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>
108 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
109 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
113 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
115 static int sock_close(struct inode *inode, struct file *file);
116 static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
118 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
120 static long compat_sock_ioctl(struct file *file,
121 unsigned int cmd, unsigned long arg);
123 static int sock_fasync(int fd, struct file *filp, int on);
124 static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
126 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
127 struct pipe_inode_info *pipe, size_t len,
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
135 static const struct file_operations socket_file_ops = {
136 .owner = THIS_MODULE,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
141 .unlocked_ioctl = sock_ioctl,
143 .compat_ioctl = compat_sock_ioctl,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
151 .splice_read = sock_splice_read,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock);
159 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
162 * Statistics counters of the socket lists
165 static DEFINE_PER_CPU(int, sockets_in_use);
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
190 if (copy_from_user(kaddr, uaddr, ulen))
192 return audit_sockaddr(ulen, kaddr);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
218 err = get_user(len, ulen);
223 if (len < 0 || len > sizeof(struct sockaddr_storage))
226 if (audit_sockaddr(klen, kaddr))
228 if (copy_to_user(uaddr, kaddr, len))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen, ulen);
238 static struct kmem_cache *sock_inode_cachep __read_mostly;
240 static struct inode *sock_alloc_inode(struct super_block *sb)
242 struct socket_alloc *ei;
243 struct socket_wq *wq;
245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
250 kmem_cache_free(sock_inode_cachep, ei);
253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 struct socket_alloc *ei;
269 struct socket_wq *wq;
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
272 wq = rcu_dereference_protected(ei->socket.wq, 1);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static int init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
293 if (sock_inode_cachep == NULL)
298 static const struct super_operations sockfs_ops = {
299 .alloc_inode = sock_alloc_inode,
300 .destroy_inode = sock_destroy_inode,
301 .statfs = simple_statfs,
305 * sockfs_dname() is called from d_path().
307 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
309 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310 dentry->d_inode->i_ino);
313 static const struct dentry_operations sockfs_dentry_operations = {
314 .d_dname = sockfs_dname,
317 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318 int flags, const char *dev_name, void *data)
320 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321 &sockfs_dentry_operations, SOCKFS_MAGIC);
324 static struct vfsmount *sock_mnt __read_mostly;
326 static struct file_system_type sock_fs_type = {
328 .mount = sockfs_mount,
329 .kill_sb = kill_anon_super,
333 * Obtains the first available file descriptor and sets it up for use.
335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
349 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
351 struct qstr name = { .name = "" };
357 name.len = strlen(name.name);
358 } else if (sock->sk) {
359 name.name = sock->sk->sk_prot_creator->name;
360 name.len = strlen(name.name);
362 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
363 if (unlikely(!path.dentry))
364 return ERR_PTR(-ENOMEM);
365 path.mnt = mntget(sock_mnt);
367 d_instantiate(path.dentry, SOCK_INODE(sock));
368 SOCK_INODE(sock)->i_fop = &socket_file_ops;
370 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
372 if (unlikely(IS_ERR(file))) {
373 /* drop dentry, keep inode */
374 ihold(path.dentry->d_inode);
380 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
381 file->private_data = sock;
384 EXPORT_SYMBOL(sock_alloc_file);
386 static int sock_map_fd(struct socket *sock, int flags)
388 struct file *newfile;
389 int fd = get_unused_fd_flags(flags);
390 if (unlikely(fd < 0))
393 newfile = sock_alloc_file(sock, flags, NULL);
394 if (likely(!IS_ERR(newfile))) {
395 fd_install(fd, newfile);
400 return PTR_ERR(newfile);
403 struct socket *sock_from_file(struct file *file, int *err)
405 if (file->f_op == &socket_file_ops)
406 return file->private_data; /* set in sock_map_fd */
411 EXPORT_SYMBOL(sock_from_file);
414 * sockfd_lookup - Go from a file number to its socket slot
416 * @err: pointer to an error code return
418 * The file handle passed in is locked and the socket it is bound
419 * too is returned. If an error occurs the err pointer is overwritten
420 * with a negative errno code and NULL is returned. The function checks
421 * for both invalid handles and passing a handle which is not a socket.
423 * On a success the socket object pointer is returned.
426 struct socket *sockfd_lookup(int fd, int *err)
437 sock = sock_from_file(file, err);
442 EXPORT_SYMBOL(sockfd_lookup);
444 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450 file = fget_light(fd, fput_needed);
452 sock = sock_from_file(file, err);
455 fput_light(file, *fput_needed);
460 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
461 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
462 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
463 static ssize_t sockfs_getxattr(struct dentry *dentry,
464 const char *name, void *value, size_t size)
466 const char *proto_name;
471 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
472 proto_name = dentry->d_name.name;
473 proto_size = strlen(proto_name);
477 if (proto_size + 1 > size)
480 strncpy(value, proto_name, proto_size + 1);
482 error = proto_size + 1;
489 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
495 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
505 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
510 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 static const struct inode_operations sockfs_inode_ops = {
518 .getxattr = sockfs_getxattr,
519 .listxattr = sockfs_listxattr,
523 * sock_alloc - allocate a socket
525 * Allocate a new inode and socket object. The two are bound together
526 * and initialised. The socket is then returned. If we are out of inodes
530 static struct socket *sock_alloc(void)
535 inode = new_inode_pseudo(sock_mnt->mnt_sb);
539 sock = SOCKET_I(inode);
541 kmemcheck_annotate_bitfield(sock, type);
542 inode->i_ino = get_next_ino();
543 inode->i_mode = S_IFSOCK | S_IRWXUGO;
544 inode->i_uid = current_fsuid();
545 inode->i_gid = current_fsgid();
546 inode->i_op = &sockfs_inode_ops;
548 this_cpu_add(sockets_in_use, 1);
553 * In theory you can't get an open on this inode, but /proc provides
554 * a back door. Remember to keep it shut otherwise you'll let the
555 * creepy crawlies in.
558 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
563 const struct file_operations bad_sock_fops = {
564 .owner = THIS_MODULE,
565 .open = sock_no_open,
566 .llseek = noop_llseek,
570 * sock_release - close a socket
571 * @sock: socket to close
573 * The socket is released from the protocol stack if it has a release
574 * callback, and the inode is then released if the socket is bound to
575 * an inode not a file.
