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 ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117 unsigned long nr_segs, loff_t pos);
118 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119 unsigned long nr_segs, loff_t pos);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
122 static int sock_close(struct inode *inode, struct file *file);
123 static unsigned int sock_poll(struct file *file,
124 struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
127 static long compat_sock_ioctl(struct file *file,
128 unsigned int cmd, unsigned long arg);
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_sendpage(struct file *file, struct page *page,
132 int offset, size_t size, loff_t *ppos, int more);
133 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134 struct pipe_inode_info *pipe, size_t len,
138 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
139 * in the operation structures but are done directly via the socketcall() multiplexor.
142 static const struct file_operations socket_file_ops = {
143 .owner = THIS_MODULE,
145 .aio_read = sock_aio_read,
146 .aio_write = sock_aio_write,
148 .unlocked_ioctl = sock_ioctl,
150 .compat_ioctl = compat_sock_ioctl,
153 .release = sock_close,
154 .fasync = sock_fasync,
155 .sendpage = sock_sendpage,
156 .splice_write = generic_splice_sendpage,
157 .splice_read = sock_splice_read,
161 * The protocol list. Each protocol is registered in here.
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
168 * Statistics counters of the socket lists
171 static DEFINE_PER_CPU(int, sockets_in_use);
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
192 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
196 if (copy_from_user(kaddr, uaddr, ulen))
198 return audit_sockaddr(ulen, kaddr);
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219 void __user *uaddr, int __user *ulen)
224 BUG_ON(klen > sizeof(struct sockaddr_storage));
225 err = get_user(len, ulen);
233 if (audit_sockaddr(klen, kaddr))
235 if (copy_to_user(uaddr, kaddr, len))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen, ulen);
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
247 static struct inode *sock_alloc_inode(struct super_block *sb)
249 struct socket_alloc *ei;
250 struct socket_wq *wq;
252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
257 kmem_cache_free(sock_inode_cachep, ei);
260 init_waitqueue_head(&wq->wait);
261 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
270 return &ei->vfs_inode;
273 static void sock_destroy_inode(struct inode *inode)
275 struct socket_alloc *ei;
276 struct socket_wq *wq;
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
279 wq = rcu_dereference_protected(ei->socket.wq, 1);
281 kmem_cache_free(sock_inode_cachep, ei);
284 static void init_once(void *foo)
286 struct socket_alloc *ei = (struct socket_alloc *)foo;
288 inode_init_once(&ei->vfs_inode);
291 static int init_inodecache(void)
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294 sizeof(struct socket_alloc),
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
300 if (sock_inode_cachep == NULL)
305 static const struct super_operations sockfs_ops = {
306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
312 * sockfs_dname() is called from d_path().
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 dentry->d_inode->i_ino);
320 static const struct dentry_operations sockfs_dentry_operations = {
321 .d_dname = sockfs_dname,
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
331 static struct vfsmount *sock_mnt __read_mostly;
333 static struct file_system_type sock_fs_type = {
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
340 * Obtains the first available file descriptor and sets it up for use.
342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
358 struct qstr name = { .name = "" };
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
372 path.mnt = mntget(sock_mnt);
374 d_instantiate(path.dentry, SOCK_INODE(sock));
375 SOCK_INODE(sock)->i_fop = &socket_file_ops;
377 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
379 if (unlikely(IS_ERR(file))) {
380 /* drop dentry, keep inode */
381 ihold(path.dentry->d_inode);
387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
388 file->private_data = sock;
391 EXPORT_SYMBOL(sock_alloc_file);
393 static int sock_map_fd(struct socket *sock, int flags)
395 struct file *newfile;
396 int fd = get_unused_fd_flags(flags);
397 if (unlikely(fd < 0))
400 newfile = sock_alloc_file(sock, flags, NULL);
401 if (likely(!IS_ERR(newfile))) {
402 fd_install(fd, newfile);
407 return PTR_ERR(newfile);
410 struct socket *sock_from_file(struct file *file, int *err)
412 if (file->f_op == &socket_file_ops)
413 return file->private_data; /* set in sock_map_fd */
418 EXPORT_SYMBOL(sock_from_file);
421 * sockfd_lookup - Go from a file number to its socket slot
423 * @err: pointer to an error code return
425 * The file handle passed in is locked and the socket it is bound
426 * too is returned. If an error occurs the err pointer is overwritten
427 * with a negative errno code and NULL is returned. The function checks
428 * for both invalid handles and passing a handle which is not a socket.
430 * On a success the socket object pointer is returned.
433 struct socket *sockfd_lookup(int fd, int *err)
444 sock = sock_from_file(file, err);
449 EXPORT_SYMBOL(sockfd_lookup);
451 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
453 struct fd f = fdget(fd);
458 sock = sock_from_file(f.file, err);
460 *fput_needed = f.flags;
468 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
469 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
470 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
471 static ssize_t sockfs_getxattr(struct dentry *dentry,
472 const char *name, void *value, size_t size)
474 const char *proto_name;
479 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
480 proto_name = dentry->d_name.name;
481 proto_size = strlen(proto_name);
485 if (proto_size + 1 > size)
488 strncpy(value, proto_name, proto_size + 1);
490 error = proto_size + 1;
497 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
503 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
513 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
518 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
525 static const struct inode_operations sockfs_inode_ops = {
526 .getxattr = sockfs_getxattr,
527 .listxattr = sockfs_listxattr,
531 * sock_alloc - allocate a socket
533 * Allocate a new inode and socket object. The two are bound together
534 * and initialised. The socket is then returned. If we are out of inodes
538 static struct socket *sock_alloc(void)
543 inode = new_inode_pseudo(sock_mnt->mnt_sb);
547 sock = SOCKET_I(inode);
549 kmemcheck_annotate_bitfield(sock, type);
550 inode->i_ino = get_next_ino();
551 inode->i_mode = S_IFSOCK | S_IRWXUGO;
552 inode->i_uid = current_fsuid();
553 inode->i_gid = current_fsgid();
554 inode->i_op = &sockfs_inode_ops;
556 this_cpu_add(sockets_in_use, 1);
561 * sock_release - close a socket
562 * @sock: socket to close
564 * The socket is released from the protocol stack if it has a release
565 * callback, and the inode is then released if the socket is bound to
566 * an inode not a file.
