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/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
87 #include <linux/nsproxy.h>
88 #include <linux/magic.h>
89 #include <linux/slab.h>
90 #include <linux/xattr.h>
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
107 #include <net/busy_poll.h>
109 #ifdef CONFIG_NET_RX_BUSY_POLL
110 unsigned int sysctl_net_busy_read __read_mostly;
111 unsigned int sysctl_net_busy_poll __read_mostly;
114 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
115 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
116 unsigned long nr_segs, loff_t pos);
117 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
118 unsigned long nr_segs, loff_t pos);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 static int sock_close(struct inode *inode, struct file *file);
122 static unsigned int sock_poll(struct file *file,
123 struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 static long compat_sock_ioctl(struct file *file,
127 unsigned int cmd, unsigned long arg);
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131 int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133 struct pipe_inode_info *pipe, size_t len,
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
141 static const struct file_operations socket_file_ops = {
142 .owner = THIS_MODULE,
144 .aio_read = sock_aio_read,
145 .aio_write = sock_aio_write,
147 .unlocked_ioctl = sock_ioctl,
149 .compat_ioctl = compat_sock_ioctl,
152 .open = sock_no_open, /* special open code to disallow open via /proc */
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));
375 SOCK_INODE(sock)->i_fop = &socket_file_ops;
377 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
379 if (unlikely(IS_ERR(file))) {
380 /* drop dentry, keep inode */
381 ihold(path.dentry->d_inode);
387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
388 file->private_data = sock;
391 EXPORT_SYMBOL(sock_alloc_file);
393 static int sock_map_fd(struct socket *sock, int flags)
395 struct file *newfile;
396 int fd = get_unused_fd_flags(flags);
397 if (unlikely(fd < 0))
400 newfile = sock_alloc_file(sock, flags, NULL);
401 if (likely(!IS_ERR(newfile))) {
402 fd_install(fd, newfile);
407 return PTR_ERR(newfile);
410 struct socket *sock_from_file(struct file *file, int *err)
412 if (file->f_op == &socket_file_ops)
413 return file->private_data; /* set in sock_map_fd */
418 EXPORT_SYMBOL(sock_from_file);
421 * sockfd_lookup - Go from a file number to its socket slot
423 * @err: pointer to an error code return
425 * The file handle passed in is locked and the socket it is bound
426 * too is returned. If an error occurs the err pointer is overwritten
427 * with a negative errno code and NULL is returned. The function checks
428 * for both invalid handles and passing a handle which is not a socket.
430 * On a success the socket object pointer is returned.
433 struct socket *sockfd_lookup(int fd, int *err)
444 sock = sock_from_file(file, err);
449 EXPORT_SYMBOL(sockfd_lookup);
451 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
457 file = fget_light(fd, fput_needed);
459 sock = sock_from_file(file, err);
462 fput_light(file, *fput_needed);
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 * In theory you can't get an open on this inode, but /proc provides
561 * a back door. Remember to keep it shut otherwise you'll let the
562 * creepy crawlies in.
565 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
570 const struct file_operations bad_sock_fops = {
571 .owner = THIS_MODULE,
572 .open = sock_no_open,
573 .llseek = noop_llseek,
577 * sock_release - close a socket
578 * @sock: socket to close
580 * The socket is released from the protocol stack if it has a release
581 * callback, and the inode is then released if the socket is bound to
582 * an inode not a file.
585 void sock_release(struct socket *sock)
588 struct module *owner = sock->ops->owner;
590 sock->ops->release(sock);
595 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
596 printk(KERN_ERR "sock_release: fasync list not empty!\n");
598 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
601 this_cpu_sub(sockets_in_use, 1);
603 iput(SOCK_INODE(sock));
608 EXPORT_SYMBOL(sock_release);
610 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
613 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
614 *tx_flags |= SKBTX_HW_TSTAMP;
615 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
616 *tx_flags |= SKBTX_SW_TSTAMP;
617 if (sock_flag(sk, SOCK_WIFI_STATUS))
618 *tx_flags |= SKBTX_WIFI_STATUS;
620 EXPORT_SYMBOL(sock_tx_timestamp);
622 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
623 struct msghdr *msg, size_t size)
625 struct sock_iocb *si = kiocb_to_siocb(iocb);
632 return sock->ops->sendmsg(iocb, sock, msg, size);
635 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
636 struct msghdr *msg, size_t size)
638 int err = security_socket_sendmsg(sock, msg, size);
640 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
643 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
646 struct sock_iocb siocb;
649 init_sync_kiocb(&iocb, NULL);
650 iocb.private = &siocb;
651 ret = __sock_sendmsg(&iocb, sock, msg, size);
652 if (-EIOCBQUEUED == ret)
653 ret = wait_on_sync_kiocb(&iocb);
656 EXPORT_SYMBOL(sock_sendmsg);
658 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
661 struct sock_iocb siocb;
664 init_sync_kiocb(&iocb, NULL);
665 iocb.private = &siocb;
666 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
667 if (-EIOCBQUEUED == ret)
668 ret = wait_on_sync_kiocb(&iocb);
672 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
673 struct kvec *vec, size_t num, size_t size)
675 mm_segment_t oldfs = get_fs();
680 * the following is safe, since for compiler definitions of kvec and
681 * iovec are identical, yielding the same in-core layout and alignment
683 msg->msg_iov = (struct iovec *)vec;
684 msg->msg_iovlen = num;
685 result = sock_sendmsg(sock, msg, size);
689 EXPORT_SYMBOL(kernel_sendmsg);
692 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
694 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
697 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
698 struct timespec ts[3];
700 struct skb_shared_hwtstamps *shhwtstamps =
703 /* Race occurred between timestamp enabling and packet
704 receiving. Fill in the current time for now. */
705 if (need_software_tstamp && skb->tstamp.tv64 == 0)
706 __net_timestamp(skb);
708 if (need_software_tstamp) {
709 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
711 skb_get_timestamp(skb, &tv);
712 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
715 skb_get_timestampns(skb, &ts[0]);
716 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
717 sizeof(ts[0]), &ts[0]);
722 memset(ts, 0, sizeof(ts));
723 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
724 ktime_to_timespec_cond(skb->tstamp, ts + 0))
727 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
728 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
730 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
731 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
735 put_cmsg(msg, SOL_SOCKET,
736 SCM_TIMESTAMPING, sizeof(ts), &ts);
738 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
740 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
745 if (!sock_flag(sk, SOCK_WIFI_STATUS))
747 if (!skb->wifi_acked_valid)
750 ack = skb->wifi_acked;
752 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
754 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
756 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
759 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
760 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
761 sizeof(__u32), &skb->dropcount);
764 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
767 sock_recv_timestamp(msg, sk, skb);
768 sock_recv_drops(msg, sk, skb);
770 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
772 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
773 struct msghdr *msg, size_t size, int flags)
775 struct sock_iocb *si = kiocb_to_siocb(iocb);
783 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
786 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
787 struct msghdr *msg, size_t size, int flags)
789 int err = security_socket_recvmsg(sock, msg, size, flags);
791 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
794 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
795 size_t size, int flags)
798 struct sock_iocb siocb;
801 init_sync_kiocb(&iocb, NULL);
802 iocb.private = &siocb;
803 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
804 if (-EIOCBQUEUED == ret)
805 ret = wait_on_sync_kiocb(&iocb);
808 EXPORT_SYMBOL(sock_recvmsg);
810 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
811 size_t size, int flags)
814 struct sock_iocb siocb;
817 init_sync_kiocb(&iocb, NULL);
818 iocb.private = &siocb;
819 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
820 if (-EIOCBQUEUED == ret)
821 ret = wait_on_sync_kiocb(&iocb);
826 * kernel_recvmsg - Receive a message from a socket (kernel space)
827 * @sock: The socket to receive the message from
828 * @msg: Received message
829 * @vec: Input s/g array for message data
830 * @num: Size of input s/g array
831 * @size: Number of bytes to read
832 * @flags: Message flags (MSG_DONTWAIT, etc...)
