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));
376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 if (unlikely(IS_ERR(file))) {
379 /* drop dentry, keep inode */
380 ihold(path.dentry->d_inode);
386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387 file->private_data = sock;
390 EXPORT_SYMBOL(sock_alloc_file);
392 static int sock_map_fd(struct socket *sock, int flags)
394 struct file *newfile;
395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
399 newfile = sock_alloc_file(sock, flags, NULL);
400 if (likely(!IS_ERR(newfile))) {
401 fd_install(fd, newfile);
406 return PTR_ERR(newfile);
409 struct socket *sock_from_file(struct file *file, int *err)
411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
417 EXPORT_SYMBOL(sock_from_file);
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket *sockfd_lookup(int fd, int *err)
443 sock = sock_from_file(file, err);
448 EXPORT_SYMBOL(sockfd_lookup);
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452 struct fd f = fdget(fd);
457 sock = sock_from_file(f.file, err);
459 *fput_needed = f.flags;
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
473 const char *proto_name;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
484 if (proto_size + 1 > size)
487 strncpy(value, proto_name, proto_size + 1);
489 error = proto_size + 1;
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
524 static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
530 * sock_alloc - allocate a socket
532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
537 static struct socket *sock_alloc(void)
542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
546 sock = SOCKET_I(inode);
548 kmemcheck_annotate_bitfield(sock, type);
549 inode->i_ino = get_next_ino();
550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
553 inode->i_op = &sockfs_inode_ops;
555 this_cpu_add(sockets_in_use, 1);
560 * sock_release - close a socket
561 * @sock: socket to close
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
568 void sock_release(struct socket *sock)
571 struct module *owner = sock->ops->owner;
573 sock->ops->release(sock);
578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579 pr_err("%s: fasync list not empty!\n", __func__);
581 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
584 this_cpu_sub(sockets_in_use, 1);
586 iput(SOCK_INODE(sock));
591 EXPORT_SYMBOL(sock_release);
593 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
595 u8 flags = *tx_flags;
597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
598 flags |= SKBTX_HW_TSTAMP;
600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
601 flags |= SKBTX_SW_TSTAMP;
603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
604 flags |= SKBTX_SCHED_TSTAMP;
606 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
607 flags |= SKBTX_ACK_TSTAMP;
611 EXPORT_SYMBOL(__sock_tx_timestamp);
613 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
614 struct msghdr *msg, size_t size)
616 return sock->ops->sendmsg(iocb, sock, msg, size);
619 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
620 struct msghdr *msg, size_t size)
622 int err = security_socket_sendmsg(sock, msg, size);
624 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
627 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
628 size_t size, bool nosec)
633 init_sync_kiocb(&iocb, NULL);
634 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
635 __sock_sendmsg(&iocb, sock, msg, size);
636 if (-EIOCBQUEUED == ret)
637 ret = wait_on_sync_kiocb(&iocb);
641 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
643 return do_sock_sendmsg(sock, msg, size, false);
645 EXPORT_SYMBOL(sock_sendmsg);
647 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
649 return do_sock_sendmsg(sock, msg, size, true);
652 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
653 struct kvec *vec, size_t num, size_t size)
655 mm_segment_t oldfs = get_fs();
660 * the following is safe, since for compiler definitions of kvec and
661 * iovec are identical, yielding the same in-core layout and alignment
663 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
664 result = sock_sendmsg(sock, msg, size);
668 EXPORT_SYMBOL(kernel_sendmsg);
671 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
673 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
676 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
677 struct scm_timestamping tss;
679 struct skb_shared_hwtstamps *shhwtstamps =
682 /* Race occurred between timestamp enabling and packet
683 receiving. Fill in the current time for now. */
684 if (need_software_tstamp && skb->tstamp.tv64 == 0)
685 __net_timestamp(skb);
687 if (need_software_tstamp) {
688 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
690 skb_get_timestamp(skb, &tv);
691 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
695 skb_get_timestampns(skb, &ts);
696 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
701 memset(&tss, 0, sizeof(tss));
702 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
703 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
706 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
707 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
710 put_cmsg(msg, SOL_SOCKET,
711 SCM_TIMESTAMPING, sizeof(tss), &tss);
713 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
715 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
720 if (!sock_flag(sk, SOCK_WIFI_STATUS))
722 if (!skb->wifi_acked_valid)
725 ack = skb->wifi_acked;
727 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
729 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
731 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
734 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
735 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
736 sizeof(__u32), &skb->dropcount);
739 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
742 sock_recv_timestamp(msg, sk, skb);
743 sock_recv_drops(msg, sk, skb);
745 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
747 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
748 struct msghdr *msg, size_t size, int flags)
750 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
753 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
754 struct msghdr *msg, size_t size, int flags)
756 int err = security_socket_recvmsg(sock, msg, size, flags);
758 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
761 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
762 size_t size, int flags)
767 init_sync_kiocb(&iocb, NULL);
768 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
769 if (-EIOCBQUEUED == ret)
770 ret = wait_on_sync_kiocb(&iocb);
773 EXPORT_SYMBOL(sock_recvmsg);
775 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
776 size_t size, int flags)
781 init_sync_kiocb(&iocb, NULL);
782 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
783 if (-EIOCBQUEUED == ret)
784 ret = wait_on_sync_kiocb(&iocb);
789 * kernel_recvmsg - Receive a message from a socket (kernel space)
790 * @sock: The socket to receive the message from
791 * @msg: Received message
792 * @vec: Input s/g array for message data
793 * @num: Size of input s/g array
794 * @size: Number of bytes to read
795 * @flags: Message flags (MSG_DONTWAIT, etc...)
797 * On return the msg structure contains the scatter/gather array passed in the
798 * vec argument. The array is modified so that it consists of the unfilled
799 * portion of the original array.
801 * The returned value is the total number of bytes received, or an error.
