2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
117 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
118 unsigned long nr_segs, loff_t pos);
119 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
120 unsigned long nr_segs, loff_t pos);
121 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
123 static int sock_close(struct inode *inode, struct file *file);
124 static unsigned int sock_poll(struct file *file,
125 struct poll_table_struct *wait);
126 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
128 static long compat_sock_ioctl(struct file *file,
129 unsigned int cmd, unsigned long arg);
131 static int sock_fasync(int fd, struct file *filp, int on);
132 static ssize_t sock_sendpage(struct file *file, struct page *page,
133 int offset, size_t size, loff_t *ppos, int more);
134 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
135 struct pipe_inode_info *pipe, size_t len,
139 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
140 * in the operation structures but are done directly via the socketcall() multiplexor.
143 static const struct file_operations socket_file_ops = {
144 .owner = THIS_MODULE,
146 .aio_read = sock_aio_read,
147 .aio_write = sock_aio_write,
149 .unlocked_ioctl = sock_ioctl,
151 .compat_ioctl = compat_sock_ioctl,
154 .open = sock_no_open, /* special open code to disallow open via /proc */
155 .release = sock_close,
156 .fasync = sock_fasync,
157 .sendpage = sock_sendpage,
158 .splice_write = generic_splice_sendpage,
159 .splice_read = sock_splice_read,
163 * The protocol list. Each protocol is registered in here.
166 static DEFINE_SPINLOCK(net_family_lock);
167 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
170 * Statistics counters of the socket lists
173 static DEFINE_PER_CPU(int, sockets_in_use);
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
198 if (copy_from_user(kaddr, uaddr, ulen))
200 return audit_sockaddr(ulen, kaddr);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221 void __user *uaddr, int __user *ulen)
226 BUG_ON(klen > sizeof(struct sockaddr_storage));
227 err = get_user(len, ulen);
235 if (audit_sockaddr(klen, kaddr))
237 if (copy_to_user(uaddr, kaddr, len))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen, ulen);
247 static struct kmem_cache *sock_inode_cachep __read_mostly;
249 static struct inode *sock_alloc_inode(struct super_block *sb)
251 struct socket_alloc *ei;
252 struct socket_wq *wq;
254 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
257 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
259 kmem_cache_free(sock_inode_cachep, ei);
262 init_waitqueue_head(&wq->wait);
263 wq->fasync_list = NULL;
264 RCU_INIT_POINTER(ei->socket.wq, wq);
266 ei->socket.state = SS_UNCONNECTED;
267 ei->socket.flags = 0;
268 ei->socket.ops = NULL;
269 ei->socket.sk = NULL;
270 ei->socket.file = NULL;
272 return &ei->vfs_inode;
275 static void sock_destroy_inode(struct inode *inode)
277 struct socket_alloc *ei;
278 struct socket_wq *wq;
280 ei = container_of(inode, struct socket_alloc, vfs_inode);
281 wq = rcu_dereference_protected(ei->socket.wq, 1);
283 kmem_cache_free(sock_inode_cachep, ei);
286 static void init_once(void *foo)
288 struct socket_alloc *ei = (struct socket_alloc *)foo;
290 inode_init_once(&ei->vfs_inode);
293 static int init_inodecache(void)
295 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
296 sizeof(struct socket_alloc),
298 (SLAB_HWCACHE_ALIGN |
299 SLAB_RECLAIM_ACCOUNT |
302 if (sock_inode_cachep == NULL)
307 static const struct super_operations sockfs_ops = {
308 .alloc_inode = sock_alloc_inode,
309 .destroy_inode = sock_destroy_inode,
310 .statfs = simple_statfs,
314 * sockfs_dname() is called from d_path().
316 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
318 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
319 dentry->d_inode->i_ino);
322 static const struct dentry_operations sockfs_dentry_operations = {
323 .d_dname = sockfs_dname,
326 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
327 int flags, const char *dev_name, void *data)
329 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
330 &sockfs_dentry_operations, SOCKFS_MAGIC);
333 static struct vfsmount *sock_mnt __read_mostly;
335 static struct file_system_type sock_fs_type = {
337 .mount = sockfs_mount,
338 .kill_sb = kill_anon_super,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
360 struct qstr name = { .name = "" };
366 name.len = strlen(name.name);
367 } else if (sock->sk) {
368 name.name = sock->sk->sk_prot_creator->name;
369 name.len = strlen(name.name);
371 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
372 if (unlikely(!path.dentry))
373 return ERR_PTR(-ENOMEM);
374 path.mnt = mntget(sock_mnt);
376 d_instantiate(path.dentry, SOCK_INODE(sock));
377 SOCK_INODE(sock)->i_fop = &socket_file_ops;
379 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
381 if (unlikely(IS_ERR(file))) {
382 /* drop dentry, keep inode */
383 ihold(path.dentry->d_inode);
389 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
390 file->private_data = sock;
393 EXPORT_SYMBOL(sock_alloc_file);
395 static int sock_map_fd(struct socket *sock, int flags)
397 struct file *newfile;
398 int fd = get_unused_fd_flags(flags);
399 if (unlikely(fd < 0))
402 newfile = sock_alloc_file(sock, flags, NULL);
403 if (likely(!IS_ERR(newfile))) {
404 fd_install(fd, newfile);
409 return PTR_ERR(newfile);
412 struct socket *sock_from_file(struct file *file, int *err)
414 if (file->f_op == &socket_file_ops)
415 return file->private_data; /* set in sock_map_fd */
420 EXPORT_SYMBOL(sock_from_file);
423 * sockfd_lookup - Go from a file number to its socket slot
425 * @err: pointer to an error code return
427 * The file handle passed in is locked and the socket it is bound
428 * too is returned. If an error occurs the err pointer is overwritten
429 * with a negative errno code and NULL is returned. The function checks
430 * for both invalid handles and passing a handle which is not a socket.
432 * On a success the socket object pointer is returned.
435 struct socket *sockfd_lookup(int fd, int *err)
446 sock = sock_from_file(file, err);
451 EXPORT_SYMBOL(sockfd_lookup);
453 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
455 struct fd f = fdget(fd);
460 sock = sock_from_file(f.file, err);
462 *fput_needed = f.flags;
470 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
471 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
472 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
473 static ssize_t sockfs_getxattr(struct dentry *dentry,
474 const char *name, void *value, size_t size)
476 const char *proto_name;
481 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
482 proto_name = dentry->d_name.name;
483 proto_size = strlen(proto_name);
487 if (proto_size + 1 > size)
490 strncpy(value, proto_name, proto_size + 1);
492 error = proto_size + 1;
499 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
505 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
515 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
520 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
527 static const struct inode_operations sockfs_inode_ops = {
528 .getxattr = sockfs_getxattr,
529 .listxattr = sockfs_listxattr,
533 * sock_alloc - allocate a socket
535 * Allocate a new inode and socket object. The two are bound together
536 * and initialised. The socket is then returned. If we are out of inodes
540 static struct socket *sock_alloc(void)
545 inode = new_inode_pseudo(sock_mnt->mnt_sb);
549 sock = SOCKET_I(inode);
551 kmemcheck_annotate_bitfield(sock, type);
552 inode->i_ino = get_next_ino();
553 inode->i_mode = S_IFSOCK | S_IRWXUGO;
554 inode->i_uid = current_fsuid();
555 inode->i_gid = current_fsgid();
556 inode->i_op = &sockfs_inode_ops;
558 this_cpu_add(sockets_in_use, 1);
563 * In theory you can't get an open on this inode, but /proc provides
564 * a back door. Remember to keep it shut otherwise you'll let the
565 * creepy crawlies in.