834 * On return the msg structure contains the scatter/gather array passed in the
835 * vec argument. The array is modified so that it consists of the unfilled
836 * portion of the original array.
838 * The returned value is the total number of bytes received, or an error.
840 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
841 struct kvec *vec, size_t num, size_t size, int flags)
843 mm_segment_t oldfs = get_fs();
848 * the following is safe, since for compiler definitions of kvec and
849 * iovec are identical, yielding the same in-core layout and alignment
851 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
852 result = sock_recvmsg(sock, msg, size, flags);
856 EXPORT_SYMBOL(kernel_recvmsg);
858 static ssize_t sock_sendpage(struct file *file, struct page *page,
859 int offset, size_t size, loff_t *ppos, int more)
864 sock = file->private_data;
866 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
867 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
870 return kernel_sendpage(sock, page, offset, size, flags);
873 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
874 struct pipe_inode_info *pipe, size_t len,
877 struct socket *sock = file->private_data;
879 if (unlikely(!sock->ops->splice_read))
882 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
885 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
886 struct sock_iocb *siocb)
888 if (!is_sync_kiocb(iocb))
892 iocb->private = siocb;
896 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
897 struct file *file, const struct iovec *iov,
898 unsigned long nr_segs)
900 struct socket *sock = file->private_data;
904 for (i = 0; i < nr_segs; i++)
905 size += iov[i].iov_len;
907 msg->msg_name = NULL;
908 msg->msg_namelen = 0;
909 msg->msg_control = NULL;
910 msg->msg_controllen = 0;
911 msg->msg_iov = (struct iovec *)iov;
912 msg->msg_iovlen = nr_segs;
913 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
915 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
918 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
919 unsigned long nr_segs, loff_t pos)
921 struct sock_iocb siocb, *x;
926 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
930 x = alloc_sock_iocb(iocb, &siocb);
933 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
936 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
937 struct file *file, const struct iovec *iov,
938 unsigned long nr_segs)
940 struct socket *sock = file->private_data;
944 for (i = 0; i < nr_segs; i++)
945 size += iov[i].iov_len;
947 msg->msg_name = NULL;
948 msg->msg_namelen = 0;
949 msg->msg_control = NULL;
950 msg->msg_controllen = 0;
951 msg->msg_iov = (struct iovec *)iov;
952 msg->msg_iovlen = nr_segs;
953 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
954 if (sock->type == SOCK_SEQPACKET)
955 msg->msg_flags |= MSG_EOR;
957 return __sock_sendmsg(iocb, sock, msg, size);
960 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
961 unsigned long nr_segs, loff_t pos)
963 struct sock_iocb siocb, *x;
968 x = alloc_sock_iocb(iocb, &siocb);
972 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
976 * Atomic setting of ioctl hooks to avoid race
977 * with module unload.
980 static DEFINE_MUTEX(br_ioctl_mutex);
981 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
983 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
985 mutex_lock(&br_ioctl_mutex);
986 br_ioctl_hook = hook;
987 mutex_unlock(&br_ioctl_mutex);
989 EXPORT_SYMBOL(brioctl_set);
991 static DEFINE_MUTEX(vlan_ioctl_mutex);
992 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
994 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
996 mutex_lock(&vlan_ioctl_mutex);
997 vlan_ioctl_hook = hook;
998 mutex_unlock(&vlan_ioctl_mutex);
1000 EXPORT_SYMBOL(vlan_ioctl_set);
1002 static DEFINE_MUTEX(dlci_ioctl_mutex);
1003 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1005 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1007 mutex_lock(&dlci_ioctl_mutex);
1008 dlci_ioctl_hook = hook;
1009 mutex_unlock(&dlci_ioctl_mutex);
1011 EXPORT_SYMBOL(dlci_ioctl_set);
1013 static long sock_do_ioctl(struct net *net, struct socket *sock,
1014 unsigned int cmd, unsigned long arg)
1017 void __user *argp = (void __user *)arg;
1019 err = sock->ops->ioctl(sock, cmd, arg);
1022 * If this ioctl is unknown try to hand it down
1023 * to the NIC driver.
1025 if (err == -ENOIOCTLCMD)
1026 err = dev_ioctl(net, cmd, argp);
1032 * With an ioctl, arg may well be a user mode pointer, but we don't know
1033 * what to do with it - that's up to the protocol still.
1036 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1038 struct socket *sock;
1040 void __user *argp = (void __user *)arg;
1044 sock = file->private_data;
1047 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1048 err = dev_ioctl(net, cmd, argp);
1050 #ifdef CONFIG_WEXT_CORE
1051 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1052 err = dev_ioctl(net, cmd, argp);
1059 if (get_user(pid, (int __user *)argp))
1061 err = f_setown(sock->file, pid, 1);
1065 err = put_user(f_getown(sock->file),
1066 (int __user *)argp);
1074 request_module("bridge");
1076 mutex_lock(&br_ioctl_mutex);
1078 err = br_ioctl_hook(net, cmd, argp);
1079 mutex_unlock(&br_ioctl_mutex);
1084 if (!vlan_ioctl_hook)
1085 request_module("8021q");
1087 mutex_lock(&vlan_ioctl_mutex);
1088 if (vlan_ioctl_hook)
1089 err = vlan_ioctl_hook(net, argp);
1090 mutex_unlock(&vlan_ioctl_mutex);
1095 if (!dlci_ioctl_hook)
1096 request_module("dlci");
1098 mutex_lock(&dlci_ioctl_mutex);
1099 if (dlci_ioctl_hook)
1100 err = dlci_ioctl_hook(cmd, argp);
1101 mutex_unlock(&dlci_ioctl_mutex);
1104 err = sock_do_ioctl(net, sock, cmd, arg);
1110 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1113 struct socket *sock = NULL;
1115 err = security_socket_create(family, type, protocol, 1);
1119 sock = sock_alloc();
1126 err = security_socket_post_create(sock, family, type, protocol, 1);
1138 EXPORT_SYMBOL(sock_create_lite);
1140 /* No kernel lock held - perfect */
1141 static unsigned int sock_poll(struct file *file, poll_table *wait)
1143 unsigned int busy_flag = 0;
1144 struct socket *sock;
1147 * We can't return errors to poll, so it's either yes or no.
