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/ll_poll.h>
109 #ifdef CONFIG_NET_LL_RX_POLL
110 unsigned int sysctl_net_ll_poll __read_mostly;
113 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
114 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
115 unsigned long nr_segs, loff_t pos);
116 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
117 unsigned long nr_segs, loff_t pos);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
140 static const struct file_operations socket_file_ops = {
141 .owner = THIS_MODULE,
143 .aio_read = sock_aio_read,
144 .aio_write = sock_aio_write,
146 .unlocked_ioctl = sock_ioctl,
148 .compat_ioctl = compat_sock_ioctl,
151 .open = sock_no_open, /* special open code to disallow open via /proc */
152 .release = sock_close,
153 .fasync = sock_fasync,
154 .sendpage = sock_sendpage,
155 .splice_write = generic_splice_sendpage,
156 .splice_read = sock_splice_read,
160 * The protocol list. Each protocol is registered in here.
163 static DEFINE_SPINLOCK(net_family_lock);
164 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
167 * Statistics counters of the socket lists
170 static DEFINE_PER_CPU(int, sockets_in_use);
174 * Move socket addresses back and forth across the kernel/user
175 * divide and look after the messy bits.
179 * move_addr_to_kernel - copy a socket address into kernel space
180 * @uaddr: Address in user space
181 * @kaddr: Address in kernel space
182 * @ulen: Length in user space
184 * The address is copied into kernel space. If the provided address is
185 * too long an error code of -EINVAL is returned. If the copy gives
186 * invalid addresses -EFAULT is returned. On a success 0 is returned.
189 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195 if (copy_from_user(kaddr, uaddr, ulen))
197 return audit_sockaddr(ulen, kaddr);
201 * move_addr_to_user - copy an address to user space
202 * @kaddr: kernel space address
203 * @klen: length of address in kernel
204 * @uaddr: user space address
205 * @ulen: pointer to user length field
207 * The value pointed to by ulen on entry is the buffer length available.
208 * This is overwritten with the buffer space used. -EINVAL is returned
209 * if an overlong buffer is specified or a negative buffer size. -EFAULT
210 * is returned if either the buffer or the length field are not
212 * After copying the data up to the limit the user specifies, the true
213 * length of the data is written over the length limit the user
214 * specified. Zero is returned for a success.
217 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
218 void __user *uaddr, int __user *ulen)
223 err = get_user(len, ulen);
228 if (len < 0 || len > sizeof(struct sockaddr_storage))
231 if (audit_sockaddr(klen, kaddr))
233 if (copy_to_user(uaddr, kaddr, len))
237 * "fromlen shall refer to the value before truncation.."
240 return __put_user(klen, ulen);
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
245 static struct inode *sock_alloc_inode(struct super_block *sb)
247 struct socket_alloc *ei;
248 struct socket_wq *wq;
250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 kmem_cache_free(sock_inode_cachep, ei);
258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
260 RCU_INIT_POINTER(ei->socket.wq, wq);
262 ei->socket.state = SS_UNCONNECTED;
263 ei->socket.flags = 0;
264 ei->socket.ops = NULL;
265 ei->socket.sk = NULL;
266 ei->socket.file = NULL;
268 return &ei->vfs_inode;
271 static void sock_destroy_inode(struct inode *inode)
273 struct socket_alloc *ei;
274 struct socket_wq *wq;
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 wq = rcu_dereference_protected(ei->socket.wq, 1);
279 kmem_cache_free(sock_inode_cachep, ei);
282 static void init_once(void *foo)
284 struct socket_alloc *ei = (struct socket_alloc *)foo;
286 inode_init_once(&ei->vfs_inode);
289 static int init_inodecache(void)
291 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
292 sizeof(struct socket_alloc),
294 (SLAB_HWCACHE_ALIGN |
295 SLAB_RECLAIM_ACCOUNT |
298 if (sock_inode_cachep == NULL)
303 static const struct super_operations sockfs_ops = {
304 .alloc_inode = sock_alloc_inode,
305 .destroy_inode = sock_destroy_inode,
306 .statfs = simple_statfs,
310 * sockfs_dname() is called from d_path().
312 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
315 dentry->d_inode->i_ino);
318 static const struct dentry_operations sockfs_dentry_operations = {
319 .d_dname = sockfs_dname,
322 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
323 int flags, const char *dev_name, void *data)
325 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
326 &sockfs_dentry_operations, SOCKFS_MAGIC);
329 static struct vfsmount *sock_mnt __read_mostly;
331 static struct file_system_type sock_fs_type = {
333 .mount = sockfs_mount,
334 .kill_sb = kill_anon_super,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356 struct qstr name = { .name = "" };
362 name.len = strlen(name.name);
363 } else if (sock->sk) {
364 name.name = sock->sk->sk_prot_creator->name;
365 name.len = strlen(name.name);
367 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
368 if (unlikely(!path.dentry))
369 return ERR_PTR(-ENOMEM);
370 path.mnt = mntget(sock_mnt);
372 d_instantiate(path.dentry, SOCK_INODE(sock));
373 SOCK_INODE(sock)->i_fop = &socket_file_ops;
375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
377 if (unlikely(IS_ERR(file))) {
378 /* drop dentry, keep inode */
379 ihold(path.dentry->d_inode);
385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
386 file->private_data = sock;
389 EXPORT_SYMBOL(sock_alloc_file);
391 static int sock_map_fd(struct socket *sock, int flags)
393 struct file *newfile;
394 int fd = get_unused_fd_flags(flags);
395 if (unlikely(fd < 0))
398 newfile = sock_alloc_file(sock, flags, NULL);
399 if (likely(!IS_ERR(newfile))) {
400 fd_install(fd, newfile);
405 return PTR_ERR(newfile);
408 struct socket *sock_from_file(struct file *file, int *err)
410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
416 EXPORT_SYMBOL(sock_from_file);
419 * sockfd_lookup - Go from a file number to its socket slot
421 * @err: pointer to an error code return
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
428 * On a success the socket object pointer is returned.
431 struct socket *sockfd_lookup(int fd, int *err)
442 sock = sock_from_file(file, err);
447 EXPORT_SYMBOL(sockfd_lookup);
449 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
455 file = fget_light(fd, fput_needed);
457 sock = sock_from_file(file, err);
460 fput_light(file, *fput_needed);
465 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
466 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
467 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
468 static ssize_t sockfs_getxattr(struct dentry *dentry,
469 const char *name, void *value, size_t size)
471 const char *proto_name;
476 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
477 proto_name = dentry->d_name.name;
478 proto_size = strlen(proto_name);
482 if (proto_size + 1 > size)
485 strncpy(value, proto_name, proto_size + 1);
487 error = proto_size + 1;
494 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
500 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
510 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
515 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
522 static const struct inode_operations sockfs_inode_ops = {
523 .getxattr = sockfs_getxattr,
524 .listxattr = sockfs_listxattr,
528 * sock_alloc - allocate a socket
530 * Allocate a new inode and socket object. The two are bound together
531 * and initialised. The socket is then returned. If we are out of inodes
535 static struct socket *sock_alloc(void)
540 inode = new_inode_pseudo(sock_mnt->mnt_sb);
544 sock = SOCKET_I(inode);
546 kmemcheck_annotate_bitfield(sock, type);
547 inode->i_ino = get_next_ino();
548 inode->i_mode = S_IFSOCK | S_IRWXUGO;
549 inode->i_uid = current_fsuid();
550 inode->i_gid = current_fsgid();
551 inode->i_op = &sockfs_inode_ops;
553 this_cpu_add(sockets_in_use, 1);
558 * In theory you can't get an open on this inode, but /proc provides
559 * a back door. Remember to keep it shut otherwise you'll let the
560 * creepy crawlies in.