578 void sock_release(struct socket *sock)
581 struct module *owner = sock->ops->owner;
583 sock->ops->release(sock);
588 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
589 printk(KERN_ERR "sock_release: fasync list not empty!\n");
591 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
594 this_cpu_sub(sockets_in_use, 1);
596 iput(SOCK_INODE(sock));
601 EXPORT_SYMBOL(sock_release);
603 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
606 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
607 *tx_flags |= SKBTX_HW_TSTAMP;
608 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
609 *tx_flags |= SKBTX_SW_TSTAMP;
610 if (sock_flag(sk, SOCK_WIFI_STATUS))
611 *tx_flags |= SKBTX_WIFI_STATUS;
613 EXPORT_SYMBOL(sock_tx_timestamp);
615 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
616 struct msghdr *msg, size_t size)
618 struct sock_iocb *si = kiocb_to_siocb(iocb);
625 return sock->ops->sendmsg(iocb, sock, msg, size);
628 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
629 struct msghdr *msg, size_t size)
631 int err = security_socket_sendmsg(sock, msg, size);
633 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
636 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
639 struct sock_iocb siocb;
642 init_sync_kiocb(&iocb, NULL);
643 iocb.private = &siocb;
644 ret = __sock_sendmsg(&iocb, sock, msg, size);
645 if (-EIOCBQUEUED == ret)
646 ret = wait_on_sync_kiocb(&iocb);
649 EXPORT_SYMBOL(sock_sendmsg);
651 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
654 struct sock_iocb siocb;
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
665 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
666 struct kvec *vec, size_t num, size_t size)
668 mm_segment_t oldfs = get_fs();
673 * the following is safe, since for compiler definitions of kvec and
674 * iovec are identical, yielding the same in-core layout and alignment
676 msg->msg_iov = (struct iovec *)vec;
677 msg->msg_iovlen = num;
678 result = sock_sendmsg(sock, msg, size);
682 EXPORT_SYMBOL(kernel_sendmsg);
685 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
687 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
690 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
691 struct timespec ts[3];
693 struct skb_shared_hwtstamps *shhwtstamps =
696 /* Race occurred between timestamp enabling and packet
697 receiving. Fill in the current time for now. */
698 if (need_software_tstamp && skb->tstamp.tv64 == 0)
699 __net_timestamp(skb);
701 if (need_software_tstamp) {
702 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
704 skb_get_timestamp(skb, &tv);
705 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
708 skb_get_timestampns(skb, &ts[0]);
709 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
710 sizeof(ts[0]), &ts[0]);
715 memset(ts, 0, sizeof(ts));
716 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
717 ktime_to_timespec_cond(skb->tstamp, ts + 0))
720 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
721 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
723 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
724 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
728 put_cmsg(msg, SOL_SOCKET,
729 SCM_TIMESTAMPING, sizeof(ts), &ts);
731 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
733 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
738 if (!sock_flag(sk, SOCK_WIFI_STATUS))
740 if (!skb->wifi_acked_valid)
743 ack = skb->wifi_acked;
745 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
747 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
749 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
752 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
753 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
754 sizeof(__u32), &skb->dropcount);
757 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
760 sock_recv_timestamp(msg, sk, skb);
761 sock_recv_drops(msg, sk, skb);
763 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
765 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
766 struct msghdr *msg, size_t size, int flags)
768 struct sock_iocb *si = kiocb_to_siocb(iocb);
776 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
779 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
780 struct msghdr *msg, size_t size, int flags)
782 int err = security_socket_recvmsg(sock, msg, size, flags);
784 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
787 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
788 size_t size, int flags)
791 struct sock_iocb siocb;
794 init_sync_kiocb(&iocb, NULL);
795 iocb.private = &siocb;
796 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
797 if (-EIOCBQUEUED == ret)
798 ret = wait_on_sync_kiocb(&iocb);
801 EXPORT_SYMBOL(sock_recvmsg);
803 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
804 size_t size, int flags)
807 struct sock_iocb siocb;
810 init_sync_kiocb(&iocb, NULL);
811 iocb.private = &siocb;
812 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
813 if (-EIOCBQUEUED == ret)
814 ret = wait_on_sync_kiocb(&iocb);
819 * kernel_recvmsg - Receive a message from a socket (kernel space)
820 * @sock: The socket to receive the message from
821 * @msg: Received message
822 * @vec: Input s/g array for message data
823 * @num: Size of input s/g array
824 * @size: Number of bytes to read
825 * @flags: Message flags (MSG_DONTWAIT, etc...)
827 * On return the msg structure contains the scatter/gather array passed in the
828 * vec argument. The array is modified so that it consists of the unfilled
829 * portion of the original array.
831 * The returned value is the total number of bytes received, or an error.
833 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
834 struct kvec *vec, size_t num, size_t size, int flags)
836 mm_segment_t oldfs = get_fs();
841 * the following is safe, since for compiler definitions of kvec and
842 * iovec are identical, yielding the same in-core layout and alignment
844 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
845 result = sock_recvmsg(sock, msg, size, flags);
849 EXPORT_SYMBOL(kernel_recvmsg);
851 static void sock_aio_dtor(struct kiocb *iocb)
853 kfree(iocb->private);
856 static ssize_t sock_sendpage(struct file *file, struct page *page,
857 int offset, size_t size, loff_t *ppos, int more)
862 sock = file->private_data;
864 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
865 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
868 return kernel_sendpage(sock, page, offset, size, flags);
871 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
872 struct pipe_inode_info *pipe, size_t len,
875 struct socket *sock = file->private_data;
877 if (unlikely(!sock->ops->splice_read))
880 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
883 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
884 struct sock_iocb *siocb)
886 if (!is_sync_kiocb(iocb)) {
887 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
890 iocb->ki_dtor = sock_aio_dtor;
894 iocb->private = siocb;
898 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
899 struct file *file, const struct iovec *iov,
900 unsigned long nr_segs)
902 struct socket *sock = file->private_data;
906 for (i = 0; i < nr_segs; i++)
907 size += iov[i].iov_len;
909 msg->msg_name = NULL;
910 msg->msg_namelen = 0;
911 msg->msg_control = NULL;
912 msg->msg_controllen = 0;
913 msg->msg_iov = (struct iovec *)iov;
914 msg->msg_iovlen = nr_segs;
915 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
917 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
920 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
921 unsigned long nr_segs, loff_t pos)
923 struct sock_iocb siocb, *x;
928 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
932 x = alloc_sock_iocb(iocb, &siocb);
935 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
938 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
939 struct file *file, const struct iovec *iov,
940 unsigned long nr_segs)
942 struct socket *sock = file->private_data;
946 for (i = 0; i < nr_segs; i++)
947 size += iov[i].iov_len;
949 msg->msg_name = NULL;
950 msg->msg_namelen = 0;
951 msg->msg_control = NULL;
952 msg->msg_controllen = 0;
953 msg->msg_iov = (struct iovec *)iov;
954 msg->msg_iovlen = nr_segs;
955 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
956 if (sock->type == SOCK_SEQPACKET)
957 msg->msg_flags |= MSG_EOR;
959 return __sock_sendmsg(iocb, sock, msg, size);
962 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
963 unsigned long nr_segs, loff_t pos)
965 struct sock_iocb siocb, *x;
970 x = alloc_sock_iocb(iocb, &siocb);
974 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
978 * Atomic setting of ioctl hooks to avoid race
979 * with module unload.