803 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
804 struct kvec *vec, size_t num, size_t size, int flags)
806 mm_segment_t oldfs = get_fs();
811 * the following is safe, since for compiler definitions of kvec and
812 * iovec are identical, yielding the same in-core layout and alignment
814 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
815 result = sock_recvmsg(sock, msg, size, flags);
819 EXPORT_SYMBOL(kernel_recvmsg);
821 static ssize_t sock_sendpage(struct file *file, struct page *page,
822 int offset, size_t size, loff_t *ppos, int more)
827 sock = file->private_data;
829 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
830 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
833 return kernel_sendpage(sock, page, offset, size, flags);
836 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
837 struct pipe_inode_info *pipe, size_t len,
840 struct socket *sock = file->private_data;
842 if (unlikely(!sock->ops->splice_read))
845 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
848 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
849 struct file *file, const struct iovec *iov,
850 unsigned long nr_segs)
852 struct socket *sock = file->private_data;
854 msg->msg_name = NULL;
855 msg->msg_namelen = 0;
856 msg->msg_control = NULL;
857 msg->msg_controllen = 0;
858 iov_iter_init(&msg->msg_iter, READ, iov, nr_segs, iocb->ki_nbytes);
859 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
861 return __sock_recvmsg(iocb, sock, msg, iocb->ki_nbytes, msg->msg_flags);
864 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
865 unsigned long nr_segs, loff_t pos)
872 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
875 return do_sock_read(&msg, iocb, iocb->ki_filp, iov, nr_segs);
878 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
879 struct file *file, const struct iovec *iov,
880 unsigned long nr_segs)
882 struct socket *sock = file->private_data;
884 msg->msg_name = NULL;
885 msg->msg_namelen = 0;
886 msg->msg_control = NULL;
887 msg->msg_controllen = 0;
888 iov_iter_init(&msg->msg_iter, WRITE, iov, nr_segs, iocb->ki_nbytes);
889 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
890 if (sock->type == SOCK_SEQPACKET)
891 msg->msg_flags |= MSG_EOR;
893 return __sock_sendmsg(iocb, sock, msg, iocb->ki_nbytes);
896 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
897 unsigned long nr_segs, loff_t pos)
904 return do_sock_write(&msg, iocb, iocb->ki_filp, iov, nr_segs);
908 * Atomic setting of ioctl hooks to avoid race
909 * with module unload.
912 static DEFINE_MUTEX(br_ioctl_mutex);
913 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
915 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
917 mutex_lock(&br_ioctl_mutex);
918 br_ioctl_hook = hook;
919 mutex_unlock(&br_ioctl_mutex);
921 EXPORT_SYMBOL(brioctl_set);
923 static DEFINE_MUTEX(vlan_ioctl_mutex);
924 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
926 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
928 mutex_lock(&vlan_ioctl_mutex);
929 vlan_ioctl_hook = hook;
930 mutex_unlock(&vlan_ioctl_mutex);
932 EXPORT_SYMBOL(vlan_ioctl_set);
934 static DEFINE_MUTEX(dlci_ioctl_mutex);
935 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
937 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
939 mutex_lock(&dlci_ioctl_mutex);
940 dlci_ioctl_hook = hook;
941 mutex_unlock(&dlci_ioctl_mutex);
943 EXPORT_SYMBOL(dlci_ioctl_set);
945 static long sock_do_ioctl(struct net *net, struct socket *sock,
946 unsigned int cmd, unsigned long arg)
949 void __user *argp = (void __user *)arg;
951 err = sock->ops->ioctl(sock, cmd, arg);
954 * If this ioctl is unknown try to hand it down
957 if (err == -ENOIOCTLCMD)
958 err = dev_ioctl(net, cmd, argp);
964 * With an ioctl, arg may well be a user mode pointer, but we don't know
965 * what to do with it - that's up to the protocol still.
968 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
972 void __user *argp = (void __user *)arg;
976 sock = file->private_data;
979 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
980 err = dev_ioctl(net, cmd, argp);
982 #ifdef CONFIG_WEXT_CORE
983 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
984 err = dev_ioctl(net, cmd, argp);
991 if (get_user(pid, (int __user *)argp))
993 f_setown(sock->file, pid, 1);
998 err = put_user(f_getown(sock->file),
1007 request_module("bridge");
1009 mutex_lock(&br_ioctl_mutex);
1011 err = br_ioctl_hook(net, cmd, argp);
1012 mutex_unlock(&br_ioctl_mutex);
1017 if (!vlan_ioctl_hook)
1018 request_module("8021q");
1020 mutex_lock(&vlan_ioctl_mutex);
1021 if (vlan_ioctl_hook)
1022 err = vlan_ioctl_hook(net, argp);
1023 mutex_unlock(&vlan_ioctl_mutex);
1028 if (!dlci_ioctl_hook)
1029 request_module("dlci");
1031 mutex_lock(&dlci_ioctl_mutex);
1032 if (dlci_ioctl_hook)
1033 err = dlci_ioctl_hook(cmd, argp);
1034 mutex_unlock(&dlci_ioctl_mutex);
1037 err = sock_do_ioctl(net, sock, cmd, arg);
1043 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1046 struct socket *sock = NULL;
1048 err = security_socket_create(family, type, protocol, 1);
1052 sock = sock_alloc();
1059 err = security_socket_post_create(sock, family, type, protocol, 1);
1071 EXPORT_SYMBOL(sock_create_lite);
1073 /* No kernel lock held - perfect */
1074 static unsigned int sock_poll(struct file *file, poll_table *wait)
1076 unsigned int busy_flag = 0;
1077 struct socket *sock;
1080 * We can't return errors to poll, so it's either yes or no.
1082 sock = file->private_data;
1084 if (sk_can_busy_loop(sock->sk)) {
1085 /* this socket can poll_ll so tell the system call */
1086 busy_flag = POLL_BUSY_LOOP;
1088 /* once, only if requested by syscall */
1089 if (wait && (wait->_key & POLL_BUSY_LOOP))
1090 sk_busy_loop(sock->sk, 1);
1093 return busy_flag | sock->ops->poll(file, sock, wait);
1096 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1098 struct socket *sock = file->private_data;
1100 return sock->ops->mmap(file, sock, vma);
1103 static int sock_close(struct inode *inode, struct file *filp)
1105 sock_release(SOCKET_I(inode));
1110 * Update the socket async list
1112 * Fasync_list locking strategy.
1114 * 1. fasync_list is modified only under process context socket lock
1115 * i.e. under semaphore.