563 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
568 const struct file_operations bad_sock_fops = {
569 .owner = THIS_MODULE,
570 .open = sock_no_open,
571 .llseek = noop_llseek,
575 * sock_release - close a socket
576 * @sock: socket to close
578 * The socket is released from the protocol stack if it has a release
579 * callback, and the inode is then released if the socket is bound to
580 * an inode not a file.
583 void sock_release(struct socket *sock)
586 struct module *owner = sock->ops->owner;
588 sock->ops->release(sock);
593 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
594 printk(KERN_ERR "sock_release: fasync list not empty!\n");
596 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
599 this_cpu_sub(sockets_in_use, 1);
601 iput(SOCK_INODE(sock));
606 EXPORT_SYMBOL(sock_release);
608 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
611 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
612 *tx_flags |= SKBTX_HW_TSTAMP;
613 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
614 *tx_flags |= SKBTX_SW_TSTAMP;
615 if (sock_flag(sk, SOCK_WIFI_STATUS))
616 *tx_flags |= SKBTX_WIFI_STATUS;
618 EXPORT_SYMBOL(sock_tx_timestamp);
620 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
621 struct msghdr *msg, size_t size)
623 struct sock_iocb *si = kiocb_to_siocb(iocb);
630 return sock->ops->sendmsg(iocb, sock, msg, size);
633 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
634 struct msghdr *msg, size_t size)
636 int err = security_socket_sendmsg(sock, msg, size);
638 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
641 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
644 struct sock_iocb siocb;
647 init_sync_kiocb(&iocb, NULL);
648 iocb.private = &siocb;
649 ret = __sock_sendmsg(&iocb, sock, msg, size);
650 if (-EIOCBQUEUED == ret)
651 ret = wait_on_sync_kiocb(&iocb);
654 EXPORT_SYMBOL(sock_sendmsg);
656 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
659 struct sock_iocb siocb;
662 init_sync_kiocb(&iocb, NULL);
663 iocb.private = &siocb;
664 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
665 if (-EIOCBQUEUED == ret)
666 ret = wait_on_sync_kiocb(&iocb);
670 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
671 struct kvec *vec, size_t num, size_t size)
673 mm_segment_t oldfs = get_fs();
678 * the following is safe, since for compiler definitions of kvec and
679 * iovec are identical, yielding the same in-core layout and alignment
681 msg->msg_iov = (struct iovec *)vec;
682 msg->msg_iovlen = num;
683 result = sock_sendmsg(sock, msg, size);
687 EXPORT_SYMBOL(kernel_sendmsg);
690 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
692 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
695 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
696 struct timespec ts[3];
698 struct skb_shared_hwtstamps *shhwtstamps =
701 /* Race occurred between timestamp enabling and packet
702 receiving. Fill in the current time for now. */
703 if (need_software_tstamp && skb->tstamp.tv64 == 0)
704 __net_timestamp(skb);
706 if (need_software_tstamp) {
707 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
709 skb_get_timestamp(skb, &tv);
710 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
713 skb_get_timestampns(skb, &ts[0]);
714 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
715 sizeof(ts[0]), &ts[0]);
720 memset(ts, 0, sizeof(ts));
721 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
722 ktime_to_timespec_cond(skb->tstamp, ts + 0))
725 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
726 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
728 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
729 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
733 put_cmsg(msg, SOL_SOCKET,
734 SCM_TIMESTAMPING, sizeof(ts), &ts);
736 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
738 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
743 if (!sock_flag(sk, SOCK_WIFI_STATUS))
745 if (!skb->wifi_acked_valid)
748 ack = skb->wifi_acked;
750 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
752 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
754 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
757 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
758 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
759 sizeof(__u32), &skb->dropcount);
762 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
765 sock_recv_timestamp(msg, sk, skb);
766 sock_recv_drops(msg, sk, skb);
768 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
770 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
771 struct msghdr *msg, size_t size, int flags)
773 struct sock_iocb *si = kiocb_to_siocb(iocb);
781 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
784 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
785 struct msghdr *msg, size_t size, int flags)
787 int err = security_socket_recvmsg(sock, msg, size, flags);
789 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
792 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
793 size_t size, int flags)
796 struct sock_iocb siocb;
799 init_sync_kiocb(&iocb, NULL);
800 iocb.private = &siocb;
801 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
802 if (-EIOCBQUEUED == ret)
803 ret = wait_on_sync_kiocb(&iocb);
806 EXPORT_SYMBOL(sock_recvmsg);
808 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
809 size_t size, int flags)
812 struct sock_iocb siocb;
815 init_sync_kiocb(&iocb, NULL);
816 iocb.private = &siocb;
817 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
818 if (-EIOCBQUEUED == ret)
819 ret = wait_on_sync_kiocb(&iocb);
824 * kernel_recvmsg - Receive a message from a socket (kernel space)
825 * @sock: The socket to receive the message from
826 * @msg: Received message
827 * @vec: Input s/g array for message data
828 * @num: Size of input s/g array
829 * @size: Number of bytes to read
830 * @flags: Message flags (MSG_DONTWAIT, etc...)
832 * On return the msg structure contains the scatter/gather array passed in the
833 * vec argument. The array is modified so that it consists of the unfilled
834 * portion of the original array.
836 * The returned value is the total number of bytes received, or an error.
838 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
839 struct kvec *vec, size_t num, size_t size, int flags)
841 mm_segment_t oldfs = get_fs();
846 * the following is safe, since for compiler definitions of kvec and
847 * iovec are identical, yielding the same in-core layout and alignment
849 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
850 result = sock_recvmsg(sock, msg, size, flags);
854 EXPORT_SYMBOL(kernel_recvmsg);
856 static void sock_aio_dtor(struct kiocb *iocb)
858 kfree(iocb->private);
861 static ssize_t sock_sendpage(struct file *file, struct page *page,
862 int offset, size_t size, loff_t *ppos, int more)
867 sock = file->private_data;
869 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
870 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
873 return kernel_sendpage(sock, page, offset, size, flags);
876 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
877 struct pipe_inode_info *pipe, size_t len,
880 struct socket *sock = file->private_data;
882 if (unlikely(!sock->ops->splice_read))
885 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
888 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
889 struct sock_iocb *siocb)
891 if (!is_sync_kiocb(iocb)) {
892 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
895 iocb->ki_dtor = sock_aio_dtor;
899 iocb->private = siocb;
903 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
904 struct file *file, const struct iovec *iov,
905 unsigned long nr_segs)
907 struct socket *sock = file->private_data;
911 for (i = 0; i < nr_segs; i++)
912 size += iov[i].iov_len;
914 msg->msg_name = NULL;
915 msg->msg_namelen = 0;
916 msg->msg_control = NULL;
917 msg->msg_controllen = 0;
918 msg->msg_iov = (struct iovec *)iov;
919 msg->msg_iovlen = nr_segs;
920 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
922 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
925 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
926 unsigned long nr_segs, loff_t pos)
928 struct sock_iocb siocb, *x;
933 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
937 x = alloc_sock_iocb(iocb, &siocb);
940 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
943 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
944 struct file *file, const struct iovec *iov,
945 unsigned long nr_segs)
947 struct socket *sock = file->private_data;
951 for (i = 0; i < nr_segs; i++)
952 size += iov[i].iov_len;
954 msg->msg_name = NULL;
955 msg->msg_namelen = 0;
956 msg->msg_control = NULL;
957 msg->msg_controllen = 0;
958 msg->msg_iov = (struct iovec *)iov;
959 msg->msg_iovlen = nr_segs;
960 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
961 if (sock->type == SOCK_SEQPACKET)
962 msg->msg_flags |= MSG_EOR;
964 return __sock_sendmsg(iocb, sock, msg, size);
967 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
968 unsigned long nr_segs, loff_t pos)
970 struct sock_iocb siocb, *x;
975 x = alloc_sock_iocb(iocb, &siocb);
979 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
983 * Atomic setting of ioctl hooks to avoid race
984 * with module unload.
