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)
453 struct fd f = fdget(fd);
458 sock = sock_from_file(f.file, err);
460 *fput_needed = f.flags;
468 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
469 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
470 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
471 static ssize_t sockfs_getxattr(struct dentry *dentry,
472 const char *name, void *value, size_t size)
474 const char *proto_name;
479 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
480 proto_name = dentry->d_name.name;
481 proto_size = strlen(proto_name);
485 if (proto_size + 1 > size)
488 strncpy(value, proto_name, proto_size + 1);
490 error = proto_size + 1;
497 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
503 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
513 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
518 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
525 static const struct inode_operations sockfs_inode_ops = {
526 .getxattr = sockfs_getxattr,
527 .listxattr = sockfs_listxattr,
531 * sock_alloc - allocate a socket
533 * Allocate a new inode and socket object. The two are bound together
534 * and initialised. The socket is then returned. If we are out of inodes
538 static struct socket *sock_alloc(void)
543 inode = new_inode_pseudo(sock_mnt->mnt_sb);
547 sock = SOCKET_I(inode);
549 kmemcheck_annotate_bitfield(sock, type);
550 inode->i_ino = get_next_ino();
551 inode->i_mode = S_IFSOCK | S_IRWXUGO;
552 inode->i_uid = current_fsuid();
553 inode->i_gid = current_fsgid();
554 inode->i_op = &sockfs_inode_ops;
556 this_cpu_add(sockets_in_use, 1);
561 * In theory you can't get an open on this inode, but /proc provides
562 * a back door. Remember to keep it shut otherwise you'll let the
563 * creepy crawlies in.
566 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
571 const struct file_operations bad_sock_fops = {
572 .owner = THIS_MODULE,
573 .open = sock_no_open,
574 .llseek = noop_llseek,
578 * sock_release - close a socket
579 * @sock: socket to close
581 * The socket is released from the protocol stack if it has a release
582 * callback, and the inode is then released if the socket is bound to
583 * an inode not a file.
586 void sock_release(struct socket *sock)
589 struct module *owner = sock->ops->owner;
591 sock->ops->release(sock);
596 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
597 printk(KERN_ERR "sock_release: fasync list not empty!\n");
599 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
602 this_cpu_sub(sockets_in_use, 1);
604 iput(SOCK_INODE(sock));
609 EXPORT_SYMBOL(sock_release);
611 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
614 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
615 *tx_flags |= SKBTX_HW_TSTAMP;
616 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
617 *tx_flags |= SKBTX_SW_TSTAMP;
618 if (sock_flag(sk, SOCK_WIFI_STATUS))
619 *tx_flags |= SKBTX_WIFI_STATUS;
621 EXPORT_SYMBOL(sock_tx_timestamp);
623 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
624 struct msghdr *msg, size_t size)
626 struct sock_iocb *si = kiocb_to_siocb(iocb);
633 return sock->ops->sendmsg(iocb, sock, msg, size);
636 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
637 struct msghdr *msg, size_t size)
639 int err = security_socket_sendmsg(sock, msg, size);
641 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
644 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
647 struct sock_iocb siocb;
650 init_sync_kiocb(&iocb, NULL);
651 iocb.private = &siocb;
652 ret = __sock_sendmsg(&iocb, sock, msg, size);
653 if (-EIOCBQUEUED == ret)
654 ret = wait_on_sync_kiocb(&iocb);
657 EXPORT_SYMBOL(sock_sendmsg);
659 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
662 struct sock_iocb siocb;
665 init_sync_kiocb(&iocb, NULL);
666 iocb.private = &siocb;
667 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
668 if (-EIOCBQUEUED == ret)
669 ret = wait_on_sync_kiocb(&iocb);
673 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
674 struct kvec *vec, size_t num, size_t size)
676 mm_segment_t oldfs = get_fs();
681 * the following is safe, since for compiler definitions of kvec and
682 * iovec are identical, yielding the same in-core layout and alignment
684 msg->msg_iov = (struct iovec *)vec;
685 msg->msg_iovlen = num;
686 result = sock_sendmsg(sock, msg, size);
690 EXPORT_SYMBOL(kernel_sendmsg);
693 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
695 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
698 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
699 struct timespec ts[3];
701 struct skb_shared_hwtstamps *shhwtstamps =
704 /* Race occurred between timestamp enabling and packet
705 receiving. Fill in the current time for now. */
706 if (need_software_tstamp && skb->tstamp.tv64 == 0)
707 __net_timestamp(skb);
709 if (need_software_tstamp) {
710 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
712 skb_get_timestamp(skb, &tv);
713 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
716 skb_get_timestampns(skb, &ts[0]);
717 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
718 sizeof(ts[0]), &ts[0]);
723 memset(ts, 0, sizeof(ts));
724 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
725 ktime_to_timespec_cond(skb->tstamp, ts + 0))
728 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
729 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
731 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
732 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
736 put_cmsg(msg, SOL_SOCKET,
737 SCM_TIMESTAMPING, sizeof(ts), &ts);
739 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
741 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
746 if (!sock_flag(sk, SOCK_WIFI_STATUS))
748 if (!skb->wifi_acked_valid)
751 ack = skb->wifi_acked;
753 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
755 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
757 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
760 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
761 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
762 sizeof(__u32), &skb->dropcount);
765 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
768 sock_recv_timestamp(msg, sk, skb);
769 sock_recv_drops(msg, sk, skb);
771 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
773 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
774 struct msghdr *msg, size_t size, int flags)
776 struct sock_iocb *si = kiocb_to_siocb(iocb);
784 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
787 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
788 struct msghdr *msg, size_t size, int flags)
790 int err = security_socket_recvmsg(sock, msg, size, flags);
792 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
795 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
796 size_t size, int flags)
799 struct sock_iocb siocb;
802 init_sync_kiocb(&iocb, NULL);
803 iocb.private = &siocb;
804 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
805 if (-EIOCBQUEUED == ret)
806 ret = wait_on_sync_kiocb(&iocb);
809 EXPORT_SYMBOL(sock_recvmsg);
811 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
812 size_t size, int flags)
815 struct sock_iocb siocb;
818 init_sync_kiocb(&iocb, NULL);
819 iocb.private = &siocb;
820 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
821 if (-EIOCBQUEUED == ret)
822 ret = wait_on_sync_kiocb(&iocb);
827 * kernel_recvmsg - Receive a message from a socket (kernel space)
828 * @sock: The socket to receive the message from
829 * @msg: Received message
830 * @vec: Input s/g array for message data
831 * @num: Size of input s/g array
832 * @size: Number of bytes to read
833 * @flags: Message flags (MSG_DONTWAIT, etc...)
