2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
117 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
118 unsigned long nr_segs, loff_t pos);
119 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
120 unsigned long nr_segs, loff_t pos);
121 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
123 static int sock_close(struct inode *inode, struct file *file);
124 static unsigned int sock_poll(struct file *file,
125 struct poll_table_struct *wait);
126 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
128 static long compat_sock_ioctl(struct file *file,
129 unsigned int cmd, unsigned long arg);
131 static int sock_fasync(int fd, struct file *filp, int on);
132 static ssize_t sock_sendpage(struct file *file, struct page *page,
133 int offset, size_t size, loff_t *ppos, int more);
134 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
135 struct pipe_inode_info *pipe, size_t len,
139 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
140 * in the operation structures but are done directly via the socketcall() multiplexor.
143 static const struct file_operations socket_file_ops = {
144 .owner = THIS_MODULE,
146 .aio_read = sock_aio_read,
147 .aio_write = sock_aio_write,
149 .unlocked_ioctl = sock_ioctl,
151 .compat_ioctl = compat_sock_ioctl,
154 .open = sock_no_open, /* special open code to disallow open via /proc */
155 .release = sock_close,
156 .fasync = sock_fasync,
157 .sendpage = sock_sendpage,
158 .splice_write = generic_splice_sendpage,
159 .splice_read = sock_splice_read,
163 * The protocol list. Each protocol is registered in here.
166 static DEFINE_SPINLOCK(net_family_lock);
167 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
170 * Statistics counters of the socket lists
173 static DEFINE_PER_CPU(int, sockets_in_use);
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
198 if (copy_from_user(kaddr, uaddr, ulen))
200 return audit_sockaddr(ulen, kaddr);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221 void __user *uaddr, int __user *ulen)
226 BUG_ON(klen > sizeof(struct sockaddr_storage));
227 err = get_user(len, ulen);
235 if (audit_sockaddr(klen, kaddr))
237 if (copy_to_user(uaddr, kaddr, len))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen, ulen);
247 static struct kmem_cache *sock_inode_cachep __read_mostly;
249 static struct inode *sock_alloc_inode(struct super_block *sb)
251 struct socket_alloc *ei;
252 struct socket_wq *wq;
254 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
257 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
259 kmem_cache_free(sock_inode_cachep, ei);
262 init_waitqueue_head(&wq->wait);
263 wq->fasync_list = NULL;
264 RCU_INIT_POINTER(ei->socket.wq, wq);
266 ei->socket.state = SS_UNCONNECTED;
267 ei->socket.flags = 0;
268 ei->socket.ops = NULL;
269 ei->socket.sk = NULL;
270 ei->socket.file = NULL;
272 return &ei->vfs_inode;
275 static void sock_destroy_inode(struct inode *inode)
277 struct socket_alloc *ei;
278 struct socket_wq *wq;
280 ei = container_of(inode, struct socket_alloc, vfs_inode);
281 wq = rcu_dereference_protected(ei->socket.wq, 1);
283 kmem_cache_free(sock_inode_cachep, ei);
286 static void init_once(void *foo)
288 struct socket_alloc *ei = (struct socket_alloc *)foo;
290 inode_init_once(&ei->vfs_inode);
293 static int init_inodecache(void)
295 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
296 sizeof(struct socket_alloc),
298 (SLAB_HWCACHE_ALIGN |
299 SLAB_RECLAIM_ACCOUNT |
302 if (sock_inode_cachep == NULL)
307 static const struct super_operations sockfs_ops = {
308 .alloc_inode = sock_alloc_inode,
309 .destroy_inode = sock_destroy_inode,
310 .statfs = simple_statfs,
314 * sockfs_dname() is called from d_path().