569 void sock_release(struct socket *sock)
572 struct module *owner = sock->ops->owner;
574 sock->ops->release(sock);
579 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
580 pr_err("%s: fasync list not empty!\n", __func__);
582 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
585 this_cpu_sub(sockets_in_use, 1);
587 iput(SOCK_INODE(sock));
592 EXPORT_SYMBOL(sock_release);
594 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
596 u8 flags = *tx_flags;
598 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
599 flags |= SKBTX_HW_TSTAMP;
601 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
602 flags |= SKBTX_SW_TSTAMP;
604 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
605 flags |= SKBTX_SCHED_TSTAMP;
607 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
608 flags |= SKBTX_ACK_TSTAMP;
612 EXPORT_SYMBOL(__sock_tx_timestamp);
614 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
615 struct msghdr *msg, size_t size)
617 struct sock_iocb *si = kiocb_to_siocb(iocb);
624 return sock->ops->sendmsg(iocb, sock, msg, size);
627 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
628 struct msghdr *msg, size_t size)
630 int err = security_socket_sendmsg(sock, msg, size);
632 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
635 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
636 size_t size, bool nosec)
639 struct sock_iocb siocb;
642 init_sync_kiocb(&iocb, NULL);
643 iocb.private = &siocb;
644 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
645 __sock_sendmsg(&iocb, sock, msg, size);
646 if (-EIOCBQUEUED == ret)
647 ret = wait_on_sync_kiocb(&iocb);
651 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
653 return do_sock_sendmsg(sock, msg, size, false);
655 EXPORT_SYMBOL(sock_sendmsg);
657 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
659 return do_sock_sendmsg(sock, msg, size, true);
662 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
663 struct kvec *vec, size_t num, size_t size)
665 mm_segment_t oldfs = get_fs();
670 * the following is safe, since for compiler definitions of kvec and
671 * iovec are identical, yielding the same in-core layout and alignment
673 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
674 result = sock_sendmsg(sock, msg, size);
678 EXPORT_SYMBOL(kernel_sendmsg);
681 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
683 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
686 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
687 struct scm_timestamping tss;
689 struct skb_shared_hwtstamps *shhwtstamps =
692 /* Race occurred between timestamp enabling and packet
693 receiving. Fill in the current time for now. */
694 if (need_software_tstamp && skb->tstamp.tv64 == 0)
695 __net_timestamp(skb);
697 if (need_software_tstamp) {
698 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
700 skb_get_timestamp(skb, &tv);
701 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
705 skb_get_timestampns(skb, &ts);
706 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
711 memset(&tss, 0, sizeof(tss));
712 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
713 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
716 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
717 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
720 put_cmsg(msg, SOL_SOCKET,
721 SCM_TIMESTAMPING, sizeof(tss), &tss);
723 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
725 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
730 if (!sock_flag(sk, SOCK_WIFI_STATUS))
732 if (!skb->wifi_acked_valid)
735 ack = skb->wifi_acked;
737 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
739 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
741 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
744 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
745 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
746 sizeof(__u32), &skb->dropcount);
749 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
752 sock_recv_timestamp(msg, sk, skb);
753 sock_recv_drops(msg, sk, skb);
755 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
757 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
758 struct msghdr *msg, size_t size, int flags)
760 struct sock_iocb *si = kiocb_to_siocb(iocb);
768 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
771 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
772 struct msghdr *msg, size_t size, int flags)
774 int err = security_socket_recvmsg(sock, msg, size, flags);
776 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
779 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
780 size_t size, int flags)
783 struct sock_iocb siocb;
786 init_sync_kiocb(&iocb, NULL);
787 iocb.private = &siocb;
788 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
789 if (-EIOCBQUEUED == ret)
790 ret = wait_on_sync_kiocb(&iocb);
793 EXPORT_SYMBOL(sock_recvmsg);
795 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
796 size_t size, int flags)
799 struct sock_iocb siocb;
802 init_sync_kiocb(&iocb, NULL);
803 iocb.private = &siocb;
804 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
805 if (-EIOCBQUEUED == ret)
806 ret = wait_on_sync_kiocb(&iocb);
811 * kernel_recvmsg - Receive a message from a socket (kernel space)
812 * @sock: The socket to receive the message from
813 * @msg: Received message
814 * @vec: Input s/g array for message data
815 * @num: Size of input s/g array
816 * @size: Number of bytes to read
817 * @flags: Message flags (MSG_DONTWAIT, etc...)
819 * On return the msg structure contains the scatter/gather array passed in the
820 * vec argument. The array is modified so that it consists of the unfilled
821 * portion of the original array.
823 * The returned value is the total number of bytes received, or an error.
825 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
826 struct kvec *vec, size_t num, size_t size, int flags)
828 mm_segment_t oldfs = get_fs();
833 * the following is safe, since for compiler definitions of kvec and
834 * iovec are identical, yielding the same in-core layout and alignment
836 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
837 result = sock_recvmsg(sock, msg, size, flags);
841 EXPORT_SYMBOL(kernel_recvmsg);
843 static ssize_t sock_sendpage(struct file *file, struct page *page,
844 int offset, size_t size, loff_t *ppos, int more)
849 sock = file->private_data;
851 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
852 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
855 return kernel_sendpage(sock, page, offset, size, flags);
858 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
859 struct pipe_inode_info *pipe, size_t len,
862 struct socket *sock = file->private_data;
864 if (unlikely(!sock->ops->splice_read))
867 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
870 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
871 struct sock_iocb *siocb)
873 if (!is_sync_kiocb(iocb))
877 iocb->private = siocb;
881 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
882 struct file *file, const struct iovec *iov,
883 unsigned long nr_segs)
885 struct socket *sock = file->private_data;
889 for (i = 0; i < nr_segs; i++)
890 size += iov[i].iov_len;
892 msg->msg_name = NULL;
893 msg->msg_namelen = 0;
894 msg->msg_control = NULL;
895 msg->msg_controllen = 0;
896 iov_iter_init(&msg->msg_iter, READ, iov, nr_segs, size);
897 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
899 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
902 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
903 unsigned long nr_segs, loff_t pos)
905 struct sock_iocb siocb, *x;
910 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
914 x = alloc_sock_iocb(iocb, &siocb);
917 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
920 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
921 struct file *file, const struct iovec *iov,
922 unsigned long nr_segs)
924 struct socket *sock = file->private_data;
928 for (i = 0; i < nr_segs; i++)
929 size += iov[i].iov_len;
931 msg->msg_name = NULL;
932 msg->msg_namelen = 0;
933 msg->msg_control = NULL;
934 msg->msg_controllen = 0;
935 iov_iter_init(&msg->msg_iter, WRITE, iov, nr_segs, size);
936 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
937 if (sock->type == SOCK_SEQPACKET)
938 msg->msg_flags |= MSG_EOR;
940 return __sock_sendmsg(iocb, sock, msg, size);
943 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
944 unsigned long nr_segs, loff_t pos)
946 struct sock_iocb siocb, *x;
951 x = alloc_sock_iocb(iocb, &siocb);
955 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
959 * Atomic setting of ioctl hooks to avoid race
960 * with module unload.
963 static DEFINE_MUTEX(br_ioctl_mutex);
964 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
966 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
968 mutex_lock(&br_ioctl_mutex);
969 br_ioctl_hook = hook;
970 mutex_unlock(&br_ioctl_mutex);
972 EXPORT_SYMBOL(brioctl_set);
974 static DEFINE_MUTEX(vlan_ioctl_mutex);
975 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
977 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
979 mutex_lock(&vlan_ioctl_mutex);
980 vlan_ioctl_hook = hook;
981 mutex_unlock(&vlan_ioctl_mutex);
983 EXPORT_SYMBOL(vlan_ioctl_set);
985 static DEFINE_MUTEX(dlci_ioctl_mutex);
986 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
988 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
990 mutex_lock(&dlci_ioctl_mutex);
991 dlci_ioctl_hook = hook;
992 mutex_unlock(&dlci_ioctl_mutex);
994 EXPORT_SYMBOL(dlci_ioctl_set);
996 static long sock_do_ioctl(struct net *net, struct socket *sock,
997 unsigned int cmd, unsigned long arg)
1000 void __user *argp = (void __user *)arg;
1002 err = sock->ops->ioctl(sock, cmd, arg);
1005 * If this ioctl is unknown try to hand it down
1006 * to the NIC driver.