568 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
573 const struct file_operations bad_sock_fops = {
574 .owner = THIS_MODULE,
575 .open = sock_no_open,
576 .llseek = noop_llseek,
580 * sock_release - close a socket
581 * @sock: socket to close
583 * The socket is released from the protocol stack if it has a release
584 * callback, and the inode is then released if the socket is bound to
585 * an inode not a file.
588 void sock_release(struct socket *sock)
591 struct module *owner = sock->ops->owner;
593 sock->ops->release(sock);
598 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
599 pr_err("%s: fasync list not empty!\n", __func__);
601 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
604 this_cpu_sub(sockets_in_use, 1);
606 iput(SOCK_INODE(sock));
611 EXPORT_SYMBOL(sock_release);
613 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
616 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
617 *tx_flags |= SKBTX_HW_TSTAMP;
618 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
619 *tx_flags |= SKBTX_SW_TSTAMP;
620 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
621 *tx_flags |= SKBTX_SCHED_TSTAMP;
623 if (sock_flag(sk, SOCK_WIFI_STATUS))
624 *tx_flags |= SKBTX_WIFI_STATUS;
626 EXPORT_SYMBOL(sock_tx_timestamp);
628 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
629 struct msghdr *msg, size_t size)
631 struct sock_iocb *si = kiocb_to_siocb(iocb);
638 return sock->ops->sendmsg(iocb, sock, msg, size);
641 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
642 struct msghdr *msg, size_t size)
644 int err = security_socket_sendmsg(sock, msg, size);
646 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
649 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
652 struct sock_iocb siocb;
655 init_sync_kiocb(&iocb, NULL);
656 iocb.private = &siocb;
657 ret = __sock_sendmsg(&iocb, sock, msg, size);
658 if (-EIOCBQUEUED == ret)
659 ret = wait_on_sync_kiocb(&iocb);
662 EXPORT_SYMBOL(sock_sendmsg);
664 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
667 struct sock_iocb siocb;
670 init_sync_kiocb(&iocb, NULL);
671 iocb.private = &siocb;
672 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
673 if (-EIOCBQUEUED == ret)
674 ret = wait_on_sync_kiocb(&iocb);
678 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
679 struct kvec *vec, size_t num, size_t size)
681 mm_segment_t oldfs = get_fs();
686 * the following is safe, since for compiler definitions of kvec and
687 * iovec are identical, yielding the same in-core layout and alignment
689 msg->msg_iov = (struct iovec *)vec;
690 msg->msg_iovlen = num;
691 result = sock_sendmsg(sock, msg, size);
695 EXPORT_SYMBOL(kernel_sendmsg);
698 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
700 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
703 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
704 struct scm_timestamping tss;
706 struct skb_shared_hwtstamps *shhwtstamps =
709 /* Race occurred between timestamp enabling and packet
710 receiving. Fill in the current time for now. */
711 if (need_software_tstamp && skb->tstamp.tv64 == 0)
712 __net_timestamp(skb);
714 if (need_software_tstamp) {
715 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
717 skb_get_timestamp(skb, &tv);
718 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
722 skb_get_timestampns(skb, &ts);
723 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
728 memset(&tss, 0, sizeof(tss));
729 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE ||
730 skb_shinfo(skb)->tx_flags & SKBTX_ANY_SW_TSTAMP) &&
731 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
734 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
735 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
738 put_cmsg(msg, SOL_SOCKET,
739 SCM_TIMESTAMPING, sizeof(tss), &tss);
741 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
743 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
748 if (!sock_flag(sk, SOCK_WIFI_STATUS))
750 if (!skb->wifi_acked_valid)
753 ack = skb->wifi_acked;
755 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
757 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
759 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
762 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
763 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
764 sizeof(__u32), &skb->dropcount);
767 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
770 sock_recv_timestamp(msg, sk, skb);
771 sock_recv_drops(msg, sk, skb);
773 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
775 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
776 struct msghdr *msg, size_t size, int flags)
778 struct sock_iocb *si = kiocb_to_siocb(iocb);
786 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
789 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
790 struct msghdr *msg, size_t size, int flags)
792 int err = security_socket_recvmsg(sock, msg, size, flags);
794 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
797 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
798 size_t size, int flags)
801 struct sock_iocb siocb;
804 init_sync_kiocb(&iocb, NULL);
805 iocb.private = &siocb;
806 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
807 if (-EIOCBQUEUED == ret)
808 ret = wait_on_sync_kiocb(&iocb);
811 EXPORT_SYMBOL(sock_recvmsg);
813 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
814 size_t size, int flags)
817 struct sock_iocb siocb;
820 init_sync_kiocb(&iocb, NULL);
821 iocb.private = &siocb;
822 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
823 if (-EIOCBQUEUED == ret)
824 ret = wait_on_sync_kiocb(&iocb);
829 * kernel_recvmsg - Receive a message from a socket (kernel space)
830 * @sock: The socket to receive the message from
831 * @msg: Received message
832 * @vec: Input s/g array for message data
833 * @num: Size of input s/g array
834 * @size: Number of bytes to read
835 * @flags: Message flags (MSG_DONTWAIT, etc...)
837 * On return the msg structure contains the scatter/gather array passed in the
838 * vec argument. The array is modified so that it consists of the unfilled
839 * portion of the original array.
841 * The returned value is the total number of bytes received, or an error.
843 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
844 struct kvec *vec, size_t num, size_t size, int flags)
846 mm_segment_t oldfs = get_fs();
851 * the following is safe, since for compiler definitions of kvec and
852 * iovec are identical, yielding the same in-core layout and alignment
854 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
855 result = sock_recvmsg(sock, msg, size, flags);
859 EXPORT_SYMBOL(kernel_recvmsg);
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))
895 iocb->private = siocb;
899 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
900 struct file *file, const struct iovec *iov,
901 unsigned long nr_segs)
903 struct socket *sock = file->private_data;
907 for (i = 0; i < nr_segs; i++)
908 size += iov[i].iov_len;
910 msg->msg_name = NULL;
911 msg->msg_namelen = 0;
912 msg->msg_control = NULL;
913 msg->msg_controllen = 0;
914 msg->msg_iov = (struct iovec *)iov;
915 msg->msg_iovlen = nr_segs;
916 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
918 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
921 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
922 unsigned long nr_segs, loff_t pos)
924 struct sock_iocb siocb, *x;
929 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
933 x = alloc_sock_iocb(iocb, &siocb);
936 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
939 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
940 struct file *file, const struct iovec *iov,
941 unsigned long nr_segs)
943 struct socket *sock = file->private_data;
947 for (i = 0; i < nr_segs; i++)
948 size += iov[i].iov_len;
950 msg->msg_name = NULL;
951 msg->msg_namelen = 0;
952 msg->msg_control = NULL;
953 msg->msg_controllen = 0;
954 msg->msg_iov = (struct iovec *)iov;
955 msg->msg_iovlen = nr_segs;
956 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
957 if (sock->type == SOCK_SEQPACKET)
958 msg->msg_flags |= MSG_EOR;
960 return __sock_sendmsg(iocb, sock, msg, size);
963 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
964 unsigned long nr_segs, loff_t pos)
966 struct sock_iocb siocb, *x;
971 x = alloc_sock_iocb(iocb, &siocb);
975 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
979 * Atomic setting of ioctl hooks to avoid race
980 * with module unload.