316 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
318 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
319 dentry->d_inode->i_ino);
322 static const struct dentry_operations sockfs_dentry_operations = {
323 .d_dname = sockfs_dname,
326 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
327 int flags, const char *dev_name, void *data)
329 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
330 &sockfs_dentry_operations, SOCKFS_MAGIC);
333 static struct vfsmount *sock_mnt __read_mostly;
335 static struct file_system_type sock_fs_type = {
337 .mount = sockfs_mount,
338 .kill_sb = kill_anon_super,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
360 struct qstr name = { .name = "" };
366 name.len = strlen(name.name);
367 } else if (sock->sk) {
368 name.name = sock->sk->sk_prot_creator->name;
369 name.len = strlen(name.name);
371 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
372 if (unlikely(!path.dentry))
373 return ERR_PTR(-ENOMEM);
374 path.mnt = mntget(sock_mnt);
376 d_instantiate(path.dentry, SOCK_INODE(sock));
377 SOCK_INODE(sock)->i_fop = &socket_file_ops;
379 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
381 if (unlikely(IS_ERR(file))) {
382 /* drop dentry, keep inode */
383 ihold(path.dentry->d_inode);
389 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
390 file->private_data = sock;
393 EXPORT_SYMBOL(sock_alloc_file);
395 static int sock_map_fd(struct socket *sock, int flags)
397 struct file *newfile;
398 int fd = get_unused_fd_flags(flags);
399 if (unlikely(fd < 0))
402 newfile = sock_alloc_file(sock, flags, NULL);
403 if (likely(!IS_ERR(newfile))) {
404 fd_install(fd, newfile);
409 return PTR_ERR(newfile);
412 struct socket *sock_from_file(struct file *file, int *err)
414 if (file->f_op == &socket_file_ops)
415 return file->private_data; /* set in sock_map_fd */
420 EXPORT_SYMBOL(sock_from_file);
423 * sockfd_lookup - Go from a file number to its socket slot
425 * @err: pointer to an error code return
427 * The file handle passed in is locked and the socket it is bound
428 * too is returned. If an error occurs the err pointer is overwritten
429 * with a negative errno code and NULL is returned. The function checks
430 * for both invalid handles and passing a handle which is not a socket.
432 * On a success the socket object pointer is returned.
435 struct socket *sockfd_lookup(int fd, int *err)
446 sock = sock_from_file(file, err);
451 EXPORT_SYMBOL(sockfd_lookup);
453 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
455 struct fd f = fdget(fd);
460 sock = sock_from_file(f.file, err);
462 *fput_needed = f.flags;
470 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
471 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
472 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
473 static ssize_t sockfs_getxattr(struct dentry *dentry,
474 const char *name, void *value, size_t size)
476 const char *proto_name;
481 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
482 proto_name = dentry->d_name.name;
483 proto_size = strlen(proto_name);
487 if (proto_size + 1 > size)
490 strncpy(value, proto_name, proto_size + 1);
492 error = proto_size + 1;
499 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
505 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
515 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
520 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
527 static const struct inode_operations sockfs_inode_ops = {
528 .getxattr = sockfs_getxattr,
529 .listxattr = sockfs_listxattr,
533 * sock_alloc - allocate a socket
535 * Allocate a new inode and socket object. The two are bound together
536 * and initialised. The socket is then returned. If we are out of inodes
540 static struct socket *sock_alloc(void)
545 inode = new_inode_pseudo(sock_mnt->mnt_sb);
549 sock = SOCKET_I(inode);
551 kmemcheck_annotate_bitfield(sock, type);
552 inode->i_ino = get_next_ino();
553 inode->i_mode = S_IFSOCK | S_IRWXUGO;
554 inode->i_uid = current_fsuid();
555 inode->i_gid = current_fsgid();
556 inode->i_op = &sockfs_inode_ops;
558 this_cpu_add(sockets_in_use, 1);
563 * In theory you can't get an open on this inode, but /proc provides
564 * a back door. Remember to keep it shut otherwise you'll let the
565 * creepy crawlies in.
568 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
573 const struct file_operations bad_sock_fops = {
574 .owner = THIS_MODULE,
575 .open = sock_no_open,
576 .llseek = noop_llseek,
580 * sock_release - close a socket
581 * @sock: socket to close
583 * The socket is released from the protocol stack if it has a release
584 * callback, and the inode is then released if the socket is bound to
585 * an inode not a file.