1116 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1117 * or under socket lock
1120 static int sock_fasync(int fd, struct file *filp, int on)
1122 struct socket *sock = filp->private_data;
1123 struct sock *sk = sock->sk;
1124 struct socket_wq *wq;
1130 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1131 fasync_helper(fd, filp, on, &wq->fasync_list);
1133 if (!wq->fasync_list)
1134 sock_reset_flag(sk, SOCK_FASYNC);
1136 sock_set_flag(sk, SOCK_FASYNC);
1142 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1144 int sock_wake_async(struct socket *sock, int how, int band)
1146 struct socket_wq *wq;
1151 wq = rcu_dereference(sock->wq);
1152 if (!wq || !wq->fasync_list) {
1157 case SOCK_WAKE_WAITD:
1158 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1161 case SOCK_WAKE_SPACE:
1162 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1167 kill_fasync(&wq->fasync_list, SIGIO, band);
1170 kill_fasync(&wq->fasync_list, SIGURG, band);
1175 EXPORT_SYMBOL(sock_wake_async);
1177 int __sock_create(struct net *net, int family, int type, int protocol,
1178 struct socket **res, int kern)
1181 struct socket *sock;
1182 const struct net_proto_family *pf;
1185 * Check protocol is in range
1187 if (family < 0 || family >= NPROTO)
1188 return -EAFNOSUPPORT;
1189 if (type < 0 || type >= SOCK_MAX)
1194 This uglymoron is moved from INET layer to here to avoid
1195 deadlock in module load.
1197 if (family == PF_INET && type == SOCK_PACKET) {
1201 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1207 err = security_socket_create(family, type, protocol, kern);
1212 * Allocate the socket and allow the family to set things up. if
1213 * the protocol is 0, the family is instructed to select an appropriate
1216 sock = sock_alloc();
1218 net_warn_ratelimited("socket: no more sockets\n");
1219 return -ENFILE; /* Not exactly a match, but its the
1220 closest posix thing */
1225 #ifdef CONFIG_MODULES
1226 /* Attempt to load a protocol module if the find failed.
1228 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1229 * requested real, full-featured networking support upon configuration.
1230 * Otherwise module support will break!
1232 if (rcu_access_pointer(net_families[family]) == NULL)
1233 request_module("net-pf-%d", family);
1237 pf = rcu_dereference(net_families[family]);
1238 err = -EAFNOSUPPORT;
1243 * We will call the ->create function, that possibly is in a loadable
1244 * module, so we have to bump that loadable module refcnt first.
1246 if (!try_module_get(pf->owner))
1249 /* Now protected by module ref count */
1252 err = pf->create(net, sock, protocol, kern);
1254 goto out_module_put;
1257 * Now to bump the refcnt of the [loadable] module that owns this
1258 * socket at sock_release time we decrement its refcnt.
1260 if (!try_module_get(sock->ops->owner))
1261 goto out_module_busy;
1264 * Now that we're done with the ->create function, the [loadable]
1265 * module can have its refcnt decremented
1267 module_put(pf->owner);
1268 err = security_socket_post_create(sock, family, type, protocol, kern);
1270 goto out_sock_release;
1276 err = -EAFNOSUPPORT;
1279 module_put(pf->owner);
1286 goto out_sock_release;
1288 EXPORT_SYMBOL(__sock_create);
1290 int sock_create(int family, int type, int protocol, struct socket **res)
1292 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1294 EXPORT_SYMBOL(sock_create);
1296 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1298 return __sock_create(&init_net, family, type, protocol, res, 1);
1300 EXPORT_SYMBOL(sock_create_kern);
1302 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1305 struct socket *sock;
1308 /* Check the SOCK_* constants for consistency. */
1309 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1310 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1311 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1312 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1314 flags = type & ~SOCK_TYPE_MASK;
1315 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1317 type &= SOCK_TYPE_MASK;
1319 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1320 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1322 retval = sock_create(family, type, protocol, &sock);
1326 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1331 /* It may be already another descriptor 8) Not kernel problem. */
1340 * Create a pair of connected sockets.
1343 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1344 int __user *, usockvec)
1346 struct socket *sock1, *sock2;
1348 struct file *newfile1, *newfile2;
1351 flags = type & ~SOCK_TYPE_MASK;
1352 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1354 type &= SOCK_TYPE_MASK;
1356 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1357 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1360 * Obtain the first socket and check if the underlying protocol
1361 * supports the socketpair call.
1364 err = sock_create(family, type, protocol, &sock1);
1368 err = sock_create(family, type, protocol, &sock2);
1372 err = sock1->ops->socketpair(sock1, sock2);
1374 goto out_release_both;
1376 fd1 = get_unused_fd_flags(flags);
1377 if (unlikely(fd1 < 0)) {
1379 goto out_release_both;
1382 fd2 = get_unused_fd_flags(flags);
1383 if (unlikely(fd2 < 0)) {
1385 goto out_put_unused_1;
1388 newfile1 = sock_alloc_file(sock1, flags, NULL);
1389 if (unlikely(IS_ERR(newfile1))) {
1390 err = PTR_ERR(newfile1);
1391 goto out_put_unused_both;
1394 newfile2 = sock_alloc_file(sock2, flags, NULL);
1395 if (IS_ERR(newfile2)) {
1396 err = PTR_ERR(newfile2);
1400 err = put_user(fd1, &usockvec[0]);
1404 err = put_user(fd2, &usockvec[1]);
1408 audit_fd_pair(fd1, fd2);
1410 fd_install(fd1, newfile1);
1411 fd_install(fd2, newfile2);
1412 /* fd1 and fd2 may be already another descriptors.
1413 * Not kernel problem.
1429 sock_release(sock2);
1432 out_put_unused_both:
1437 sock_release(sock2);
1439 sock_release(sock1);
1445 * Bind a name to a socket. Nothing much to do here since it's
1446 * the protocol's responsibility to handle the local address.
1448 * We move the socket address to kernel space before we call
1449 * the protocol layer (having also checked the address is ok).
1452 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1454 struct socket *sock;
1455 struct sockaddr_storage address;
1456 int err, fput_needed;
1458 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1460 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1462 err = security_socket_bind(sock,
1463 (struct sockaddr *)&address,
1466 err = sock->ops->bind(sock,
1470 fput_light(sock->file, fput_needed);
1476 * Perform a listen. Basically, we allow the protocol to do anything
1477 * necessary for a listen, and if that works, we mark the socket as
1478 * ready for listening.
1481 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1483 struct socket *sock;
1484 int err, fput_needed;
1487 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1489 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1490 if ((unsigned int)backlog > somaxconn)
1491 backlog = somaxconn;
1493 err = security_socket_listen(sock, backlog);
1495 err = sock->ops->listen(sock, backlog);
1497 fput_light(sock->file, fput_needed);
1503 * For accept, we attempt to create a new socket, set up the link
1504 * with the client, wake up the client, then return the new
1505 * connected fd. We collect the address of the connector in kernel
1506 * space and move it to user at the very end. This is unclean because
1507 * we open the socket then return an error.
1509 * 1003.1g adds the ability to recvmsg() to query connection pending
1510 * status to recvmsg. We need to add that support in a way thats
1511 * clean when we restucture accept also.