835 * On return the msg structure contains the scatter/gather array passed in the
836 * vec argument. The array is modified so that it consists of the unfilled
837 * portion of the original array.
839 * The returned value is the total number of bytes received, or an error.
841 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
842 struct kvec *vec, size_t num, size_t size, int flags)
844 mm_segment_t oldfs = get_fs();
849 * the following is safe, since for compiler definitions of kvec and
850 * iovec are identical, yielding the same in-core layout and alignment
852 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
853 result = sock_recvmsg(sock, msg, size, flags);
857 EXPORT_SYMBOL(kernel_recvmsg);
859 static ssize_t sock_sendpage(struct file *file, struct page *page,
860 int offset, size_t size, loff_t *ppos, int more)
865 sock = file->private_data;
867 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
868 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
871 return kernel_sendpage(sock, page, offset, size, flags);
874 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
875 struct pipe_inode_info *pipe, size_t len,
878 struct socket *sock = file->private_data;
880 if (unlikely(!sock->ops->splice_read))
883 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
886 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
887 struct sock_iocb *siocb)
889 if (!is_sync_kiocb(iocb))
893 iocb->private = siocb;
897 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
898 struct file *file, const struct iovec *iov,
899 unsigned long nr_segs)
901 struct socket *sock = file->private_data;
905 for (i = 0; i < nr_segs; i++)
906 size += iov[i].iov_len;
908 msg->msg_name = NULL;
909 msg->msg_namelen = 0;
910 msg->msg_control = NULL;
911 msg->msg_controllen = 0;
912 msg->msg_iov = (struct iovec *)iov;
913 msg->msg_iovlen = nr_segs;
914 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
916 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
919 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
920 unsigned long nr_segs, loff_t pos)
922 struct sock_iocb siocb, *x;
927 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
931 x = alloc_sock_iocb(iocb, &siocb);
934 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
937 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
938 struct file *file, const struct iovec *iov,
939 unsigned long nr_segs)
941 struct socket *sock = file->private_data;
945 for (i = 0; i < nr_segs; i++)
946 size += iov[i].iov_len;
948 msg->msg_name = NULL;
949 msg->msg_namelen = 0;
950 msg->msg_control = NULL;
951 msg->msg_controllen = 0;
952 msg->msg_iov = (struct iovec *)iov;
953 msg->msg_iovlen = nr_segs;
954 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
955 if (sock->type == SOCK_SEQPACKET)
956 msg->msg_flags |= MSG_EOR;
958 return __sock_sendmsg(iocb, sock, msg, size);
961 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
962 unsigned long nr_segs, loff_t pos)
964 struct sock_iocb siocb, *x;
969 x = alloc_sock_iocb(iocb, &siocb);
973 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
977 * Atomic setting of ioctl hooks to avoid race
978 * with module unload.
981 static DEFINE_MUTEX(br_ioctl_mutex);
982 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
984 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
986 mutex_lock(&br_ioctl_mutex);
987 br_ioctl_hook = hook;
988 mutex_unlock(&br_ioctl_mutex);
990 EXPORT_SYMBOL(brioctl_set);
992 static DEFINE_MUTEX(vlan_ioctl_mutex);
993 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
995 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
997 mutex_lock(&vlan_ioctl_mutex);
998 vlan_ioctl_hook = hook;
999 mutex_unlock(&vlan_ioctl_mutex);
1001 EXPORT_SYMBOL(vlan_ioctl_set);
1003 static DEFINE_MUTEX(dlci_ioctl_mutex);
1004 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1006 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1008 mutex_lock(&dlci_ioctl_mutex);
1009 dlci_ioctl_hook = hook;
1010 mutex_unlock(&dlci_ioctl_mutex);
1012 EXPORT_SYMBOL(dlci_ioctl_set);
1014 static long sock_do_ioctl(struct net *net, struct socket *sock,
1015 unsigned int cmd, unsigned long arg)
1018 void __user *argp = (void __user *)arg;
1020 err = sock->ops->ioctl(sock, cmd, arg);
1023 * If this ioctl is unknown try to hand it down
1024 * to the NIC driver.
1026 if (err == -ENOIOCTLCMD)
1027 err = dev_ioctl(net, cmd, argp);
1033 * With an ioctl, arg may well be a user mode pointer, but we don't know
1034 * what to do with it - that's up to the protocol still.
1037 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1039 struct socket *sock;
1041 void __user *argp = (void __user *)arg;
1045 sock = file->private_data;
1048 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1049 err = dev_ioctl(net, cmd, argp);
1051 #ifdef CONFIG_WEXT_CORE
1052 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1053 err = dev_ioctl(net, cmd, argp);
1060 if (get_user(pid, (int __user *)argp))
1062 err = f_setown(sock->file, pid, 1);
1066 err = put_user(f_getown(sock->file),
1067 (int __user *)argp);
1075 request_module("bridge");
1077 mutex_lock(&br_ioctl_mutex);
1079 err = br_ioctl_hook(net, cmd, argp);
1080 mutex_unlock(&br_ioctl_mutex);
1085 if (!vlan_ioctl_hook)
1086 request_module("8021q");
1088 mutex_lock(&vlan_ioctl_mutex);
1089 if (vlan_ioctl_hook)
1090 err = vlan_ioctl_hook(net, argp);
1091 mutex_unlock(&vlan_ioctl_mutex);
1096 if (!dlci_ioctl_hook)
1097 request_module("dlci");
1099 mutex_lock(&dlci_ioctl_mutex);
1100 if (dlci_ioctl_hook)
1101 err = dlci_ioctl_hook(cmd, argp);
1102 mutex_unlock(&dlci_ioctl_mutex);
1105 err = sock_do_ioctl(net, sock, cmd, arg);
1111 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1114 struct socket *sock = NULL;
1116 err = security_socket_create(family, type, protocol, 1);
1120 sock = sock_alloc();
1127 err = security_socket_post_create(sock, family, type, protocol, 1);
1139 EXPORT_SYMBOL(sock_create_lite);
1141 /* No kernel lock held - perfect */
1142 static unsigned int sock_poll(struct file *file, poll_table *wait)
1144 unsigned int busy_flag = 0;
1145 struct socket *sock;
1148 * We can't return errors to poll, so it's either yes or no.
1150 sock = file->private_data;
1152 if (sk_can_busy_loop(sock->sk)) {
1153 /* this socket can poll_ll so tell the system call */
1154 busy_flag = POLL_BUSY_LOOP;
1156 /* once, only if requested by syscall */
1157 if (wait && (wait->_key & POLL_BUSY_LOOP))
1158 sk_busy_loop(sock->sk, 1);
1161 return busy_flag | sock->ops->poll(file, sock, wait);
1164 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1166 struct socket *sock = file->private_data;
1168 return sock->ops->mmap(file, sock, vma);
1171 static int sock_close(struct inode *inode, struct file *filp)
1173 sock_release(SOCKET_I(inode));
1178 * Update the socket async list
1180 * Fasync_list locking strategy.