982 static DEFINE_MUTEX(br_ioctl_mutex);
983 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
985 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
987 mutex_lock(&br_ioctl_mutex);
988 br_ioctl_hook = hook;
989 mutex_unlock(&br_ioctl_mutex);
991 EXPORT_SYMBOL(brioctl_set);
993 static DEFINE_MUTEX(vlan_ioctl_mutex);
994 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
996 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
998 mutex_lock(&vlan_ioctl_mutex);
999 vlan_ioctl_hook = hook;
1000 mutex_unlock(&vlan_ioctl_mutex);
1002 EXPORT_SYMBOL(vlan_ioctl_set);
1004 static DEFINE_MUTEX(dlci_ioctl_mutex);
1005 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1007 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1009 mutex_lock(&dlci_ioctl_mutex);
1010 dlci_ioctl_hook = hook;
1011 mutex_unlock(&dlci_ioctl_mutex);
1013 EXPORT_SYMBOL(dlci_ioctl_set);
1015 static long sock_do_ioctl(struct net *net, struct socket *sock,
1016 unsigned int cmd, unsigned long arg)
1019 void __user *argp = (void __user *)arg;
1021 err = sock->ops->ioctl(sock, cmd, arg);
1024 * If this ioctl is unknown try to hand it down
1025 * to the NIC driver.
1027 if (err == -ENOIOCTLCMD)
1028 err = dev_ioctl(net, cmd, argp);
1034 * With an ioctl, arg may well be a user mode pointer, but we don't know
1035 * what to do with it - that's up to the protocol still.
1038 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1040 struct socket *sock;
1042 void __user *argp = (void __user *)arg;
1046 sock = file->private_data;
1049 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1050 err = dev_ioctl(net, cmd, argp);
1052 #ifdef CONFIG_WEXT_CORE
1053 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1054 err = dev_ioctl(net, cmd, argp);
1061 if (get_user(pid, (int __user *)argp))
1063 err = f_setown(sock->file, pid, 1);
1067 err = put_user(f_getown(sock->file),
1068 (int __user *)argp);
1076 request_module("bridge");
1078 mutex_lock(&br_ioctl_mutex);
1080 err = br_ioctl_hook(net, cmd, argp);
1081 mutex_unlock(&br_ioctl_mutex);
1086 if (!vlan_ioctl_hook)
1087 request_module("8021q");
1089 mutex_lock(&vlan_ioctl_mutex);
1090 if (vlan_ioctl_hook)
1091 err = vlan_ioctl_hook(net, argp);
1092 mutex_unlock(&vlan_ioctl_mutex);
1097 if (!dlci_ioctl_hook)
1098 request_module("dlci");
1100 mutex_lock(&dlci_ioctl_mutex);
1101 if (dlci_ioctl_hook)
1102 err = dlci_ioctl_hook(cmd, argp);
1103 mutex_unlock(&dlci_ioctl_mutex);
1106 err = sock_do_ioctl(net, sock, cmd, arg);
1112 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1115 struct socket *sock = NULL;
1117 err = security_socket_create(family, type, protocol, 1);
1121 sock = sock_alloc();
1128 err = security_socket_post_create(sock, family, type, protocol, 1);
1140 EXPORT_SYMBOL(sock_create_lite);
1142 /* No kernel lock held - perfect */
1143 static unsigned int sock_poll(struct file *file, poll_table *wait)
1145 struct socket *sock;
1148 * We can't return errors to poll, so it's either yes or no.
1150 sock = file->private_data;
1151 return sock->ops->poll(file, sock, wait);
1154 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1156 struct socket *sock = file->private_data;
1158 return sock->ops->mmap(file, sock, vma);
1161 static int sock_close(struct inode *inode, struct file *filp)
1164 * It was possible the inode is NULL we were
1165 * closing an unfinished socket.
1169 printk(KERN_DEBUG "sock_close: NULL inode\n");
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;
1448 fd2 = get_unused_fd_flags(flags);
1449 if (unlikely(fd2 < 0)) {
1452 goto out_release_both;
1455 newfile1 = sock_alloc_file(sock1, flags, NULL);
1456 if (unlikely(IS_ERR(newfile1))) {
1457 err = PTR_ERR(newfile1);
1460 goto out_release_both;
1463 newfile2 = sock_alloc_file(sock2, flags, NULL);
1464 if (IS_ERR(newfile2)) {
1465 err = PTR_ERR(newfile2);
1469 sock_release(sock2);
1473 audit_fd_pair(fd1, fd2);
1474 fd_install(fd1, newfile1);
1475 fd_install(fd2, newfile2);
1476 /* fd1 and fd2 may be already another descriptors.
1477 * Not kernel problem.
1480 err = put_user(fd1, &usockvec[0]);
1482 err = put_user(fd2, &usockvec[1]);
1491 sock_release(sock2);
1493 sock_release(sock1);
1499 * Bind a name to a socket. Nothing much to do here since it's
1500 * the protocol's responsibility to handle the local address.
1502 * We move the socket address to kernel space before we call
1503 * the protocol layer (having also checked the address is ok).
1506 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1508 struct socket *sock;
1509 struct sockaddr_storage address;
1510 int err, fput_needed;
1512 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1514 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1516 err = security_socket_bind(sock,
1517 (struct sockaddr *)&address,
1520 err = sock->ops->bind(sock,
1524 fput_light(sock->file, fput_needed);
1530 * Perform a listen. Basically, we allow the protocol to do anything
1531 * necessary for a listen, and if that works, we mark the socket as
1532 * ready for listening.