1008 if (err == -ENOIOCTLCMD)
1009 err = dev_ioctl(net, cmd, argp);
1015 * With an ioctl, arg may well be a user mode pointer, but we don't know
1016 * what to do with it - that's up to the protocol still.
1019 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1021 struct socket *sock;
1023 void __user *argp = (void __user *)arg;
1027 sock = file->private_data;
1030 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1031 err = dev_ioctl(net, cmd, argp);
1033 #ifdef CONFIG_WEXT_CORE
1034 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1035 err = dev_ioctl(net, cmd, argp);
1042 if (get_user(pid, (int __user *)argp))
1044 f_setown(sock->file, pid, 1);
1049 err = put_user(f_getown(sock->file),
1050 (int __user *)argp);
1058 request_module("bridge");
1060 mutex_lock(&br_ioctl_mutex);
1062 err = br_ioctl_hook(net, cmd, argp);
1063 mutex_unlock(&br_ioctl_mutex);
1068 if (!vlan_ioctl_hook)
1069 request_module("8021q");
1071 mutex_lock(&vlan_ioctl_mutex);
1072 if (vlan_ioctl_hook)
1073 err = vlan_ioctl_hook(net, argp);
1074 mutex_unlock(&vlan_ioctl_mutex);
1079 if (!dlci_ioctl_hook)
1080 request_module("dlci");
1082 mutex_lock(&dlci_ioctl_mutex);
1083 if (dlci_ioctl_hook)
1084 err = dlci_ioctl_hook(cmd, argp);
1085 mutex_unlock(&dlci_ioctl_mutex);
1088 err = sock_do_ioctl(net, sock, cmd, arg);
1094 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1097 struct socket *sock = NULL;
1099 err = security_socket_create(family, type, protocol, 1);
1103 sock = sock_alloc();
1110 err = security_socket_post_create(sock, family, type, protocol, 1);
1122 EXPORT_SYMBOL(sock_create_lite);
1124 /* No kernel lock held - perfect */
1125 static unsigned int sock_poll(struct file *file, poll_table *wait)
1127 unsigned int busy_flag = 0;
1128 struct socket *sock;
1131 * We can't return errors to poll, so it's either yes or no.
1133 sock = file->private_data;
1135 if (sk_can_busy_loop(sock->sk)) {
1136 /* this socket can poll_ll so tell the system call */
1137 busy_flag = POLL_BUSY_LOOP;
1139 /* once, only if requested by syscall */
1140 if (wait && (wait->_key & POLL_BUSY_LOOP))
1141 sk_busy_loop(sock->sk, 1);
1144 return busy_flag | sock->ops->poll(file, sock, wait);
1147 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1149 struct socket *sock = file->private_data;
1151 return sock->ops->mmap(file, sock, vma);
1154 static int sock_close(struct inode *inode, struct file *filp)
1156 sock_release(SOCKET_I(inode));
1161 * Update the socket async list
1163 * Fasync_list locking strategy.
1165 * 1. fasync_list is modified only under process context socket lock
1166 * i.e. under semaphore.
1167 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1168 * or under socket lock
1171 static int sock_fasync(int fd, struct file *filp, int on)
1173 struct socket *sock = filp->private_data;
1174 struct sock *sk = sock->sk;
1175 struct socket_wq *wq;
1181 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1182 fasync_helper(fd, filp, on, &wq->fasync_list);
1184 if (!wq->fasync_list)
1185 sock_reset_flag(sk, SOCK_FASYNC);
1187 sock_set_flag(sk, SOCK_FASYNC);
1193 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1195 int sock_wake_async(struct socket *sock, int how, int band)
1197 struct socket_wq *wq;
1202 wq = rcu_dereference(sock->wq);
1203 if (!wq || !wq->fasync_list) {
1208 case SOCK_WAKE_WAITD:
1209 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1212 case SOCK_WAKE_SPACE:
1213 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1218 kill_fasync(&wq->fasync_list, SIGIO, band);
1221 kill_fasync(&wq->fasync_list, SIGURG, band);
1226 EXPORT_SYMBOL(sock_wake_async);
1228 int __sock_create(struct net *net, int family, int type, int protocol,
1229 struct socket **res, int kern)
1232 struct socket *sock;
1233 const struct net_proto_family *pf;
1236 * Check protocol is in range
1238 if (family < 0 || family >= NPROTO)
1239 return -EAFNOSUPPORT;
1240 if (type < 0 || type >= SOCK_MAX)
1245 This uglymoron is moved from INET layer to here to avoid
1246 deadlock in module load.
1248 if (family == PF_INET && type == SOCK_PACKET) {
1252 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1258 err = security_socket_create(family, type, protocol, kern);
1263 * Allocate the socket and allow the family to set things up. if
1264 * the protocol is 0, the family is instructed to select an appropriate
1267 sock = sock_alloc();
1269 net_warn_ratelimited("socket: no more sockets\n");
1270 return -ENFILE; /* Not exactly a match, but its the
1271 closest posix thing */
1276 #ifdef CONFIG_MODULES
1277 /* Attempt to load a protocol module if the find failed.
1279 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1280 * requested real, full-featured networking support upon configuration.
1281 * Otherwise module support will break!
1283 if (rcu_access_pointer(net_families[family]) == NULL)
1284 request_module("net-pf-%d", family);
1288 pf = rcu_dereference(net_families[family]);
1289 err = -EAFNOSUPPORT;
1294 * We will call the ->create function, that possibly is in a loadable
1295 * module, so we have to bump that loadable module refcnt first.
1297 if (!try_module_get(pf->owner))
1300 /* Now protected by module ref count */
1303 err = pf->create(net, sock, protocol, kern);
1305 goto out_module_put;
1308 * Now to bump the refcnt of the [loadable] module that owns this
1309 * socket at sock_release time we decrement its refcnt.
1311 if (!try_module_get(sock->ops->owner))
1312 goto out_module_busy;
1315 * Now that we're done with the ->create function, the [loadable]
1316 * module can have its refcnt decremented
1318 module_put(pf->owner);
1319 err = security_socket_post_create(sock, family, type, protocol, kern);
1321 goto out_sock_release;
1327 err = -EAFNOSUPPORT;
1330 module_put(pf->owner);
1337 goto out_sock_release;
1339 EXPORT_SYMBOL(__sock_create);
1341 int sock_create(int family, int type, int protocol, struct socket **res)
1343 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1345 EXPORT_SYMBOL(sock_create);
1347 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1349 return __sock_create(&init_net, family, type, protocol, res, 1);
1351 EXPORT_SYMBOL(sock_create_kern);
1353 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1356 struct socket *sock;
1359 /* Check the SOCK_* constants for consistency. */
1360 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1361 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1362 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1363 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1365 flags = type & ~SOCK_TYPE_MASK;
1366 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1368 type &= SOCK_TYPE_MASK;
1370 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1371 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1373 retval = sock_create(family, type, protocol, &sock);
1377 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1382 /* It may be already another descriptor 8) Not kernel problem. */
1391 * Create a pair of connected sockets.