987 static DEFINE_MUTEX(br_ioctl_mutex);
988 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
990 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
992 mutex_lock(&br_ioctl_mutex);
993 br_ioctl_hook = hook;
994 mutex_unlock(&br_ioctl_mutex);
996 EXPORT_SYMBOL(brioctl_set);
998 static DEFINE_MUTEX(vlan_ioctl_mutex);
999 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1001 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1003 mutex_lock(&vlan_ioctl_mutex);
1004 vlan_ioctl_hook = hook;
1005 mutex_unlock(&vlan_ioctl_mutex);
1007 EXPORT_SYMBOL(vlan_ioctl_set);
1009 static DEFINE_MUTEX(dlci_ioctl_mutex);
1010 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1012 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1014 mutex_lock(&dlci_ioctl_mutex);
1015 dlci_ioctl_hook = hook;
1016 mutex_unlock(&dlci_ioctl_mutex);
1018 EXPORT_SYMBOL(dlci_ioctl_set);
1020 static long sock_do_ioctl(struct net *net, struct socket *sock,
1021 unsigned int cmd, unsigned long arg)
1024 void __user *argp = (void __user *)arg;
1026 err = sock->ops->ioctl(sock, cmd, arg);
1029 * If this ioctl is unknown try to hand it down
1030 * to the NIC driver.
1032 if (err == -ENOIOCTLCMD)
1033 err = dev_ioctl(net, cmd, argp);
1039 * With an ioctl, arg may well be a user mode pointer, but we don't know
1040 * what to do with it - that's up to the protocol still.
1043 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1045 struct socket *sock;
1047 void __user *argp = (void __user *)arg;
1051 sock = file->private_data;
1054 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1055 err = dev_ioctl(net, cmd, argp);
1057 #ifdef CONFIG_WEXT_CORE
1058 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1059 err = dev_ioctl(net, cmd, argp);
1066 if (get_user(pid, (int __user *)argp))
1068 err = f_setown(sock->file, pid, 1);
1072 err = put_user(f_getown(sock->file),
1073 (int __user *)argp);
1081 request_module("bridge");
1083 mutex_lock(&br_ioctl_mutex);
1085 err = br_ioctl_hook(net, cmd, argp);
1086 mutex_unlock(&br_ioctl_mutex);
1091 if (!vlan_ioctl_hook)
1092 request_module("8021q");
1094 mutex_lock(&vlan_ioctl_mutex);
1095 if (vlan_ioctl_hook)
1096 err = vlan_ioctl_hook(net, argp);
1097 mutex_unlock(&vlan_ioctl_mutex);
1102 if (!dlci_ioctl_hook)
1103 request_module("dlci");
1105 mutex_lock(&dlci_ioctl_mutex);
1106 if (dlci_ioctl_hook)
1107 err = dlci_ioctl_hook(cmd, argp);
1108 mutex_unlock(&dlci_ioctl_mutex);
1111 err = sock_do_ioctl(net, sock, cmd, arg);
1117 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1120 struct socket *sock = NULL;
1122 err = security_socket_create(family, type, protocol, 1);
1126 sock = sock_alloc();
1133 err = security_socket_post_create(sock, family, type, protocol, 1);
1145 EXPORT_SYMBOL(sock_create_lite);
1147 /* No kernel lock held - perfect */
1148 static unsigned int sock_poll(struct file *file, poll_table *wait)
1150 struct socket *sock;
1153 * We can't return errors to poll, so it's either yes or no.
1155 sock = file->private_data;
1156 return sock->ops->poll(file, sock, wait);
1159 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1161 struct socket *sock = file->private_data;
1163 return sock->ops->mmap(file, sock, vma);
1166 static int sock_close(struct inode *inode, struct file *filp)
1168 sock_release(SOCKET_I(inode));
1173 * Update the socket async list
1175 * Fasync_list locking strategy.
1177 * 1. fasync_list is modified only under process context socket lock
1178 * i.e. under semaphore.
1179 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1180 * or under socket lock
1183 static int sock_fasync(int fd, struct file *filp, int on)
1185 struct socket *sock = filp->private_data;
1186 struct sock *sk = sock->sk;
1187 struct socket_wq *wq;
1193 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1194 fasync_helper(fd, filp, on, &wq->fasync_list);
1196 if (!wq->fasync_list)
1197 sock_reset_flag(sk, SOCK_FASYNC);
1199 sock_set_flag(sk, SOCK_FASYNC);
1205 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1207 int sock_wake_async(struct socket *sock, int how, int band)
1209 struct socket_wq *wq;
1214 wq = rcu_dereference(sock->wq);
1215 if (!wq || !wq->fasync_list) {
1220 case SOCK_WAKE_WAITD:
1221 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1224 case SOCK_WAKE_SPACE:
1225 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1230 kill_fasync(&wq->fasync_list, SIGIO, band);
1233 kill_fasync(&wq->fasync_list, SIGURG, band);
1238 EXPORT_SYMBOL(sock_wake_async);
1240 int __sock_create(struct net *net, int family, int type, int protocol,
1241 struct socket **res, int kern)
1244 struct socket *sock;
1245 const struct net_proto_family *pf;
1248 * Check protocol is in range
1250 if (family < 0 || family >= NPROTO)
1251 return -EAFNOSUPPORT;
1252 if (type < 0 || type >= SOCK_MAX)
1257 This uglymoron is moved from INET layer to here to avoid
1258 deadlock in module load.
1260 if (family == PF_INET && type == SOCK_PACKET) {
1264 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1270 err = security_socket_create(family, type, protocol, kern);
1275 * Allocate the socket and allow the family to set things up. if
1276 * the protocol is 0, the family is instructed to select an appropriate
1279 sock = sock_alloc();
1281 net_warn_ratelimited("socket: no more sockets\n");
1282 return -ENFILE; /* Not exactly a match, but its the
1283 closest posix thing */
1288 #ifdef CONFIG_MODULES
1289 /* Attempt to load a protocol module if the find failed.
1291 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1292 * requested real, full-featured networking support upon configuration.
1293 * Otherwise module support will break!
1295 if (rcu_access_pointer(net_families[family]) == NULL)
1296 request_module("net-pf-%d", family);
1300 pf = rcu_dereference(net_families[family]);
1301 err = -EAFNOSUPPORT;
1306 * We will call the ->create function, that possibly is in a loadable
1307 * module, so we have to bump that loadable module refcnt first.
1309 if (!try_module_get(pf->owner))
1312 /* Now protected by module ref count */
1315 err = pf->create(net, sock, protocol, kern);
1317 goto out_module_put;
1320 * Now to bump the refcnt of the [loadable] module that owns this
1321 * socket at sock_release time we decrement its refcnt.