1149 sock = file->private_data;
1151 if (sk_can_busy_loop(sock->sk)) {
1152 /* this socket can poll_ll so tell the system call */
1153 busy_flag = POLL_BUSY_LOOP;
1155 /* once, only if requested by syscall */
1156 if (wait && (wait->_key & POLL_BUSY_LOOP))
1157 sk_busy_loop(sock->sk, 1);
1160 return busy_flag | sock->ops->poll(file, sock, wait);
1163 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1165 struct socket *sock = file->private_data;
1167 return sock->ops->mmap(file, sock, vma);
1170 static int sock_close(struct inode *inode, struct file *filp)
1172 sock_release(SOCKET_I(inode));
1177 * Update the socket async list
1179 * Fasync_list locking strategy.
1181 * 1. fasync_list is modified only under process context socket lock
1182 * i.e. under semaphore.
1183 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1184 * or under socket lock
1187 static int sock_fasync(int fd, struct file *filp, int on)
1189 struct socket *sock = filp->private_data;
1190 struct sock *sk = sock->sk;
1191 struct socket_wq *wq;
1197 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1198 fasync_helper(fd, filp, on, &wq->fasync_list);
1200 if (!wq->fasync_list)
1201 sock_reset_flag(sk, SOCK_FASYNC);
1203 sock_set_flag(sk, SOCK_FASYNC);
1209 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1211 int sock_wake_async(struct socket *sock, int how, int band)
1213 struct socket_wq *wq;
1218 wq = rcu_dereference(sock->wq);
1219 if (!wq || !wq->fasync_list) {
1224 case SOCK_WAKE_WAITD:
1225 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1228 case SOCK_WAKE_SPACE:
1229 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1234 kill_fasync(&wq->fasync_list, SIGIO, band);
1237 kill_fasync(&wq->fasync_list, SIGURG, band);
1242 EXPORT_SYMBOL(sock_wake_async);
1244 int __sock_create(struct net *net, int family, int type, int protocol,
1245 struct socket **res, int kern)
1248 struct socket *sock;
1249 const struct net_proto_family *pf;
1252 * Check protocol is in range
1254 if (family < 0 || family >= NPROTO)
1255 return -EAFNOSUPPORT;
1256 if (type < 0 || type >= SOCK_MAX)
1261 This uglymoron is moved from INET layer to here to avoid
1262 deadlock in module load.
1264 if (family == PF_INET && type == SOCK_PACKET) {
1268 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1274 err = security_socket_create(family, type, protocol, kern);
1279 * Allocate the socket and allow the family to set things up. if
1280 * the protocol is 0, the family is instructed to select an appropriate
1283 sock = sock_alloc();
1285 net_warn_ratelimited("socket: no more sockets\n");
1286 return -ENFILE; /* Not exactly a match, but its the
1287 closest posix thing */
1292 #ifdef CONFIG_MODULES
1293 /* Attempt to load a protocol module if the find failed.
1295 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1296 * requested real, full-featured networking support upon configuration.
1297 * Otherwise module support will break!
1299 if (rcu_access_pointer(net_families[family]) == NULL)
1300 request_module("net-pf-%d", family);
1304 pf = rcu_dereference(net_families[family]);
1305 err = -EAFNOSUPPORT;
1310 * We will call the ->create function, that possibly is in a loadable
1311 * module, so we have to bump that loadable module refcnt first.
1313 if (!try_module_get(pf->owner))
1316 /* Now protected by module ref count */
1319 err = pf->create(net, sock, protocol, kern);
1321 goto out_module_put;
1324 * Now to bump the refcnt of the [loadable] module that owns this
1325 * socket at sock_release time we decrement its refcnt.
1327 if (!try_module_get(sock->ops->owner))
1328 goto out_module_busy;
1331 * Now that we're done with the ->create function, the [loadable]
1332 * module can have its refcnt decremented
1334 module_put(pf->owner);
1335 err = security_socket_post_create(sock, family, type, protocol, kern);
1337 goto out_sock_release;
1343 err = -EAFNOSUPPORT;
1346 module_put(pf->owner);
1353 goto out_sock_release;
1355 EXPORT_SYMBOL(__sock_create);
1357 int sock_create(int family, int type, int protocol, struct socket **res)
1359 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1361 EXPORT_SYMBOL(sock_create);
1363 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1365 return __sock_create(&init_net, family, type, protocol, res, 1);
1367 EXPORT_SYMBOL(sock_create_kern);
1369 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1372 struct socket *sock;
1375 /* Check the SOCK_* constants for consistency. */
1376 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1377 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1378 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1379 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1381 flags = type & ~SOCK_TYPE_MASK;
1382 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1384 type &= SOCK_TYPE_MASK;
1386 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1387 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1389 retval = sock_create(family, type, protocol, &sock);
1393 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1398 /* It may be already another descriptor 8) Not kernel problem. */
1407 * Create a pair of connected sockets.
1410 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1411 int __user *, usockvec)
1413 struct socket *sock1, *sock2;
1415 struct file *newfile1, *newfile2;
1418 flags = type & ~SOCK_TYPE_MASK;
1419 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1421 type &= SOCK_TYPE_MASK;
1423 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1424 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1427 * Obtain the first socket and check if the underlying protocol
1428 * supports the socketpair call.
1431 err = sock_create(family, type, protocol, &sock1);
1435 err = sock_create(family, type, protocol, &sock2);
1439 err = sock1->ops->socketpair(sock1, sock2);
1441 goto out_release_both;
1443 fd1 = get_unused_fd_flags(flags);
1444 if (unlikely(fd1 < 0)) {
1446 goto out_release_both;
1449 fd2 = get_unused_fd_flags(flags);
1450 if (unlikely(fd2 < 0)) {
1452 goto out_put_unused_1;
1455 newfile1 = sock_alloc_file(sock1, flags, NULL);
1456 if (unlikely(IS_ERR(newfile1))) {
1457 err = PTR_ERR(newfile1);
1458 goto out_put_unused_both;
1461 newfile2 = sock_alloc_file(sock2, flags, NULL);
1462 if (IS_ERR(newfile2)) {
1463 err = PTR_ERR(newfile2);
1467 err = put_user(fd1, &usockvec[0]);
1471 err = put_user(fd2, &usockvec[1]);
1475 audit_fd_pair(fd1, fd2);
1477 fd_install(fd1, newfile1);
1478 fd_install(fd2, newfile2);
1479 /* fd1 and fd2 may be already another descriptors.