1182 * 1. fasync_list is modified only under process context socket lock
1183 * i.e. under semaphore.
1184 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1185 * or under socket lock
1188 static int sock_fasync(int fd, struct file *filp, int on)
1190 struct socket *sock = filp->private_data;
1191 struct sock *sk = sock->sk;
1192 struct socket_wq *wq;
1198 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1199 fasync_helper(fd, filp, on, &wq->fasync_list);
1201 if (!wq->fasync_list)
1202 sock_reset_flag(sk, SOCK_FASYNC);
1204 sock_set_flag(sk, SOCK_FASYNC);
1210 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1212 int sock_wake_async(struct socket *sock, int how, int band)
1214 struct socket_wq *wq;
1219 wq = rcu_dereference(sock->wq);
1220 if (!wq || !wq->fasync_list) {
1225 case SOCK_WAKE_WAITD:
1226 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1229 case SOCK_WAKE_SPACE:
1230 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1235 kill_fasync(&wq->fasync_list, SIGIO, band);
1238 kill_fasync(&wq->fasync_list, SIGURG, band);
1243 EXPORT_SYMBOL(sock_wake_async);
1245 int __sock_create(struct net *net, int family, int type, int protocol,
1246 struct socket **res, int kern)
1249 struct socket *sock;
1250 const struct net_proto_family *pf;
1253 * Check protocol is in range
1255 if (family < 0 || family >= NPROTO)
1256 return -EAFNOSUPPORT;
1257 if (type < 0 || type >= SOCK_MAX)
1262 This uglymoron is moved from INET layer to here to avoid
1263 deadlock in module load.
1265 if (family == PF_INET && type == SOCK_PACKET) {
1269 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1275 err = security_socket_create(family, type, protocol, kern);
1280 * Allocate the socket and allow the family to set things up. if
1281 * the protocol is 0, the family is instructed to select an appropriate
1284 sock = sock_alloc();
1286 net_warn_ratelimited("socket: no more sockets\n");
1287 return -ENFILE; /* Not exactly a match, but its the
1288 closest posix thing */
1293 #ifdef CONFIG_MODULES
1294 /* Attempt to load a protocol module if the find failed.
1296 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1297 * requested real, full-featured networking support upon configuration.
1298 * Otherwise module support will break!
1300 if (rcu_access_pointer(net_families[family]) == NULL)
1301 request_module("net-pf-%d", family);
1305 pf = rcu_dereference(net_families[family]);
1306 err = -EAFNOSUPPORT;
1311 * We will call the ->create function, that possibly is in a loadable
1312 * module, so we have to bump that loadable module refcnt first.
1314 if (!try_module_get(pf->owner))
1317 /* Now protected by module ref count */
1320 err = pf->create(net, sock, protocol, kern);
1322 goto out_module_put;
1325 * Now to bump the refcnt of the [loadable] module that owns this
1326 * socket at sock_release time we decrement its refcnt.
1328 if (!try_module_get(sock->ops->owner))
1329 goto out_module_busy;
1332 * Now that we're done with the ->create function, the [loadable]
1333 * module can have its refcnt decremented
1335 module_put(pf->owner);
1336 err = security_socket_post_create(sock, family, type, protocol, kern);
1338 goto out_sock_release;
1344 err = -EAFNOSUPPORT;
1347 module_put(pf->owner);
1354 goto out_sock_release;
1356 EXPORT_SYMBOL(__sock_create);
1358 int sock_create(int family, int type, int protocol, struct socket **res)
1360 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1362 EXPORT_SYMBOL(sock_create);
1364 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1366 return __sock_create(&init_net, family, type, protocol, res, 1);
1368 EXPORT_SYMBOL(sock_create_kern);
1370 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1373 struct socket *sock;
1376 /* Check the SOCK_* constants for consistency. */
1377 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1378 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1379 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1380 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1382 flags = type & ~SOCK_TYPE_MASK;
1383 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1385 type &= SOCK_TYPE_MASK;
1387 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1388 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1390 retval = sock_create(family, type, protocol, &sock);
1394 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1399 /* It may be already another descriptor 8) Not kernel problem. */
1408 * Create a pair of connected sockets.
1411 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1412 int __user *, usockvec)
1414 struct socket *sock1, *sock2;
1416 struct file *newfile1, *newfile2;
1419 flags = type & ~SOCK_TYPE_MASK;
1420 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1422 type &= SOCK_TYPE_MASK;
1424 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1425 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1428 * Obtain the first socket and check if the underlying protocol
1429 * supports the socketpair call.
1432 err = sock_create(family, type, protocol, &sock1);
1436 err = sock_create(family, type, protocol, &sock2);
1440 err = sock1->ops->socketpair(sock1, sock2);
1442 goto out_release_both;
1444 fd1 = get_unused_fd_flags(flags);
1445 if (unlikely(fd1 < 0)) {
1447 goto out_release_both;
1450 fd2 = get_unused_fd_flags(flags);
1451 if (unlikely(fd2 < 0)) {
1453 goto out_put_unused_1;
1456 newfile1 = sock_alloc_file(sock1, flags, NULL);
1457 if (unlikely(IS_ERR(newfile1))) {
1458 err = PTR_ERR(newfile1);
1459 goto out_put_unused_both;
1462 newfile2 = sock_alloc_file(sock2, flags, NULL);
1463 if (IS_ERR(newfile2)) {
1464 err = PTR_ERR(newfile2);
1468 err = put_user(fd1, &usockvec[0]);
1472 err = put_user(fd2, &usockvec[1]);
1476 audit_fd_pair(fd1, fd2);
1478 fd_install(fd1, newfile1);
1479 fd_install(fd2, newfile2);
1480 /* fd1 and fd2 may be already another descriptors.
1481 * Not kernel problem.
1497 sock_release(sock2);
1500 out_put_unused_both:
1505 sock_release(sock2);
1507 sock_release(sock1);
1513 * Bind a name to a socket. Nothing much to do here since it's
1514 * the protocol's responsibility to handle the local address.
1516 * We move the socket address to kernel space before we call
1517 * the protocol layer (having also checked the address is ok).
1520 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1522 struct socket *sock;
1523 struct sockaddr_storage address;
1524 int err, fput_needed;
1526 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1528 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1530 err = security_socket_bind(sock,
1531 (struct sockaddr *)&address,
1534 err = sock->ops->bind(sock,
1538 fput_light(sock->file, fput_needed);
1544 * Perform a listen. Basically, we allow the protocol to do anything
1545 * necessary for a listen, and if that works, we mark the socket as
1546 * ready for listening.