1514 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1515 int __user *, upeer_addrlen, int, flags)
1517 struct socket *sock, *newsock;
1518 struct file *newfile;
1519 int err, len, newfd, fput_needed;
1520 struct sockaddr_storage address;
1522 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1525 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1526 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1528 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1533 newsock = sock_alloc();
1537 newsock->type = sock->type;
1538 newsock->ops = sock->ops;
1541 * We don't need try_module_get here, as the listening socket (sock)
1542 * has the protocol module (sock->ops->owner) held.
1544 __module_get(newsock->ops->owner);
1546 newfd = get_unused_fd_flags(flags);
1547 if (unlikely(newfd < 0)) {
1549 sock_release(newsock);
1552 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1553 if (unlikely(IS_ERR(newfile))) {
1554 err = PTR_ERR(newfile);
1555 put_unused_fd(newfd);
1556 sock_release(newsock);
1560 err = security_socket_accept(sock, newsock);
1564 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1568 if (upeer_sockaddr) {
1569 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1571 err = -ECONNABORTED;
1574 err = move_addr_to_user(&address,
1575 len, upeer_sockaddr, upeer_addrlen);
1580 /* File flags are not inherited via accept() unlike another OSes. */
1582 fd_install(newfd, newfile);
1586 fput_light(sock->file, fput_needed);
1591 put_unused_fd(newfd);
1595 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1596 int __user *, upeer_addrlen)
1598 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1602 * Attempt to connect to a socket with the server address. The address
1603 * is in user space so we verify it is OK and move it to kernel space.
1605 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1608 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1609 * other SEQPACKET protocols that take time to connect() as it doesn't
1610 * include the -EINPROGRESS status for such sockets.
1613 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1616 struct socket *sock;
1617 struct sockaddr_storage address;
1618 int err, fput_needed;
1620 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1623 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1628 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1632 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1633 sock->file->f_flags);
1635 fput_light(sock->file, fput_needed);
1641 * Get the local address ('name') of a socket object. Move the obtained
1642 * name to user space.
1645 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1646 int __user *, usockaddr_len)
1648 struct socket *sock;
1649 struct sockaddr_storage address;
1650 int len, err, fput_needed;
1652 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1656 err = security_socket_getsockname(sock);
1660 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1663 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1666 fput_light(sock->file, fput_needed);
1672 * Get the remote address ('name') of a socket object. Move the obtained
1673 * name to user space.
1676 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1677 int __user *, usockaddr_len)
1679 struct socket *sock;
1680 struct sockaddr_storage address;
1681 int len, err, fput_needed;
1683 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1685 err = security_socket_getpeername(sock);
1687 fput_light(sock->file, fput_needed);
1692 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1695 err = move_addr_to_user(&address, len, usockaddr,
1697 fput_light(sock->file, fput_needed);
1703 * Send a datagram to a given address. We move the address into kernel
1704 * space and check the user space data area is readable before invoking
1708 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1709 unsigned int, flags, struct sockaddr __user *, addr,
1712 struct socket *sock;
1713 struct sockaddr_storage address;
1721 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1725 iov.iov_base = buff;
1727 msg.msg_name = NULL;
1728 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1729 msg.msg_control = NULL;
1730 msg.msg_controllen = 0;
1731 msg.msg_namelen = 0;
1733 err = move_addr_to_kernel(addr, addr_len, &address);
1736 msg.msg_name = (struct sockaddr *)&address;
1737 msg.msg_namelen = addr_len;
1739 if (sock->file->f_flags & O_NONBLOCK)
1740 flags |= MSG_DONTWAIT;
1741 msg.msg_flags = flags;
1742 err = sock_sendmsg(sock, &msg, len);
1745 fput_light(sock->file, fput_needed);
1751 * Send a datagram down a socket.
1754 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1755 unsigned int, flags)
1757 return sys_sendto(fd, buff, len, flags, NULL, 0);
1761 * Receive a frame from the socket and optionally record the address of the
1762 * sender. We verify the buffers are writable and if needed move the
1763 * sender address from kernel to user space.
1766 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1767 unsigned int, flags, struct sockaddr __user *, addr,
1768 int __user *, addr_len)
1770 struct socket *sock;
1773 struct sockaddr_storage address;
1779 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1783 msg.msg_control = NULL;
1784 msg.msg_controllen = 0;
1786 iov.iov_base = ubuf;
1787 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1788 /* Save some cycles and don't copy the address if not needed */
1789 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1790 /* We assume all kernel code knows the size of sockaddr_storage */
1791 msg.msg_namelen = 0;
1792 if (sock->file->f_flags & O_NONBLOCK)
1793 flags |= MSG_DONTWAIT;
1794 err = sock_recvmsg(sock, &msg, size, flags);
1796 if (err >= 0 && addr != NULL) {
1797 err2 = move_addr_to_user(&address,
1798 msg.msg_namelen, addr, addr_len);
1803 fput_light(sock->file, fput_needed);
1809 * Receive a datagram from a socket.
1812 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1813 unsigned int, flags)
1815 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1819 * Set a socket option. Because we don't know the option lengths we have
1820 * to pass the user mode parameter for the protocols to sort out.
1823 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1824 char __user *, optval, int, optlen)
1826 int err, fput_needed;
1827 struct socket *sock;
1832 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1834 err = security_socket_setsockopt(sock, level, optname);
1838 if (level == SOL_SOCKET)
1840 sock_setsockopt(sock, level, optname, optval,
1844 sock->ops->setsockopt(sock, level, optname, optval,
1847 fput_light(sock->file, fput_needed);
1853 * Get a socket option. Because we don't know the option lengths we have
1854 * to pass a user mode parameter for the protocols to sort out.
1857 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1858 char __user *, optval, int __user *, optlen)
1860 int err, fput_needed;
1861 struct socket *sock;
1863 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1865 err = security_socket_getsockopt(sock, level, optname);
1869 if (level == SOL_SOCKET)
1871 sock_getsockopt(sock, level, optname, optval,
1875 sock->ops->getsockopt(sock, level, optname, optval,
1878 fput_light(sock->file, fput_needed);
1884 * Shutdown a socket.
1887 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1889 int err, fput_needed;
1890 struct socket *sock;
1892 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1894 err = security_socket_shutdown(sock, how);
1896 err = sock->ops->shutdown(sock, how);
1897 fput_light(sock->file, fput_needed);
1902 /* A couple of helpful macros for getting the address of the 32/64 bit
1903 * fields which are the same type (int / unsigned) on our platforms.