1323 if (!try_module_get(sock->ops->owner))
1324 goto out_module_busy;
1327 * Now that we're done with the ->create function, the [loadable]
1328 * module can have its refcnt decremented
1330 module_put(pf->owner);
1331 err = security_socket_post_create(sock, family, type, protocol, kern);
1333 goto out_sock_release;
1339 err = -EAFNOSUPPORT;
1342 module_put(pf->owner);
1349 goto out_sock_release;
1351 EXPORT_SYMBOL(__sock_create);
1353 int sock_create(int family, int type, int protocol, struct socket **res)
1355 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1357 EXPORT_SYMBOL(sock_create);
1359 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1361 return __sock_create(&init_net, family, type, protocol, res, 1);
1363 EXPORT_SYMBOL(sock_create_kern);
1365 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1368 struct socket *sock;
1371 /* Check the SOCK_* constants for consistency. */
1372 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1373 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1374 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1375 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1377 flags = type & ~SOCK_TYPE_MASK;
1378 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1380 type &= SOCK_TYPE_MASK;
1382 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1383 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1385 retval = sock_create(family, type, protocol, &sock);
1389 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1394 /* It may be already another descriptor 8) Not kernel problem. */
1403 * Create a pair of connected sockets.
1406 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1407 int __user *, usockvec)
1409 struct socket *sock1, *sock2;
1411 struct file *newfile1, *newfile2;
1414 flags = type & ~SOCK_TYPE_MASK;
1415 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1417 type &= SOCK_TYPE_MASK;
1419 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1420 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1423 * Obtain the first socket and check if the underlying protocol
1424 * supports the socketpair call.
1427 err = sock_create(family, type, protocol, &sock1);
1431 err = sock_create(family, type, protocol, &sock2);
1435 err = sock1->ops->socketpair(sock1, sock2);
1437 goto out_release_both;
1439 fd1 = get_unused_fd_flags(flags);
1440 if (unlikely(fd1 < 0)) {
1442 goto out_release_both;
1444 fd2 = get_unused_fd_flags(flags);
1445 if (unlikely(fd2 < 0)) {
1448 goto out_release_both;
1451 newfile1 = sock_alloc_file(sock1, flags, NULL);
1452 if (unlikely(IS_ERR(newfile1))) {
1453 err = PTR_ERR(newfile1);
1456 goto out_release_both;
1459 newfile2 = sock_alloc_file(sock2, flags, NULL);
1460 if (IS_ERR(newfile2)) {
1461 err = PTR_ERR(newfile2);
1465 sock_release(sock2);
1469 audit_fd_pair(fd1, fd2);
1470 fd_install(fd1, newfile1);
1471 fd_install(fd2, newfile2);
1472 /* fd1 and fd2 may be already another descriptors.
1473 * Not kernel problem.
1476 err = put_user(fd1, &usockvec[0]);
1478 err = put_user(fd2, &usockvec[1]);
1487 sock_release(sock2);
1489 sock_release(sock1);
1495 * Bind a name to a socket. Nothing much to do here since it's
1496 * the protocol's responsibility to handle the local address.
1498 * We move the socket address to kernel space before we call
1499 * the protocol layer (having also checked the address is ok).
1502 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1504 struct socket *sock;
1505 struct sockaddr_storage address;
1506 int err, fput_needed;
1508 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1510 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1512 err = security_socket_bind(sock,
1513 (struct sockaddr *)&address,
1516 err = sock->ops->bind(sock,
1520 fput_light(sock->file, fput_needed);
1526 * Perform a listen. Basically, we allow the protocol to do anything
1527 * necessary for a listen, and if that works, we mark the socket as
1528 * ready for listening.
1531 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1533 struct socket *sock;
1534 int err, fput_needed;
1537 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1539 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1540 if ((unsigned int)backlog > somaxconn)
1541 backlog = somaxconn;
1543 err = security_socket_listen(sock, backlog);
1545 err = sock->ops->listen(sock, backlog);
1547 fput_light(sock->file, fput_needed);
1553 * For accept, we attempt to create a new socket, set up the link
1554 * with the client, wake up the client, then return the new
1555 * connected fd. We collect the address of the connector in kernel
1556 * space and move it to user at the very end. This is unclean because
1557 * we open the socket then return an error.
1559 * 1003.1g adds the ability to recvmsg() to query connection pending
1560 * status to recvmsg. We need to add that support in a way thats
1561 * clean when we restucture accept also.
1564 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1565 int __user *, upeer_addrlen, int, flags)
1567 struct socket *sock, *newsock;
1568 struct file *newfile;
1569 int err, len, newfd, fput_needed;
1570 struct sockaddr_storage address;
1572 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1575 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1576 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1578 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1583 newsock = sock_alloc();
1587 newsock->type = sock->type;
1588 newsock->ops = sock->ops;
1591 * We don't need try_module_get here, as the listening socket (sock)
1592 * has the protocol module (sock->ops->owner) held.
1594 __module_get(newsock->ops->owner);
1596 newfd = get_unused_fd_flags(flags);
1597 if (unlikely(newfd < 0)) {
1599 sock_release(newsock);
1602 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1603 if (unlikely(IS_ERR(newfile))) {
1604 err = PTR_ERR(newfile);
1605 put_unused_fd(newfd);
1606 sock_release(newsock);
1610 err = security_socket_accept(sock, newsock);
1614 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1618 if (upeer_sockaddr) {
1619 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1621 err = -ECONNABORTED;
1624 err = move_addr_to_user(&address,
1625 len, upeer_sockaddr, upeer_addrlen);
1630 /* File flags are not inherited via accept() unlike another OSes. */
1632 fd_install(newfd, newfile);
1636 fput_light(sock->file, fput_needed);
1641 put_unused_fd(newfd);
1645 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1646 int __user *, upeer_addrlen)
1648 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1652 * Attempt to connect to a socket with the server address. The address
1653 * is in user space so we verify it is OK and move it to kernel space.
1655 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1658 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1659 * other SEQPACKET protocols that take time to connect() as it doesn't
1660 * include the -EINPROGRESS status for such sockets.
1663 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1666 struct socket *sock;
1667 struct sockaddr_storage address;
1668 int err, fput_needed;
1670 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1673 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1678 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1682 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1683 sock->file->f_flags);
1685 fput_light(sock->file, fput_needed);
1691 * Get the local address ('name') of a socket object. Move the obtained
1692 * name to user space.
1695 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1696 int __user *, usockaddr_len)
1698 struct socket *sock;
1699 struct sockaddr_storage address;
1700 int len, err, fput_needed;
1702 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706 err = security_socket_getsockname(sock);
1710 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1713 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1716 fput_light(sock->file, fput_needed);
1722 * Get the remote address ('name') of a socket object. Move the obtained
1723 * name to user space.