983 static DEFINE_MUTEX(br_ioctl_mutex);
984 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
986 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
988 mutex_lock(&br_ioctl_mutex);
989 br_ioctl_hook = hook;
990 mutex_unlock(&br_ioctl_mutex);
992 EXPORT_SYMBOL(brioctl_set);
994 static DEFINE_MUTEX(vlan_ioctl_mutex);
995 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
997 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
999 mutex_lock(&vlan_ioctl_mutex);
1000 vlan_ioctl_hook = hook;
1001 mutex_unlock(&vlan_ioctl_mutex);
1003 EXPORT_SYMBOL(vlan_ioctl_set);
1005 static DEFINE_MUTEX(dlci_ioctl_mutex);
1006 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1008 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1010 mutex_lock(&dlci_ioctl_mutex);
1011 dlci_ioctl_hook = hook;
1012 mutex_unlock(&dlci_ioctl_mutex);
1014 EXPORT_SYMBOL(dlci_ioctl_set);
1016 static long sock_do_ioctl(struct net *net, struct socket *sock,
1017 unsigned int cmd, unsigned long arg)
1020 void __user *argp = (void __user *)arg;
1022 err = sock->ops->ioctl(sock, cmd, arg);
1025 * If this ioctl is unknown try to hand it down
1026 * to the NIC driver.
1028 if (err == -ENOIOCTLCMD)
1029 err = dev_ioctl(net, cmd, argp);
1035 * With an ioctl, arg may well be a user mode pointer, but we don't know
1036 * what to do with it - that's up to the protocol still.
1039 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1041 struct socket *sock;
1043 void __user *argp = (void __user *)arg;
1047 sock = file->private_data;
1050 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1051 err = dev_ioctl(net, cmd, argp);
1053 #ifdef CONFIG_WEXT_CORE
1054 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1055 err = dev_ioctl(net, cmd, argp);
1062 if (get_user(pid, (int __user *)argp))
1064 err = f_setown(sock->file, pid, 1);
1068 err = put_user(f_getown(sock->file),
1069 (int __user *)argp);
1077 request_module("bridge");
1079 mutex_lock(&br_ioctl_mutex);
1081 err = br_ioctl_hook(net, cmd, argp);
1082 mutex_unlock(&br_ioctl_mutex);
1087 if (!vlan_ioctl_hook)
1088 request_module("8021q");
1090 mutex_lock(&vlan_ioctl_mutex);
1091 if (vlan_ioctl_hook)
1092 err = vlan_ioctl_hook(net, argp);
1093 mutex_unlock(&vlan_ioctl_mutex);
1098 if (!dlci_ioctl_hook)
1099 request_module("dlci");
1101 mutex_lock(&dlci_ioctl_mutex);
1102 if (dlci_ioctl_hook)
1103 err = dlci_ioctl_hook(cmd, argp);
1104 mutex_unlock(&dlci_ioctl_mutex);
1107 err = sock_do_ioctl(net, sock, cmd, arg);
1113 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1116 struct socket *sock = NULL;
1118 err = security_socket_create(family, type, protocol, 1);
1122 sock = sock_alloc();
1129 err = security_socket_post_create(sock, family, type, protocol, 1);
1141 EXPORT_SYMBOL(sock_create_lite);
1143 /* No kernel lock held - perfect */
1144 static unsigned int sock_poll(struct file *file, poll_table *wait)
1146 unsigned int busy_flag = 0;
1147 struct socket *sock;
1150 * We can't return errors to poll, so it's either yes or no.
1152 sock = file->private_data;
1154 if (sk_can_busy_loop(sock->sk)) {
1155 /* this socket can poll_ll so tell the system call */
1156 busy_flag = POLL_BUSY_LOOP;
1158 /* once, only if requested by syscall */
1159 if (wait && (wait->_key & POLL_BUSY_LOOP))
1160 sk_busy_loop(sock->sk, 1);
1163 return busy_flag | sock->ops->poll(file, sock, wait);
1166 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1168 struct socket *sock = file->private_data;
1170 return sock->ops->mmap(file, sock, vma);
1173 static int sock_close(struct inode *inode, struct file *filp)
1175 sock_release(SOCKET_I(inode));
1180 * Update the socket async list
1182 * Fasync_list locking strategy.
1184 * 1. fasync_list is modified only under process context socket lock
1185 * i.e. under semaphore.
1186 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1187 * or under socket lock
1190 static int sock_fasync(int fd, struct file *filp, int on)
1192 struct socket *sock = filp->private_data;
1193 struct sock *sk = sock->sk;
1194 struct socket_wq *wq;
1200 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1201 fasync_helper(fd, filp, on, &wq->fasync_list);
1203 if (!wq->fasync_list)
1204 sock_reset_flag(sk, SOCK_FASYNC);
1206 sock_set_flag(sk, SOCK_FASYNC);
1212 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1214 int sock_wake_async(struct socket *sock, int how, int band)
1216 struct socket_wq *wq;
1221 wq = rcu_dereference(sock->wq);
1222 if (!wq || !wq->fasync_list) {
1227 case SOCK_WAKE_WAITD:
1228 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1231 case SOCK_WAKE_SPACE:
1232 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1237 kill_fasync(&wq->fasync_list, SIGIO, band);
1240 kill_fasync(&wq->fasync_list, SIGURG, band);
1245 EXPORT_SYMBOL(sock_wake_async);
1247 int __sock_create(struct net *net, int family, int type, int protocol,
1248 struct socket **res, int kern)
1251 struct socket *sock;
1252 const struct net_proto_family *pf;
1255 * Check protocol is in range
1257 if (family < 0 || family >= NPROTO)
1258 return -EAFNOSUPPORT;
1259 if (type < 0 || type >= SOCK_MAX)
1264 This uglymoron is moved from INET layer to here to avoid
1265 deadlock in module load.
1267 if (family == PF_INET && type == SOCK_PACKET) {
1271 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1277 err = security_socket_create(family, type, protocol, kern);
1282 * Allocate the socket and allow the family to set things up. if
1283 * the protocol is 0, the family is instructed to select an appropriate
1286 sock = sock_alloc();
1288 net_warn_ratelimited("socket: no more sockets\n");
1289 return -ENFILE; /* Not exactly a match, but its the
1290 closest posix thing */
1295 #ifdef CONFIG_MODULES
1296 /* Attempt to load a protocol module if the find failed.