1549 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1551 struct socket *sock;
1552 int err, fput_needed;
1555 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1558 if ((unsigned int)backlog > somaxconn)
1559 backlog = somaxconn;
1561 err = security_socket_listen(sock, backlog);
1563 err = sock->ops->listen(sock, backlog);
1565 fput_light(sock->file, fput_needed);
1571 * For accept, we attempt to create a new socket, set up the link
1572 * with the client, wake up the client, then return the new
1573 * connected fd. We collect the address of the connector in kernel
1574 * space and move it to user at the very end. This is unclean because
1575 * we open the socket then return an error.
1577 * 1003.1g adds the ability to recvmsg() to query connection pending
1578 * status to recvmsg. We need to add that support in a way thats
1579 * clean when we restucture accept also.
1582 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1583 int __user *, upeer_addrlen, int, flags)
1585 struct socket *sock, *newsock;
1586 struct file *newfile;
1587 int err, len, newfd, fput_needed;
1588 struct sockaddr_storage address;
1590 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1593 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1594 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1596 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1601 newsock = sock_alloc();
1605 newsock->type = sock->type;
1606 newsock->ops = sock->ops;
1609 * We don't need try_module_get here, as the listening socket (sock)
1610 * has the protocol module (sock->ops->owner) held.
1612 __module_get(newsock->ops->owner);
1614 newfd = get_unused_fd_flags(flags);
1615 if (unlikely(newfd < 0)) {
1617 sock_release(newsock);
1620 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1621 if (unlikely(IS_ERR(newfile))) {
1622 err = PTR_ERR(newfile);
1623 put_unused_fd(newfd);
1624 sock_release(newsock);
1628 err = security_socket_accept(sock, newsock);
1632 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1636 if (upeer_sockaddr) {
1637 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1639 err = -ECONNABORTED;
1642 err = move_addr_to_user(&address,
1643 len, upeer_sockaddr, upeer_addrlen);
1648 /* File flags are not inherited via accept() unlike another OSes. */
1650 fd_install(newfd, newfile);
1654 fput_light(sock->file, fput_needed);
1659 put_unused_fd(newfd);
1663 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1664 int __user *, upeer_addrlen)
1666 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1670 * Attempt to connect to a socket with the server address. The address
1671 * is in user space so we verify it is OK and move it to kernel space.
1673 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1676 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1677 * other SEQPACKET protocols that take time to connect() as it doesn't
1678 * include the -EINPROGRESS status for such sockets.
1681 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1684 struct socket *sock;
1685 struct sockaddr_storage address;
1686 int err, fput_needed;
1688 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1691 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1696 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1700 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1701 sock->file->f_flags);
1703 fput_light(sock->file, fput_needed);
1709 * Get the local address ('name') of a socket object. Move the obtained
1710 * name to user space.
1713 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1714 int __user *, usockaddr_len)
1716 struct socket *sock;
1717 struct sockaddr_storage address;
1718 int len, err, fput_needed;
1720 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1724 err = security_socket_getsockname(sock);
1728 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1731 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1734 fput_light(sock->file, fput_needed);
1740 * Get the remote address ('name') of a socket object. Move the obtained
1741 * name to user space.
1744 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1745 int __user *, usockaddr_len)
1747 struct socket *sock;
1748 struct sockaddr_storage address;
1749 int len, err, fput_needed;
1751 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1753 err = security_socket_getpeername(sock);
1755 fput_light(sock->file, fput_needed);
1760 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1763 err = move_addr_to_user(&address, len, usockaddr,
1765 fput_light(sock->file, fput_needed);
1771 * Send a datagram to a given address. We move the address into kernel
1772 * space and check the user space data area is readable before invoking
1776 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1777 unsigned int, flags, struct sockaddr __user *, addr,
1780 struct socket *sock;
1781 struct sockaddr_storage address;
1789 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1793 iov.iov_base = buff;
1795 msg.msg_name = NULL;
1798 msg.msg_control = NULL;
1799 msg.msg_controllen = 0;
1800 msg.msg_namelen = 0;
1802 err = move_addr_to_kernel(addr, addr_len, &address);
1805 msg.msg_name = (struct sockaddr *)&address;
1806 msg.msg_namelen = addr_len;
1808 if (sock->file->f_flags & O_NONBLOCK)
1809 flags |= MSG_DONTWAIT;
1810 msg.msg_flags = flags;
1811 err = sock_sendmsg(sock, &msg, len);
1814 fput_light(sock->file, fput_needed);
1820 * Send a datagram down a socket.
1823 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1824 unsigned int, flags)
1826 return sys_sendto(fd, buff, len, flags, NULL, 0);
1830 * Receive a frame from the socket and optionally record the address of the
1831 * sender. We verify the buffers are writable and if needed move the
1832 * sender address from kernel to user space.
1835 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1836 unsigned int, flags, struct sockaddr __user *, addr,
1837 int __user *, addr_len)
1839 struct socket *sock;
1842 struct sockaddr_storage address;
1848 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1852 msg.msg_control = NULL;
1853 msg.msg_controllen = 0;
1857 iov.iov_base = ubuf;
1858 /* Save some cycles and don't copy the address if not needed */
1859 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1860 /* We assume all kernel code knows the size of sockaddr_storage */
1861 msg.msg_namelen = 0;
1862 if (sock->file->f_flags & O_NONBLOCK)
1863 flags |= MSG_DONTWAIT;
1864 err = sock_recvmsg(sock, &msg, size, flags);
1866 if (err >= 0 && addr != NULL) {
1867 err2 = move_addr_to_user(&address,
1868 msg.msg_namelen, addr, addr_len);
1873 fput_light(sock->file, fput_needed);
1879 * Receive a datagram from a socket.
1882 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1885 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1889 * Set a socket option. Because we don't know the option lengths we have
1890 * to pass the user mode parameter for the protocols to sort out.
1893 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1894 char __user *, optval, int, optlen)
1896 int err, fput_needed;
1897 struct socket *sock;
1902 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1904 err = security_socket_setsockopt(sock, level, optname);
1908 if (level == SOL_SOCKET)
1910 sock_setsockopt(sock, level, optname, optval,
1914 sock->ops->setsockopt(sock, level, optname, optval,
1917 fput_light(sock->file, fput_needed);
1923 * Get a socket option. Because we don't know the option lengths we have
1924 * to pass a user mode parameter for the protocols to sort out.