1726 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1727 int __user *, usockaddr_len)
1729 struct socket *sock;
1730 struct sockaddr_storage address;
1731 int len, err, fput_needed;
1733 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1735 err = security_socket_getpeername(sock);
1737 fput_light(sock->file, fput_needed);
1742 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1745 err = move_addr_to_user(&address, len, usockaddr,
1747 fput_light(sock->file, fput_needed);
1753 * Send a datagram to a given address. We move the address into kernel
1754 * space and check the user space data area is readable before invoking
1758 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1759 unsigned int, flags, struct sockaddr __user *, addr,
1762 struct socket *sock;
1763 struct sockaddr_storage address;
1771 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1775 iov.iov_base = buff;
1777 msg.msg_name = NULL;
1780 msg.msg_control = NULL;
1781 msg.msg_controllen = 0;
1782 msg.msg_namelen = 0;
1784 err = move_addr_to_kernel(addr, addr_len, &address);
1787 msg.msg_name = (struct sockaddr *)&address;
1788 msg.msg_namelen = addr_len;
1790 if (sock->file->f_flags & O_NONBLOCK)
1791 flags |= MSG_DONTWAIT;
1792 msg.msg_flags = flags;
1793 err = sock_sendmsg(sock, &msg, len);
1796 fput_light(sock->file, fput_needed);
1802 * Send a datagram down a socket.
1805 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1806 unsigned int, flags)
1808 return sys_sendto(fd, buff, len, flags, NULL, 0);
1812 * Receive a frame from the socket and optionally record the address of the
1813 * sender. We verify the buffers are writable and if needed move the
1814 * sender address from kernel to user space.
1817 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1818 unsigned int, flags, struct sockaddr __user *, addr,
1819 int __user *, addr_len)
1821 struct socket *sock;
1824 struct sockaddr_storage address;
1830 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1834 msg.msg_control = NULL;
1835 msg.msg_controllen = 0;
1839 iov.iov_base = ubuf;
1840 msg.msg_name = (struct sockaddr *)&address;
1841 msg.msg_namelen = sizeof(address);
1842 if (sock->file->f_flags & O_NONBLOCK)
1843 flags |= MSG_DONTWAIT;
1844 err = sock_recvmsg(sock, &msg, size, flags);
1846 if (err >= 0 && addr != NULL) {
1847 err2 = move_addr_to_user(&address,
1848 msg.msg_namelen, addr, addr_len);
1853 fput_light(sock->file, fput_needed);
1859 * Receive a datagram from a socket.
1862 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1865 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1869 * Set a socket option. Because we don't know the option lengths we have
1870 * to pass the user mode parameter for the protocols to sort out.
1873 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1874 char __user *, optval, int, optlen)
1876 int err, fput_needed;
1877 struct socket *sock;
1882 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1884 err = security_socket_setsockopt(sock, level, optname);
1888 if (level == SOL_SOCKET)
1890 sock_setsockopt(sock, level, optname, optval,
1894 sock->ops->setsockopt(sock, level, optname, optval,
1897 fput_light(sock->file, fput_needed);
1903 * Get a socket option. Because we don't know the option lengths we have
1904 * to pass a user mode parameter for the protocols to sort out.
1907 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1908 char __user *, optval, int __user *, optlen)
1910 int err, fput_needed;
1911 struct socket *sock;
1913 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1915 err = security_socket_getsockopt(sock, level, optname);
1919 if (level == SOL_SOCKET)
1921 sock_getsockopt(sock, level, optname, optval,
1925 sock->ops->getsockopt(sock, level, optname, optval,
1928 fput_light(sock->file, fput_needed);
1934 * Shutdown a socket.
1937 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1939 int err, fput_needed;
1940 struct socket *sock;
1942 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1944 err = security_socket_shutdown(sock, how);
1946 err = sock->ops->shutdown(sock, how);
1947 fput_light(sock->file, fput_needed);
1952 /* A couple of helpful macros for getting the address of the 32/64 bit
1953 * fields which are the same type (int / unsigned) on our platforms.
1955 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1956 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1957 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1959 struct used_address {
1960 struct sockaddr_storage name;
1961 unsigned int name_len;
1964 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1965 struct msghdr *msg_sys, unsigned int flags,
1966 struct used_address *used_address)
1968 struct compat_msghdr __user *msg_compat =
1969 (struct compat_msghdr __user *)msg;
1970 struct sockaddr_storage address;
1971 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1972 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1973 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1974 /* 20 is size of ipv6_pktinfo */
1975 unsigned char *ctl_buf = ctl;
1976 int err, ctl_len, total_len;
1979 if (MSG_CMSG_COMPAT & flags) {
1980 if (get_compat_msghdr(msg_sys, msg_compat))
1982 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1985 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1987 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1990 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
1996 /* This will also move the address data into kernel space */
1997 if (MSG_CMSG_COMPAT & flags) {
1998 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2000 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2007 if (msg_sys->msg_controllen > INT_MAX)
2009 ctl_len = msg_sys->msg_controllen;
2010 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2012 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2016 ctl_buf = msg_sys->msg_control;
2017 ctl_len = msg_sys->msg_controllen;
2018 } else if (ctl_len) {
2019 if (ctl_len > sizeof(ctl)) {
2020 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2021 if (ctl_buf == NULL)
2026 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2027 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2028 * checking falls down on this.
2030 if (copy_from_user(ctl_buf,
2031 (void __user __force *)msg_sys->msg_control,
2034 msg_sys->msg_control = ctl_buf;
2036 msg_sys->msg_flags = flags;
2038 if (sock->file->f_flags & O_NONBLOCK)
2039 msg_sys->msg_flags |= MSG_DONTWAIT;
2041 * If this is sendmmsg() and current destination address is same as
2042 * previously succeeded address, omit asking LSM's decision.
2043 * used_address->name_len is initialized to UINT_MAX so that the first
2044 * destination address never matches.
2046 if (used_address && msg_sys->msg_name &&
2047 used_address->name_len == msg_sys->msg_namelen &&
2048 !memcmp(&used_address->name, msg_sys->msg_name,
2049 used_address->name_len)) {
2050 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2053 err = sock_sendmsg(sock, msg_sys, total_len);
2055 * If this is sendmmsg() and sending to current destination address was
2056 * successful, remember it.