1394 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1395 int __user *, usockvec)
1397 struct socket *sock1, *sock2;
1399 struct file *newfile1, *newfile2;
1402 flags = type & ~SOCK_TYPE_MASK;
1403 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1405 type &= SOCK_TYPE_MASK;
1407 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1408 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1411 * Obtain the first socket and check if the underlying protocol
1412 * supports the socketpair call.
1415 err = sock_create(family, type, protocol, &sock1);
1419 err = sock_create(family, type, protocol, &sock2);
1423 err = sock1->ops->socketpair(sock1, sock2);
1425 goto out_release_both;
1427 fd1 = get_unused_fd_flags(flags);
1428 if (unlikely(fd1 < 0)) {
1430 goto out_release_both;
1433 fd2 = get_unused_fd_flags(flags);
1434 if (unlikely(fd2 < 0)) {
1436 goto out_put_unused_1;
1439 newfile1 = sock_alloc_file(sock1, flags, NULL);
1440 if (unlikely(IS_ERR(newfile1))) {
1441 err = PTR_ERR(newfile1);
1442 goto out_put_unused_both;
1445 newfile2 = sock_alloc_file(sock2, flags, NULL);
1446 if (IS_ERR(newfile2)) {
1447 err = PTR_ERR(newfile2);
1451 err = put_user(fd1, &usockvec[0]);
1455 err = put_user(fd2, &usockvec[1]);
1459 audit_fd_pair(fd1, fd2);
1461 fd_install(fd1, newfile1);
1462 fd_install(fd2, newfile2);
1463 /* fd1 and fd2 may be already another descriptors.
1464 * Not kernel problem.
1480 sock_release(sock2);
1483 out_put_unused_both:
1488 sock_release(sock2);
1490 sock_release(sock1);
1496 * Bind a name to a socket. Nothing much to do here since it's
1497 * the protocol's responsibility to handle the local address.
1499 * We move the socket address to kernel space before we call
1500 * the protocol layer (having also checked the address is ok).
1503 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1505 struct socket *sock;
1506 struct sockaddr_storage address;
1507 int err, fput_needed;
1509 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1511 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1513 err = security_socket_bind(sock,
1514 (struct sockaddr *)&address,
1517 err = sock->ops->bind(sock,
1521 fput_light(sock->file, fput_needed);
1527 * Perform a listen. Basically, we allow the protocol to do anything
1528 * necessary for a listen, and if that works, we mark the socket as
1529 * ready for listening.
1532 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1534 struct socket *sock;
1535 int err, fput_needed;
1538 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1540 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1541 if ((unsigned int)backlog > somaxconn)
1542 backlog = somaxconn;
1544 err = security_socket_listen(sock, backlog);
1546 err = sock->ops->listen(sock, backlog);
1548 fput_light(sock->file, fput_needed);
1554 * For accept, we attempt to create a new socket, set up the link
1555 * with the client, wake up the client, then return the new
1556 * connected fd. We collect the address of the connector in kernel
1557 * space and move it to user at the very end. This is unclean because
1558 * we open the socket then return an error.
1560 * 1003.1g adds the ability to recvmsg() to query connection pending
1561 * status to recvmsg. We need to add that support in a way thats
1562 * clean when we restucture accept also.
1565 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1566 int __user *, upeer_addrlen, int, flags)
1568 struct socket *sock, *newsock;
1569 struct file *newfile;
1570 int err, len, newfd, fput_needed;
1571 struct sockaddr_storage address;
1573 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1576 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1577 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1579 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1584 newsock = sock_alloc();
1588 newsock->type = sock->type;
1589 newsock->ops = sock->ops;
1592 * We don't need try_module_get here, as the listening socket (sock)
1593 * has the protocol module (sock->ops->owner) held.
1595 __module_get(newsock->ops->owner);
1597 newfd = get_unused_fd_flags(flags);
1598 if (unlikely(newfd < 0)) {
1600 sock_release(newsock);
1603 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1604 if (unlikely(IS_ERR(newfile))) {
1605 err = PTR_ERR(newfile);
1606 put_unused_fd(newfd);
1607 sock_release(newsock);
1611 err = security_socket_accept(sock, newsock);
1615 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1619 if (upeer_sockaddr) {
1620 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1622 err = -ECONNABORTED;
1625 err = move_addr_to_user(&address,
1626 len, upeer_sockaddr, upeer_addrlen);
1631 /* File flags are not inherited via accept() unlike another OSes. */
1633 fd_install(newfd, newfile);
1637 fput_light(sock->file, fput_needed);
1642 put_unused_fd(newfd);
1646 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1647 int __user *, upeer_addrlen)
1649 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1653 * Attempt to connect to a socket with the server address. The address
1654 * is in user space so we verify it is OK and move it to kernel space.
1656 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1659 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1660 * other SEQPACKET protocols that take time to connect() as it doesn't
1661 * include the -EINPROGRESS status for such sockets.
1664 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1667 struct socket *sock;
1668 struct sockaddr_storage address;
1669 int err, fput_needed;
1671 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1674 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1679 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1683 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1684 sock->file->f_flags);
1686 fput_light(sock->file, fput_needed);
1692 * Get the local address ('name') of a socket object. Move the obtained
1693 * name to user space.
1696 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1697 int __user *, usockaddr_len)
1699 struct socket *sock;
1700 struct sockaddr_storage address;
1701 int len, err, fput_needed;
1703 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707 err = security_socket_getsockname(sock);
1711 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1714 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1717 fput_light(sock->file, fput_needed);
1723 * Get the remote address ('name') of a socket object. Move the obtained
1724 * name to user space.
1727 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1728 int __user *, usockaddr_len)
1730 struct socket *sock;
1731 struct sockaddr_storage address;
1732 int len, err, fput_needed;
1734 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1736 err = security_socket_getpeername(sock);
1738 fput_light(sock->file, fput_needed);
1743 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1746 err = move_addr_to_user(&address, len, usockaddr,
1748 fput_light(sock->file, fput_needed);
1754 * Send a datagram to a given address. We move the address into kernel
1755 * space and check the user space data area is readable before invoking
1759 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1760 unsigned int, flags, struct sockaddr __user *, addr,
1763 struct socket *sock;
1764 struct sockaddr_storage address;
1772 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1776 iov.iov_base = buff;
1778 msg.msg_name = NULL;
1779 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
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;
1837 iov.iov_base = ubuf;
1838 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1839 /* Save some cycles and don't copy the address if not needed */
1840 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1841 /* We assume all kernel code knows the size of sockaddr_storage */
1842 msg.msg_namelen = 0;
1843 if (sock->file->f_flags & O_NONBLOCK)
1844 flags |= MSG_DONTWAIT;
1845 err = sock_recvmsg(sock, &msg, size, flags);
1847 if (err >= 0 && addr != NULL) {
1848 err2 = move_addr_to_user(&address,
1849 msg.msg_namelen, addr, addr_len);
1854 fput_light(sock->file, fput_needed);
1860 * Receive a datagram from a socket.