1298 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1299 * requested real, full-featured networking support upon configuration.
1300 * Otherwise module support will break!
1302 if (rcu_access_pointer(net_families[family]) == NULL)
1303 request_module("net-pf-%d", family);
1307 pf = rcu_dereference(net_families[family]);
1308 err = -EAFNOSUPPORT;
1313 * We will call the ->create function, that possibly is in a loadable
1314 * module, so we have to bump that loadable module refcnt first.
1316 if (!try_module_get(pf->owner))
1319 /* Now protected by module ref count */
1322 err = pf->create(net, sock, protocol, kern);
1324 goto out_module_put;
1327 * Now to bump the refcnt of the [loadable] module that owns this
1328 * socket at sock_release time we decrement its refcnt.
1330 if (!try_module_get(sock->ops->owner))
1331 goto out_module_busy;
1334 * Now that we're done with the ->create function, the [loadable]
1335 * module can have its refcnt decremented
1337 module_put(pf->owner);
1338 err = security_socket_post_create(sock, family, type, protocol, kern);
1340 goto out_sock_release;
1346 err = -EAFNOSUPPORT;
1349 module_put(pf->owner);
1356 goto out_sock_release;
1358 EXPORT_SYMBOL(__sock_create);
1360 int sock_create(int family, int type, int protocol, struct socket **res)
1362 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1364 EXPORT_SYMBOL(sock_create);
1366 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1368 return __sock_create(&init_net, family, type, protocol, res, 1);
1370 EXPORT_SYMBOL(sock_create_kern);
1372 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1375 struct socket *sock;
1378 /* Check the SOCK_* constants for consistency. */
1379 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1380 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1381 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1382 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1384 flags = type & ~SOCK_TYPE_MASK;
1385 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1387 type &= SOCK_TYPE_MASK;
1389 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1390 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1392 retval = sock_create(family, type, protocol, &sock);
1396 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1401 /* It may be already another descriptor 8) Not kernel problem. */
1410 * Create a pair of connected sockets.
1413 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1414 int __user *, usockvec)
1416 struct socket *sock1, *sock2;
1418 struct file *newfile1, *newfile2;
1421 flags = type & ~SOCK_TYPE_MASK;
1422 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1424 type &= SOCK_TYPE_MASK;
1426 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1427 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1430 * Obtain the first socket and check if the underlying protocol
1431 * supports the socketpair call.
1434 err = sock_create(family, type, protocol, &sock1);
1438 err = sock_create(family, type, protocol, &sock2);
1442 err = sock1->ops->socketpair(sock1, sock2);
1444 goto out_release_both;
1446 fd1 = get_unused_fd_flags(flags);
1447 if (unlikely(fd1 < 0)) {
1449 goto out_release_both;
1452 fd2 = get_unused_fd_flags(flags);
1453 if (unlikely(fd2 < 0)) {
1455 goto out_put_unused_1;
1458 newfile1 = sock_alloc_file(sock1, flags, NULL);
1459 if (unlikely(IS_ERR(newfile1))) {
1460 err = PTR_ERR(newfile1);
1461 goto out_put_unused_both;
1464 newfile2 = sock_alloc_file(sock2, flags, NULL);
1465 if (IS_ERR(newfile2)) {
1466 err = PTR_ERR(newfile2);
1470 err = put_user(fd1, &usockvec[0]);
1474 err = put_user(fd2, &usockvec[1]);
1478 audit_fd_pair(fd1, fd2);
1480 fd_install(fd1, newfile1);
1481 fd_install(fd2, newfile2);
1482 /* fd1 and fd2 may be already another descriptors.
1483 * Not kernel problem.
1499 sock_release(sock2);
1502 out_put_unused_both:
1507 sock_release(sock2);
1509 sock_release(sock1);
1515 * Bind a name to a socket. Nothing much to do here since it's
1516 * the protocol's responsibility to handle the local address.
1518 * We move the socket address to kernel space before we call
1519 * the protocol layer (having also checked the address is ok).
1522 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1524 struct socket *sock;
1525 struct sockaddr_storage address;
1526 int err, fput_needed;
1528 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1530 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1532 err = security_socket_bind(sock,
1533 (struct sockaddr *)&address,
1536 err = sock->ops->bind(sock,
1540 fput_light(sock->file, fput_needed);
1546 * Perform a listen. Basically, we allow the protocol to do anything
1547 * necessary for a listen, and if that works, we mark the socket as
1548 * ready for listening.
1551 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1553 struct socket *sock;
1554 int err, fput_needed;
1557 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1559 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1560 if ((unsigned int)backlog > somaxconn)
1561 backlog = somaxconn;
1563 err = security_socket_listen(sock, backlog);
1565 err = sock->ops->listen(sock, backlog);
1567 fput_light(sock->file, fput_needed);
1573 * For accept, we attempt to create a new socket, set up the link
1574 * with the client, wake up the client, then return the new
1575 * connected fd. We collect the address of the connector in kernel
1576 * space and move it to user at the very end. This is unclean because
1577 * we open the socket then return an error.
1579 * 1003.1g adds the ability to recvmsg() to query connection pending
1580 * status to recvmsg. We need to add that support in a way thats
1581 * clean when we restucture accept also.
1584 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1585 int __user *, upeer_addrlen, int, flags)
1587 struct socket *sock, *newsock;
1588 struct file *newfile;
1589 int err, len, newfd, fput_needed;
1590 struct sockaddr_storage address;
1592 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1595 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1596 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1598 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1603 newsock = sock_alloc();
1607 newsock->type = sock->type;
1608 newsock->ops = sock->ops;
1611 * We don't need try_module_get here, as the listening socket (sock)
1612 * has the protocol module (sock->ops->owner) held.
1614 __module_get(newsock->ops->owner);
1616 newfd = get_unused_fd_flags(flags);
1617 if (unlikely(newfd < 0)) {
1619 sock_release(newsock);
1622 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1623 if (unlikely(IS_ERR(newfile))) {
1624 err = PTR_ERR(newfile);
1625 put_unused_fd(newfd);
1626 sock_release(newsock);
1630 err = security_socket_accept(sock, newsock);
1634 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1638 if (upeer_sockaddr) {
1639 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1641 err = -ECONNABORTED;
1644 err = move_addr_to_user(&address,
1645 len, upeer_sockaddr, upeer_addrlen);
1650 /* File flags are not inherited via accept() unlike another OSes. */
1652 fd_install(newfd, newfile);
1656 fput_light(sock->file, fput_needed);
1661 put_unused_fd(newfd);
1665 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1666 int __user *, upeer_addrlen)
1668 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1672 * Attempt to connect to a socket with the server address. The address
1673 * is in user space so we verify it is OK and move it to kernel space.
1675 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1678 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1679 * other SEQPACKET protocols that take time to connect() as it doesn't
1680 * include the -EINPROGRESS status for such sockets.