1480 * Not kernel problem.
1496 sock_release(sock2);
1499 out_put_unused_both:
1504 sock_release(sock2);
1506 sock_release(sock1);
1512 * Bind a name to a socket. Nothing much to do here since it's
1513 * the protocol's responsibility to handle the local address.
1515 * We move the socket address to kernel space before we call
1516 * the protocol layer (having also checked the address is ok).
1519 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1521 struct socket *sock;
1522 struct sockaddr_storage address;
1523 int err, fput_needed;
1525 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1527 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1529 err = security_socket_bind(sock,
1530 (struct sockaddr *)&address,
1533 err = sock->ops->bind(sock,
1537 fput_light(sock->file, fput_needed);
1543 * Perform a listen. Basically, we allow the protocol to do anything
1544 * necessary for a listen, and if that works, we mark the socket as
1545 * ready for listening.
1548 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1550 struct socket *sock;
1551 int err, fput_needed;
1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1556 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1557 if ((unsigned int)backlog > somaxconn)
1558 backlog = somaxconn;
1560 err = security_socket_listen(sock, backlog);
1562 err = sock->ops->listen(sock, backlog);
1564 fput_light(sock->file, fput_needed);
1570 * For accept, we attempt to create a new socket, set up the link
1571 * with the client, wake up the client, then return the new
1572 * connected fd. We collect the address of the connector in kernel
1573 * space and move it to user at the very end. This is unclean because
1574 * we open the socket then return an error.
1576 * 1003.1g adds the ability to recvmsg() to query connection pending
1577 * status to recvmsg. We need to add that support in a way thats
1578 * clean when we restucture accept also.
1581 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1582 int __user *, upeer_addrlen, int, flags)
1584 struct socket *sock, *newsock;
1585 struct file *newfile;
1586 int err, len, newfd, fput_needed;
1587 struct sockaddr_storage address;
1589 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1592 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1593 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1595 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1600 newsock = sock_alloc();
1604 newsock->type = sock->type;
1605 newsock->ops = sock->ops;
1608 * We don't need try_module_get here, as the listening socket (sock)
1609 * has the protocol module (sock->ops->owner) held.
1611 __module_get(newsock->ops->owner);
1613 newfd = get_unused_fd_flags(flags);
1614 if (unlikely(newfd < 0)) {
1616 sock_release(newsock);
1619 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1620 if (unlikely(IS_ERR(newfile))) {
1621 err = PTR_ERR(newfile);
1622 put_unused_fd(newfd);
1623 sock_release(newsock);
1627 err = security_socket_accept(sock, newsock);
1631 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1635 if (upeer_sockaddr) {
1636 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1638 err = -ECONNABORTED;
1641 err = move_addr_to_user(&address,
1642 len, upeer_sockaddr, upeer_addrlen);
1647 /* File flags are not inherited via accept() unlike another OSes. */
1649 fd_install(newfd, newfile);
1653 fput_light(sock->file, fput_needed);
1658 put_unused_fd(newfd);
1662 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1663 int __user *, upeer_addrlen)
1665 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1669 * Attempt to connect to a socket with the server address. The address
1670 * is in user space so we verify it is OK and move it to kernel space.
1672 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1675 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1676 * other SEQPACKET protocols that take time to connect() as it doesn't
1677 * include the -EINPROGRESS status for such sockets.
1680 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1683 struct socket *sock;
1684 struct sockaddr_storage address;
1685 int err, fput_needed;
1687 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1695 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1699 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1700 sock->file->f_flags);
1702 fput_light(sock->file, fput_needed);
1708 * Get the local address ('name') of a socket object. Move the obtained
1709 * name to user space.
1712 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1713 int __user *, usockaddr_len)
1715 struct socket *sock;
1716 struct sockaddr_storage address;
1717 int len, err, fput_needed;
1719 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1723 err = security_socket_getsockname(sock);
1727 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1730 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1733 fput_light(sock->file, fput_needed);
1739 * Get the remote address ('name') of a socket object. Move the obtained
1740 * name to user space.
1743 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1744 int __user *, usockaddr_len)
1746 struct socket *sock;
1747 struct sockaddr_storage address;
1748 int len, err, fput_needed;
1750 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1752 err = security_socket_getpeername(sock);
1754 fput_light(sock->file, fput_needed);
1759 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1762 err = move_addr_to_user(&address, len, usockaddr,
1764 fput_light(sock->file, fput_needed);
1770 * Send a datagram to a given address. We move the address into kernel
1771 * space and check the user space data area is readable before invoking
1775 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1776 unsigned int, flags, struct sockaddr __user *, addr,
1779 struct socket *sock;
1780 struct sockaddr_storage address;
1788 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1792 iov.iov_base = buff;
1794 msg.msg_name = NULL;
1797 msg.msg_control = NULL;
1798 msg.msg_controllen = 0;
1799 msg.msg_namelen = 0;
1801 err = move_addr_to_kernel(addr, addr_len, &address);
1804 msg.msg_name = (struct sockaddr *)&address;
1805 msg.msg_namelen = addr_len;
1807 if (sock->file->f_flags & O_NONBLOCK)
1808 flags |= MSG_DONTWAIT;
1809 msg.msg_flags = flags;
1810 err = sock_sendmsg(sock, &msg, len);
1813 fput_light(sock->file, fput_needed);
1819 * Send a datagram down a socket.
1822 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1823 unsigned int, flags)
1825 return sys_sendto(fd, buff, len, flags, NULL, 0);
1829 * Receive a frame from the socket and optionally record the address of the
1830 * sender. We verify the buffers are writable and if needed move the
1831 * sender address from kernel to user space.
1834 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1835 unsigned int, flags, struct sockaddr __user *, addr,
1836 int __user *, addr_len)
1838 struct socket *sock;
1841 struct sockaddr_storage address;
1847 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1851 msg.msg_control = NULL;
1852 msg.msg_controllen = 0;
1856 iov.iov_base = ubuf;
1857 /* Save some cycles and don't copy the address if not needed */
1858 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1859 /* We assume all kernel code knows the size of sockaddr_storage */
1860 msg.msg_namelen = 0;
1861 if (sock->file->f_flags & O_NONBLOCK)
1862 flags |= MSG_DONTWAIT;
1863 err = sock_recvmsg(sock, &msg, size, flags);
1865 if (err >= 0 && addr != NULL) {
1866 err2 = move_addr_to_user(&address,
1867 msg.msg_namelen, addr, addr_len);
1872 fput_light(sock->file, fput_needed);
1878 * Receive a datagram from a socket.
1881 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1884 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1888 * Set a socket option. Because we don't know the option lengths we have
1889 * to pass the user mode parameter for the protocols to sort out.