1863 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1864 unsigned int, flags)
1866 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1870 * Set a socket option. Because we don't know the option lengths we have
1871 * to pass the user mode parameter for the protocols to sort out.
1874 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1875 char __user *, optval, int, optlen)
1877 int err, fput_needed;
1878 struct socket *sock;
1883 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1885 err = security_socket_setsockopt(sock, level, optname);
1889 if (level == SOL_SOCKET)
1891 sock_setsockopt(sock, level, optname, optval,
1895 sock->ops->setsockopt(sock, level, optname, optval,
1898 fput_light(sock->file, fput_needed);
1904 * Get a socket option. Because we don't know the option lengths we have
1905 * to pass a user mode parameter for the protocols to sort out.
1908 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1909 char __user *, optval, int __user *, optlen)
1911 int err, fput_needed;
1912 struct socket *sock;
1914 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1916 err = security_socket_getsockopt(sock, level, optname);
1920 if (level == SOL_SOCKET)
1922 sock_getsockopt(sock, level, optname, optval,
1926 sock->ops->getsockopt(sock, level, optname, optval,
1929 fput_light(sock->file, fput_needed);
1935 * Shutdown a socket.
1938 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1940 int err, fput_needed;
1941 struct socket *sock;
1943 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1945 err = security_socket_shutdown(sock, how);
1947 err = sock->ops->shutdown(sock, how);
1948 fput_light(sock->file, fput_needed);
1953 /* A couple of helpful macros for getting the address of the 32/64 bit
1954 * fields which are the same type (int / unsigned) on our platforms.
1956 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1957 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1958 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1960 struct used_address {
1961 struct sockaddr_storage name;
1962 unsigned int name_len;
1965 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1966 struct user_msghdr __user *umsg,
1967 struct sockaddr __user **save_addr,
1970 struct sockaddr __user *uaddr;
1971 struct iovec __user *uiov;
1975 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1976 __get_user(uaddr, &umsg->msg_name) ||
1977 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1978 __get_user(uiov, &umsg->msg_iov) ||
1979 __get_user(nr_segs, &umsg->msg_iovlen) ||
1980 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1981 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1982 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1986 kmsg->msg_namelen = 0;
1988 if (kmsg->msg_namelen < 0)
1991 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1992 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1997 if (uaddr && kmsg->msg_namelen) {
1999 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
2005 kmsg->msg_name = NULL;
2006 kmsg->msg_namelen = 0;
2009 if (nr_segs > UIO_MAXIOV)
2012 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
2014 UIO_FASTIOV, *iov, iov);
2016 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
2017 *iov, nr_segs, err);
2021 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2022 struct msghdr *msg_sys, unsigned int flags,
2023 struct used_address *used_address)
2025 struct compat_msghdr __user *msg_compat =
2026 (struct compat_msghdr __user *)msg;
2027 struct sockaddr_storage address;
2028 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2029 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2030 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2031 /* 20 is size of ipv6_pktinfo */
2032 unsigned char *ctl_buf = ctl;
2033 int ctl_len, total_len;
2036 msg_sys->msg_name = &address;
2038 if (MSG_CMSG_COMPAT & flags)
2039 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2041 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2048 if (msg_sys->msg_controllen > INT_MAX)
2050 ctl_len = msg_sys->msg_controllen;
2051 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2053 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2057 ctl_buf = msg_sys->msg_control;
2058 ctl_len = msg_sys->msg_controllen;
2059 } else if (ctl_len) {
2060 if (ctl_len > sizeof(ctl)) {
2061 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2062 if (ctl_buf == NULL)
2067 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2068 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2069 * checking falls down on this.
2071 if (copy_from_user(ctl_buf,
2072 (void __user __force *)msg_sys->msg_control,
2075 msg_sys->msg_control = ctl_buf;
2077 msg_sys->msg_flags = flags;
2079 if (sock->file->f_flags & O_NONBLOCK)
2080 msg_sys->msg_flags |= MSG_DONTWAIT;
2082 * If this is sendmmsg() and current destination address is same as
2083 * previously succeeded address, omit asking LSM's decision.
2084 * used_address->name_len is initialized to UINT_MAX so that the first
2085 * destination address never matches.
2087 if (used_address && msg_sys->msg_name &&
2088 used_address->name_len == msg_sys->msg_namelen &&
2089 !memcmp(&used_address->name, msg_sys->msg_name,
2090 used_address->name_len)) {
2091 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2094 err = sock_sendmsg(sock, msg_sys, total_len);
2096 * If this is sendmmsg() and sending to current destination address was
2097 * successful, remember it.