1535 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1537 struct socket *sock;
1538 int err, fput_needed;
1541 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1543 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1544 if ((unsigned int)backlog > somaxconn)
1545 backlog = somaxconn;
1547 err = security_socket_listen(sock, backlog);
1549 err = sock->ops->listen(sock, backlog);
1551 fput_light(sock->file, fput_needed);
1557 * For accept, we attempt to create a new socket, set up the link
1558 * with the client, wake up the client, then return the new
1559 * connected fd. We collect the address of the connector in kernel
1560 * space and move it to user at the very end. This is unclean because
1561 * we open the socket then return an error.
1563 * 1003.1g adds the ability to recvmsg() to query connection pending
1564 * status to recvmsg. We need to add that support in a way thats
1565 * clean when we restucture accept also.
1568 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1569 int __user *, upeer_addrlen, int, flags)
1571 struct socket *sock, *newsock;
1572 struct file *newfile;
1573 int err, len, newfd, fput_needed;
1574 struct sockaddr_storage address;
1576 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1579 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1580 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1582 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1587 newsock = sock_alloc();
1591 newsock->type = sock->type;
1592 newsock->ops = sock->ops;
1595 * We don't need try_module_get here, as the listening socket (sock)
1596 * has the protocol module (sock->ops->owner) held.
1598 __module_get(newsock->ops->owner);
1600 newfd = get_unused_fd_flags(flags);
1601 if (unlikely(newfd < 0)) {
1603 sock_release(newsock);
1606 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1607 if (unlikely(IS_ERR(newfile))) {
1608 err = PTR_ERR(newfile);
1609 put_unused_fd(newfd);
1610 sock_release(newsock);
1614 err = security_socket_accept(sock, newsock);
1618 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1622 if (upeer_sockaddr) {
1623 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1625 err = -ECONNABORTED;
1628 err = move_addr_to_user(&address,
1629 len, upeer_sockaddr, upeer_addrlen);
1634 /* File flags are not inherited via accept() unlike another OSes. */
1636 fd_install(newfd, newfile);
1640 fput_light(sock->file, fput_needed);
1645 put_unused_fd(newfd);
1649 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1650 int __user *, upeer_addrlen)
1652 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1656 * Attempt to connect to a socket with the server address. The address
1657 * is in user space so we verify it is OK and move it to kernel space.
1659 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1662 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1663 * other SEQPACKET protocols that take time to connect() as it doesn't
1664 * include the -EINPROGRESS status for such sockets.
1667 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1670 struct socket *sock;
1671 struct sockaddr_storage address;
1672 int err, fput_needed;
1674 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1677 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1682 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1686 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1687 sock->file->f_flags);
1689 fput_light(sock->file, fput_needed);
1695 * Get the local address ('name') of a socket object. Move the obtained
1696 * name to user space.
1699 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1700 int __user *, usockaddr_len)
1702 struct socket *sock;
1703 struct sockaddr_storage address;
1704 int len, err, fput_needed;
1706 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1710 err = security_socket_getsockname(sock);
1714 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1717 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1720 fput_light(sock->file, fput_needed);
1726 * Get the remote address ('name') of a socket object. Move the obtained
1727 * name to user space.
1730 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1731 int __user *, usockaddr_len)
1733 struct socket *sock;
1734 struct sockaddr_storage address;
1735 int len, err, fput_needed;
1737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1739 err = security_socket_getpeername(sock);
1741 fput_light(sock->file, fput_needed);
1746 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1749 err = move_addr_to_user(&address, len, usockaddr,
1751 fput_light(sock->file, fput_needed);
1757 * Send a datagram to a given address. We move the address into kernel
1758 * space and check the user space data area is readable before invoking
1762 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1763 unsigned int, flags, struct sockaddr __user *, addr,
1766 struct socket *sock;
1767 struct sockaddr_storage address;
1775 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1779 iov.iov_base = buff;
1781 msg.msg_name = NULL;
1784 msg.msg_control = NULL;
1785 msg.msg_controllen = 0;
1786 msg.msg_namelen = 0;
1788 err = move_addr_to_kernel(addr, addr_len, &address);
1791 msg.msg_name = (struct sockaddr *)&address;
1792 msg.msg_namelen = addr_len;
1794 if (sock->file->f_flags & O_NONBLOCK)
1795 flags |= MSG_DONTWAIT;
1796 msg.msg_flags = flags;
1797 err = sock_sendmsg(sock, &msg, len);
1800 fput_light(sock->file, fput_needed);
1806 * Send a datagram down a socket.
1809 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1810 unsigned int, flags)
1812 return sys_sendto(fd, buff, len, flags, NULL, 0);
1816 * Receive a frame from the socket and optionally record the address of the
1817 * sender. We verify the buffers are writable and if needed move the
1818 * sender address from kernel to user space.
1821 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1822 unsigned int, flags, struct sockaddr __user *, addr,
1823 int __user *, addr_len)
1825 struct socket *sock;
1828 struct sockaddr_storage address;
1834 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1838 msg.msg_control = NULL;
1839 msg.msg_controllen = 0;
1843 iov.iov_base = ubuf;
1844 msg.msg_name = (struct sockaddr *)&address;
1845 msg.msg_namelen = sizeof(address);
1846 if (sock->file->f_flags & O_NONBLOCK)
1847 flags |= MSG_DONTWAIT;
1848 err = sock_recvmsg(sock, &msg, size, flags);
1850 if (err >= 0 && addr != NULL) {
1851 err2 = move_addr_to_user(&address,
1852 msg.msg_namelen, addr, addr_len);
1857 fput_light(sock->file, fput_needed);
1863 * Receive a datagram from a socket.
1866 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1869 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1873 * Set a socket option. Because we don't know the option lengths we have
1874 * to pass the user mode parameter for the protocols to sort out.
1877 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1878 char __user *, optval, int, optlen)
1880 int err, fput_needed;
1881 struct socket *sock;
1886 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1888 err = security_socket_setsockopt(sock, level, optname);
1892 if (level == SOL_SOCKET)
1894 sock_setsockopt(sock, level, optname, optval,
1898 sock->ops->setsockopt(sock, level, optname, optval,
1901 fput_light(sock->file, fput_needed);
1907 * Get a socket option. Because we don't know the option lengths we have
1908 * to pass a user mode parameter for the protocols to sort out.
1911 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1912 char __user *, optval, int __user *, optlen)
1914 int err, fput_needed;
1915 struct socket *sock;
1917 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1919 err = security_socket_getsockopt(sock, level, optname);
1923 if (level == SOL_SOCKET)
1925 sock_getsockopt(sock, level, optname, optval,
1929 sock->ops->getsockopt(sock, level, optname, optval,
1932 fput_light(sock->file, fput_needed);
1938 * Shutdown a socket.