1892 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1893 char __user *, optval, int, optlen)
1895 int err, fput_needed;
1896 struct socket *sock;
1901 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1903 err = security_socket_setsockopt(sock, level, optname);
1907 if (level == SOL_SOCKET)
1909 sock_setsockopt(sock, level, optname, optval,
1913 sock->ops->setsockopt(sock, level, optname, optval,
1916 fput_light(sock->file, fput_needed);
1922 * Get a socket option. Because we don't know the option lengths we have
1923 * to pass a user mode parameter for the protocols to sort out.
1926 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1927 char __user *, optval, int __user *, optlen)
1929 int err, fput_needed;
1930 struct socket *sock;
1932 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1934 err = security_socket_getsockopt(sock, level, optname);
1938 if (level == SOL_SOCKET)
1940 sock_getsockopt(sock, level, optname, optval,
1944 sock->ops->getsockopt(sock, level, optname, optval,
1947 fput_light(sock->file, fput_needed);
1953 * Shutdown a socket.
1956 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1958 int err, fput_needed;
1959 struct socket *sock;
1961 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1963 err = security_socket_shutdown(sock, how);
1965 err = sock->ops->shutdown(sock, how);
1966 fput_light(sock->file, fput_needed);
1971 /* A couple of helpful macros for getting the address of the 32/64 bit
1972 * fields which are the same type (int / unsigned) on our platforms.
1974 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1975 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1976 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1978 struct used_address {
1979 struct sockaddr_storage name;
1980 unsigned int name_len;
1983 static int copy_msghdr_from_user(struct msghdr *kmsg,
1984 struct msghdr __user *umsg)
1986 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1988 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1989 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1993 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1994 struct msghdr *msg_sys, unsigned int flags,
1995 struct used_address *used_address)
1997 struct compat_msghdr __user *msg_compat =
1998 (struct compat_msghdr __user *)msg;
1999 struct sockaddr_storage address;
2000 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2001 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2002 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2003 /* 20 is size of ipv6_pktinfo */
2004 unsigned char *ctl_buf = ctl;
2005 int err, ctl_len, total_len;
2008 if (MSG_CMSG_COMPAT & flags) {
2009 if (get_compat_msghdr(msg_sys, msg_compat))
2012 err = copy_msghdr_from_user(msg_sys, msg);
2017 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2019 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2022 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2028 /* This will also move the address data into kernel space */
2029 if (MSG_CMSG_COMPAT & flags) {
2030 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2032 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2039 if (msg_sys->msg_controllen > INT_MAX)
2041 ctl_len = msg_sys->msg_controllen;
2042 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2044 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2048 ctl_buf = msg_sys->msg_control;
2049 ctl_len = msg_sys->msg_controllen;
2050 } else if (ctl_len) {
2051 if (ctl_len > sizeof(ctl)) {
2052 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2053 if (ctl_buf == NULL)
2058 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2059 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2060 * checking falls down on this.
2062 if (copy_from_user(ctl_buf,
2063 (void __user __force *)msg_sys->msg_control,
2066 msg_sys->msg_control = ctl_buf;
2068 msg_sys->msg_flags = flags;
2070 if (sock->file->f_flags & O_NONBLOCK)
2071 msg_sys->msg_flags |= MSG_DONTWAIT;
2073 * If this is sendmmsg() and current destination address is same as
2074 * previously succeeded address, omit asking LSM's decision.
2075 * used_address->name_len is initialized to UINT_MAX so that the first
2076 * destination address never matches.
2078 if (used_address && msg_sys->msg_name &&
2079 used_address->name_len == msg_sys->msg_namelen &&
2080 !memcmp(&used_address->name, msg_sys->msg_name,
2081 used_address->name_len)) {
2082 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2085 err = sock_sendmsg(sock, msg_sys, total_len);
2087 * If this is sendmmsg() and sending to current destination address was
2088 * successful, remember it.
2090 if (used_address && err >= 0) {
2091 used_address->name_len = msg_sys->msg_namelen;
2092 if (msg_sys->msg_name)
2093 memcpy(&used_address->name, msg_sys->msg_name,
2094 used_address->name_len);
2099 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2101 if (iov != iovstack)
2108 * BSD sendmsg interface
2111 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2113 int fput_needed, err;
2114 struct msghdr msg_sys;
2115 struct socket *sock;
2117 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2121 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2123 fput_light(sock->file, fput_needed);
2128 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2130 if (flags & MSG_CMSG_COMPAT)
2132 return __sys_sendmsg(fd, msg, flags);
2136 * Linux sendmmsg interface
2139 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2142 int fput_needed, err, datagrams;
2143 struct socket *sock;
2144 struct mmsghdr __user *entry;
2145 struct compat_mmsghdr __user *compat_entry;
2146 struct msghdr msg_sys;
2147 struct used_address used_address;
2149 if (vlen > UIO_MAXIOV)
2154 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2158 used_address.name_len = UINT_MAX;
2160 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2163 while (datagrams < vlen) {
2164 if (MSG_CMSG_COMPAT & flags) {
2165 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2166 &msg_sys, flags, &used_address);
2169 err = __put_user(err, &compat_entry->msg_len);
2172 err = ___sys_sendmsg(sock,
2173 (struct msghdr __user *)entry,
2174 &msg_sys, flags, &used_address);
2177 err = put_user(err, &entry->msg_len);
2186 fput_light(sock->file, fput_needed);
2188 /* We only return an error if no datagrams were able to be sent */
2195 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2196 unsigned int, vlen, unsigned int, flags)
2198 if (flags & MSG_CMSG_COMPAT)
2200 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2203 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2204 struct msghdr *msg_sys, unsigned int flags, int nosec)
2206 struct compat_msghdr __user *msg_compat =
2207 (struct compat_msghdr __user *)msg;
2208 struct iovec iovstack[UIO_FASTIOV];
2209 struct iovec *iov = iovstack;
2210 unsigned long cmsg_ptr;
2211 int err, total_len, len;
2213 /* kernel mode address */
2214 struct sockaddr_storage addr;
2216 /* user mode address pointers */
2217 struct sockaddr __user *uaddr;
2218 int __user *uaddr_len;
2220 if (MSG_CMSG_COMPAT & flags) {
2221 if (get_compat_msghdr(msg_sys, msg_compat))
2224 err = copy_msghdr_from_user(msg_sys, msg);
2229 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2231 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2234 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2240 /* Save the user-mode address (verify_iovec will change the
2241 * kernel msghdr to use the kernel address space)
2243 uaddr = (__force void __user *)msg_sys->msg_name;
2244 uaddr_len = COMPAT_NAMELEN(msg);
2245 if (MSG_CMSG_COMPAT & flags)
2246 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2248 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2253 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2254 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2256 /* We assume all kernel code knows the size of sockaddr_storage */
2257 msg_sys->msg_namelen = 0;
2259 if (sock->file->f_flags & O_NONBLOCK)
2260 flags |= MSG_DONTWAIT;
2261 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2267 if (uaddr != NULL) {
2268 err = move_addr_to_user(&addr,
2269 msg_sys->msg_namelen, uaddr,
2274 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2278 if (MSG_CMSG_COMPAT & flags)
2279 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2280 &msg_compat->msg_controllen);
2282 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2283 &msg->msg_controllen);
2289 if (iov != iovstack)
2296 * BSD recvmsg interface
2299 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2301 int fput_needed, err;
2302 struct msghdr msg_sys;
2303 struct socket *sock;
2305 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2309 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2311 fput_light(sock->file, fput_needed);
2316 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2317 unsigned int, flags)
2319 if (flags & MSG_CMSG_COMPAT)
2321 return __sys_recvmsg(fd, msg, flags);
2325 * Linux recvmmsg interface
2328 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2329 unsigned int flags, struct timespec *timeout)
2331 int fput_needed, err, datagrams;
2332 struct socket *sock;
2333 struct mmsghdr __user *entry;
2334 struct compat_mmsghdr __user *compat_entry;
2335 struct msghdr msg_sys;
2336 struct timespec end_time;
2339 poll_select_set_timeout(&end_time, timeout->tv_sec,
2345 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2349 err = sock_error(sock->sk);
2354 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2356 while (datagrams < vlen) {
2358 * No need to ask LSM for more than the first datagram.