2099 if (used_address && err >= 0) {
2100 used_address->name_len = msg_sys->msg_namelen;
2101 if (msg_sys->msg_name)
2102 memcpy(&used_address->name, msg_sys->msg_name,
2103 used_address->name_len);
2108 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2110 if (iov != iovstack)
2116 * BSD sendmsg interface
2119 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2121 int fput_needed, err;
2122 struct msghdr msg_sys;
2123 struct socket *sock;
2125 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2129 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2131 fput_light(sock->file, fput_needed);
2136 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2138 if (flags & MSG_CMSG_COMPAT)
2140 return __sys_sendmsg(fd, msg, flags);
2144 * Linux sendmmsg interface
2147 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2150 int fput_needed, err, datagrams;
2151 struct socket *sock;
2152 struct mmsghdr __user *entry;
2153 struct compat_mmsghdr __user *compat_entry;
2154 struct msghdr msg_sys;
2155 struct used_address used_address;
2157 if (vlen > UIO_MAXIOV)
2162 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2166 used_address.name_len = UINT_MAX;
2168 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2171 while (datagrams < vlen) {
2172 if (MSG_CMSG_COMPAT & flags) {
2173 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2174 &msg_sys, flags, &used_address);
2177 err = __put_user(err, &compat_entry->msg_len);
2180 err = ___sys_sendmsg(sock,
2181 (struct user_msghdr __user *)entry,
2182 &msg_sys, flags, &used_address);
2185 err = put_user(err, &entry->msg_len);
2194 fput_light(sock->file, fput_needed);
2196 /* We only return an error if no datagrams were able to be sent */
2203 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2204 unsigned int, vlen, unsigned int, flags)
2206 if (flags & MSG_CMSG_COMPAT)
2208 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2211 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2212 struct msghdr *msg_sys, unsigned int flags, int nosec)
2214 struct compat_msghdr __user *msg_compat =
2215 (struct compat_msghdr __user *)msg;
2216 struct iovec iovstack[UIO_FASTIOV];
2217 struct iovec *iov = iovstack;
2218 unsigned long cmsg_ptr;
2222 /* kernel mode address */
2223 struct sockaddr_storage addr;
2225 /* user mode address pointers */
2226 struct sockaddr __user *uaddr;
2227 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2229 msg_sys->msg_name = &addr;
2231 if (MSG_CMSG_COMPAT & flags)
2232 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2234 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2239 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2240 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2242 /* We assume all kernel code knows the size of sockaddr_storage */
2243 msg_sys->msg_namelen = 0;
2245 if (sock->file->f_flags & O_NONBLOCK)
2246 flags |= MSG_DONTWAIT;
2247 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2253 if (uaddr != NULL) {
2254 err = move_addr_to_user(&addr,
2255 msg_sys->msg_namelen, uaddr,
2260 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2264 if (MSG_CMSG_COMPAT & flags)
2265 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2266 &msg_compat->msg_controllen);
2268 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2269 &msg->msg_controllen);
2275 if (iov != iovstack)
2281 * BSD recvmsg interface
2284 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2286 int fput_needed, err;
2287 struct msghdr msg_sys;
2288 struct socket *sock;
2290 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2294 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2296 fput_light(sock->file, fput_needed);
2301 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2302 unsigned int, flags)
2304 if (flags & MSG_CMSG_COMPAT)
2306 return __sys_recvmsg(fd, msg, flags);
2310 * Linux recvmmsg interface
2313 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2314 unsigned int flags, struct timespec *timeout)
2316 int fput_needed, err, datagrams;
2317 struct socket *sock;
2318 struct mmsghdr __user *entry;
2319 struct compat_mmsghdr __user *compat_entry;
2320 struct msghdr msg_sys;
2321 struct timespec end_time;
2324 poll_select_set_timeout(&end_time, timeout->tv_sec,
2330 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2334 err = sock_error(sock->sk);
2339 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2341 while (datagrams < vlen) {
2343 * No need to ask LSM for more than the first datagram.
2345 if (MSG_CMSG_COMPAT & flags) {
2346 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2347 &msg_sys, flags & ~MSG_WAITFORONE,
2351 err = __put_user(err, &compat_entry->msg_len);
2354 err = ___sys_recvmsg(sock,
2355 (struct user_msghdr __user *)entry,
2356 &msg_sys, flags & ~MSG_WAITFORONE,
2360 err = put_user(err, &entry->msg_len);
2368 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2369 if (flags & MSG_WAITFORONE)
2370 flags |= MSG_DONTWAIT;
2373 ktime_get_ts(timeout);
2374 *timeout = timespec_sub(end_time, *timeout);
2375 if (timeout->tv_sec < 0) {
2376 timeout->tv_sec = timeout->tv_nsec = 0;
2380 /* Timeout, return less than vlen datagrams */
2381 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2385 /* Out of band data, return right away */
2386 if (msg_sys.msg_flags & MSG_OOB)
2391 fput_light(sock->file, fput_needed);
2396 if (datagrams != 0) {
2398 * We may return less entries than requested (vlen) if the
2399 * sock is non block and there aren't enough datagrams...
2401 if (err != -EAGAIN) {
2403 * ... or if recvmsg returns an error after we
2404 * received some datagrams, where we record the
2405 * error to return on the next call or if the
2406 * app asks about it using getsockopt(SO_ERROR).
2408 sock->sk->sk_err = -err;
2417 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2418 unsigned int, vlen, unsigned int, flags,
2419 struct timespec __user *, timeout)
2422 struct timespec timeout_sys;
2424 if (flags & MSG_CMSG_COMPAT)
2428 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2430 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2433 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2435 if (datagrams > 0 &&
2436 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2437 datagrams = -EFAULT;
2442 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2443 /* Argument list sizes for sys_socketcall */
2444 #define AL(x) ((x) * sizeof(unsigned long))
2445 static const unsigned char nargs[21] = {
2446 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2447 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2448 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2455 * System call vectors.
2457 * Argument checking cleaned up. Saved 20% in size.
2458 * This function doesn't need to set the kernel lock because
2459 * it is set by the callees.
2462 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2464 unsigned long a[AUDITSC_ARGS];
2465 unsigned long a0, a1;
2469 if (call < 1 || call > SYS_SENDMMSG)
2473 if (len > sizeof(a))
2476 /* copy_from_user should be SMP safe. */
2477 if (copy_from_user(a, args, len))
2480 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2489 err = sys_socket(a0, a1, a[2]);
2492 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2495 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2498 err = sys_listen(a0, a1);
2501 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2502 (int __user *)a[2], 0);
2504 case SYS_GETSOCKNAME:
2506 sys_getsockname(a0, (struct sockaddr __user *)a1,
2507 (int __user *)a[2]);
2509 case SYS_GETPEERNAME:
2511 sys_getpeername(a0, (struct sockaddr __user *)a1,
2512 (int __user *)a[2]);
2514 case SYS_SOCKETPAIR:
2515 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2518 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2521 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2522 (struct sockaddr __user *)a[4], a[5]);
2525 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2528 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2529 (struct sockaddr __user *)a[4],
2530 (int __user *)a[5]);
2533 err = sys_shutdown(a0, a1);
2535 case SYS_SETSOCKOPT:
2536 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2538 case SYS_GETSOCKOPT:
2540 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2541 (int __user *)a[4]);
2544 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2547 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2550 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2553 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2554 (struct timespec __user *)a[4]);
2557 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2558 (int __user *)a[2], a[3]);
2567 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2570 * sock_register - add a socket protocol handler
2571 * @ops: description of protocol
2573 * This function is called by a protocol handler that wants to
2574 * advertise its address family, and have it linked into the
2575 * socket interface. The value ops->family corresponds to the
2576 * socket system call protocol family.
2578 int sock_register(const struct net_proto_family *ops)
2582 if (ops->family >= NPROTO) {
2583 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2587 spin_lock(&net_family_lock);
2588 if (rcu_dereference_protected(net_families[ops->family],
2589 lockdep_is_held(&net_family_lock)))
2592 rcu_assign_pointer(net_families[ops->family], ops);
2595 spin_unlock(&net_family_lock);
2597 pr_info("NET: Registered protocol family %d\n", ops->family);
2600 EXPORT_SYMBOL(sock_register);
2603 * sock_unregister - remove a protocol handler
2604 * @family: protocol family to remove
2606 * This function is called by a protocol handler that wants to
2607 * remove its address family, and have it unlinked from the
2608 * new socket creation.
2610 * If protocol handler is a module, then it can use module reference
2611 * counts to protect against new references. If protocol handler is not
2612 * a module then it needs to provide its own protection in
2613 * the ops->create routine.
2615 void sock_unregister(int family)
2617 BUG_ON(family < 0 || family >= NPROTO);
2619 spin_lock(&net_family_lock);
2620 RCU_INIT_POINTER(net_families[family], NULL);
2621 spin_unlock(&net_family_lock);
2625 pr_info("NET: Unregistered protocol family %d\n", family);
2627 EXPORT_SYMBOL(sock_unregister);
2629 static int __init sock_init(void)
2633 * Initialize the network sysctl infrastructure.