1683 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1686 struct socket *sock;
1687 struct sockaddr_storage address;
1688 int err, fput_needed;
1690 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1693 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1698 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1702 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1703 sock->file->f_flags);
1705 fput_light(sock->file, fput_needed);
1711 * Get the local address ('name') of a socket object. Move the obtained
1712 * name to user space.
1715 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1716 int __user *, usockaddr_len)
1718 struct socket *sock;
1719 struct sockaddr_storage address;
1720 int len, err, fput_needed;
1722 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1726 err = security_socket_getsockname(sock);
1730 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1733 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1736 fput_light(sock->file, fput_needed);
1742 * Get the remote address ('name') of a socket object. Move the obtained
1743 * name to user space.
1746 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1747 int __user *, usockaddr_len)
1749 struct socket *sock;
1750 struct sockaddr_storage address;
1751 int len, err, fput_needed;
1753 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1755 err = security_socket_getpeername(sock);
1757 fput_light(sock->file, fput_needed);
1762 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1765 err = move_addr_to_user(&address, len, usockaddr,
1767 fput_light(sock->file, fput_needed);
1773 * Send a datagram to a given address. We move the address into kernel
1774 * space and check the user space data area is readable before invoking
1778 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1779 unsigned int, flags, struct sockaddr __user *, addr,
1782 struct socket *sock;
1783 struct sockaddr_storage address;
1791 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1795 iov.iov_base = buff;
1797 msg.msg_name = NULL;
1800 msg.msg_control = NULL;
1801 msg.msg_controllen = 0;
1802 msg.msg_namelen = 0;
1804 err = move_addr_to_kernel(addr, addr_len, &address);
1807 msg.msg_name = (struct sockaddr *)&address;
1808 msg.msg_namelen = addr_len;
1810 if (sock->file->f_flags & O_NONBLOCK)
1811 flags |= MSG_DONTWAIT;
1812 msg.msg_flags = flags;
1813 err = sock_sendmsg(sock, &msg, len);
1816 fput_light(sock->file, fput_needed);
1822 * Send a datagram down a socket.
1825 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1826 unsigned int, flags)
1828 return sys_sendto(fd, buff, len, flags, NULL, 0);
1832 * Receive a frame from the socket and optionally record the address of the
1833 * sender. We verify the buffers are writable and if needed move the
1834 * sender address from kernel to user space.
1837 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1838 unsigned int, flags, struct sockaddr __user *, addr,
1839 int __user *, addr_len)
1841 struct socket *sock;
1844 struct sockaddr_storage address;
1850 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1854 msg.msg_control = NULL;
1855 msg.msg_controllen = 0;
1859 iov.iov_base = ubuf;
1860 /* Save some cycles and don't copy the address if not needed */
1861 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1862 /* We assume all kernel code knows the size of sockaddr_storage */
1863 msg.msg_namelen = 0;
1864 if (sock->file->f_flags & O_NONBLOCK)
1865 flags |= MSG_DONTWAIT;
1866 err = sock_recvmsg(sock, &msg, size, flags);
1868 if (err >= 0 && addr != NULL) {
1869 err2 = move_addr_to_user(&address,
1870 msg.msg_namelen, addr, addr_len);
1875 fput_light(sock->file, fput_needed);
1881 * Receive a datagram from a socket.
1884 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1885 unsigned int, flags)
1887 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1891 * Set a socket option. Because we don't know the option lengths we have
1892 * to pass the user mode parameter for the protocols to sort out.
1895 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1896 char __user *, optval, int, optlen)
1898 int err, fput_needed;
1899 struct socket *sock;
1904 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1906 err = security_socket_setsockopt(sock, level, optname);
1910 if (level == SOL_SOCKET)
1912 sock_setsockopt(sock, level, optname, optval,
1916 sock->ops->setsockopt(sock, level, optname, optval,
1919 fput_light(sock->file, fput_needed);
1925 * Get a socket option. Because we don't know the option lengths we have
1926 * to pass a user mode parameter for the protocols to sort out.
1929 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1930 char __user *, optval, int __user *, optlen)
1932 int err, fput_needed;
1933 struct socket *sock;
1935 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1937 err = security_socket_getsockopt(sock, level, optname);
1941 if (level == SOL_SOCKET)
1943 sock_getsockopt(sock, level, optname, optval,
1947 sock->ops->getsockopt(sock, level, optname, optval,
1950 fput_light(sock->file, fput_needed);
1956 * Shutdown a socket.
1959 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1961 int err, fput_needed;
1962 struct socket *sock;
1964 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1966 err = security_socket_shutdown(sock, how);
1968 err = sock->ops->shutdown(sock, how);
1969 fput_light(sock->file, fput_needed);
1974 /* A couple of helpful macros for getting the address of the 32/64 bit
1975 * fields which are the same type (int / unsigned) on our platforms.
1977 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1978 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1979 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1981 struct used_address {
1982 struct sockaddr_storage name;
1983 unsigned int name_len;
1986 static int copy_msghdr_from_user(struct msghdr *kmsg,
1987 struct msghdr __user *umsg)
1989 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1992 if (kmsg->msg_namelen < 0)
1995 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1996 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2000 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
2001 struct msghdr *msg_sys, unsigned int flags,
2002 struct used_address *used_address)
2004 struct compat_msghdr __user *msg_compat =
2005 (struct compat_msghdr __user *)msg;
2006 struct sockaddr_storage address;
2007 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2008 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2009 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2010 /* 20 is size of ipv6_pktinfo */
2011 unsigned char *ctl_buf = ctl;
2012 int err, ctl_len, total_len;
2015 if (MSG_CMSG_COMPAT & flags) {
2016 if (get_compat_msghdr(msg_sys, msg_compat))
2019 err = copy_msghdr_from_user(msg_sys, msg);
2024 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2026 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2029 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2035 /* This will also move the address data into kernel space */
2036 if (MSG_CMSG_COMPAT & flags) {
2037 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2039 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2046 if (msg_sys->msg_controllen > INT_MAX)
2048 ctl_len = msg_sys->msg_controllen;
2049 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2051 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2055 ctl_buf = msg_sys->msg_control;
2056 ctl_len = msg_sys->msg_controllen;
2057 } else if (ctl_len) {
2058 if (ctl_len > sizeof(ctl)) {
2059 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2060 if (ctl_buf == NULL)
2065 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2066 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2067 * checking falls down on this.
2069 if (copy_from_user(ctl_buf,
2070 (void __user __force *)msg_sys->msg_control,
2073 msg_sys->msg_control = ctl_buf;
2075 msg_sys->msg_flags = flags;
2077 if (sock->file->f_flags & O_NONBLOCK)
2078 msg_sys->msg_flags |= MSG_DONTWAIT;
2080 * If this is sendmmsg() and current destination address is same as
2081 * previously succeeded address, omit asking LSM's decision.
2082 * used_address->name_len is initialized to UINT_MAX so that the first
2083 * destination address never matches.
2085 if (used_address && msg_sys->msg_name &&
2086 used_address->name_len == msg_sys->msg_namelen &&
2087 !memcmp(&used_address->name, msg_sys->msg_name,
2088 used_address->name_len)) {
2089 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2092 err = sock_sendmsg(sock, msg_sys, total_len);
2094 * If this is sendmmsg() and sending to current destination address was
2095 * successful, remember it.