1927 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1928 char __user *, optval, int __user *, optlen)
1930 int err, fput_needed;
1931 struct socket *sock;
1933 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1935 err = security_socket_getsockopt(sock, level, optname);
1939 if (level == SOL_SOCKET)
1941 sock_getsockopt(sock, level, optname, optval,
1945 sock->ops->getsockopt(sock, level, optname, optval,
1948 fput_light(sock->file, fput_needed);
1954 * Shutdown a socket.
1957 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1959 int err, fput_needed;
1960 struct socket *sock;
1962 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1964 err = security_socket_shutdown(sock, how);
1966 err = sock->ops->shutdown(sock, how);
1967 fput_light(sock->file, fput_needed);
1972 /* A couple of helpful macros for getting the address of the 32/64 bit
1973 * fields which are the same type (int / unsigned) on our platforms.
1975 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1976 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1977 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1979 struct used_address {
1980 struct sockaddr_storage name;
1981 unsigned int name_len;
1984 static int copy_msghdr_from_user(struct msghdr *kmsg,
1985 struct msghdr __user *umsg)
1987 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1990 if (kmsg->msg_namelen < 0)
1993 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1994 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1998 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1999 struct msghdr *msg_sys, unsigned int flags,
2000 struct used_address *used_address)
2002 struct compat_msghdr __user *msg_compat =
2003 (struct compat_msghdr __user *)msg;
2004 struct sockaddr_storage address;
2005 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2006 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2007 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2008 /* 20 is size of ipv6_pktinfo */
2009 unsigned char *ctl_buf = ctl;
2010 int err, ctl_len, total_len;
2013 if (MSG_CMSG_COMPAT & flags) {
2014 if (get_compat_msghdr(msg_sys, msg_compat))
2017 err = copy_msghdr_from_user(msg_sys, msg);
2022 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2024 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2027 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2033 /* This will also move the address data into kernel space */
2034 if (MSG_CMSG_COMPAT & flags) {
2035 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2037 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2044 if (msg_sys->msg_controllen > INT_MAX)
2046 ctl_len = msg_sys->msg_controllen;
2047 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2049 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2053 ctl_buf = msg_sys->msg_control;
2054 ctl_len = msg_sys->msg_controllen;
2055 } else if (ctl_len) {
2056 if (ctl_len > sizeof(ctl)) {
2057 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2058 if (ctl_buf == NULL)
2063 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2064 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2065 * checking falls down on this.
2067 if (copy_from_user(ctl_buf,
2068 (void __user __force *)msg_sys->msg_control,
2071 msg_sys->msg_control = ctl_buf;
2073 msg_sys->msg_flags = flags;
2075 if (sock->file->f_flags & O_NONBLOCK)
2076 msg_sys->msg_flags |= MSG_DONTWAIT;
2078 * If this is sendmmsg() and current destination address is same as
2079 * previously succeeded address, omit asking LSM's decision.
2080 * used_address->name_len is initialized to UINT_MAX so that the first
2081 * destination address never matches.
2083 if (used_address && msg_sys->msg_name &&
2084 used_address->name_len == msg_sys->msg_namelen &&
2085 !memcmp(&used_address->name, msg_sys->msg_name,
2086 used_address->name_len)) {
2087 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2090 err = sock_sendmsg(sock, msg_sys, total_len);
2092 * If this is sendmmsg() and sending to current destination address was
2093 * successful, remember it.
2095 if (used_address && err >= 0) {
2096 used_address->name_len = msg_sys->msg_namelen;
2097 if (msg_sys->msg_name)
2098 memcpy(&used_address->name, msg_sys->msg_name,
2099 used_address->name_len);
2104 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2106 if (iov != iovstack)
2113 * BSD sendmsg interface
2116 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2118 int fput_needed, err;
2119 struct msghdr msg_sys;
2120 struct socket *sock;
2122 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2126 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2128 fput_light(sock->file, fput_needed);
2133 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2135 if (flags & MSG_CMSG_COMPAT)
2137 return __sys_sendmsg(fd, msg, flags);
2141 * Linux sendmmsg interface
2144 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2147 int fput_needed, err, datagrams;
2148 struct socket *sock;
2149 struct mmsghdr __user *entry;
2150 struct compat_mmsghdr __user *compat_entry;
2151 struct msghdr msg_sys;
2152 struct used_address used_address;
2154 if (vlen > UIO_MAXIOV)
2159 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2163 used_address.name_len = UINT_MAX;
2165 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2168 while (datagrams < vlen) {
2169 if (MSG_CMSG_COMPAT & flags) {
2170 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2171 &msg_sys, flags, &used_address);
2174 err = __put_user(err, &compat_entry->msg_len);
2177 err = ___sys_sendmsg(sock,
2178 (struct msghdr __user *)entry,
2179 &msg_sys, flags, &used_address);
2182 err = put_user(err, &entry->msg_len);
2191 fput_light(sock->file, fput_needed);
2193 /* We only return an error if no datagrams were able to be sent */
2200 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2201 unsigned int, vlen, unsigned int, flags)
2203 if (flags & MSG_CMSG_COMPAT)
2205 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2208 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2209 struct msghdr *msg_sys, unsigned int flags, int nosec)
2211 struct compat_msghdr __user *msg_compat =
2212 (struct compat_msghdr __user *)msg;
2213 struct iovec iovstack[UIO_FASTIOV];
2214 struct iovec *iov = iovstack;
2215 unsigned long cmsg_ptr;
2216 int err, total_len, len;
2218 /* kernel mode address */
2219 struct sockaddr_storage addr;
2221 /* user mode address pointers */
2222 struct sockaddr __user *uaddr;
2223 int __user *uaddr_len;
2225 if (MSG_CMSG_COMPAT & flags) {
2226 if (get_compat_msghdr(msg_sys, msg_compat))
2229 err = copy_msghdr_from_user(msg_sys, msg);
2234 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2236 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2239 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2245 /* Save the user-mode address (verify_iovec will change the
2246 * kernel msghdr to use the kernel address space)
2248 uaddr = (__force void __user *)msg_sys->msg_name;
2249 uaddr_len = COMPAT_NAMELEN(msg);
2250 if (MSG_CMSG_COMPAT & flags)
2251 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2253 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2258 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2259 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2261 /* We assume all kernel code knows the size of sockaddr_storage */
2262 msg_sys->msg_namelen = 0;
2264 if (sock->file->f_flags & O_NONBLOCK)
2265 flags |= MSG_DONTWAIT;
2266 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2272 if (uaddr != NULL) {
2273 err = move_addr_to_user(&addr,
2274 msg_sys->msg_namelen, uaddr,
2279 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2283 if (MSG_CMSG_COMPAT & flags)
2284 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2285 &msg_compat->msg_controllen);
2287 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2288 &msg->msg_controllen);
2294 if (iov != iovstack)
2301 * BSD recvmsg interface
2304 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2306 int fput_needed, err;
2307 struct msghdr msg_sys;
2308 struct socket *sock;
2310 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2314 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2316 fput_light(sock->file, fput_needed);
2321 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2322 unsigned int, flags)
2324 if (flags & MSG_CMSG_COMPAT)
2326 return __sys_recvmsg(fd, msg, flags);
2330 * Linux recvmmsg interface
2333 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2334 unsigned int flags, struct timespec *timeout)
2336 int fput_needed, err, datagrams;
2337 struct socket *sock;
2338 struct mmsghdr __user *entry;
2339 struct compat_mmsghdr __user *compat_entry;
2340 struct msghdr msg_sys;
2341 struct timespec end_time;
2344 poll_select_set_timeout(&end_time, timeout->tv_sec,
2350 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2354 err = sock_error(sock->sk);
2359 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2361 while (datagrams < vlen) {
2363 * No need to ask LSM for more than the first datagram.