2635 err = net_sysctl_init();
2640 * Initialize skbuff SLAB cache
2645 * Initialize the protocols module.
2650 err = register_filesystem(&sock_fs_type);
2653 sock_mnt = kern_mount(&sock_fs_type);
2654 if (IS_ERR(sock_mnt)) {
2655 err = PTR_ERR(sock_mnt);
2659 /* The real protocol initialization is performed in later initcalls.
2662 #ifdef CONFIG_NETFILTER
2663 err = netfilter_init();
2668 ptp_classifier_init();
2674 unregister_filesystem(&sock_fs_type);
2679 core_initcall(sock_init); /* early initcall */
2681 #ifdef CONFIG_PROC_FS
2682 void socket_seq_show(struct seq_file *seq)
2687 for_each_possible_cpu(cpu)
2688 counter += per_cpu(sockets_in_use, cpu);
2690 /* It can be negative, by the way. 8) */
2694 seq_printf(seq, "sockets: used %d\n", counter);
2696 #endif /* CONFIG_PROC_FS */
2698 #ifdef CONFIG_COMPAT
2699 static int do_siocgstamp(struct net *net, struct socket *sock,
2700 unsigned int cmd, void __user *up)
2702 mm_segment_t old_fs = get_fs();
2707 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2710 err = compat_put_timeval(&ktv, up);
2715 static int do_siocgstampns(struct net *net, struct socket *sock,
2716 unsigned int cmd, void __user *up)
2718 mm_segment_t old_fs = get_fs();
2719 struct timespec kts;
2723 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2726 err = compat_put_timespec(&kts, up);
2731 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2733 struct ifreq __user *uifr;
2736 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2737 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2740 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2744 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2750 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2752 struct compat_ifconf ifc32;
2754 struct ifconf __user *uifc;
2755 struct compat_ifreq __user *ifr32;
2756 struct ifreq __user *ifr;
2760 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2763 memset(&ifc, 0, sizeof(ifc));
2764 if (ifc32.ifcbuf == 0) {
2768 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2770 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2771 sizeof(struct ifreq);
2772 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2774 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2775 ifr32 = compat_ptr(ifc32.ifcbuf);
2776 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2777 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2783 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2786 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2790 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2794 ifr32 = compat_ptr(ifc32.ifcbuf);
2796 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2797 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2798 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2804 if (ifc32.ifcbuf == 0) {
2805 /* Translate from 64-bit structure multiple to
2809 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2814 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2820 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2822 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2823 bool convert_in = false, convert_out = false;
2824 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2825 struct ethtool_rxnfc __user *rxnfc;
2826 struct ifreq __user *ifr;
2827 u32 rule_cnt = 0, actual_rule_cnt;
2832 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2835 compat_rxnfc = compat_ptr(data);
2837 if (get_user(ethcmd, &compat_rxnfc->cmd))
2840 /* Most ethtool structures are defined without padding.
2841 * Unfortunately struct ethtool_rxnfc is an exception.
2846 case ETHTOOL_GRXCLSRLALL:
2847 /* Buffer size is variable */
2848 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2850 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2852 buf_size += rule_cnt * sizeof(u32);
2854 case ETHTOOL_GRXRINGS:
2855 case ETHTOOL_GRXCLSRLCNT:
2856 case ETHTOOL_GRXCLSRULE:
2857 case ETHTOOL_SRXCLSRLINS:
2860 case ETHTOOL_SRXCLSRLDEL:
2861 buf_size += sizeof(struct ethtool_rxnfc);
2866 ifr = compat_alloc_user_space(buf_size);
2867 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2869 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2872 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2873 &ifr->ifr_ifru.ifru_data))
2877 /* We expect there to be holes between fs.m_ext and
2878 * fs.ring_cookie and at the end of fs, but nowhere else.
2880 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2881 sizeof(compat_rxnfc->fs.m_ext) !=
2882 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2883 sizeof(rxnfc->fs.m_ext));
2885 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2886 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2887 offsetof(struct ethtool_rxnfc, fs.location) -
2888 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2890 if (copy_in_user(rxnfc, compat_rxnfc,
2891 (void __user *)(&rxnfc->fs.m_ext + 1) -
2892 (void __user *)rxnfc) ||
2893 copy_in_user(&rxnfc->fs.ring_cookie,
2894 &compat_rxnfc->fs.ring_cookie,
2895 (void __user *)(&rxnfc->fs.location + 1) -
2896 (void __user *)&rxnfc->fs.ring_cookie) ||
2897 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2898 sizeof(rxnfc->rule_cnt)))
2902 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2907 if (copy_in_user(compat_rxnfc, rxnfc,
2908 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2909 (const void __user *)rxnfc) ||
2910 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2911 &rxnfc->fs.ring_cookie,
2912 (const void __user *)(&rxnfc->fs.location + 1) -
2913 (const void __user *)&rxnfc->fs.ring_cookie) ||
2914 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2915 sizeof(rxnfc->rule_cnt)))
2918 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2919 /* As an optimisation, we only copy the actual
2920 * number of rules that the underlying
2921 * function returned. Since Mallory might
2922 * change the rule count in user memory, we
2923 * check that it is less than the rule count
2924 * originally given (as the user buffer size),
2925 * which has been range-checked.