1941 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1943 int err, fput_needed;
1944 struct socket *sock;
1946 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1948 err = security_socket_shutdown(sock, how);
1950 err = sock->ops->shutdown(sock, how);
1951 fput_light(sock->file, fput_needed);
1956 /* A couple of helpful macros for getting the address of the 32/64 bit
1957 * fields which are the same type (int / unsigned) on our platforms.
1959 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1960 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1961 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1963 struct used_address {
1964 struct sockaddr_storage name;
1965 unsigned int name_len;
1968 static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1969 struct msghdr *msg_sys, unsigned int flags,
1970 struct used_address *used_address)
1972 struct compat_msghdr __user *msg_compat =
1973 (struct compat_msghdr __user *)msg;
1974 struct sockaddr_storage address;
1975 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1976 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1977 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1978 /* 20 is size of ipv6_pktinfo */
1979 unsigned char *ctl_buf = ctl;
1980 int err, ctl_len, total_len;
1983 if (MSG_CMSG_COMPAT & flags) {
1984 if (get_compat_msghdr(msg_sys, msg_compat))
1986 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1989 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1991 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1994 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2000 /* This will also move the address data into kernel space */
2001 if (MSG_CMSG_COMPAT & flags) {
2002 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2004 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2011 if (msg_sys->msg_controllen > INT_MAX)
2013 ctl_len = msg_sys->msg_controllen;
2014 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2016 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2020 ctl_buf = msg_sys->msg_control;
2021 ctl_len = msg_sys->msg_controllen;
2022 } else if (ctl_len) {
2023 if (ctl_len > sizeof(ctl)) {
2024 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2025 if (ctl_buf == NULL)
2030 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2031 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2032 * checking falls down on this.
2034 if (copy_from_user(ctl_buf,
2035 (void __user __force *)msg_sys->msg_control,
2038 msg_sys->msg_control = ctl_buf;
2040 msg_sys->msg_flags = flags;
2042 if (sock->file->f_flags & O_NONBLOCK)
2043 msg_sys->msg_flags |= MSG_DONTWAIT;
2045 * If this is sendmmsg() and current destination address is same as
2046 * previously succeeded address, omit asking LSM's decision.
2047 * used_address->name_len is initialized to UINT_MAX so that the first
2048 * destination address never matches.
2050 if (used_address && msg_sys->msg_name &&
2051 used_address->name_len == msg_sys->msg_namelen &&
2052 !memcmp(&used_address->name, msg_sys->msg_name,
2053 used_address->name_len)) {
2054 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2057 err = sock_sendmsg(sock, msg_sys, total_len);
2059 * If this is sendmmsg() and sending to current destination address was
2060 * successful, remember it.
2062 if (used_address && err >= 0) {
2063 used_address->name_len = msg_sys->msg_namelen;
2064 if (msg_sys->msg_name)
2065 memcpy(&used_address->name, msg_sys->msg_name,
2066 used_address->name_len);
2071 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2073 if (iov != iovstack)
2080 * BSD sendmsg interface
2083 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2085 int fput_needed, err;
2086 struct msghdr msg_sys;
2087 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2092 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2094 fput_light(sock->file, fput_needed);
2100 * Linux sendmmsg interface
2103 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2106 int fput_needed, err, datagrams;
2107 struct socket *sock;
2108 struct mmsghdr __user *entry;
2109 struct compat_mmsghdr __user *compat_entry;
2110 struct msghdr msg_sys;
2111 struct used_address used_address;
2113 if (vlen > UIO_MAXIOV)
2118 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2122 used_address.name_len = UINT_MAX;
2124 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2127 while (datagrams < vlen) {
2128 if (MSG_CMSG_COMPAT & flags) {
2129 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2130 &msg_sys, flags, &used_address);
2133 err = __put_user(err, &compat_entry->msg_len);
2136 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
2137 &msg_sys, flags, &used_address);
2140 err = put_user(err, &entry->msg_len);
2149 fput_light(sock->file, fput_needed);
2151 /* We only return an error if no datagrams were able to be sent */
2158 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2159 unsigned int, vlen, unsigned int, flags)
2161 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2164 static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2165 struct msghdr *msg_sys, unsigned int flags, int nosec)
2167 struct compat_msghdr __user *msg_compat =
2168 (struct compat_msghdr __user *)msg;
2169 struct iovec iovstack[UIO_FASTIOV];
2170 struct iovec *iov = iovstack;
2171 unsigned long cmsg_ptr;
2172 int err, total_len, len;
2174 /* kernel mode address */
2175 struct sockaddr_storage addr;
2177 /* user mode address pointers */
2178 struct sockaddr __user *uaddr;
2179 int __user *uaddr_len;
2181 if (MSG_CMSG_COMPAT & flags) {
2182 if (get_compat_msghdr(msg_sys, msg_compat))
2184 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2187 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2189 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2192 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2199 * Save the user-mode address (verify_iovec will change the
2200 * kernel msghdr to use the kernel address space)
2203 uaddr = (__force void __user *)msg_sys->msg_name;
2204 uaddr_len = COMPAT_NAMELEN(msg);
2205 if (MSG_CMSG_COMPAT & flags) {
2206 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2208 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2213 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2214 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2216 if (sock->file->f_flags & O_NONBLOCK)
2217 flags |= MSG_DONTWAIT;
2218 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2224 if (uaddr != NULL) {
2225 err = move_addr_to_user(&addr,
2226 msg_sys->msg_namelen, uaddr,
2231 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2235 if (MSG_CMSG_COMPAT & flags)
2236 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2237 &msg_compat->msg_controllen);
2239 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2240 &msg->msg_controllen);
2246 if (iov != iovstack)
2253 * BSD recvmsg interface
2256 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2257 unsigned int, flags)
2259 int fput_needed, err;
2260 struct msghdr msg_sys;
2261 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2266 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2268 fput_light(sock->file, fput_needed);
2274 * Linux recvmmsg interface
2277 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2278 unsigned int flags, struct timespec *timeout)
2280 int fput_needed, err, datagrams;
2281 struct socket *sock;
2282 struct mmsghdr __user *entry;
2283 struct compat_mmsghdr __user *compat_entry;
2284 struct msghdr msg_sys;
2285 struct timespec end_time;
2288 poll_select_set_timeout(&end_time, timeout->tv_sec,
2294 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2298 err = sock_error(sock->sk);
2303 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2305 while (datagrams < vlen) {
2307 * No need to ask LSM for more than the first datagram.