2360 if (MSG_CMSG_COMPAT & flags) {
2361 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2362 &msg_sys, flags & ~MSG_WAITFORONE,
2366 err = __put_user(err, &compat_entry->msg_len);
2369 err = ___sys_recvmsg(sock,
2370 (struct msghdr __user *)entry,
2371 &msg_sys, flags & ~MSG_WAITFORONE,
2375 err = put_user(err, &entry->msg_len);
2383 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2384 if (flags & MSG_WAITFORONE)
2385 flags |= MSG_DONTWAIT;
2388 ktime_get_ts(timeout);
2389 *timeout = timespec_sub(end_time, *timeout);
2390 if (timeout->tv_sec < 0) {
2391 timeout->tv_sec = timeout->tv_nsec = 0;
2395 /* Timeout, return less than vlen datagrams */
2396 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2400 /* Out of band data, return right away */
2401 if (msg_sys.msg_flags & MSG_OOB)
2406 fput_light(sock->file, fput_needed);
2411 if (datagrams != 0) {
2413 * We may return less entries than requested (vlen) if the
2414 * sock is non block and there aren't enough datagrams...
2416 if (err != -EAGAIN) {
2418 * ... or if recvmsg returns an error after we
2419 * received some datagrams, where we record the
2420 * error to return on the next call or if the
2421 * app asks about it using getsockopt(SO_ERROR).
2423 sock->sk->sk_err = -err;
2432 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2433 unsigned int, vlen, unsigned int, flags,
2434 struct timespec __user *, timeout)
2437 struct timespec timeout_sys;
2439 if (flags & MSG_CMSG_COMPAT)
2443 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2445 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2448 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2450 if (datagrams > 0 &&
2451 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2452 datagrams = -EFAULT;
2457 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2458 /* Argument list sizes for sys_socketcall */
2459 #define AL(x) ((x) * sizeof(unsigned long))
2460 static const unsigned char nargs[21] = {
2461 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2462 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2463 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2470 * System call vectors.
2472 * Argument checking cleaned up. Saved 20% in size.
2473 * This function doesn't need to set the kernel lock because
2474 * it is set by the callees.
2477 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2479 unsigned long a[AUDITSC_ARGS];
2480 unsigned long a0, a1;
2484 if (call < 1 || call > SYS_SENDMMSG)
2488 if (len > sizeof(a))
2491 /* copy_from_user should be SMP safe. */
2492 if (copy_from_user(a, args, len))
2495 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2504 err = sys_socket(a0, a1, a[2]);
2507 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2510 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2513 err = sys_listen(a0, a1);
2516 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2517 (int __user *)a[2], 0);
2519 case SYS_GETSOCKNAME:
2521 sys_getsockname(a0, (struct sockaddr __user *)a1,
2522 (int __user *)a[2]);
2524 case SYS_GETPEERNAME:
2526 sys_getpeername(a0, (struct sockaddr __user *)a1,
2527 (int __user *)a[2]);
2529 case SYS_SOCKETPAIR:
2530 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2533 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2536 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2537 (struct sockaddr __user *)a[4], a[5]);
2540 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2543 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2544 (struct sockaddr __user *)a[4],
2545 (int __user *)a[5]);
2548 err = sys_shutdown(a0, a1);
2550 case SYS_SETSOCKOPT:
2551 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2553 case SYS_GETSOCKOPT:
2555 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2556 (int __user *)a[4]);
2559 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2562 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2565 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2568 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2569 (struct timespec __user *)a[4]);
2572 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2573 (int __user *)a[2], a[3]);
2582 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2585 * sock_register - add a socket protocol handler
2586 * @ops: description of protocol
2588 * This function is called by a protocol handler that wants to
2589 * advertise its address family, and have it linked into the
2590 * socket interface. The value ops->family coresponds to the
2591 * socket system call protocol family.
2593 int sock_register(const struct net_proto_family *ops)
2597 if (ops->family >= NPROTO) {
2598 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2603 spin_lock(&net_family_lock);
2604 if (rcu_dereference_protected(net_families[ops->family],
2605 lockdep_is_held(&net_family_lock)))
2608 rcu_assign_pointer(net_families[ops->family], ops);
2611 spin_unlock(&net_family_lock);
2613 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2616 EXPORT_SYMBOL(sock_register);
2619 * sock_unregister - remove a protocol handler
2620 * @family: protocol family to remove
2622 * This function is called by a protocol handler that wants to
2623 * remove its address family, and have it unlinked from the
2624 * new socket creation.
2626 * If protocol handler is a module, then it can use module reference
2627 * counts to protect against new references. If protocol handler is not
2628 * a module then it needs to provide its own protection in
2629 * the ops->create routine.
2631 void sock_unregister(int family)
2633 BUG_ON(family < 0 || family >= NPROTO);
2635 spin_lock(&net_family_lock);
2636 RCU_INIT_POINTER(net_families[family], NULL);
2637 spin_unlock(&net_family_lock);
2641 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2643 EXPORT_SYMBOL(sock_unregister);
2645 static int __init sock_init(void)
2649 * Initialize the network sysctl infrastructure.
2651 err = net_sysctl_init();
2656 * Initialize skbuff SLAB cache
2661 * Initialize the protocols module.
