp))
3122 if (tmp == BRCTL_GET_VERSION)
3123 return BRCTL_VERSION + 1;
3127 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3128 unsigned int cmd, unsigned long arg)
3130 void __user *argp = compat_ptr(arg);
3131 struct sock *sk = sock->sk;
3132 struct net *net = sock_net(sk);
3134 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3135 return compat_ifr_data_ioctl(net, cmd, argp);
3140 return old_bridge_ioctl(argp);
3142 return dev_ifname32(net, argp);
3144 return dev_ifconf(net, argp);
3146 return ethtool_ioctl(net, argp);
3148 return compat_siocwandev(net, argp);
3151 return compat_sioc_ifmap(net, cmd, argp);
3152 case SIOCBONDENSLAVE:
3153 case SIOCBONDRELEASE:
3154 case SIOCBONDSETHWADDR:
3155 case SIOCBONDCHANGEACTIVE:
3156 return bond_ioctl(net, cmd, argp);
3159 return routing_ioctl(net, sock, cmd, argp);
3161 return do_siocgstamp(net, sock, cmd, argp);
3163 return do_siocgstampns(net, sock, cmd, argp);
3164 case SIOCBONDSLAVEINFOQUERY:
3165 case SIOCBONDINFOQUERY:
3168 return compat_ifr_data_ioctl(net, cmd, argp);
3180 return sock_ioctl(file, cmd, arg);
3197 case SIOCSIFHWBROADCAST:
3199 case SIOCGIFBRDADDR:
3200 case SIOCSIFBRDADDR:
3201 case SIOCGIFDSTADDR:
3202 case SIOCSIFDSTADDR:
3203 case SIOCGIFNETMASK:
3204 case SIOCSIFNETMASK:
3215 return dev_ifsioc(net, sock, cmd, argp);
3221 return sock_do_ioctl(net, sock, cmd, arg);
3224 return -ENOIOCTLCMD;
3227 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3230 struct socket *sock = file->private_data;
3231 int ret = -ENOIOCTLCMD;
3238 if (sock->ops->compat_ioctl)
3239 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3241 if (ret == -ENOIOCTLCMD &&
3242 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3243 ret = compat_wext_handle_ioctl(net, cmd, arg);
3245 if (ret == -ENOIOCTLCMD)
3246 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3252 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3254 return sock->ops->bind(sock, addr, addrlen);
3256 EXPORT_SYMBOL(kernel_bind);
3258 int kernel_listen(struct socket *sock, int backlog)
3260 return sock->ops->listen(sock, backlog);
3262 EXPORT_SYMBOL(kernel_listen);
3264 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3266 struct sock *sk = sock->sk;
3269 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3274 err = sock->ops->accept(sock, *newsock, flags);
3276 sock_release(*newsock);
3281 (*newsock)->ops = sock->ops;
3282 __module_get((*newsock)->ops->owner);
3287 EXPORT_SYMBOL(kernel_accept);
3289 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3292 return sock->ops->connect(sock, addr, addrlen, flags);
3294 EXPORT_SYMBOL(kernel_connect);
3296 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3299 return sock->ops->getname(sock, addr, addrlen, 0);
3301 EXPORT_SYMBOL(kernel_getsockname);
3303 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3306 return sock->ops->getname(sock, addr, addrlen, 1);
3308 EXPORT_SYMBOL(kernel_getpeername);
3310 int kernel_getsockopt(struct socket *sock, int level, int optname,
3311 char *optval, int *optlen)
3313 mm_segment_t oldfs = get_fs();
3314 char __user *uoptval;
3315 int __user *uoptlen;
3318 uoptval = (char __user __force *) optval;
3319 uoptlen = (int __user __force *) optlen;
3322 if (level == SOL_SOCKET)
3323 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3325 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3330 EXPORT_SYMBOL(kernel_getsockopt);
3332 int kernel_setsockopt(struct socket *sock, int level, int optname,
3333 char *optval, unsigned int optlen)
3335 mm_segment_t oldfs = get_fs();
3336 char __user *uoptval;
3339 uoptval = (char __user __force *) optval;
3342 if (level == SOL_SOCKET)
3343 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3345 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3350 EXPORT_SYMBOL(kernel_setsockopt);
3352 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3353 size_t size, int flags)
3355 if (sock->ops->sendpage)
3356 return sock->ops->sendpage(sock, page, offset, size, flags);
3358 return sock_no_sendpage(sock, page, offset, size, flags);
3360 EXPORT_SYMBOL(kernel_sendpage);
3362 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3364 mm_segment_t oldfs = get_fs();
3368 err = sock->ops->ioctl(sock, cmd, arg);
3373 EXPORT_SYMBOL(kernel_sock_ioctl);
3375 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3377 return sock->ops->shutdown(sock, how);
3379 EXPORT_SYMBOL(kernel_sock_shutdown);