2058 if (used_address && err >= 0) {
2059 used_address->name_len = msg_sys->msg_namelen;
2060 if (msg_sys->msg_name)
2061 memcpy(&used_address->name, msg_sys->msg_name,
2062 used_address->name_len);
2067 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2069 if (iov != iovstack)
2076 * BSD sendmsg interface
2079 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2081 int fput_needed, err;
2082 struct msghdr msg_sys;
2083 struct socket *sock;
2085 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2089 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2091 fput_light(sock->file, fput_needed);
2096 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2098 if (flags & MSG_CMSG_COMPAT)
2100 return __sys_sendmsg(fd, msg, flags);
2104 * Linux sendmmsg interface
2107 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2110 int fput_needed, err, datagrams;
2111 struct socket *sock;
2112 struct mmsghdr __user *entry;
2113 struct compat_mmsghdr __user *compat_entry;
2114 struct msghdr msg_sys;
2115 struct used_address used_address;
2117 if (vlen > UIO_MAXIOV)
2122 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2126 used_address.name_len = UINT_MAX;
2128 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2131 while (datagrams < vlen) {
2132 if (MSG_CMSG_COMPAT & flags) {
2133 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2134 &msg_sys, flags, &used_address);
2137 err = __put_user(err, &compat_entry->msg_len);
2140 err = ___sys_sendmsg(sock,
2141 (struct msghdr __user *)entry,
2142 &msg_sys, flags, &used_address);
2145 err = put_user(err, &entry->msg_len);
2154 fput_light(sock->file, fput_needed);
2156 /* We only return an error if no datagrams were able to be sent */
2163 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2164 unsigned int, vlen, unsigned int, flags)
2166 if (flags & MSG_CMSG_COMPAT)
2168 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2171 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2172 struct msghdr *msg_sys, unsigned int flags, int nosec)
2174 struct compat_msghdr __user *msg_compat =
2175 (struct compat_msghdr __user *)msg;
2176 struct iovec iovstack[UIO_FASTIOV];
2177 struct iovec *iov = iovstack;
2178 unsigned long cmsg_ptr;
2179 int err, total_len, len;
2181 /* kernel mode address */
2182 struct sockaddr_storage addr;
2184 /* user mode address pointers */
2185 struct sockaddr __user *uaddr;
2186 int __user *uaddr_len;
2188 if (MSG_CMSG_COMPAT & flags) {
2189 if (get_compat_msghdr(msg_sys, msg_compat))
2191 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2194 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2196 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2199 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2206 * Save the user-mode address (verify_iovec will change the
2207 * kernel msghdr to use the kernel address space)
2210 uaddr = (__force void __user *)msg_sys->msg_name;
2211 uaddr_len = COMPAT_NAMELEN(msg);
2212 if (MSG_CMSG_COMPAT & flags) {
2213 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2215 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2220 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2221 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2223 if (sock->file->f_flags & O_NONBLOCK)
2224 flags |= MSG_DONTWAIT;
2225 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2231 if (uaddr != NULL) {
2232 err = move_addr_to_user(&addr,
2233 msg_sys->msg_namelen, uaddr,
2238 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2242 if (MSG_CMSG_COMPAT & flags)
2243 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2244 &msg_compat->msg_controllen);
2246 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2247 &msg->msg_controllen);
2253 if (iov != iovstack)
2260 * BSD recvmsg interface
2263 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2265 int fput_needed, err;
2266 struct msghdr msg_sys;
2267 struct socket *sock;
2269 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2273 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2275 fput_light(sock->file, fput_needed);
2280 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2281 unsigned int, flags)
2283 if (flags & MSG_CMSG_COMPAT)
2285 return __sys_recvmsg(fd, msg, flags);
2289 * Linux recvmmsg interface
2292 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2293 unsigned int flags, struct timespec *timeout)
2295 int fput_needed, err, datagrams;
2296 struct socket *sock;
2297 struct mmsghdr __user *entry;
2298 struct compat_mmsghdr __user *compat_entry;
2299 struct msghdr msg_sys;
2300 struct timespec end_time;
2303 poll_select_set_timeout(&end_time, timeout->tv_sec,
2309 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2313 err = sock_error(sock->sk);
2318 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2320 while (datagrams < vlen) {
2322 * No need to ask LSM for more than the first datagram.
2324 if (MSG_CMSG_COMPAT & flags) {
2325 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2326 &msg_sys, flags & ~MSG_WAITFORONE,
2330 err = __put_user(err, &compat_entry->msg_len);
2333 err = ___sys_recvmsg(sock,
2334 (struct msghdr __user *)entry,
2335 &msg_sys, flags & ~MSG_WAITFORONE,
2339 err = put_user(err, &entry->msg_len);
2347 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2348 if (flags & MSG_WAITFORONE)
2349 flags |= MSG_DONTWAIT;
2352 ktime_get_ts(timeout);
2353 *timeout = timespec_sub(end_time, *timeout);
2354 if (timeout->tv_sec < 0) {
2355 timeout->tv_sec = timeout->tv_nsec = 0;
2359 /* Timeout, return less than vlen datagrams */
2360 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2364 /* Out of band data, return right away */
2365 if (msg_sys.msg_flags & MSG_OOB)
2370 fput_light(sock->file, fput_needed);
2375 if (datagrams != 0) {
2377 * We may return less entries than requested (vlen) if the
2378 * sock is non block and there aren't enough datagrams...
2380 if (err != -EAGAIN) {
2382 * ... or if recvmsg returns an error after we
2383 * received some datagrams, where we record the
2384 * error to return on the next call or if the
2385 * app asks about it using getsockopt(SO_ERROR).
2387 sock->sk->sk_err = -err;
2396 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2397 unsigned int, vlen, unsigned int, flags,
2398 struct timespec __user *, timeout)
2401 struct timespec timeout_sys;
2403 if (flags & MSG_CMSG_COMPAT)
2407 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2409 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2412 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2414 if (datagrams > 0 &&
2415 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2416 datagrams = -EFAULT;
2421 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2422 /* Argument list sizes for sys_socketcall */
2423 #define AL(x) ((x) * sizeof(unsigned long))
2424 static const unsigned char nargs[21] = {
2425 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2426 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2427 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2434 * System call vectors.
2436 * Argument checking cleaned up. Saved 20% in size.
2437 * This function doesn't need to set the kernel lock because
2438 * it is set by the callees.
2441 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2443 unsigned long a[AUDITSC_ARGS];
2444 unsigned long a0, a1;
2448 if (call < 1 || call > SYS_SENDMMSG)
2452 if (len > sizeof(a))
2455 /* copy_from_user should be SMP safe. */
2456 if (copy_from_user(a, args, len))
2459 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2468 err = sys_socket(a0, a1, a[2]);
2471 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2474 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2477 err = sys_listen(a0, a1);
2480 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2481 (int __user *)a[2], 0);
2483 case SYS_GETSOCKNAME:
2485 sys_getsockname(a0, (struct sockaddr __user *)a1,
2486 (int __user *)a[2]);
2488 case SYS_GETPEERNAME:
2490 sys_getpeername(a0, (struct sockaddr __user *)a1,
2491 (int __user *)a[2]);
2493 case SYS_SOCKETPAIR:
2494 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2497 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2500 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2501 (struct sockaddr __user *)a[4], a[5]);
2504 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2507 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2508 (struct sockaddr __user *)a[4],
2509 (int __user *)a[5]);
2512 err = sys_shutdown(a0, a1);
2514 case SYS_SETSOCKOPT:
2515 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2517 case SYS_GETSOCKOPT:
2519 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2520 (int __user *)a[4]);
2523 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2526 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2529 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2532 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2533 (struct timespec __user *)a[4]);
2536 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2537 (int __user *)a[2], a[3]);
2546 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2549 * sock_register - add a socket protocol handler
2550 * @ops: description of protocol
2552 * This function is called by a protocol handler that wants to
2553 * advertise its address family, and have it linked into the
2554 * socket interface. The value ops->family coresponds to the
2555 * socket system call protocol family.
2557 int sock_register(const struct net_proto_family *ops)
2561 if (ops->family >= NPROTO) {
2562 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2567 spin_lock(&net_family_lock);
2568 if (rcu_dereference_protected(net_families[ops->family],
2569 lockdep_is_held(&net_family_lock)))
2572 rcu_assign_pointer(net_families[ops->family], ops);
2575 spin_unlock(&net_family_lock);
2577 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2580 EXPORT_SYMBOL(sock_register);
2583 * sock_unregister - remove a protocol handler
2584 * @family: protocol family to remove
2586 * This function is called by a protocol handler that wants to
2587 * remove its address family, and have it unlinked from the
2588 * new socket creation.
2590 * If protocol handler is a module, then it can use module reference
2591 * counts to protect against new references. If protocol handler is not
2592 * a module then it needs to provide its own protection in
2593 * the ops->create routine.
2595 void sock_unregister(int family)
2597 BUG_ON(family < 0 || family >= NPROTO);
2599 spin_lock(&net_family_lock);
2600 RCU_INIT_POINTER(net_families[family], NULL);
2601 spin_unlock(&net_family_lock);
2605 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2607 EXPORT_SYMBOL(sock_unregister);
2609 static int __init sock_init(void)
2613 * Initialize the network sysctl infrastructure.