2666 err = register_filesystem(&sock_fs_type);
2669 sock_mnt = kern_mount(&sock_fs_type);
2670 if (IS_ERR(sock_mnt)) {
2671 err = PTR_ERR(sock_mnt);
2675 /* The real protocol initialization is performed in later initcalls.
2678 #ifdef CONFIG_NETFILTER
2679 err = netfilter_init();
2684 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2685 skb_timestamping_init();
2692 unregister_filesystem(&sock_fs_type);
2697 core_initcall(sock_init); /* early initcall */
2699 #ifdef CONFIG_PROC_FS
2700 void socket_seq_show(struct seq_file *seq)
2705 for_each_possible_cpu(cpu)
2706 counter += per_cpu(sockets_in_use, cpu);
2708 /* It can be negative, by the way. 8) */
2712 seq_printf(seq, "sockets: used %d\n", counter);
2714 #endif /* CONFIG_PROC_FS */
2716 #ifdef CONFIG_COMPAT
2717 static int do_siocgstamp(struct net *net, struct socket *sock,
2718 unsigned int cmd, void __user *up)
2720 mm_segment_t old_fs = get_fs();
2725 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2728 err = compat_put_timeval(&ktv, up);
2733 static int do_siocgstampns(struct net *net, struct socket *sock,
2734 unsigned int cmd, void __user *up)
2736 mm_segment_t old_fs = get_fs();
2737 struct timespec kts;
2741 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2744 err = compat_put_timespec(&kts, up);
2749 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2751 struct ifreq __user *uifr;
2754 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2755 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2758 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2762 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2768 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2770 struct compat_ifconf ifc32;
2772 struct ifconf __user *uifc;
2773 struct compat_ifreq __user *ifr32;
2774 struct ifreq __user *ifr;
2778 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2781 memset(&ifc, 0, sizeof(ifc));
2782 if (ifc32.ifcbuf == 0) {
2786 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2788 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2789 sizeof(struct ifreq);
2790 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2792 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2793 ifr32 = compat_ptr(ifc32.ifcbuf);
2794 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2795 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2801 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2804 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2808 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2812 ifr32 = compat_ptr(ifc32.ifcbuf);
2814 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2815 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2816 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2822 if (ifc32.ifcbuf == 0) {
2823 /* Translate from 64-bit structure multiple to
2827 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2832 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2838 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2840 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2841 bool convert_in = false, convert_out = false;
2842 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2843 struct ethtool_rxnfc __user *rxnfc;
2844 struct ifreq __user *ifr;
2845 u32 rule_cnt = 0, actual_rule_cnt;
2850 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2853 compat_rxnfc = compat_ptr(data);
2855 if (get_user(ethcmd, &compat_rxnfc->cmd))
2858 /* Most ethtool structures are defined without padding.
2859 * Unfortunately struct ethtool_rxnfc is an exception.
2864 case ETHTOOL_GRXCLSRLALL:
2865 /* Buffer size is variable */
2866 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2868 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2870 buf_size += rule_cnt * sizeof(u32);
2872 case ETHTOOL_GRXRINGS:
2873 case ETHTOOL_GRXCLSRLCNT:
2874 case ETHTOOL_GRXCLSRULE:
2875 case ETHTOOL_SRXCLSRLINS:
2878 case ETHTOOL_SRXCLSRLDEL:
2879 buf_size += sizeof(struct ethtool_rxnfc);
2884 ifr = compat_alloc_user_space(buf_size);
2885 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2887 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2890 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2891 &ifr->ifr_ifru.ifru_data))
2895 /* We expect there to be holes between fs.m_ext and
2896 * fs.ring_cookie and at the end of fs, but nowhere else.
2898 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2899 sizeof(compat_rxnfc->fs.m_ext) !=
2900 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2901 sizeof(rxnfc->fs.m_ext));
2903 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2904 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2905 offsetof(struct ethtool_rxnfc, fs.location) -
2906 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2908 if (copy_in_user(rxnfc, compat_rxnfc,
2909 (void __user *)(&rxnfc->fs.m_ext + 1) -
2910 (void __user *)rxnfc) ||
2911 copy_in_user(&rxnfc->fs.ring_cookie,
2912 &compat_rxnfc->fs.ring_cookie,
2913 (void __user *)(&rxnfc->fs.location + 1) -
2914 (void __user *)&rxnfc->fs.ring_cookie) ||
2915 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2916 sizeof(rxnfc->rule_cnt)))
2920 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2925 if (copy_in_user(compat_rxnfc, rxnfc,
2926 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2927 (const void __user *)rxnfc) ||
2928 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2929 &rxnfc->fs.ring_cookie,
2930 (const void __user *)(&rxnfc->fs.location + 1) -
2931 (const void __user *)&rxnfc->fs.ring_cookie) ||
2932 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2933 sizeof(rxnfc->rule_cnt)))
2936 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2937 /* As an optimisation, we only copy the actual
2938 * number of rules that the underlying
2939 * function returned. Since Mallory might
2940 * change the rule count in user memory, we
2941 * check that it is less than the rule count
2942 * originally given (as the user buffer size),
2943 * which has been range-checked.