2309 if (MSG_CMSG_COMPAT & flags) {
2310 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2311 &msg_sys, flags & ~MSG_WAITFORONE,
2315 err = __put_user(err, &compat_entry->msg_len);
2318 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2319 &msg_sys, flags & ~MSG_WAITFORONE,
2323 err = put_user(err, &entry->msg_len);
2331 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2332 if (flags & MSG_WAITFORONE)
2333 flags |= MSG_DONTWAIT;
2336 ktime_get_ts(timeout);
2337 *timeout = timespec_sub(end_time, *timeout);
2338 if (timeout->tv_sec < 0) {
2339 timeout->tv_sec = timeout->tv_nsec = 0;
2343 /* Timeout, return less than vlen datagrams */
2344 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2348 /* Out of band data, return right away */
2349 if (msg_sys.msg_flags & MSG_OOB)
2354 fput_light(sock->file, fput_needed);
2359 if (datagrams != 0) {
2361 * We may return less entries than requested (vlen) if the
2362 * sock is non block and there aren't enough datagrams...
2364 if (err != -EAGAIN) {
2366 * ... or if recvmsg returns an error after we
2367 * received some datagrams, where we record the
2368 * error to return on the next call or if the
2369 * app asks about it using getsockopt(SO_ERROR).
2371 sock->sk->sk_err = -err;
2380 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2381 unsigned int, vlen, unsigned int, flags,
2382 struct timespec __user *, timeout)
2385 struct timespec timeout_sys;
2388 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2390 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2393 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2395 if (datagrams > 0 &&
2396 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2397 datagrams = -EFAULT;
2402 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2403 /* Argument list sizes for sys_socketcall */
2404 #define AL(x) ((x) * sizeof(unsigned long))
2405 static const unsigned char nargs[21] = {
2406 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2407 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2408 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2415 * System call vectors.
2417 * Argument checking cleaned up. Saved 20% in size.
2418 * This function doesn't need to set the kernel lock because
2419 * it is set by the callees.
2422 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2425 unsigned long a0, a1;
2429 if (call < 1 || call > SYS_SENDMMSG)
2433 if (len > sizeof(a))
2436 /* copy_from_user should be SMP safe. */
2437 if (copy_from_user(a, args, len))
2440 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2447 err = sys_socket(a0, a1, a[2]);
2450 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2453 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2456 err = sys_listen(a0, a1);
2459 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2460 (int __user *)a[2], 0);
2462 case SYS_GETSOCKNAME:
2464 sys_getsockname(a0, (struct sockaddr __user *)a1,
2465 (int __user *)a[2]);
2467 case SYS_GETPEERNAME:
2469 sys_getpeername(a0, (struct sockaddr __user *)a1,
2470 (int __user *)a[2]);
2472 case SYS_SOCKETPAIR:
2473 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2476 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2479 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2480 (struct sockaddr __user *)a[4], a[5]);
2483 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2486 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2487 (struct sockaddr __user *)a[4],
2488 (int __user *)a[5]);
2491 err = sys_shutdown(a0, a1);
2493 case SYS_SETSOCKOPT:
2494 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2496 case SYS_GETSOCKOPT:
2498 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2499 (int __user *)a[4]);
2502 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2505 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2508 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2511 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2512 (struct timespec __user *)a[4]);
2515 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2516 (int __user *)a[2], a[3]);
2525 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2528 * sock_register - add a socket protocol handler
2529 * @ops: description of protocol
2531 * This function is called by a protocol handler that wants to
2532 * advertise its address family, and have it linked into the
2533 * socket interface. The value ops->family coresponds to the
2534 * socket system call protocol family.
2536 int sock_register(const struct net_proto_family *ops)
2540 if (ops->family >= NPROTO) {
2541 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2546 spin_lock(&net_family_lock);
2547 if (rcu_dereference_protected(net_families[ops->family],
2548 lockdep_is_held(&net_family_lock)))
2551 rcu_assign_pointer(net_families[ops->family], ops);
2554 spin_unlock(&net_family_lock);
2556 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2559 EXPORT_SYMBOL(sock_register);
2562 * sock_unregister - remove a protocol handler
2563 * @family: protocol family to remove
2565 * This function is called by a protocol handler that wants to
2566 * remove its address family, and have it unlinked from the
2567 * new socket creation.
2569 * If protocol handler is a module, then it can use module reference
2570 * counts to protect against new references. If protocol handler is not
2571 * a module then it needs to provide its own protection in
2572 * the ops->create routine.
2574 void sock_unregister(int family)
2576 BUG_ON(family < 0 || family >= NPROTO);
2578 spin_lock(&net_family_lock);
2579 RCU_INIT_POINTER(net_families[family], NULL);
2580 spin_unlock(&net_family_lock);
2584 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2586 EXPORT_SYMBOL(sock_unregister);
2588 static int __init sock_init(void)
2592 * Initialize the network sysctl infrastructure.
2594 err = net_sysctl_init();
2599 * Initialize skbuff SLAB cache
2604 * Initialize the protocols module.
2609 err = register_filesystem(&sock_fs_type);
2612 sock_mnt = kern_mount(&sock_fs_type);
2613 if (IS_ERR(sock_mnt)) {
2614 err = PTR_ERR(sock_mnt);
2618 /* The real protocol initialization is performed in later initcalls.
2621 #ifdef CONFIG_NETFILTER
2625 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2626 skb_timestamping_init();
2633 unregister_filesystem(&sock_fs_type);
2638 core_initcall(sock_init); /* early initcall */
2640 #ifdef CONFIG_PROC_FS
2641 void socket_seq_show(struct seq_file *seq)
2646 for_each_possible_cpu(cpu)
2647 counter += per_cpu(sockets_in_use, cpu);
2649 /* It can be negative, by the way. 8) */
2653 seq_printf(seq, "sockets: used %d\n", counter);
2655 #endif /* CONFIG_PROC_FS */
2657 #ifdef CONFIG_COMPAT
2658 static int do_siocgstamp(struct net *net, struct socket *sock,
2659 unsigned int cmd, void __user *up)
2661 mm_segment_t old_fs = get_fs();
2666 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2669 err = compat_put_timeval(&ktv, up);
2674 static int do_siocgstampns(struct net *net, struct socket *sock,
2675 unsigned int cmd, void __user *up)
2677 mm_segment_t old_fs = get_fs();
2678 struct timespec kts;
2682 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2685 err = compat_put_timespec(&kts, up);
2690 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2692 struct ifreq __user *uifr;
2695 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2696 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2699 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2703 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2709 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2711 struct compat_ifconf ifc32;
2713 struct ifconf __user *uifc;
2714 struct compat_ifreq __user *ifr32;
2715 struct ifreq __user *ifr;
2719 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2722 memset(&ifc, 0, sizeof(ifc));
2723 if (ifc32.ifcbuf == 0) {
2727 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2729 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2730 sizeof(struct ifreq);
2731 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2733 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2734 ifr32 = compat_ptr(ifc32.ifcbuf);
2735 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2736 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2742 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2745 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2749 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2753 ifr32 = compat_ptr(ifc32.ifcbuf);
2755 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2756 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2757 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2763 if (ifc32.ifcbuf == 0) {
2764 /* Translate from 64-bit structure multiple to
2768 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2773 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2779 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2781 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2782 bool convert_in = false, convert_out = false;
2783 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2784 struct ethtool_rxnfc __user *rxnfc;
2785 struct ifreq __user *ifr;
2786 u32 rule_cnt = 0, actual_rule_cnt;
2791 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2794 compat_rxnfc = compat_ptr(data);
2796 if (get_user(ethcmd, &compat_rxnfc->cmd))
2799 /* Most ethtool structures are defined without padding.