2615 err = net_sysctl_init();
2620 * Initialize skbuff SLAB cache
2625 * Initialize the protocols module.
2630 err = register_filesystem(&sock_fs_type);
2633 sock_mnt = kern_mount(&sock_fs_type);
2634 if (IS_ERR(sock_mnt)) {
2635 err = PTR_ERR(sock_mnt);
2639 /* The real protocol initialization is performed in later initcalls.
2642 #ifdef CONFIG_NETFILTER
2643 err = netfilter_init();
2648 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2649 skb_timestamping_init();
2656 unregister_filesystem(&sock_fs_type);
2661 core_initcall(sock_init); /* early initcall */
2663 #ifdef CONFIG_PROC_FS
2664 void socket_seq_show(struct seq_file *seq)
2669 for_each_possible_cpu(cpu)
2670 counter += per_cpu(sockets_in_use, cpu);
2672 /* It can be negative, by the way. 8) */
2676 seq_printf(seq, "sockets: used %d\n", counter);
2678 #endif /* CONFIG_PROC_FS */
2680 #ifdef CONFIG_COMPAT
2681 static int do_siocgstamp(struct net *net, struct socket *sock,
2682 unsigned int cmd, void __user *up)
2684 mm_segment_t old_fs = get_fs();
2689 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2692 err = compat_put_timeval(&ktv, up);
2697 static int do_siocgstampns(struct net *net, struct socket *sock,
2698 unsigned int cmd, void __user *up)
2700 mm_segment_t old_fs = get_fs();
2701 struct timespec kts;
2705 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2708 err = compat_put_timespec(&kts, up);
2713 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2715 struct ifreq __user *uifr;
2718 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2719 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2722 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2726 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2732 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2734 struct compat_ifconf ifc32;
2736 struct ifconf __user *uifc;
2737 struct compat_ifreq __user *ifr32;
2738 struct ifreq __user *ifr;
2742 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2745 memset(&ifc, 0, sizeof(ifc));
2746 if (ifc32.ifcbuf == 0) {
2750 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2752 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2753 sizeof(struct ifreq);
2754 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2756 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2757 ifr32 = compat_ptr(ifc32.ifcbuf);
2758 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2759 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2765 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2768 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2772 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2776 ifr32 = compat_ptr(ifc32.ifcbuf);
2778 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2779 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2780 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2786 if (ifc32.ifcbuf == 0) {
2787 /* Translate from 64-bit structure multiple to
2791 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2796 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2802 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2804 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2805 bool convert_in = false, convert_out = false;
2806 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2807 struct ethtool_rxnfc __user *rxnfc;
2808 struct ifreq __user *ifr;
2809 u32 rule_cnt = 0, actual_rule_cnt;
2814 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2817 compat_rxnfc = compat_ptr(data);
2819 if (get_user(ethcmd, &compat_rxnfc->cmd))
2822 /* Most ethtool structures are defined without padding.
2823 * Unfortunately struct ethtool_rxnfc is an exception.
2828 case ETHTOOL_GRXCLSRLALL:
2829 /* Buffer size is variable */
2830 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2832 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2834 buf_size += rule_cnt * sizeof(u32);
2836 case ETHTOOL_GRXRINGS:
2837 case ETHTOOL_GRXCLSRLCNT:
2838 case ETHTOOL_GRXCLSRULE:
2839 case ETHTOOL_SRXCLSRLINS:
2842 case ETHTOOL_SRXCLSRLDEL:
2843 buf_size += sizeof(struct ethtool_rxnfc);
2848 ifr = compat_alloc_user_space(buf_size);
2849 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2851 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2854 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2855 &ifr->ifr_ifru.ifru_data))
2859 /* We expect there to be holes between fs.m_ext and
2860 * fs.ring_cookie and at the end of fs, but nowhere else.
2862 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2863 sizeof(compat_rxnfc->fs.m_ext) !=
2864 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2865 sizeof(rxnfc->fs.m_ext));
2867 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2868 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2869 offsetof(struct ethtool_rxnfc, fs.location) -
2870 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2872 if (copy_in_user(rxnfc, compat_rxnfc,
2873 (void __user *)(&rxnfc->fs.m_ext + 1) -
2874 (void __user *)rxnfc) ||
2875 copy_in_user(&rxnfc->fs.ring_cookie,
2876 &compat_rxnfc->fs.ring_cookie,
2877 (void __user *)(&rxnfc->fs.location + 1) -
2878 (void __user *)&rxnfc->fs.ring_cookie) ||
2879 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2880 sizeof(rxnfc->rule_cnt)))
2884 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2889 if (copy_in_user(compat_rxnfc, rxnfc,
2890 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2891 (const void __user *)rxnfc) ||
2892 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2893 &rxnfc->fs.ring_cookie,
2894 (const void __user *)(&rxnfc->fs.location + 1) -
2895 (const void __user *)&rxnfc->fs.ring_cookie) ||
2896 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2897 sizeof(rxnfc->rule_cnt)))
2900 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2901 /* As an optimisation, we only copy the actual
2902 * number of rules that the underlying
2903 * function returned. Since Mallory might
2904 * change the rule count in user memory, we
2905 * check that it is less than the rule count
2906 * originally given (as the user buffer size),
2907 * which has been range-checked.
2909 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2911 if (actual_rule_cnt < rule_cnt)
2912 rule_cnt = actual_rule_cnt;
2913 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2914 &rxnfc->rule_locs[0],
2915 rule_cnt * sizeof(u32)))
2923 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2926 compat_uptr_t uptr32;
2927 struct ifreq __user *uifr;
2929 uifr = compat_alloc_user_space(sizeof(*uifr));
2930 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2933 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2936 uptr = compat_ptr(uptr32);
2938 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2941 return dev_ioctl(net, SIOCWANDEV, uifr);
2944 static int bond_ioctl(struct net *net, unsigned int cmd,
2945 struct compat_ifreq __user *ifr32)
2948 struct ifreq __user *uifr;
2949 mm_segment_t old_fs;
2955 case SIOCBONDENSLAVE:
2956 case SIOCBONDRELEASE:
2957 case SIOCBONDSETHWADDR:
2958 case SIOCBONDCHANGEACTIVE:
2959 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2964 err = dev_ioctl(net, cmd,
2965 (struct ifreq __user __force *) &kifr);
2969 case SIOCBONDSLAVEINFOQUERY:
2970 case SIOCBONDINFOQUERY:
2971 uifr = compat_alloc_user_space(sizeof(*uifr));
2972 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2975 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2978 datap = compat_ptr(data);
2979 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2982 return dev_ioctl(net, cmd, uifr);
2984 return -ENOIOCTLCMD;
2988 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2989 struct compat_ifreq __user *u_ifreq32)
2991 struct ifreq __user *u_ifreq64;
2992 char tmp_buf[IFNAMSIZ];
2993 void __user *data64;
2996 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2999 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3001 data64 = compat_ptr(data32);
3003 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3005 /* Don't check these user accesses, just let that get trapped
3006 * in the ioctl handler instead.