2097 if (used_address && err >= 0) {
2098 used_address->name_len = msg_sys->msg_namelen;
2099 if (msg_sys->msg_name)
2100 memcpy(&used_address->name, msg_sys->msg_name,
2101 used_address->name_len);
2106 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2108 if (iov != iovstack)
2115 * BSD sendmsg interface
2118 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2120 int fput_needed, err;
2121 struct msghdr msg_sys;
2122 struct socket *sock;
2124 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2128 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2130 fput_light(sock->file, fput_needed);
2135 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2137 if (flags & MSG_CMSG_COMPAT)
2139 return __sys_sendmsg(fd, msg, flags);
2143 * Linux sendmmsg interface
2146 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2149 int fput_needed, err, datagrams;
2150 struct socket *sock;
2151 struct mmsghdr __user *entry;
2152 struct compat_mmsghdr __user *compat_entry;
2153 struct msghdr msg_sys;
2154 struct used_address used_address;
2156 if (vlen > UIO_MAXIOV)
2161 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2165 used_address.name_len = UINT_MAX;
2167 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2170 while (datagrams < vlen) {
2171 if (MSG_CMSG_COMPAT & flags) {
2172 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2173 &msg_sys, flags, &used_address);
2176 err = __put_user(err, &compat_entry->msg_len);
2179 err = ___sys_sendmsg(sock,
2180 (struct msghdr __user *)entry,
2181 &msg_sys, flags, &used_address);
2184 err = put_user(err, &entry->msg_len);
2193 fput_light(sock->file, fput_needed);
2195 /* We only return an error if no datagrams were able to be sent */
2202 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2203 unsigned int, vlen, unsigned int, flags)
2205 if (flags & MSG_CMSG_COMPAT)
2207 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2210 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2211 struct msghdr *msg_sys, unsigned int flags, int nosec)
2213 struct compat_msghdr __user *msg_compat =
2214 (struct compat_msghdr __user *)msg;
2215 struct iovec iovstack[UIO_FASTIOV];
2216 struct iovec *iov = iovstack;
2217 unsigned long cmsg_ptr;
2218 int err, total_len, len;
2220 /* kernel mode address */
2221 struct sockaddr_storage addr;
2223 /* user mode address pointers */
2224 struct sockaddr __user *uaddr;
2225 int __user *uaddr_len;
2227 if (MSG_CMSG_COMPAT & flags) {
2228 if (get_compat_msghdr(msg_sys, msg_compat))
2231 err = copy_msghdr_from_user(msg_sys, msg);
2236 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2238 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2241 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2247 /* Save the user-mode address (verify_iovec will change the
2248 * kernel msghdr to use the kernel address space)
2250 uaddr = (__force void __user *)msg_sys->msg_name;
2251 uaddr_len = COMPAT_NAMELEN(msg);
2252 if (MSG_CMSG_COMPAT & flags)
2253 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2255 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2260 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2261 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2263 /* We assume all kernel code knows the size of sockaddr_storage */
2264 msg_sys->msg_namelen = 0;
2266 if (sock->file->f_flags & O_NONBLOCK)
2267 flags |= MSG_DONTWAIT;
2268 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2274 if (uaddr != NULL) {
2275 err = move_addr_to_user(&addr,
2276 msg_sys->msg_namelen, uaddr,
2281 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2285 if (MSG_CMSG_COMPAT & flags)
2286 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2287 &msg_compat->msg_controllen);
2289 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2290 &msg->msg_controllen);
2296 if (iov != iovstack)
2303 * BSD recvmsg interface
2306 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2308 int fput_needed, err;
2309 struct msghdr msg_sys;
2310 struct socket *sock;
2312 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2316 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2318 fput_light(sock->file, fput_needed);
2323 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2324 unsigned int, flags)
2326 if (flags & MSG_CMSG_COMPAT)
2328 return __sys_recvmsg(fd, msg, flags);
2332 * Linux recvmmsg interface
2335 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2336 unsigned int flags, struct timespec *timeout)
2338 int fput_needed, err, datagrams;
2339 struct socket *sock;
2340 struct mmsghdr __user *entry;
2341 struct compat_mmsghdr __user *compat_entry;
2342 struct msghdr msg_sys;
2343 struct timespec end_time;
2346 poll_select_set_timeout(&end_time, timeout->tv_sec,
2352 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2356 err = sock_error(sock->sk);
2361 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2363 while (datagrams < vlen) {
2365 * No need to ask LSM for more than the first datagram.
2367 if (MSG_CMSG_COMPAT & flags) {
2368 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2369 &msg_sys, flags & ~MSG_WAITFORONE,
2373 err = __put_user(err, &compat_entry->msg_len);
2376 err = ___sys_recvmsg(sock,
2377 (struct msghdr __user *)entry,
2378 &msg_sys, flags & ~MSG_WAITFORONE,
2382 err = put_user(err, &entry->msg_len);
2390 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2391 if (flags & MSG_WAITFORONE)
2392 flags |= MSG_DONTWAIT;
2395 ktime_get_ts(timeout);
2396 *timeout = timespec_sub(end_time, *timeout);
2397 if (timeout->tv_sec < 0) {
2398 timeout->tv_sec = timeout->tv_nsec = 0;
2402 /* Timeout, return less than vlen datagrams */
2403 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2407 /* Out of band data, return right away */
2408 if (msg_sys.msg_flags & MSG_OOB)
2413 fput_light(sock->file, fput_needed);
2418 if (datagrams != 0) {
2420 * We may return less entries than requested (vlen) if the
2421 * sock is non block and there aren't enough datagrams...
2423 if (err != -EAGAIN) {
2425 * ... or if recvmsg returns an error after we
2426 * received some datagrams, where we record the
2427 * error to return on the next call or if the
2428 * app asks about it using getsockopt(SO_ERROR).
2430 sock->sk->sk_err = -err;
2439 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2440 unsigned int, vlen, unsigned int, flags,
2441 struct timespec __user *, timeout)
2444 struct timespec timeout_sys;
2446 if (flags & MSG_CMSG_COMPAT)
2450 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2452 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2455 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2457 if (datagrams > 0 &&
2458 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2459 datagrams = -EFAULT;
2464 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2465 /* Argument list sizes for sys_socketcall */
2466 #define AL(x) ((x) * sizeof(unsigned long))
2467 static const unsigned char nargs[21] = {
2468 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2469 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2470 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2477 * System call vectors.
2479 * Argument checking cleaned up. Saved 20% in size.
2480 * This function doesn't need to set the kernel lock because
2481 * it is set by the callees.