2800 * Unfortunately struct ethtool_rxnfc is an exception.
2805 case ETHTOOL_GRXCLSRLALL:
2806 /* Buffer size is variable */
2807 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2809 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2811 buf_size += rule_cnt * sizeof(u32);
2813 case ETHTOOL_GRXRINGS:
2814 case ETHTOOL_GRXCLSRLCNT:
2815 case ETHTOOL_GRXCLSRULE:
2816 case ETHTOOL_SRXCLSRLINS:
2819 case ETHTOOL_SRXCLSRLDEL:
2820 buf_size += sizeof(struct ethtool_rxnfc);
2825 ifr = compat_alloc_user_space(buf_size);
2826 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2828 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2831 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2832 &ifr->ifr_ifru.ifru_data))
2836 /* We expect there to be holes between fs.m_ext and
2837 * fs.ring_cookie and at the end of fs, but nowhere else.
2839 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2840 sizeof(compat_rxnfc->fs.m_ext) !=
2841 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2842 sizeof(rxnfc->fs.m_ext));
2844 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2845 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2846 offsetof(struct ethtool_rxnfc, fs.location) -
2847 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2849 if (copy_in_user(rxnfc, compat_rxnfc,
2850 (void __user *)(&rxnfc->fs.m_ext + 1) -
2851 (void __user *)rxnfc) ||
2852 copy_in_user(&rxnfc->fs.ring_cookie,
2853 &compat_rxnfc->fs.ring_cookie,
2854 (void __user *)(&rxnfc->fs.location + 1) -
2855 (void __user *)&rxnfc->fs.ring_cookie) ||
2856 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2857 sizeof(rxnfc->rule_cnt)))
2861 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2866 if (copy_in_user(compat_rxnfc, rxnfc,
2867 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2868 (const void __user *)rxnfc) ||
2869 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2870 &rxnfc->fs.ring_cookie,
2871 (const void __user *)(&rxnfc->fs.location + 1) -
2872 (const void __user *)&rxnfc->fs.ring_cookie) ||
2873 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2874 sizeof(rxnfc->rule_cnt)))
2877 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2878 /* As an optimisation, we only copy the actual
2879 * number of rules that the underlying
2880 * function returned. Since Mallory might
2881 * change the rule count in user memory, we
2882 * check that it is less than the rule count
2883 * originally given (as the user buffer size),
2884 * which has been range-checked.
2886 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2888 if (actual_rule_cnt < rule_cnt)
2889 rule_cnt = actual_rule_cnt;
2890 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2891 &rxnfc->rule_locs[0],
2892 rule_cnt * sizeof(u32)))
2900 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2903 compat_uptr_t uptr32;
2904 struct ifreq __user *uifr;
2906 uifr = compat_alloc_user_space(sizeof(*uifr));
2907 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2910 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2913 uptr = compat_ptr(uptr32);
2915 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2918 return dev_ioctl(net, SIOCWANDEV, uifr);
2921 static int bond_ioctl(struct net *net, unsigned int cmd,
2922 struct compat_ifreq __user *ifr32)
2925 struct ifreq __user *uifr;
2926 mm_segment_t old_fs;
2932 case SIOCBONDENSLAVE:
2933 case SIOCBONDRELEASE:
2934 case SIOCBONDSETHWADDR:
2935 case SIOCBONDCHANGEACTIVE:
2936 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2941 err = dev_ioctl(net, cmd,
2942 (struct ifreq __user __force *) &kifr);
2946 case SIOCBONDSLAVEINFOQUERY:
2947 case SIOCBONDINFOQUERY:
2948 uifr = compat_alloc_user_space(sizeof(*uifr));
2949 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2952 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2955 datap = compat_ptr(data);
2956 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2959 return dev_ioctl(net, cmd, uifr);
2961 return -ENOIOCTLCMD;
2965 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2966 struct compat_ifreq __user *u_ifreq32)
2968 struct ifreq __user *u_ifreq64;
2969 char tmp_buf[IFNAMSIZ];
2970 void __user *data64;
2973 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2976 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2978 data64 = compat_ptr(data32);
2980 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2982 /* Don't check these user accesses, just let that get trapped
2983 * in the ioctl handler instead.