2365 if (MSG_CMSG_COMPAT & flags) {
2366 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2367 &msg_sys, flags & ~MSG_WAITFORONE,
2371 err = __put_user(err, &compat_entry->msg_len);
2374 err = ___sys_recvmsg(sock,
2375 (struct msghdr __user *)entry,
2376 &msg_sys, flags & ~MSG_WAITFORONE,
2380 err = put_user(err, &entry->msg_len);
2388 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2389 if (flags & MSG_WAITFORONE)
2390 flags |= MSG_DONTWAIT;
2393 ktime_get_ts(timeout);
2394 *timeout = timespec_sub(end_time, *timeout);
2395 if (timeout->tv_sec < 0) {
2396 timeout->tv_sec = timeout->tv_nsec = 0;
2400 /* Timeout, return less than vlen datagrams */
2401 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2405 /* Out of band data, return right away */
2406 if (msg_sys.msg_flags & MSG_OOB)
2411 fput_light(sock->file, fput_needed);
2416 if (datagrams != 0) {
2418 * We may return less entries than requested (vlen) if the
2419 * sock is non block and there aren't enough datagrams...
2421 if (err != -EAGAIN) {
2423 * ... or if recvmsg returns an error after we
2424 * received some datagrams, where we record the
2425 * error to return on the next call or if the
2426 * app asks about it using getsockopt(SO_ERROR).
2428 sock->sk->sk_err = -err;
2437 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2438 unsigned int, vlen, unsigned int, flags,
2439 struct timespec __user *, timeout)
2442 struct timespec timeout_sys;
2444 if (flags & MSG_CMSG_COMPAT)
2448 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2450 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2453 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2455 if (datagrams > 0 &&
2456 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2457 datagrams = -EFAULT;
2462 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2463 /* Argument list sizes for sys_socketcall */
2464 #define AL(x) ((x) * sizeof(unsigned long))
2465 static const unsigned char nargs[21] = {
2466 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2467 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2468 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2475 * System call vectors.
2477 * Argument checking cleaned up. Saved 20% in size.
2478 * This function doesn't need to set the kernel lock because
2479 * it is set by the callees.
2482 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2484 unsigned long a[AUDITSC_ARGS];
2485 unsigned long a0, a1;
2489 if (call < 1 || call > SYS_SENDMMSG)
2493 if (len > sizeof(a))
2496 /* copy_from_user should be SMP safe. */
2497 if (copy_from_user(a, args, len))
2500 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2509 err = sys_socket(a0, a1, a[2]);
2512 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2515 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2518 err = sys_listen(a0, a1);
2521 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2522 (int __user *)a[2], 0);
2524 case SYS_GETSOCKNAME:
2526 sys_getsockname(a0, (struct sockaddr __user *)a1,
2527 (int __user *)a[2]);
2529 case SYS_GETPEERNAME:
2531 sys_getpeername(a0, (struct sockaddr __user *)a1,
2532 (int __user *)a[2]);
2534 case SYS_SOCKETPAIR:
2535 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2538 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2541 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2542 (struct sockaddr __user *)a[4], a[5]);
2545 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2548 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2549 (struct sockaddr __user *)a[4],
2550 (int __user *)a[5]);
2553 err = sys_shutdown(a0, a1);
2555 case SYS_SETSOCKOPT:
2556 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2558 case SYS_GETSOCKOPT:
2560 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2561 (int __user *)a[4]);
2564 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2567 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2570 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2573 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2574 (struct timespec __user *)a[4]);
2577 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2578 (int __user *)a[2], a[3]);
2587 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2590 * sock_register - add a socket protocol handler
2591 * @ops: description of protocol
2593 * This function is called by a protocol handler that wants to
2594 * advertise its address family, and have it linked into the
2595 * socket interface. The value ops->family coresponds to the
2596 * socket system call protocol family.
2598 int sock_register(const struct net_proto_family *ops)
2602 if (ops->family >= NPROTO) {
2603 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2608 spin_lock(&net_family_lock);
2609 if (rcu_dereference_protected(net_families[ops->family],
2610 lockdep_is_held(&net_family_lock)))
2613 rcu_assign_pointer(net_families[ops->family], ops);
2616 spin_unlock(&net_family_lock);
2618 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2621 EXPORT_SYMBOL(sock_register);
2624 * sock_unregister - remove a protocol handler
2625 * @family: protocol family to remove
2627 * This function is called by a protocol handler that wants to
2628 * remove its address family, and have it unlinked from the
2629 * new socket creation.
2631 * If protocol handler is a module, then it can use module reference
2632 * counts to protect against new references. If protocol handler is not
2633 * a module then it needs to provide its own protection in
2634 * the ops->create routine.
2636 void sock_unregister(int family)
2638 BUG_ON(family < 0 || family >= NPROTO);
2640 spin_lock(&net_family_lock);
2641 RCU_INIT_POINTER(net_families[family], NULL);
2642 spin_unlock(&net_family_lock);
2646 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2648 EXPORT_SYMBOL(sock_unregister);
2650 static int __init sock_init(void)
2654 * Initialize the network sysctl infrastructure.
2656 err = net_sysctl_init();
2661 * Initialize skbuff SLAB cache
2666 * Initialize the protocols module.
2671 err = register_filesystem(&sock_fs_type);
2674 sock_mnt = kern_mount(&sock_fs_type);
2675 if (IS_ERR(sock_mnt)) {
2676 err = PTR_ERR(sock_mnt);
2680 /* The real protocol initialization is performed in later initcalls.