588 void sock_release(struct socket *sock)
591 struct module *owner = sock->ops->owner;
593 sock->ops->release(sock);
598 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
599 pr_err("%s: fasync list not empty!\n", __func__);
601 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
604 this_cpu_sub(sockets_in_use, 1);
606 iput(SOCK_INODE(sock));
611 EXPORT_SYMBOL(sock_release);
613 void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
615 u8 flags = *tx_flags;
617 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
618 flags |= SKBTX_HW_TSTAMP;
620 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
621 flags |= SKBTX_SW_TSTAMP;
623 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
624 flags |= SKBTX_SCHED_TSTAMP;
626 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
627 flags |= SKBTX_ACK_TSTAMP;
629 if (sock_flag(sk, SOCK_WIFI_STATUS))
630 flags |= SKBTX_WIFI_STATUS;
634 EXPORT_SYMBOL(sock_tx_timestamp);
636 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
637 struct msghdr *msg, size_t size)
639 struct sock_iocb *si = kiocb_to_siocb(iocb);
646 return sock->ops->sendmsg(iocb, sock, msg, size);
649 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
650 struct msghdr *msg, size_t size)
652 int err = security_socket_sendmsg(sock, msg, size);
654 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
657 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
660 struct sock_iocb siocb;
663 init_sync_kiocb(&iocb, NULL);
664 iocb.private = &siocb;
665 ret = __sock_sendmsg(&iocb, sock, msg, size);
666 if (-EIOCBQUEUED == ret)
667 ret = wait_on_sync_kiocb(&iocb);
670 EXPORT_SYMBOL(sock_sendmsg);
672 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
675 struct sock_iocb siocb;
678 init_sync_kiocb(&iocb, NULL);
679 iocb.private = &siocb;
680 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
681 if (-EIOCBQUEUED == ret)
682 ret = wait_on_sync_kiocb(&iocb);
686 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
687 struct kvec *vec, size_t num, size_t size)
689 mm_segment_t oldfs = get_fs();
694 * the following is safe, since for compiler definitions of kvec and
695 * iovec are identical, yielding the same in-core layout and alignment
697 msg->msg_iov = (struct iovec *)vec;
698 msg->msg_iovlen = num;
699 result = sock_sendmsg(sock, msg, size);
703 EXPORT_SYMBOL(kernel_sendmsg);
706 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
708 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
711 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
712 struct scm_timestamping tss;
714 struct skb_shared_hwtstamps *shhwtstamps =
717 /* Race occurred between timestamp enabling and packet
718 receiving. Fill in the current time for now. */
719 if (need_software_tstamp && skb->tstamp.tv64 == 0)
720 __net_timestamp(skb);
722 if (need_software_tstamp) {
723 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
725 skb_get_timestamp(skb, &tv);
726 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
730 skb_get_timestampns(skb, &ts);
731 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
736 memset(&tss, 0, sizeof(tss));
737 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
738 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
741 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
742 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
745 put_cmsg(msg, SOL_SOCKET,
746 SCM_TIMESTAMPING, sizeof(tss), &tss);
748 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
750 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
755 if (!sock_flag(sk, SOCK_WIFI_STATUS))
757 if (!skb->wifi_acked_valid)
760 ack = skb->wifi_acked;
762 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
764 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
766 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
769 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
770 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
771 sizeof(__u32), &skb->dropcount);
774 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
777 sock_recv_timestamp(msg, sk, skb);
778 sock_recv_drops(msg, sk, skb);
780 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
782 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
783 struct msghdr *msg, size_t size, int flags)
785 struct sock_iocb *si = kiocb_to_siocb(iocb);
793 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
796 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
797 struct msghdr *msg, size_t size, int flags)
799 int err = security_socket_recvmsg(sock, msg, size, flags);
801 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
804 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
805 size_t size, int flags)
808 struct sock_iocb siocb;
811 init_sync_kiocb(&iocb, NULL);
812 iocb.private = &siocb;
813 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
814 if (-EIOCBQUEUED == ret)
815 ret = wait_on_sync_kiocb(&iocb);
818 EXPORT_SYMBOL(sock_recvmsg);
820 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
821 size_t size, int flags)
824 struct sock_iocb siocb;
827 init_sync_kiocb(&iocb, NULL);
828 iocb.private = &siocb;
829 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
830 if (-EIOCBQUEUED == ret)
831 ret = wait_on_sync_kiocb(&iocb);
836 * kernel_recvmsg - Receive a message from a socket (kernel space)
837 * @sock: The socket to receive the message from
838 * @msg: Received message
839 * @vec: Input s/g array for message data
840 * @num: Size of input s/g array
841 * @size: Number of bytes to read
842 * @flags: Message flags (MSG_DONTWAIT, etc...)