1905 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1906 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1907 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1909 struct used_address {
1910 struct sockaddr_storage name;
1911 unsigned int name_len;
1914 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1915 struct user_msghdr __user *umsg,
1916 struct sockaddr __user **save_addr,
1919 struct sockaddr __user *uaddr;
1920 struct iovec __user *uiov;
1924 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1925 __get_user(uaddr, &umsg->msg_name) ||
1926 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1927 __get_user(uiov, &umsg->msg_iov) ||
1928 __get_user(nr_segs, &umsg->msg_iovlen) ||
1929 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1930 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1931 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1935 kmsg->msg_namelen = 0;
1937 if (kmsg->msg_namelen < 0)
1940 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1941 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1946 if (uaddr && kmsg->msg_namelen) {
1948 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1954 kmsg->msg_name = NULL;
1955 kmsg->msg_namelen = 0;
1958 if (nr_segs > UIO_MAXIOV)
1961 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
1963 UIO_FASTIOV, *iov, iov);
1965 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1966 *iov, nr_segs, err);
1970 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1971 struct msghdr *msg_sys, unsigned int flags,
1972 struct used_address *used_address)
1974 struct compat_msghdr __user *msg_compat =
1975 (struct compat_msghdr __user *)msg;
1976 struct sockaddr_storage address;
1977 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1978 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1979 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1980 /* 20 is size of ipv6_pktinfo */
1981 unsigned char *ctl_buf = ctl;
1982 int ctl_len, total_len;
1985 msg_sys->msg_name = &address;
1987 if (MSG_CMSG_COMPAT & flags)
1988 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1990 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1997 if (msg_sys->msg_controllen > INT_MAX)
1999 ctl_len = msg_sys->msg_controllen;
2000 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2002 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2006 ctl_buf = msg_sys->msg_control;
2007 ctl_len = msg_sys->msg_controllen;
2008 } else if (ctl_len) {
2009 if (ctl_len > sizeof(ctl)) {
2010 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2011 if (ctl_buf == NULL)
2016 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2017 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2018 * checking falls down on this.
2020 if (copy_from_user(ctl_buf,
2021 (void __user __force *)msg_sys->msg_control,
2024 msg_sys->msg_control = ctl_buf;
2026 msg_sys->msg_flags = flags;
2028 if (sock->file->f_flags & O_NONBLOCK)
2029 msg_sys->msg_flags |= MSG_DONTWAIT;
2031 * If this is sendmmsg() and current destination address is same as
2032 * previously succeeded address, omit asking LSM's decision.
2033 * used_address->name_len is initialized to UINT_MAX so that the first
2034 * destination address never matches.
2036 if (used_address && msg_sys->msg_name &&
2037 used_address->name_len == msg_sys->msg_namelen &&
2038 !memcmp(&used_address->name, msg_sys->msg_name,
2039 used_address->name_len)) {
2040 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2043 err = sock_sendmsg(sock, msg_sys, total_len);
2045 * If this is sendmmsg() and sending to current destination address was
2046 * successful, remember it.
2048 if (used_address && err >= 0) {
2049 used_address->name_len = msg_sys->msg_namelen;
2050 if (msg_sys->msg_name)
2051 memcpy(&used_address->name, msg_sys->msg_name,
2052 used_address->name_len);
2057 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2059 if (iov != iovstack)
2065 * BSD sendmsg interface
2068 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2070 int fput_needed, err;
2071 struct msghdr msg_sys;
2072 struct socket *sock;
2074 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2078 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2080 fput_light(sock->file, fput_needed);
2085 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2087 if (flags & MSG_CMSG_COMPAT)
2089 return __sys_sendmsg(fd, msg, flags);
2093 * Linux sendmmsg interface
2096 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2099 int fput_needed, err, datagrams;
2100 struct socket *sock;
2101 struct mmsghdr __user *entry;
2102 struct compat_mmsghdr __user *compat_entry;
2103 struct msghdr msg_sys;
2104 struct used_address used_address;
2106 if (vlen > UIO_MAXIOV)
2111 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2115 used_address.name_len = UINT_MAX;
2117 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2120 while (datagrams < vlen) {
2121 if (MSG_CMSG_COMPAT & flags) {
2122 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2123 &msg_sys, flags, &used_address);
2126 err = __put_user(err, &compat_entry->msg_len);
2129 err = ___sys_sendmsg(sock,
2130 (struct user_msghdr __user *)entry,
2131 &msg_sys, flags, &used_address);
2134 err = put_user(err, &entry->msg_len);
2143 fput_light(sock->file, fput_needed);
2145 /* We only return an error if no datagrams were able to be sent */
2152 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2153 unsigned int, vlen, unsigned int, flags)
2155 if (flags & MSG_CMSG_COMPAT)
2157 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2160 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2161 struct msghdr *msg_sys, unsigned int flags, int nosec)
2163 struct compat_msghdr __user *msg_compat =
2164 (struct compat_msghdr __user *)msg;
2165 struct iovec iovstack[UIO_FASTIOV];
2166 struct iovec *iov = iovstack;
2167 unsigned long cmsg_ptr;
2171 /* kernel mode address */
2172 struct sockaddr_storage addr;
2174 /* user mode address pointers */
2175 struct sockaddr __user *uaddr;
2176 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2178 msg_sys->msg_name = &addr;
2180 if (MSG_CMSG_COMPAT & flags)
2181 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2183 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2188 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2189 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2191 /* We assume all kernel code knows the size of sockaddr_storage */
2192 msg_sys->msg_namelen = 0;
2194 if (sock->file->f_flags & O_NONBLOCK)
2195 flags |= MSG_DONTWAIT;
2196 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2202 if (uaddr != NULL) {
2203 err = move_addr_to_user(&addr,
2204 msg_sys->msg_namelen, uaddr,
2209 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2213 if (MSG_CMSG_COMPAT & flags)
2214 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2215 &msg_compat->msg_controllen);
2217 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2218 &msg->msg_controllen);
2224 if (iov != iovstack)
2230 * BSD recvmsg interface
2233 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2235 int fput_needed, err;
2236 struct msghdr msg_sys;
2237 struct socket *sock;
2239 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2243 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2245 fput_light(sock->file, fput_needed);
2250 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2251 unsigned int, flags)
2253 if (flags & MSG_CMSG_COMPAT)
2255 return __sys_recvmsg(fd, msg, flags);
2259 * Linux recvmmsg interface
2262 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2263 unsigned int flags, struct timespec *timeout)
2265 int fput_needed, err, datagrams;
2266 struct socket *sock;
2267 struct mmsghdr __user *entry;
2268 struct compat_mmsghdr __user *compat_entry;
2269 struct msghdr msg_sys;
2270 struct timespec end_time;
2273 poll_select_set_timeout(&end_time, timeout->tv_sec,
2279 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2283 err = sock_error(sock->sk);
2288 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2290 while (datagrams < vlen) {
2292 * No need to ask LSM for more than the first datagram.