2985 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2988 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2991 return dev_ioctl(net, cmd, u_ifreq64);
2994 static int dev_ifsioc(struct net *net, struct socket *sock,
2995 unsigned int cmd, struct compat_ifreq __user *uifr32)
2997 struct ifreq __user *uifr;
3000 uifr = compat_alloc_user_space(sizeof(*uifr));
3001 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3004 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3015 case SIOCGIFBRDADDR:
3016 case SIOCGIFDSTADDR:
3017 case SIOCGIFNETMASK:
3022 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3030 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3031 struct compat_ifreq __user *uifr32)
3034 struct compat_ifmap __user *uifmap32;
3035 mm_segment_t old_fs;
3038 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3039 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3040 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3041 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3042 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3043 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3044 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3045 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3051 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3054 if (cmd == SIOCGIFMAP && !err) {
3055 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3056 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3057 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3058 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3059 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3060 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3061 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3068 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3071 compat_uptr_t uptr32;
3072 struct ifreq __user *uifr;
3074 uifr = compat_alloc_user_space(sizeof(*uifr));
3075 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3078 if (get_user(uptr32, &uifr32->ifr_data))
3081 uptr = compat_ptr(uptr32);
3083 if (put_user(uptr, &uifr->ifr_data))
3086 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3091 struct sockaddr rt_dst; /* target address */
3092 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3093 struct sockaddr rt_genmask; /* target network mask (IP) */
3094 unsigned short rt_flags;
3097 unsigned char rt_tos;
3098 unsigned char rt_class;
3100 short rt_metric; /* +1 for binary compatibility! */
3101 /* char * */ u32 rt_dev; /* forcing the device at add */
3102 u32 rt_mtu; /* per route MTU/Window */
3103 u32 rt_window; /* Window clamping */
3104 unsigned short rt_irtt; /* Initial RTT */
3107 struct in6_rtmsg32 {
3108 struct in6_addr rtmsg_dst;
3109 struct in6_addr rtmsg_src;
3110 struct in6_addr rtmsg_gateway;
3120 static int routing_ioctl(struct net *net, struct socket *sock,
3121 unsigned int cmd, void __user *argp)
3125 struct in6_rtmsg r6;
3129 mm_segment_t old_fs = get_fs();
3131 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3132 struct in6_rtmsg32 __user *ur6 = argp;
3133 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3134 3 * sizeof(struct in6_addr));
3135 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3136 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3137 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3138 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3139 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3140 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3141 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3145 struct rtentry32 __user *ur4 = argp;
3146 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3147 3 * sizeof(struct sockaddr));
3148 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3149 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3150 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3151 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3152 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3153 ret |= __get_user(rtdev, &(ur4->rt_dev));
3155 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3156 r4.rt_dev = (char __user __force *)devname;
3170 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3177 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3178 * for some operations; this forces use of the newer bridge-utils that
3179 * use compatible ioctls
3181 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3185 if (get_user(tmp, argp))
3187 if (tmp == BRCTL_GET_VERSION)
3188 return BRCTL_VERSION + 1;
3192 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3193 unsigned int cmd, unsigned long arg)
3195 void __user *argp = compat_ptr(arg);
3196 struct sock *sk = sock->sk;
3197 struct net *net = sock_net(sk);
3199 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3200 return siocdevprivate_ioctl(net, cmd, argp);
3205 return old_bridge_ioctl(argp);
3207 return dev_ifname32(net, argp);
3209 return dev_ifconf(net, argp);
3211 return ethtool_ioctl(net, argp);
3213 return compat_siocwandev(net, argp);
3216 return compat_sioc_ifmap(net, cmd, argp);
3217 case SIOCBONDENSLAVE:
3218 case SIOCBONDRELEASE:
3219 case SIOCBONDSETHWADDR:
3220 case SIOCBONDSLAVEINFOQUERY:
3221 case SIOCBONDINFOQUERY:
3222 case SIOCBONDCHANGEACTIVE:
3223 return bond_ioctl(net, cmd, argp);
3226 return routing_ioctl(net, sock, cmd, argp);
3228 return do_siocgstamp(net, sock, cmd, argp);
3230 return do_siocgstampns(net, sock, cmd, argp);
3232 return compat_siocshwtstamp(net, argp);
3244 return sock_ioctl(file, cmd, arg);
3261 case SIOCSIFHWBROADCAST:
3263 case SIOCGIFBRDADDR:
3264 case SIOCSIFBRDADDR:
3265 case SIOCGIFDSTADDR:
3266 case SIOCSIFDSTADDR:
3267 case SIOCGIFNETMASK:
3268 case SIOCSIFNETMASK:
3279 return dev_ifsioc(net, sock, cmd, argp);
3285 return sock_do_ioctl(net, sock, cmd, arg);
3288 return -ENOIOCTLCMD;
3291 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3294 struct socket *sock = file->private_data;
3295 int ret = -ENOIOCTLCMD;
3302 if (sock->ops->compat_ioctl)
3303 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3305 if (ret == -ENOIOCTLCMD &&
3306 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3307 ret = compat_wext_handle_ioctl(net, cmd, arg);
3309 if (ret == -ENOIOCTLCMD)
3310 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3316 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3318 return sock->ops->bind(sock, addr, addrlen);
3320 EXPORT_SYMBOL(kernel_bind);
3322 int kernel_listen(struct socket *sock, int backlog)
3324 return sock->ops->listen(sock, backlog);
3326 EXPORT_SYMBOL(kernel_listen);
3328 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3330 struct sock *sk = sock->sk;
3333 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3338 err = sock->ops->accept(sock, *newsock, flags);
3340 sock_release(*newsock);
3345 (*newsock)->ops = sock->ops;
3346 __module_get((*newsock)->ops->owner);
3351 EXPORT_SYMBOL(kernel_accept);
3353 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3356 return sock->ops->connect(sock, addr, addrlen, flags);
3358 EXPORT_SYMBOL(kernel_connect);
3360 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3363 return sock->ops->getname(sock, addr, addrlen, 0);
3365 EXPORT_SYMBOL(kernel_getsockname);
3367 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3370 return sock->ops->getname(sock, addr, addrlen, 1);
3372 EXPORT_SYMBOL(kernel_getpeername);
3374 int kernel_getsockopt(struct socket *sock, int level, int optname,
3375 char *optval, int *optlen)
3377 mm_segment_t oldfs = get_fs();
3378 char __user *uoptval;
3379 int __user *uoptlen;
3382 uoptval = (char __user __force *) optval;
3383 uoptlen = (int __user __force *) optlen;
3386 if (level == SOL_SOCKET)
3387 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3389 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3394 EXPORT_SYMBOL(kernel_getsockopt);
3396 int kernel_setsockopt(struct socket *sock, int level, int optname,
3397 char *optval, unsigned int optlen)
3399 mm_segment_t oldfs = get_fs();
3400 char __user *uoptval;
3403 uoptval = (char __user __force *) optval;
3406 if (level == SOL_SOCKET)
3407 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3409 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3414 EXPORT_SYMBOL(kernel_setsockopt);
3416 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3417 size_t size, int flags)
3419 if (sock->ops->sendpage)
3420 return sock->ops->sendpage(sock, page, offset, size, flags);
3422 return sock_no_sendpage(sock, page, offset, size, flags);
3424 EXPORT_SYMBOL(kernel_sendpage);
3426 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3428 mm_segment_t oldfs = get_fs();
3432 err = sock->ops->ioctl(sock, cmd, arg);
3437 EXPORT_SYMBOL(kernel_sock_ioctl);
3439 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3441 return sock->ops->shutdown(sock, how);
3443 EXPORT_SYMBOL(kernel_sock_shutdown);