2683 #ifdef CONFIG_NETFILTER
2684 err = netfilter_init();
2689 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2690 skb_timestamping_init();
2697 unregister_filesystem(&sock_fs_type);
2702 core_initcall(sock_init); /* early initcall */
2704 #ifdef CONFIG_PROC_FS
2705 void socket_seq_show(struct seq_file *seq)
2710 for_each_possible_cpu(cpu)
2711 counter += per_cpu(sockets_in_use, cpu);
2713 /* It can be negative, by the way. 8) */
2717 seq_printf(seq, "sockets: used %d\n", counter);
2719 #endif /* CONFIG_PROC_FS */
2721 #ifdef CONFIG_COMPAT
2722 static int do_siocgstamp(struct net *net, struct socket *sock,
2723 unsigned int cmd, void __user *up)
2725 mm_segment_t old_fs = get_fs();
2730 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2733 err = compat_put_timeval(&ktv, up);
2738 static int do_siocgstampns(struct net *net, struct socket *sock,
2739 unsigned int cmd, void __user *up)
2741 mm_segment_t old_fs = get_fs();
2742 struct timespec kts;
2746 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2749 err = compat_put_timespec(&kts, up);
2754 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2756 struct ifreq __user *uifr;
2759 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2760 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2763 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2767 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2773 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2775 struct compat_ifconf ifc32;
2777 struct ifconf __user *uifc;
2778 struct compat_ifreq __user *ifr32;
2779 struct ifreq __user *ifr;
2783 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2786 memset(&ifc, 0, sizeof(ifc));
2787 if (ifc32.ifcbuf == 0) {
2791 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2793 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2794 sizeof(struct ifreq);
2795 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2797 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2798 ifr32 = compat_ptr(ifc32.ifcbuf);
2799 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2800 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2806 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2809 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2813 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2817 ifr32 = compat_ptr(ifc32.ifcbuf);
2819 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2820 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2821 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2827 if (ifc32.ifcbuf == 0) {
2828 /* Translate from 64-bit structure multiple to
2832 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2837 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2843 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2845 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2846 bool convert_in = false, convert_out = false;
2847 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2848 struct ethtool_rxnfc __user *rxnfc;
2849 struct ifreq __user *ifr;
2850 u32 rule_cnt = 0, actual_rule_cnt;
2855 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2858 compat_rxnfc = compat_ptr(data);
2860 if (get_user(ethcmd, &compat_rxnfc->cmd))
2863 /* Most ethtool structures are defined without padding.
2864 * Unfortunately struct ethtool_rxnfc is an exception.
2869 case ETHTOOL_GRXCLSRLALL:
2870 /* Buffer size is variable */
2871 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2873 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2875 buf_size += rule_cnt * sizeof(u32);
2877 case ETHTOOL_GRXRINGS:
2878 case ETHTOOL_GRXCLSRLCNT:
2879 case ETHTOOL_GRXCLSRULE:
2880 case ETHTOOL_SRXCLSRLINS:
2883 case ETHTOOL_SRXCLSRLDEL:
2884 buf_size += sizeof(struct ethtool_rxnfc);
2889 ifr = compat_alloc_user_space(buf_size);
2890 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2892 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2895 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2896 &ifr->ifr_ifru.ifru_data))
2900 /* We expect there to be holes between fs.m_ext and
2901 * fs.ring_cookie and at the end of fs, but nowhere else.
2903 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2904 sizeof(compat_rxnfc->fs.m_ext) !=
2905 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2906 sizeof(rxnfc->fs.m_ext));
2908 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2909 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2910 offsetof(struct ethtool_rxnfc, fs.location) -
2911 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2913 if (copy_in_user(rxnfc, compat_rxnfc,
2914 (void __user *)(&rxnfc->fs.m_ext + 1) -
2915 (void __user *)rxnfc) ||
2916 copy_in_user(&rxnfc->fs.ring_cookie,
2917 &compat_rxnfc->fs.ring_cookie,
2918 (void __user *)(&rxnfc->fs.location + 1) -
2919 (void __user *)&rxnfc->fs.ring_cookie) ||
2920 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2921 sizeof(rxnfc->rule_cnt)))
2925 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2930 if (copy_in_user(compat_rxnfc, rxnfc,
2931 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2932 (const void __user *)rxnfc) ||
2933 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2934 &rxnfc->fs.ring_cookie,
2935 (const void __user *)(&rxnfc->fs.location + 1) -
2936 (const void __user *)&rxnfc->fs.ring_cookie) ||
2937 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2938 sizeof(rxnfc->rule_cnt)))
2941 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2942 /* As an optimisation, we only copy the actual
2943 * number of rules that the underlying
2944 * function returned. Since Mallory might
2945 * change the rule count in user memory, we
2946 * check that it is less than the rule count
2947 * originally given (as the user buffer size),
2948 * which has been range-checked.
2950 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2952 if (actual_rule_cnt < rule_cnt)
2953 rule_cnt = actual_rule_cnt;
2954 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2955 &rxnfc->rule_locs[0],
2956 rule_cnt * sizeof(u32)))
2964 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2967 compat_uptr_t uptr32;
2968 struct ifreq __user *uifr;
2970 uifr = compat_alloc_user_space(sizeof(*uifr));
2971 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2974 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2977 uptr = compat_ptr(uptr32);
2979 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2982 return dev_ioctl(net, SIOCWANDEV, uifr);
2985 static int bond_ioctl(struct net *net, unsigned int cmd,
2986 struct compat_ifreq __user *ifr32)
2989 mm_segment_t old_fs;
2993 case SIOCBONDENSLAVE:
2994 case SIOCBONDRELEASE:
2995 case SIOCBONDSETHWADDR:
2996 case SIOCBONDCHANGEACTIVE:
2997 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
3002 err = dev_ioctl(net, cmd,
3003 (struct ifreq __user __force *) &kifr);
3008 return -ENOIOCTLCMD;
3012 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3013 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3014 struct compat_ifreq __user *u_ifreq32)
3016 struct ifreq __user *u_ifreq64;
3017 char tmp_buf[IFNAMSIZ];
3018 void __user *data64;
3021 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3024 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3026 data64 = compat_ptr(data32);
3028 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3030 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3033 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3036 return dev_ioctl(net, cmd, u_ifreq64);
3039 static int dev_ifsioc(struct net *net, struct socket *sock,
3040 unsigned int cmd, struct compat_ifreq __user *uifr32)
3042 struct ifreq __user *uifr;
3045 uifr = compat_alloc_user_space(sizeof(*uifr));
3046 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3049 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3060 case SIOCGIFBRDADDR:
3061 case SIOCGIFDSTADDR:
3062 case SIOCGIFNETMASK:
3067 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3075 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3076 struct compat_ifreq __user *uifr32)
3079 struct compat_ifmap __user *uifmap32;
3080 mm_segment_t old_fs;
3083 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3084 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3085 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3086 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3087 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3088 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3089 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3090 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3096 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3099 if (cmd == SIOCGIFMAP && !err) {
3100 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3101 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3102 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3103 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3104 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3105 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3106 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3115 struct sockaddr rt_dst; /* target address */
3116 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3117 struct sockaddr rt_genmask; /* target network mask (IP) */
3118 unsigned short rt_flags;
3121 unsigned char rt_tos;
3122 unsigned char rt_class;
3124 short rt_metric; /* +1 for binary compatibility! */
3125 /* char * */ u32 rt_dev; /* forcing the device at add */
3126 u32 rt_mtu; /* per route MTU/Window */
3127 u32 rt_window; /* Window clamping */
3128 unsigned short rt_irtt; /* Initial RTT */
3131 struct in6_rtmsg32 {
3132 struct in6_addr rtmsg_dst;
3133 struct in6_addr rtmsg_src;
3134 struct in6_addr rtmsg_gateway;
3144 static int routing_ioctl(struct net *net, struct socket *sock,
3145 unsigned int cmd, void __user *argp)
3149 struct in6_rtmsg r6;
3153 mm_segment_t old_fs = get_fs();
3155 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3156 struct in6_rtmsg32 __user *ur6 = argp;
3157 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3158 3 * sizeof(struct in6_addr));
3159 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3160 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3161 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3162 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3163 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3164 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3165 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3169 struct rtentry32 __user *ur4 = argp;
3170 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3171 3 * sizeof(struct sockaddr));
3172 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3173 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3174 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3175 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3176 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3177 ret |= get_user(rtdev, &(ur4->rt_dev));
3179 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3180 r4.rt_dev = (char __user __force *)devname;
3194 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3201 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3202 * for some operations; this forces use of the newer bridge-utils that
3203 * use compatible ioctls
3205 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3209 if (get_user(tmp, argp))
3211 if (tmp == BRCTL_GET_VERSION)
3212 return BRCTL_VERSION + 1;
3216 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3217 unsigned int cmd, unsigned long arg)
3219 void __user *argp = compat_ptr(arg);
3220 struct sock *sk = sock->sk;
3221 struct net *net = sock_net(sk);
3223 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3224 return compat_ifr_data_ioctl(net, cmd, argp);
3229 return old_bridge_ioctl(argp);
3231 return dev_ifname32(net, argp);
3233 return dev_ifconf(net, argp);
3235 return ethtool_ioctl(net, argp);
3237 return compat_siocwandev(net, argp);
3240 return compat_sioc_ifmap(net, cmd, argp);
3241 case SIOCBONDENSLAVE:
3242 case SIOCBONDRELEASE:
3243 case SIOCBONDSETHWADDR:
3244 case SIOCBONDCHANGEACTIVE:
3245 return bond_ioctl(net, cmd, argp);
3248 return routing_ioctl(net, sock, cmd, argp);
3250 return do_siocgstamp(net, sock, cmd, argp);
3252 return do_siocgstampns(net, sock, cmd, argp);
3253 case SIOCBONDSLAVEINFOQUERY:
3254 case SIOCBONDINFOQUERY:
3257 return compat_ifr_data_ioctl(net, cmd, argp);
3269 return sock_ioctl(file, cmd, arg);
3286 case SIOCSIFHWBROADCAST:
3288 case SIOCGIFBRDADDR:
3289 case SIOCSIFBRDADDR:
3290 case SIOCGIFDSTADDR:
3291 case SIOCSIFDSTADDR:
3292 case SIOCGIFNETMASK:
3293 case SIOCSIFNETMASK:
3304 return dev_ifsioc(net, sock, cmd, argp);
3310 return sock_do_ioctl(net, sock, cmd, arg);
3313 return -ENOIOCTLCMD;
3316 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3319 struct socket *sock = file->private_data;
3320 int ret = -ENOIOCTLCMD;
3327 if (sock->ops->compat_ioctl)
3328 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3330 if (ret == -ENOIOCTLCMD &&
3331 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3332 ret = compat_wext_handle_ioctl(net, cmd, arg);
3334 if (ret == -ENOIOCTLCMD)
3335 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3341 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3343 return sock->ops->bind(sock, addr, addrlen);
3345 EXPORT_SYMBOL(kernel_bind);
3347 int kernel_listen(struct socket *sock, int backlog)
3349 return sock->ops->listen(sock, backlog);
3351 EXPORT_SYMBOL(kernel_listen);
3353 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3355 struct sock *sk = sock->sk;
3358 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3363 err = sock->ops->accept(sock, *newsock, flags);
3365 sock_release(*newsock);
3370 (*newsock)->ops = sock->ops;
3371 __module_get((*newsock)->ops->owner);
3376 EXPORT_SYMBOL(kernel_accept);
3378 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3381 return sock->ops->connect(sock, addr, addrlen, flags);
3383 EXPORT_SYMBOL(kernel_connect);
3385 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3388 return sock->ops->getname(sock, addr, addrlen, 0);
3390 EXPORT_SYMBOL(kernel_getsockname);
3392 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3395 return sock->ops->getname(sock, addr, addrlen, 1);
3397 EXPORT_SYMBOL(kernel_getpeername);
3399 int kernel_getsockopt(struct socket *sock, int level, int optname,
3400 char *optval, int *optlen)
3402 mm_segment_t oldfs = get_fs();
3403 char __user *uoptval;
3404 int __user *uoptlen;
3407 uoptval = (char __user __force *) optval;
3408 uoptlen = (int __user __force *) optlen;
3411 if (level == SOL_SOCKET)
3412 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3414 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3419 EXPORT_SYMBOL(kernel_getsockopt);
3421 int kernel_setsockopt(struct socket *sock, int level, int optname,
3422 char *optval, unsigned int optlen)
3424 mm_segment_t oldfs = get_fs();
3425 char __user *uoptval;
3428 uoptval = (char __user __force *) optval;
3431 if (level == SOL_SOCKET)
3432 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3434 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3439 EXPORT_SYMBOL(kernel_setsockopt);
3441 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3442 size_t size, int flags)
3444 if (sock->ops->sendpage)
3445 return sock->ops->sendpage(sock, page, offset, size, flags);
3447 return sock_no_sendpage(sock, page, offset, size, flags);
3449 EXPORT_SYMBOL(kernel_sendpage);
3451 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3453 mm_segment_t oldfs = get_fs();
3457 err = sock->ops->ioctl(sock, cmd, arg);
3462 EXPORT_SYMBOL(kernel_sock_ioctl);
3464 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3466 return sock->ops->shutdown(sock, how);
3468 EXPORT_SYMBOL(kernel_sock_shutdown);