2294 if (MSG_CMSG_COMPAT & flags) {
2295 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2296 &msg_sys, flags & ~MSG_WAITFORONE,
2300 err = __put_user(err, &compat_entry->msg_len);
2303 err = ___sys_recvmsg(sock,
2304 (struct user_msghdr __user *)entry,
2305 &msg_sys, flags & ~MSG_WAITFORONE,
2309 err = put_user(err, &entry->msg_len);
2317 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2318 if (flags & MSG_WAITFORONE)
2319 flags |= MSG_DONTWAIT;
2322 ktime_get_ts(timeout);
2323 *timeout = timespec_sub(end_time, *timeout);
2324 if (timeout->tv_sec < 0) {
2325 timeout->tv_sec = timeout->tv_nsec = 0;
2329 /* Timeout, return less than vlen datagrams */
2330 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2334 /* Out of band data, return right away */
2335 if (msg_sys.msg_flags & MSG_OOB)
2340 fput_light(sock->file, fput_needed);
2345 if (datagrams != 0) {
2347 * We may return less entries than requested (vlen) if the
2348 * sock is non block and there aren't enough datagrams...
2350 if (err != -EAGAIN) {
2352 * ... or if recvmsg returns an error after we
2353 * received some datagrams, where we record the
2354 * error to return on the next call or if the
2355 * app asks about it using getsockopt(SO_ERROR).
2357 sock->sk->sk_err = -err;
2366 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2367 unsigned int, vlen, unsigned int, flags,
2368 struct timespec __user *, timeout)
2371 struct timespec timeout_sys;
2373 if (flags & MSG_CMSG_COMPAT)
2377 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2379 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2382 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2384 if (datagrams > 0 &&
2385 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2386 datagrams = -EFAULT;
2391 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2392 /* Argument list sizes for sys_socketcall */
2393 #define AL(x) ((x) * sizeof(unsigned long))
2394 static const unsigned char nargs[21] = {
2395 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2396 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2397 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2404 * System call vectors.
2406 * Argument checking cleaned up. Saved 20% in size.
2407 * This function doesn't need to set the kernel lock because
2408 * it is set by the callees.
2411 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2413 unsigned long a[AUDITSC_ARGS];
2414 unsigned long a0, a1;
2418 if (call < 1 || call > SYS_SENDMMSG)
2422 if (len > sizeof(a))
2425 /* copy_from_user should be SMP safe. */
2426 if (copy_from_user(a, args, len))
2429 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2438 err = sys_socket(a0, a1, a[2]);
2441 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2444 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2447 err = sys_listen(a0, a1);
2450 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2451 (int __user *)a[2], 0);
2453 case SYS_GETSOCKNAME:
2455 sys_getsockname(a0, (struct sockaddr __user *)a1,
2456 (int __user *)a[2]);
2458 case SYS_GETPEERNAME:
2460 sys_getpeername(a0, (struct sockaddr __user *)a1,
2461 (int __user *)a[2]);
2463 case SYS_SOCKETPAIR:
2464 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2467 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2470 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2471 (struct sockaddr __user *)a[4], a[5]);
2474 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2477 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2478 (struct sockaddr __user *)a[4],
2479 (int __user *)a[5]);
2482 err = sys_shutdown(a0, a1);
2484 case SYS_SETSOCKOPT:
2485 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2487 case SYS_GETSOCKOPT:
2489 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2490 (int __user *)a[4]);
2493 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2496 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2499 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2502 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2503 (struct timespec __user *)a[4]);
2506 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2507 (int __user *)a[2], a[3]);
2516 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2519 * sock_register - add a socket protocol handler
2520 * @ops: description of protocol
2522 * This function is called by a protocol handler that wants to
2523 * advertise its address family, and have it linked into the
2524 * socket interface. The value ops->family corresponds to the
2525 * socket system call protocol family.
2527 int sock_register(const struct net_proto_family *ops)
2531 if (ops->family >= NPROTO) {
2532 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2536 spin_lock(&net_family_lock);
2537 if (rcu_dereference_protected(net_families[ops->family],
2538 lockdep_is_held(&net_family_lock)))
2541 rcu_assign_pointer(net_families[ops->family], ops);
2544 spin_unlock(&net_family_lock);
2546 pr_info("NET: Registered protocol family %d\n", ops->family);
2549 EXPORT_SYMBOL(sock_register);
2552 * sock_unregister - remove a protocol handler
2553 * @family: protocol family to remove
2555 * This function is called by a protocol handler that wants to
2556 * remove its address family, and have it unlinked from the
2557 * new socket creation.
2559 * If protocol handler is a module, then it can use module reference
2560 * counts to protect against new references. If protocol handler is not
2561 * a module then it needs to provide its own protection in
2562 * the ops->create routine.
2564 void sock_unregister(int family)
2566 BUG_ON(family < 0 || family >= NPROTO);
2568 spin_lock(&net_family_lock);
2569 RCU_INIT_POINTER(net_families[family], NULL);
2570 spin_unlock(&net_family_lock);
2574 pr_info("NET: Unregistered protocol family %d\n", family);
2576 EXPORT_SYMBOL(sock_unregister);
2578 static int __init sock_init(void)
2582 * Initialize the network sysctl infrastructure.
2584 err = net_sysctl_init();
2589 * Initialize skbuff SLAB cache
2594 * Initialize the protocols module.
2599 err = register_filesystem(&sock_fs_type);
2602 sock_mnt = kern_mount(&sock_fs_type);
2603 if (IS_ERR(sock_mnt)) {
2604 err = PTR_ERR(sock_mnt);
2608 /* The real protocol initialization is performed in later initcalls.