2927 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2929 if (actual_rule_cnt < rule_cnt)
2930 rule_cnt = actual_rule_cnt;
2931 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2932 &rxnfc->rule_locs[0],
2933 rule_cnt * sizeof(u32)))
2941 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2944 compat_uptr_t uptr32;
2945 struct ifreq __user *uifr;
2947 uifr = compat_alloc_user_space(sizeof(*uifr));
2948 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2951 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2954 uptr = compat_ptr(uptr32);
2956 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2959 return dev_ioctl(net, SIOCWANDEV, uifr);
2962 static int bond_ioctl(struct net *net, unsigned int cmd,
2963 struct compat_ifreq __user *ifr32)
2966 mm_segment_t old_fs;
2970 case SIOCBONDENSLAVE:
2971 case SIOCBONDRELEASE:
2972 case SIOCBONDSETHWADDR:
2973 case SIOCBONDCHANGEACTIVE:
2974 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2979 err = dev_ioctl(net, cmd,
2980 (struct ifreq __user __force *) &kifr);
2985 return -ENOIOCTLCMD;
2989 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2990 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2991 struct compat_ifreq __user *u_ifreq32)
2993 struct ifreq __user *u_ifreq64;
2994 char tmp_buf[IFNAMSIZ];
2995 void __user *data64;
2998 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3001 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3003 data64 = compat_ptr(data32);
3005 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3007 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3010 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3013 return dev_ioctl(net, cmd, u_ifreq64);
3016 static int dev_ifsioc(struct net *net, struct socket *sock,
3017 unsigned int cmd, struct compat_ifreq __user *uifr32)
3019 struct ifreq __user *uifr;
3022 uifr = compat_alloc_user_space(sizeof(*uifr));
3023 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3026 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3037 case SIOCGIFBRDADDR:
3038 case SIOCGIFDSTADDR:
3039 case SIOCGIFNETMASK:
3044 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3052 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3053 struct compat_ifreq __user *uifr32)
3056 struct compat_ifmap __user *uifmap32;
3057 mm_segment_t old_fs;
3060 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3061 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3062 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3063 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3064 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3065 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3066 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3067 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3073 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3076 if (cmd == SIOCGIFMAP && !err) {
3077 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3078 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3079 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3080 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3081 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3082 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3083 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3092 struct sockaddr rt_dst; /* target address */
3093 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3094 struct sockaddr rt_genmask; /* target network mask (IP) */
3095 unsigned short rt_flags;
3098 unsigned char rt_tos;
3099 unsigned char rt_class;
3101 short rt_metric; /* +1 for binary compatibility! */
3102 /* char * */ u32 rt_dev; /* forcing the device at add */
3103 u32 rt_mtu; /* per route MTU/Window */
3104 u32 rt_window; /* Window clamping */
3105 unsigned short rt_irtt; /* Initial RTT */
3108 struct in6_rtmsg32 {
3109 struct in6_addr rtmsg_dst;
3110 struct in6_addr rtmsg_src;
3111 struct in6_addr rtmsg_gateway;
3121 static int routing_ioctl(struct net *net, struct socket *sock,
3122 unsigned int cmd, void __user *argp)
3126 struct in6_rtmsg r6;
3130 mm_segment_t old_fs = get_fs();
3132 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3133 struct in6_rtmsg32 __user *ur6 = argp;
3134 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3135 3 * sizeof(struct in6_addr));
3136 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3137 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3138 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3139 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3140 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3141 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3142 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3146 struct rtentry32 __user *ur4 = argp;
3147 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3148 3 * sizeof(struct sockaddr));
3149 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3150 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3151 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3152 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3153 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3154 ret |= get_user(rtdev, &(ur4->rt_dev));
3156 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3157 r4.rt_dev = (char __user __force *)devname;
3171 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3178 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3179 * for some operations; this forces use of the newer bridge-utils that
3180 * use compatible ioctls
3182 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3186 if (get_user(tmp, argp))
3188 if (tmp == BRCTL_GET_VERSION)
3189 return BRCTL_VERSION + 1;
3193 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3194 unsigned int cmd, unsigned long arg)
3196 void __user *argp = compat_ptr(arg);
3197 struct sock *sk = sock->sk;
3198 struct net *net = sock_net(sk);
3200 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3201 return compat_ifr_data_ioctl(net, cmd, argp);
3206 return old_bridge_ioctl(argp);
3208 return dev_ifname32(net, argp);
3210 return dev_ifconf(net, argp);
3212 return ethtool_ioctl(net, argp);
3214 return compat_siocwandev(net, argp);
3217 return compat_sioc_ifmap(net, cmd, argp);
3218 case SIOCBONDENSLAVE:
3219 case SIOCBONDRELEASE:
3220 case SIOCBONDSETHWADDR:
3221 case SIOCBONDCHANGEACTIVE:
3222 return bond_ioctl(net, cmd, argp);
3225 return routing_ioctl(net, sock, cmd, argp);
3227 return do_siocgstamp(net, sock, cmd, argp);
3229 return do_siocgstampns(net, sock, cmd, argp);
3230 case SIOCBONDSLAVEINFOQUERY:
3231 case SIOCBONDINFOQUERY:
3234 return compat_ifr_data_ioctl(net, cmd, argp);
3246 return sock_ioctl(file, cmd, arg);
3263 case SIOCSIFHWBROADCAST:
3265 case SIOCGIFBRDADDR:
3266 case SIOCSIFBRDADDR:
3267 case SIOCGIFDSTADDR:
3268 case SIOCSIFDSTADDR:
3269 case SIOCGIFNETMASK:
3270 case SIOCSIFNETMASK:
3281 return dev_ifsioc(net, sock, cmd, argp);
3287 return sock_do_ioctl(net, sock, cmd, arg);
3290 return -ENOIOCTLCMD;
3293 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3296 struct socket *sock = file->private_data;
3297 int ret = -ENOIOCTLCMD;
3304 if (sock->ops->compat_ioctl)
3305 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3307 if (ret == -ENOIOCTLCMD &&
3308 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3309 ret = compat_wext_handle_ioctl(net, cmd, arg);
3311 if (ret == -ENOIOCTLCMD)
3312 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3318 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3320 return sock->ops->bind(sock, addr, addrlen);
3322 EXPORT_SYMBOL(kernel_bind);
3324 int kernel_listen(struct socket *sock, int backlog)
3326 return sock->ops->listen(sock, backlog);
3328 EXPORT_SYMBOL(kernel_listen);
3330 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3332 struct sock *sk = sock->sk;
3335 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3340 err = sock->ops->accept(sock, *newsock, flags);
3342 sock_release(*newsock);
3347 (*newsock)->ops = sock->ops;
3348 __module_get((*newsock)->ops->owner);
3353 EXPORT_SYMBOL(kernel_accept);
3355 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3358 return sock->ops->connect(sock, addr, addrlen, flags);
3360 EXPORT_SYMBOL(kernel_connect);
3362 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3365 return sock->ops->getname(sock, addr, addrlen, 0);
3367 EXPORT_SYMBOL(kernel_getsockname);
3369 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3372 return sock->ops->getname(sock, addr, addrlen, 1);
3374 EXPORT_SYMBOL(kernel_getpeername);
3376 int kernel_getsockopt(struct socket *sock, int level, int optname,
3377 char *optval, int *optlen)
3379 mm_segment_t oldfs = get_fs();
3380 char __user *uoptval;
3381 int __user *uoptlen;
3384 uoptval = (char __user __force *) optval;
3385 uoptlen = (int __user __force *) optlen;
3388 if (level == SOL_SOCKET)
3389 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3391 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3396 EXPORT_SYMBOL(kernel_getsockopt);
3398 int kernel_setsockopt(struct socket *sock, int level, int optname,
3399 char *optval, unsigned int optlen)
3401 mm_segment_t oldfs = get_fs();
3402 char __user *uoptval;
3405 uoptval = (char __user __force *) optval;
3408 if (level == SOL_SOCKET)
3409 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3411 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3416 EXPORT_SYMBOL(kernel_setsockopt);
3418 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3419 size_t size, int flags)
3421 if (sock->ops->sendpage)
3422 return sock->ops->sendpage(sock, page, offset, size, flags);
3424 return sock_no_sendpage(sock, page, offset, size, flags);
3426 EXPORT_SYMBOL(kernel_sendpage);
3428 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3430 mm_segment_t oldfs = get_fs();
3434 err = sock->ops->ioctl(sock, cmd, arg);
3439 EXPORT_SYMBOL(kernel_sock_ioctl);
3441 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3443 return sock->ops->shutdown(sock, how);
3445 EXPORT_SYMBOL(kernel_sock_shutdown);