2945 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2947 if (actual_rule_cnt < rule_cnt)
2948 rule_cnt = actual_rule_cnt;
2949 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2950 &rxnfc->rule_locs[0],
2951 rule_cnt * sizeof(u32)))
2959 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2962 compat_uptr_t uptr32;
2963 struct ifreq __user *uifr;
2965 uifr = compat_alloc_user_space(sizeof(*uifr));
2966 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2969 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2972 uptr = compat_ptr(uptr32);
2974 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2977 return dev_ioctl(net, SIOCWANDEV, uifr);
2980 static int bond_ioctl(struct net *net, unsigned int cmd,
2981 struct compat_ifreq __user *ifr32)
2984 mm_segment_t old_fs;
2988 case SIOCBONDENSLAVE:
2989 case SIOCBONDRELEASE:
2990 case SIOCBONDSETHWADDR:
2991 case SIOCBONDCHANGEACTIVE:
2992 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2997 err = dev_ioctl(net, cmd,
2998 (struct ifreq __user __force *) &kifr);
3003 return -ENOIOCTLCMD;
3007 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3008 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3009 struct compat_ifreq __user *u_ifreq32)
3011 struct ifreq __user *u_ifreq64;
3012 char tmp_buf[IFNAMSIZ];
3013 void __user *data64;
3016 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3019 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3021 data64 = compat_ptr(data32);
3023 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3025 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3028 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3031 return dev_ioctl(net, cmd, u_ifreq64);
3034 static int dev_ifsioc(struct net *net, struct socket *sock,
3035 unsigned int cmd, struct compat_ifreq __user *uifr32)
3037 struct ifreq __user *uifr;
3040 uifr = compat_alloc_user_space(sizeof(*uifr));
3041 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3044 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3055 case SIOCGIFBRDADDR:
3056 case SIOCGIFDSTADDR:
3057 case SIOCGIFNETMASK:
3062 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3070 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3071 struct compat_ifreq __user *uifr32)
3074 struct compat_ifmap __user *uifmap32;
3075 mm_segment_t old_fs;
3078 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3079 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3080 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3081 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3082 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3083 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3084 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3085 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3091 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3094 if (cmd == SIOCGIFMAP && !err) {
3095 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3096 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3097 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3098 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3099 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3100 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3101 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3110 struct sockaddr rt_dst; /* target address */
3111 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3112 struct sockaddr rt_genmask; /* target network mask (IP) */
3113 unsigned short rt_flags;
3116 unsigned char rt_tos;
3117 unsigned char rt_class;
3119 short rt_metric; /* +1 for binary compatibility! */
3120 /* char * */ u32 rt_dev; /* forcing the device at add */
3121 u32 rt_mtu; /* per route MTU/Window */
3122 u32 rt_window; /* Window clamping */
3123 unsigned short rt_irtt; /* Initial RTT */
3126 struct in6_rtmsg32 {
3127 struct in6_addr rtmsg_dst;
3128 struct in6_addr rtmsg_src;
3129 struct in6_addr rtmsg_gateway;
3139 static int routing_ioctl(struct net *net, struct socket *sock,
3140 unsigned int cmd, void __user *argp)
3144 struct in6_rtmsg r6;
3148 mm_segment_t old_fs = get_fs();
3150 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3151 struct in6_rtmsg32 __user *ur6 = argp;
3152 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3153 3 * sizeof(struct in6_addr));
3154 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3155 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3156 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3157 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3158 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3159 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3160 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3164 struct rtentry32 __user *ur4 = argp;
3165 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3166 3 * sizeof(struct sockaddr));
3167 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3168 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3169 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3170 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3171 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3172 ret |= get_user(rtdev, &(ur4->rt_dev));
3174 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3175 r4.rt_dev = (char __user __force *)devname;
3189 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3196 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3197 * for some operations; this forces use of the newer bridge-utils that
3198 * use compatible ioctls
3200 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3204 if (get_user(tmp, argp))
3206 if (tmp == BRCTL_GET_VERSION)
3207 return BRCTL_VERSION + 1;
3211 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3212 unsigned int cmd, unsigned long arg)
3214 void __user *argp = compat_ptr(arg);
3215 struct sock *sk = sock->sk;
3216 struct net *net = sock_net(sk);
3218 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3219 return compat_ifr_data_ioctl(net, cmd, argp);
3224 return old_bridge_ioctl(argp);
3226 return dev_ifname32(net, argp);
3228 return dev_ifconf(net, argp);
3230 return ethtool_ioctl(net, argp);
3232 return compat_siocwandev(net, argp);
3235 return compat_sioc_ifmap(net, cmd, argp);
3236 case SIOCBONDENSLAVE:
3237 case SIOCBONDRELEASE:
3238 case SIOCBONDSETHWADDR:
3239 case SIOCBONDCHANGEACTIVE:
3240 return bond_ioctl(net, cmd, argp);
3243 return routing_ioctl(net, sock, cmd, argp);
3245 return do_siocgstamp(net, sock, cmd, argp);
3247 return do_siocgstampns(net, sock, cmd, argp);
3248 case SIOCBONDSLAVEINFOQUERY:
3249 case SIOCBONDINFOQUERY:
3252 return compat_ifr_data_ioctl(net, cmd, argp);
3264 return sock_ioctl(file, cmd, arg);
3281 case SIOCSIFHWBROADCAST:
3283 case SIOCGIFBRDADDR:
3284 case SIOCSIFBRDADDR:
3285 case SIOCGIFDSTADDR:
3286 case SIOCSIFDSTADDR:
3287 case SIOCGIFNETMASK:
3288 case SIOCSIFNETMASK:
3299 return dev_ifsioc(net, sock, cmd, argp);
3305 return sock_do_ioctl(net, sock, cmd, arg);
3308 return -ENOIOCTLCMD;
3311 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3314 struct socket *sock = file->private_data;
3315 int ret = -ENOIOCTLCMD;
3322 if (sock->ops->compat_ioctl)
3323 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3325 if (ret == -ENOIOCTLCMD &&
3326 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3327 ret = compat_wext_handle_ioctl(net, cmd, arg);
3329 if (ret == -ENOIOCTLCMD)
3330 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3336 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3338 return sock->ops->bind(sock, addr, addrlen);
3340 EXPORT_SYMBOL(kernel_bind);
3342 int kernel_listen(struct socket *sock, int backlog)
3344 return sock->ops->listen(sock, backlog);
3346 EXPORT_SYMBOL(kernel_listen);
3348 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3350 struct sock *sk = sock->sk;
3353 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3358 err = sock->ops->accept(sock, *newsock, flags);
3360 sock_release(*newsock);
3365 (*newsock)->ops = sock->ops;
3366 __module_get((*newsock)->ops->owner);
3371 EXPORT_SYMBOL(kernel_accept);
3373 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3376 return sock->ops->connect(sock, addr, addrlen, flags);
3378 EXPORT_SYMBOL(kernel_connect);
3380 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3383 return sock->ops->getname(sock, addr, addrlen, 0);
3385 EXPORT_SYMBOL(kernel_getsockname);
3387 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3390 return sock->ops->getname(sock, addr, addrlen, 1);
3392 EXPORT_SYMBOL(kernel_getpeername);
3394 int kernel_getsockopt(struct socket *sock, int level, int optname,
3395 char *optval, int *optlen)
3397 mm_segment_t oldfs = get_fs();
3398 char __user *uoptval;
3399 int __user *uoptlen;
3402 uoptval = (char __user __force *) optval;
3403 uoptlen = (int __user __force *) optlen;
3406 if (level == SOL_SOCKET)
3407 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3409 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3414 EXPORT_SYMBOL(kernel_getsockopt);
3416 int kernel_setsockopt(struct socket *sock, int level, int optname,
3417 char *optval, unsigned int optlen)
3419 mm_segment_t oldfs = get_fs();
3420 char __user *uoptval;
3423 uoptval = (char __user __force *) optval;
3426 if (level == SOL_SOCKET)
3427 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3429 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3434 EXPORT_SYMBOL(kernel_setsockopt);
3436 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3437 size_t size, int flags)
3439 if (sock->ops->sendpage)
3440 return sock->ops->sendpage(sock, page, offset, size, flags);
3442 return sock_no_sendpage(sock, page, offset, size, flags);
3444 EXPORT_SYMBOL(kernel_sendpage);
3446 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3448 mm_segment_t oldfs = get_fs();
3452 err = sock->ops->ioctl(sock, cmd, arg);
3457 EXPORT_SYMBOL(kernel_sock_ioctl);
3459 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3461 return sock->ops->shutdown(sock, how);
3463 EXPORT_SYMBOL(kernel_sock_shutdown);