2611 #ifdef CONFIG_NETFILTER
2612 err = netfilter_init();
2617 ptp_classifier_init();
2623 unregister_filesystem(&sock_fs_type);
2628 core_initcall(sock_init); /* early initcall */
2630 #ifdef CONFIG_PROC_FS
2631 void socket_seq_show(struct seq_file *seq)
2636 for_each_possible_cpu(cpu)
2637 counter += per_cpu(sockets_in_use, cpu);
2639 /* It can be negative, by the way. 8) */
2643 seq_printf(seq, "sockets: used %d\n", counter);
2645 #endif /* CONFIG_PROC_FS */
2647 #ifdef CONFIG_COMPAT
2648 static int do_siocgstamp(struct net *net, struct socket *sock,
2649 unsigned int cmd, void __user *up)
2651 mm_segment_t old_fs = get_fs();
2656 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2659 err = compat_put_timeval(&ktv, up);
2664 static int do_siocgstampns(struct net *net, struct socket *sock,
2665 unsigned int cmd, void __user *up)
2667 mm_segment_t old_fs = get_fs();
2668 struct timespec kts;
2672 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2675 err = compat_put_timespec(&kts, up);
2680 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2682 struct ifreq __user *uifr;
2685 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2686 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2689 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2693 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2699 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2701 struct compat_ifconf ifc32;
2703 struct ifconf __user *uifc;
2704 struct compat_ifreq __user *ifr32;
2705 struct ifreq __user *ifr;
2709 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2712 memset(&ifc, 0, sizeof(ifc));
2713 if (ifc32.ifcbuf == 0) {
2717 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2719 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2720 sizeof(struct ifreq);
2721 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2723 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2724 ifr32 = compat_ptr(ifc32.ifcbuf);
2725 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2726 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2732 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2735 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2739 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2743 ifr32 = compat_ptr(ifc32.ifcbuf);
2745 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2746 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2747 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2753 if (ifc32.ifcbuf == 0) {
2754 /* Translate from 64-bit structure multiple to
2758 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2763 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2769 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2771 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2772 bool convert_in = false, convert_out = false;
2773 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2774 struct ethtool_rxnfc __user *rxnfc;
2775 struct ifreq __user *ifr;
2776 u32 rule_cnt = 0, actual_rule_cnt;
2781 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2784 compat_rxnfc = compat_ptr(data);
2786 if (get_user(ethcmd, &compat_rxnfc->cmd))
2789 /* Most ethtool structures are defined without padding.
2790 * Unfortunately struct ethtool_rxnfc is an exception.
2795 case ETHTOOL_GRXCLSRLALL:
2796 /* Buffer size is variable */
2797 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2799 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2801 buf_size += rule_cnt * sizeof(u32);
2803 case ETHTOOL_GRXRINGS:
2804 case ETHTOOL_GRXCLSRLCNT:
2805 case ETHTOOL_GRXCLSRULE:
2806 case ETHTOOL_SRXCLSRLINS:
2809 case ETHTOOL_SRXCLSRLDEL:
2810 buf_size += sizeof(struct ethtool_rxnfc);
2815 ifr = compat_alloc_user_space(buf_size);
2816 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2818 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2821 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2822 &ifr->ifr_ifru.ifru_data))
2826 /* We expect there to be holes between fs.m_ext and
2827 * fs.ring_cookie and at the end of fs, but nowhere else.
2829 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2830 sizeof(compat_rxnfc->fs.m_ext) !=
2831 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2832 sizeof(rxnfc->fs.m_ext));
2834 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2835 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2836 offsetof(struct ethtool_rxnfc, fs.location) -
2837 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2839 if (copy_in_user(rxnfc, compat_rxnfc,
2840 (void __user *)(&rxnfc->fs.m_ext + 1) -
2841 (void __user *)rxnfc) ||
2842 copy_in_user(&rxnfc->fs.ring_cookie,
2843 &compat_rxnfc->fs.ring_cookie,
2844 (void __user *)(&rxnfc->fs.location + 1) -
2845 (void __user *)&rxnfc->fs.ring_cookie) ||
2846 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2847 sizeof(rxnfc->rule_cnt)))
2851 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2856 if (copy_in_user(compat_rxnfc, rxnfc,
2857 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2858 (const void __user *)rxnfc) ||
2859 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2860 &rxnfc->fs.ring_cookie,
2861 (const void __user *)(&rxnfc->fs.location + 1) -
2862 (const void __user *)&rxnfc->fs.ring_cookie) ||
2863 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2864 sizeof(rxnfc->rule_cnt)))
2867 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2868 /* As an optimisation, we only copy the actual
2869 * number of rules that the underlying
2870 * function returned. Since Mallory might
2871 * change the rule count in user memory, we
2872 * check that it is less than the rule count
2873 * originally given (as the user buffer size),
2874 * which has been range-checked.
2876 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2878 if (actual_rule_cnt < rule_cnt)
2879 rule_cnt = actual_rule_cnt;
2880 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2881 &rxnfc->rule_locs[0],
2882 rule_cnt * sizeof(u32)))
2890 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2893 compat_uptr_t uptr32;
2894 struct ifreq __user *uifr;
2896 uifr = compat_alloc_user_space(sizeof(*uifr));
2897 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2900 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2903 uptr = compat_ptr(uptr32);
2905 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2908 return dev_ioctl(net, SIOCWANDEV, uifr);
2911 static int bond_ioctl(struct net *net, unsigned int cmd,
2912 struct compat_ifreq __user *ifr32)
2915 mm_segment_t old_fs;
2919 case SIOCBONDENSLAVE:
2920 case SIOCBONDRELEASE:
2921 case SIOCBONDSETHWADDR:
2922 case SIOCBONDCHANGEACTIVE:
2923 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2928 err = dev_ioctl(net, cmd,
2929 (struct ifreq __user __force *) &kifr);
2934 return -ENOIOCTLCMD;
2938 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2939 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2940 struct compat_ifreq __user *u_ifreq32)
2942 struct ifreq __user *u_ifreq64;
2943 char tmp_buf[IFNAMSIZ];
2944 void __user *data64;
2947 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2950 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2952 data64 = compat_ptr(data32);
2954 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2956 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2959 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2962 return dev_ioctl(net, cmd, u_ifreq64);
2965 static int dev_ifsioc(struct net *net, struct socket *sock,
2966 unsigned int cmd, struct compat_ifreq __user *uifr32)
2968 struct ifreq __user *uifr;
2971 uifr = compat_alloc_user_space(sizeof(*uifr));
2972 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2975 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2986 case SIOCGIFBRDADDR:
2987 case SIOCGIFDSTADDR:
2988 case SIOCGIFNETMASK:
2993 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3001 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3002 struct compat_ifreq __user *uifr32)
3005 struct compat_ifmap __user *uifmap32;
3006 mm_segment_t old_fs;
3009 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3010 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3011 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3012 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3013 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3014 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3015 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3016 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3022 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3025 if (cmd == SIOCGIFMAP && !err) {
3026 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3027 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3028 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3029 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3030 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3031 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3032 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3041 struct sockaddr rt_dst; /* target address */
3042 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3043 struct sockaddr rt_genmask; /* target network mask (IP) */
3044 unsigned short rt_flags;
3047 unsigned char rt_tos;
3048 unsigned char rt_class;
3050 short rt_metric; /* +1 for binary compatibility! */
3051 /* char * */ u32 rt_dev; /* forcing the device at add */
3052 u32 rt_mtu; /* per route MTU/Window */
3053 u32 rt_window; /* Window clamping */
3054 unsigned short rt_irtt; /* Initial RTT */
3057 struct in6_rtmsg32 {
3058 struct in6_addr rtmsg_dst;
3059 struct in6_addr rtmsg_src;
3060 struct in6_addr rtmsg_gateway;
3070 static int routing_ioctl(struct net *net, struct socket *sock,
3071 unsigned int cmd, void __user *argp)
3075 struct in6_rtmsg r6;
3079 mm_segment_t old_fs = get_fs();
3081 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3082 struct in6_rtmsg32 __user *ur6 = argp;
3083 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3084 3 * sizeof(struct in6_addr));
3085 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3086 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3087 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3088 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3089 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3090 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3091 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3095 struct rtentry32 __user *ur4 = argp;
3096 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3097 3 * sizeof(struct sockaddr));
3098 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3099 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3100 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3101 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3102 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3103 ret |= get_user(rtdev, &(ur4->rt_dev));
3105 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3106 r4.rt_dev = (char __user __force *)devname;
3120 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3127 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3128 * for some operations; this forces use of the newer bridge-utils that
3129 * use compatible ioctls
3131 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3135 if (get_user(tmp, argp))
3137 if (tmp == BRCTL_GET_VERSION)
3138 return BRCTL_VERSION + 1;
3142 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3143 unsigned int cmd, unsigned long arg)
3145 void __user *argp = compat_ptr(arg);
3146 struct sock *sk = sock->sk;
3147 struct net *net = sock_net(sk);
3149 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3150 return compat_ifr_data_ioctl(net, cmd, argp);
3155 return old_bridge_ioctl(argp);
3157 return dev_ifname32(net, argp);
3159 return dev_ifconf(net, argp);
3161 return ethtool_ioctl(net, argp);
3163 return compat_siocwandev(net, argp);
3166 return compat_sioc_ifmap(net, cmd, argp);
3167 case SIOCBONDENSLAVE:
3168 case SIOCBONDRELEASE:
3169 case SIOCBONDSETHWADDR:
3170 case SIOCBONDCHANGEACTIVE:
3171 return bond_ioctl(net, cmd, argp);
3174 return routing_ioctl(net, sock, cmd, argp);
3176 return do_siocgstamp(net, sock, cmd, argp);
3178 return do_siocgstampns(net, sock, cmd, argp);
3179 case SIOCBONDSLAVEINFOQUERY:
3180 case SIOCBONDINFOQUERY:
3183 return compat_ifr_data_ioctl(net, cmd, argp);
3195 return sock_ioctl(file, cmd, arg);
3212 case SIOCSIFHWBROADCAST:
3214 case SIOCGIFBRDADDR:
3215 case SIOCSIFBRDADDR:
3216 case SIOCGIFDSTADDR:
3217 case SIOCSIFDSTADDR:
3218 case SIOCGIFNETMASK:
3219 case SIOCSIFNETMASK:
3230 return dev_ifsioc(net, sock, cmd, argp);
3236 return sock_do_ioctl(net, sock, cmd, arg);
3239 return -ENOIOCTLCMD;
3242 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3245 struct socket *sock = file->private_data;
3246 int ret = -ENOIOCTLCMD;
3253 if (sock->ops->compat_ioctl)
3254 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3256 if (ret == -ENOIOCTLCMD &&
3257 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3258 ret = compat_wext_handle_ioctl(net, cmd, arg);
3260 if (ret == -ENOIOCTLCMD)
3261 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3267 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3269 return sock->ops->bind(sock, addr, addrlen);
3271 EXPORT_SYMBOL(kernel_bind);
3273 int kernel_listen(struct socket *sock, int backlog)
3275 return sock->ops->listen(sock, backlog);
3277 EXPORT_SYMBOL(kernel_listen);
3279 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3281 struct sock *sk = sock->sk;
3284 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3289 err = sock->ops->accept(sock, *newsock, flags);
3291 sock_release(*newsock);
3296 (*newsock)->ops = sock->ops;
3297 __module_get((*newsock)->ops->owner);
3302 EXPORT_SYMBOL(kernel_accept);
3304 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3307 return sock->ops->connect(sock, addr, addrlen, flags);
3309 EXPORT_SYMBOL(kernel_connect);
3311 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3314 return sock->ops->getname(sock, addr, addrlen, 0);
3316 EXPORT_SYMBOL(kernel_getsockname);
3318 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3321 return sock->ops->getname(sock, addr, addrlen, 1);
3323 EXPORT_SYMBOL(kernel_getpeername);
3325 int kernel_getsockopt(struct socket *sock, int level, int optname,
3326 char *optval, int *optlen)
3328 mm_segment_t oldfs = get_fs();
3329 char __user *uoptval;
3330 int __user *uoptlen;
3333 uoptval = (char __user __force *) optval;
3334 uoptlen = (int __user __force *) optlen;
3337 if (level == SOL_SOCKET)
3338 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3340 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3345 EXPORT_SYMBOL(kernel_getsockopt);
3347 int kernel_setsockopt(struct socket *sock, int level, int optname,
3348 char *optval, unsigned int optlen)
3350 mm_segment_t oldfs = get_fs();
3351 char __user *uoptval;
3354 uoptval = (char __user __force *) optval;
3357 if (level == SOL_SOCKET)
3358 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3360 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3365 EXPORT_SYMBOL(kernel_setsockopt);
3367 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3368 size_t size, int flags)
3370 if (sock->ops->sendpage)
3371 return sock->ops->sendpage(sock, page, offset, size, flags);
3373 return sock_no_sendpage(sock, page, offset, size, flags);
3375 EXPORT_SYMBOL(kernel_sendpage);
3377 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3379 mm_segment_t oldfs = get_fs();
3383 err = sock->ops->ioctl(sock, cmd, arg);
3388 EXPORT_SYMBOL(kernel_sock_ioctl);
3390 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3392 return sock->ops->shutdown(sock, how);
3394 EXPORT_SYMBOL(kernel_sock_shutdown);