3008 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3011 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3014 return dev_ioctl(net, cmd, u_ifreq64);
3017 static int dev_ifsioc(struct net *net, struct socket *sock,
3018 unsigned int cmd, struct compat_ifreq __user *uifr32)
3020 struct ifreq __user *uifr;
3023 uifr = compat_alloc_user_space(sizeof(*uifr));
3024 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3027 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3038 case SIOCGIFBRDADDR:
3039 case SIOCGIFDSTADDR:
3040 case SIOCGIFNETMASK:
3045 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3053 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3054 struct compat_ifreq __user *uifr32)
3057 struct compat_ifmap __user *uifmap32;
3058 mm_segment_t old_fs;
3061 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3062 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3063 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3064 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3065 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3066 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3067 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3068 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3074 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3077 if (cmd == SIOCGIFMAP && !err) {
3078 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3079 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3080 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3081 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3082 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3083 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3084 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3091 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3094 compat_uptr_t uptr32;
3095 struct ifreq __user *uifr;
3097 uifr = compat_alloc_user_space(sizeof(*uifr));
3098 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3101 if (get_user(uptr32, &uifr32->ifr_data))
3104 uptr = compat_ptr(uptr32);
3106 if (put_user(uptr, &uifr->ifr_data))
3109 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3114 struct sockaddr rt_dst; /* target address */
3115 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3116 struct sockaddr rt_genmask; /* target network mask (IP) */
3117 unsigned short rt_flags;
3120 unsigned char rt_tos;
3121 unsigned char rt_class;
3123 short rt_metric; /* +1 for binary compatibility! */
3124 /* char * */ u32 rt_dev; /* forcing the device at add */
3125 u32 rt_mtu; /* per route MTU/Window */
3126 u32 rt_window; /* Window clamping */
3127 unsigned short rt_irtt; /* Initial RTT */
3130 struct in6_rtmsg32 {
3131 struct in6_addr rtmsg_dst;
3132 struct in6_addr rtmsg_src;
3133 struct in6_addr rtmsg_gateway;
3143 static int routing_ioctl(struct net *net, struct socket *sock,
3144 unsigned int cmd, void __user *argp)
3148 struct in6_rtmsg r6;
3152 mm_segment_t old_fs = get_fs();
3154 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3155 struct in6_rtmsg32 __user *ur6 = argp;
3156 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3157 3 * sizeof(struct in6_addr));
3158 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3159 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3160 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3161 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3162 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3163 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3164 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3168 struct rtentry32 __user *ur4 = argp;
3169 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3170 3 * sizeof(struct sockaddr));
3171 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3172 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3173 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3174 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3175 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3176 ret |= __get_user(rtdev, &(ur4->rt_dev));
3178 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3179 r4.rt_dev = (char __user __force *)devname;
3193 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3200 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3201 * for some operations; this forces use of the newer bridge-utils that
3202 * use compatible ioctls
3204 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3208 if (get_user(tmp, argp))
3210 if (tmp == BRCTL_GET_VERSION)
3211 return BRCTL_VERSION + 1;
3215 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3216 unsigned int cmd, unsigned long arg)
3218 void __user *argp = compat_ptr(arg);
3219 struct sock *sk = sock->sk;
3220 struct net *net = sock_net(sk);
3222 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3223 return siocdevprivate_ioctl(net, cmd, argp);
3228 return old_bridge_ioctl(argp);
3230 return dev_ifname32(net, argp);
3232 return dev_ifconf(net, argp);
3234 return ethtool_ioctl(net, argp);
3236 return compat_siocwandev(net, argp);
3239 return compat_sioc_ifmap(net, cmd, argp);
3240 case SIOCBONDENSLAVE:
3241 case SIOCBONDRELEASE:
3242 case SIOCBONDSETHWADDR:
3243 case SIOCBONDSLAVEINFOQUERY:
3244 case SIOCBONDINFOQUERY:
3245 case SIOCBONDCHANGEACTIVE:
3246 return bond_ioctl(net, cmd, argp);
3249 return routing_ioctl(net, sock, cmd, argp);
3251 return do_siocgstamp(net, sock, cmd, argp);
3253 return do_siocgstampns(net, sock, cmd, argp);
3255 return compat_siocshwtstamp(net, argp);
3267 return sock_ioctl(file, cmd, arg);
3284 case SIOCSIFHWBROADCAST:
3286 case SIOCGIFBRDADDR:
3287 case SIOCSIFBRDADDR:
3288 case SIOCGIFDSTADDR:
3289 case SIOCSIFDSTADDR:
3290 case SIOCGIFNETMASK:
3291 case SIOCSIFNETMASK:
3302 return dev_ifsioc(net, sock, cmd, argp);
3308 return sock_do_ioctl(net, sock, cmd, arg);
3311 return -ENOIOCTLCMD;
3314 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3317 struct socket *sock = file->private_data;
3318 int ret = -ENOIOCTLCMD;
3325 if (sock->ops->compat_ioctl)
3326 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3328 if (ret == -ENOIOCTLCMD &&
3329 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3330 ret = compat_wext_handle_ioctl(net, cmd, arg);
3332 if (ret == -ENOIOCTLCMD)
3333 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3339 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3341 return sock->ops->bind(sock, addr, addrlen);
3343 EXPORT_SYMBOL(kernel_bind);
3345 int kernel_listen(struct socket *sock, int backlog)
3347 return sock->ops->listen(sock, backlog);
3349 EXPORT_SYMBOL(kernel_listen);
3351 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3353 struct sock *sk = sock->sk;
3356 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3361 err = sock->ops->accept(sock, *newsock, flags);
3363 sock_release(*newsock);
3368 (*newsock)->ops = sock->ops;
3369 __module_get((*newsock)->ops->owner);
3374 EXPORT_SYMBOL(kernel_accept);
3376 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3379 return sock->ops->connect(sock, addr, addrlen, flags);
3381 EXPORT_SYMBOL(kernel_connect);
3383 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3386 return sock->ops->getname(sock, addr, addrlen, 0);
3388 EXPORT_SYMBOL(kernel_getsockname);
3390 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3393 return sock->ops->getname(sock, addr, addrlen, 1);
3395 EXPORT_SYMBOL(kernel_getpeername);
3397 int kernel_getsockopt(struct socket *sock, int level, int optname,
3398 char *optval, int *optlen)
3400 mm_segment_t oldfs = get_fs();
3401 char __user *uoptval;
3402 int __user *uoptlen;
3405 uoptval = (char __user __force *) optval;
3406 uoptlen = (int __user __force *) optlen;
3409 if (level == SOL_SOCKET)
3410 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3412 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3417 EXPORT_SYMBOL(kernel_getsockopt);
3419 int kernel_setsockopt(struct socket *sock, int level, int optname,
3420 char *optval, unsigned int optlen)
3422 mm_segment_t oldfs = get_fs();
3423 char __user *uoptval;
3426 uoptval = (char __user __force *) optval;
3429 if (level == SOL_SOCKET)
3430 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3432 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3437 EXPORT_SYMBOL(kernel_setsockopt);
3439 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3440 size_t size, int flags)
3442 if (sock->ops->sendpage)
3443 return sock->ops->sendpage(sock, page, offset, size, flags);
3445 return sock_no_sendpage(sock, page, offset, size, flags);
3447 EXPORT_SYMBOL(kernel_sendpage);
3449 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3451 mm_segment_t oldfs = get_fs();
3455 err = sock->ops->ioctl(sock, cmd, arg);
3460 EXPORT_SYMBOL(kernel_sock_ioctl);
3462 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3464 return sock->ops->shutdown(sock, how);
3466 EXPORT_SYMBOL(kernel_sock_shutdown);