2484 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2486 unsigned long a[AUDITSC_ARGS];
2487 unsigned long a0, a1;
2491 if (call < 1 || call > SYS_SENDMMSG)
2495 if (len > sizeof(a))
2498 /* copy_from_user should be SMP safe. */
2499 if (copy_from_user(a, args, len))
2502 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2511 err = sys_socket(a0, a1, a[2]);
2514 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2517 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2520 err = sys_listen(a0, a1);
2523 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2524 (int __user *)a[2], 0);
2526 case SYS_GETSOCKNAME:
2528 sys_getsockname(a0, (struct sockaddr __user *)a1,
2529 (int __user *)a[2]);
2531 case SYS_GETPEERNAME:
2533 sys_getpeername(a0, (struct sockaddr __user *)a1,
2534 (int __user *)a[2]);
2536 case SYS_SOCKETPAIR:
2537 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2540 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2543 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2544 (struct sockaddr __user *)a[4], a[5]);
2547 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2550 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2551 (struct sockaddr __user *)a[4],
2552 (int __user *)a[5]);
2555 err = sys_shutdown(a0, a1);
2557 case SYS_SETSOCKOPT:
2558 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2560 case SYS_GETSOCKOPT:
2562 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2563 (int __user *)a[4]);
2566 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2569 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2572 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2575 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2576 (struct timespec __user *)a[4]);
2579 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2580 (int __user *)a[2], a[3]);
2589 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2592 * sock_register - add a socket protocol handler
2593 * @ops: description of protocol
2595 * This function is called by a protocol handler that wants to
2596 * advertise its address family, and have it linked into the
2597 * socket interface. The value ops->family coresponds to the
2598 * socket system call protocol family.
2600 int sock_register(const struct net_proto_family *ops)
2604 if (ops->family >= NPROTO) {
2605 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2609 spin_lock(&net_family_lock);
2610 if (rcu_dereference_protected(net_families[ops->family],
2611 lockdep_is_held(&net_family_lock)))
2614 rcu_assign_pointer(net_families[ops->family], ops);
2617 spin_unlock(&net_family_lock);
2619 pr_info("NET: Registered protocol family %d\n", ops->family);
2622 EXPORT_SYMBOL(sock_register);
2625 * sock_unregister - remove a protocol handler
2626 * @family: protocol family to remove
2628 * This function is called by a protocol handler that wants to
2629 * remove its address family, and have it unlinked from the
2630 * new socket creation.
2632 * If protocol handler is a module, then it can use module reference
2633 * counts to protect against new references. If protocol handler is not
2634 * a module then it needs to provide its own protection in
2635 * the ops->create routine.
2637 void sock_unregister(int family)
2639 BUG_ON(family < 0 || family >= NPROTO);
2641 spin_lock(&net_family_lock);
2642 RCU_INIT_POINTER(net_families[family], NULL);
2643 spin_unlock(&net_family_lock);
2647 pr_info("NET: Unregistered protocol family %d\n", family);
2649 EXPORT_SYMBOL(sock_unregister);
2651 static int __init sock_init(void)
2655 * Initialize the network sysctl infrastructure.
2657 err = net_sysctl_init();
2662 * Initialize skbuff SLAB cache
2667 * Initialize the protocols module.
2672 err = register_filesystem(&sock_fs_type);
2675 sock_mnt = kern_mount(&sock_fs_type);
2676 if (IS_ERR(sock_mnt)) {
2677 err = PTR_ERR(sock_mnt);
2681 /* The real protocol initialization is performed in later initcalls.
2684 #ifdef CONFIG_NETFILTER
2685 err = netfilter_init();
2690 ptp_classifier_init();
2696 unregister_filesystem(&sock_fs_type);
2701 core_initcall(sock_init); /* early initcall */
2703 #ifdef CONFIG_PROC_FS
2704 void socket_seq_show(struct seq_file *seq)
2709 for_each_possible_cpu(cpu)
2710 counter += per_cpu(sockets_in_use, cpu);
2712 /* It can be negative, by the way. 8) */
2716 seq_printf(seq, "sockets: used %d\n", counter);
2718 #endif /* CONFIG_PROC_FS */
2720 #ifdef CONFIG_COMPAT
2721 static int do_siocgstamp(struct net *net, struct socket *sock,
2722 unsigned int cmd, void __user *up)
2724 mm_segment_t old_fs = get_fs();
2729 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2732 err = compat_put_timeval(&ktv, up);
2737 static int do_siocgstampns(struct net *net, struct socket *sock,
2738 unsigned int cmd, void __user *up)
2740 mm_segment_t old_fs = get_fs();
2741 struct timespec kts;
2745 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2748 err = compat_put_timespec(&kts, up);
2753 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2755 struct ifreq __user *uifr;
2758 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2759 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2762 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2766 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2772 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2774 struct compat_ifconf ifc32;
2776 struct ifconf __user *uifc;
2777 struct compat_ifreq __user *ifr32;
2778 struct ifreq __user *ifr;
2782 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2785 memset(&ifc, 0, sizeof(ifc));
2786 if (ifc32.ifcbuf == 0) {
2790 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2792 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2793 sizeof(struct ifreq);
2794 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2796 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2797 ifr32 = compat_ptr(ifc32.ifcbuf);
2798 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2799 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2805 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2808 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2812 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2816 ifr32 = compat_ptr(ifc32.ifcbuf);
2818 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2819 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2820 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2826 if (ifc32.ifcbuf == 0) {
2827 /* Translate from 64-bit structure multiple to
2831 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2836 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2842 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2844 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2845 bool convert_in = false, convert_out = false;
2846 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2847 struct ethtool_rxnfc __user *rxnfc;
2848 struct ifreq __user *ifr;
2849 u32 rule_cnt = 0, actual_rule_cnt;
2854 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2857 compat_rxnfc = compat_ptr(data);
2859 if (get_user(ethcmd, &compat_rxnfc->cmd))
2862 /* Most ethtool structures are defined without padding.
2863 * Unfortunately struct ethtool_rxnfc is an exception.
2868 case ETHTOOL_GRXCLSRLALL:
2869 /* Buffer size is variable */
2870 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2872 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2874 buf_size += rule_cnt * sizeof(u32);
2876 case ETHTOOL_GRXRINGS:
2877 case ETHTOOL_GRXCLSRLCNT:
2878 case ETHTOOL_GRXCLSRULE:
2879 case ETHTOOL_SRXCLSRLINS:
2882 case ETHTOOL_SRXCLSRLDEL:
2883 buf_size += sizeof(struct ethtool_rxnfc);
2888 ifr = compat_alloc_user_space(buf_size);
2889 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2891 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2894 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2895 &ifr->ifr_ifru.ifru_data))
2899 /* We expect there to be holes between fs.m_ext and
2900 * fs.ring_cookie and at the end of fs, but nowhere else.
2902 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2903 sizeof(compat_rxnfc->fs.m_ext) !=
2904 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2905 sizeof(rxnfc->fs.m_ext));
2907 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2908 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2909 offsetof(struct ethtool_rxnfc, fs.location) -
2910 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2912 if (copy_in_user(rxnfc, compat_rxnfc,
2913 (void __user *)(&rxnfc->fs.m_ext + 1) -
2914 (void __user *)rxnfc) ||
2915 copy_in_user(&rxnfc->fs.ring_cookie,
2916 &compat_rxnfc->fs.ring_cookie,
2917 (void __user *)(&rxnfc->fs.location + 1) -
2918 (void __user *)&rxnfc->fs.ring_cookie) ||
2919 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2920 sizeof(rxnfc->rule_cnt)))
2924 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2929 if (copy_in_user(compat_rxnfc, rxnfc,
2930 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2931 (const void __user *)rxnfc) ||
2932 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2933 &rxnfc->fs.ring_cookie,
2934 (const void __user *)(&rxnfc->fs.location + 1) -
2935 (const void __user *)&rxnfc->fs.ring_cookie) ||
2936 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2937 sizeof(rxnfc->rule_cnt)))
2940 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2941 /* As an optimisation, we only copy the actual
2942 * number of rules that the underlying
2943 * function returned. Since Mallory might
2944 * change the rule count in user memory, we
2945 * check that it is less than the rule count
2946 * originally given (as the user buffer size),
2947 * which has been range-checked.
2949 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2951 if (actual_rule_cnt < rule_cnt)
2952 rule_cnt = actual_rule_cnt;
2953 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2954 &rxnfc->rule_locs[0],
2955 rule_cnt * sizeof(u32)))
2963 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2966 compat_uptr_t uptr32;
2967 struct ifreq __user *uifr;
2969 uifr = compat_alloc_user_space(sizeof(*uifr));
2970 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2973 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2976 uptr = compat_ptr(uptr32);
2978 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2981 return dev_ioctl(net, SIOCWANDEV, uifr);
2984 static int bond_ioctl(struct net *net, unsigned int cmd,
2985 struct compat_ifreq __user *ifr32)
2988 mm_segment_t old_fs;
2992 case SIOCBONDENSLAVE:
2993 case SIOCBONDRELEASE:
2994 case SIOCBONDSETHWADDR:
2995 case SIOCBONDCHANGEACTIVE:
2996 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
3001 err = dev_ioctl(net, cmd,
3002 (struct ifreq __user __force *) &kifr);
3007 return -ENOIOCTLCMD;
3011 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3012 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3013 struct compat_ifreq __user *u_ifreq32)
3015 struct ifreq __user *u_ifreq64;
3016 char tmp_buf[IFNAMSIZ];
3017 void __user *data64;
3020 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3023 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3025 data64 = compat_ptr(data32);
3027 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3029 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3032 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3035 return dev_ioctl(net, cmd, u_ifreq64);
3038 static int dev_ifsioc(struct net *net, struct socket *sock,
3039 unsigned int cmd, struct compat_ifreq __user *uifr32)
3041 struct ifreq __user *uifr;
3044 uifr = compat_alloc_user_space(sizeof(*uifr));
3045 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3048 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3059 case SIOCGIFBRDADDR:
3060 case SIOCGIFDSTADDR:
3061 case SIOCGIFNETMASK:
3066 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3074 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3075 struct compat_ifreq __user *uifr32)
3078 struct compat_ifmap __user *uifmap32;
3079 mm_segment_t old_fs;
3082 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3083 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3084 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3085 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3086 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3087 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3088 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3089 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3095 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3098 if (cmd == SIOCGIFMAP && !err) {
3099 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3100 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3101 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3102 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3103 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3104 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3105 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
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 compat_ifr_data_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 SIOCBONDCHANGEACTIVE:
3244 return bond_ioctl(net, cmd, argp);
3247 return routing_ioctl(net, sock, cmd, argp);
3249 return do_siocgstamp(net, sock, cmd, argp);
3251 return do_siocgstampns(net, sock, cmd, argp);
3252 case SIOCBONDSLAVEINFOQUERY:
3253 case SIOCBONDINFOQUERY:
3256 return compat_ifr_data_ioctl(net, cmd, argp);
3268 return sock_ioctl(file, cmd, arg);
3285 case SIOCSIFHWBROADCAST:
3287 case SIOCGIFBRDADDR:
3288 case SIOCSIFBRDADDR:
3289 case SIOCGIFDSTADDR:
3290 case SIOCSIFDSTADDR:
3291 case SIOCGIFNETMASK:
3292 case SIOCSIFNETMASK:
3303 return dev_ifsioc(net, sock, cmd, argp);
3309 return sock_do_ioctl(net, sock, cmd, arg);
3312 return -ENOIOCTLCMD;
3315 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3318 struct socket *sock = file->private_data;
3319 int ret = -ENOIOCTLCMD;
3326 if (sock->ops->compat_ioctl)
3327 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3329 if (ret == -ENOIOCTLCMD &&
3330 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3331 ret = compat_wext_handle_ioctl(net, cmd, arg);
3333 if (ret == -ENOIOCTLCMD)
3334 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3340 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3342 return sock->ops->bind(sock, addr, addrlen);
3344 EXPORT_SYMBOL(kernel_bind);
3346 int kernel_listen(struct socket *sock, int backlog)
3348 return sock->ops->listen(sock, backlog);
3350 EXPORT_SYMBOL(kernel_listen);
3352 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3354 struct sock *sk = sock->sk;
3357 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3362 err = sock->ops->accept(sock, *newsock, flags);
3364 sock_release(*newsock);
3369 (*newsock)->ops = sock->ops;
3370 __module_get((*newsock)->ops->owner);
3375 EXPORT_SYMBOL(kernel_accept);
3377 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3380 return sock->ops->connect(sock, addr, addrlen, flags);
3382 EXPORT_SYMBOL(kernel_connect);
3384 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3387 return sock->ops->getname(sock, addr, addrlen, 0);
3389 EXPORT_SYMBOL(kernel_getsockname);
3391 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3394 return sock->ops->getname(sock, addr, addrlen, 1);
3396 EXPORT_SYMBOL(kernel_getpeername);
3398 int kernel_getsockopt(struct socket *sock, int level, int optname,
3399 char *optval, int *optlen)
3401 mm_segment_t oldfs = get_fs();
3402 char __user *uoptval;
3403 int __user *uoptlen;
3406 uoptval = (char __user __force *) optval;
3407 uoptlen = (int __user __force *) optlen;
3410 if (level == SOL_SOCKET)
3411 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3413 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3418 EXPORT_SYMBOL(kernel_getsockopt);
3420 int kernel_setsockopt(struct socket *sock, int level, int optname,
3421 char *optval, unsigned int optlen)
3423 mm_segment_t oldfs = get_fs();
3424 char __user *uoptval;
3427 uoptval = (char __user __force *) optval;
3430 if (level == SOL_SOCKET)
3431 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3433 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3438 EXPORT_SYMBOL(kernel_setsockopt);
3440 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3441 size_t size, int flags)
3443 if (sock->ops->sendpage)
3444 return sock->ops->sendpage(sock, page, offset, size, flags);
3446 return sock_no_sendpage(sock, page, offset, size, flags);
3448 EXPORT_SYMBOL(kernel_sendpage);
3450 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3452 mm_segment_t oldfs = get_fs();
3456 err = sock->ops->ioctl(sock, cmd, arg);
3461 EXPORT_SYMBOL(kernel_sock_ioctl);
3463 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3465 return sock->ops->shutdown(sock, how);
3467 EXPORT_SYMBOL(kernel_sock_shutdown);