844 * On return the msg structure contains the scatter/gather array passed in the
845 * vec argument. The array is modified so that it consists of the unfilled
846 * portion of the original array.
848 * The returned value is the total number of bytes received, or an error.
850 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
851 struct kvec *vec, size_t num, size_t size, int flags)
853 mm_segment_t oldfs = get_fs();
858 * the following is safe, since for compiler definitions of kvec and
859 * iovec are identical, yielding the same in-core layout and alignment
861 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
862 result = sock_recvmsg(sock, msg, size, flags);
866 EXPORT_SYMBOL(kernel_recvmsg);
868 static ssize_t sock_sendpage(struct file *file, struct page *page,
869 int offset, size_t size, loff_t *ppos, int more)
874 sock = file->private_data;
876 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
877 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
880 return kernel_sendpage(sock, page, offset, size, flags);
883 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
884 struct pipe_inode_info *pipe, size_t len,
887 struct socket *sock = file->private_data;
889 if (unlikely(!sock->ops->splice_read))
892 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
895 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
896 struct sock_iocb *siocb)
898 if (!is_sync_kiocb(iocb))
902 iocb->private = siocb;
906 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
907 struct file *file, const struct iovec *iov,
908 unsigned long nr_segs)
910 struct socket *sock = file->private_data;
914 for (i = 0; i < nr_segs; i++)
915 size += iov[i].iov_len;
917 msg->msg_name = NULL;
918 msg->msg_namelen = 0;
919 msg->msg_control = NULL;
920 msg->msg_controllen = 0;
921 msg->msg_iov = (struct iovec *)iov;
922 msg->msg_iovlen = nr_segs;
923 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
925 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
928 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
929 unsigned long nr_segs, loff_t pos)
931 struct sock_iocb siocb, *x;
936 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
940 x = alloc_sock_iocb(iocb, &siocb);
943 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
946 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
947 struct file *file, const struct iovec *iov,
948 unsigned long nr_segs)
950 struct socket *sock = file->private_data;
954 for (i = 0; i < nr_segs; i++)
955 size += iov[i].iov_len;
957 msg->msg_name = NULL;
958 msg->msg_namelen = 0;
959 msg->msg_control = NULL;
960 msg->msg_controllen = 0;
961 msg->msg_iov = (struct iovec *)iov;
962 msg->msg_iovlen = nr_segs;
963 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
964 if (sock->type == SOCK_SEQPACKET)
965 msg->msg_flags |= MSG_EOR;
967 return __sock_sendmsg(iocb, sock, msg, size);
970 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
971 unsigned long nr_segs, loff_t pos)
973 struct sock_iocb siocb, *x;
978 x = alloc_sock_iocb(iocb, &siocb);
982 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
986 * Atomic setting of ioctl hooks to avoid race
987 * with module unload.
990 static DEFINE_MUTEX(br_ioctl_mutex);
991 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
993 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
995 mutex_lock(&br_ioctl_mutex);
996 br_ioctl_hook = hook;
997 mutex_unlock(&br_ioctl_mutex);
999 EXPORT_SYMBOL(brioctl_set);
1001 static DEFINE_MUTEX(vlan_ioctl_mutex);
1002 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1004 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1006 mutex_lock(&vlan_ioctl_mutex);
1007 vlan_ioctl_hook = hook;
1008 mutex_unlock(&vlan_ioctl_mutex);
1010 EXPORT_SYMBOL(vlan_ioctl_set);
1012 static DEFINE_MUTEX(dlci_ioctl_mutex);
1013 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1015 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1017 mutex_lock(&dlci_ioctl_mutex);
1018 dlci_ioctl_hook = hook;
1019 mutex_unlock(&dlci_ioctl_mutex);
1021 EXPORT_SYMBOL(dlci_ioctl_set);
1023 static long sock_do_ioctl(struct net *net, struct socket *sock,
1024 unsigned int cmd, unsigned long arg)
1027 void __user *argp = (void __user *)arg;
1029 err = sock->ops->ioctl(sock, cmd, arg);
1032 * If this ioctl is unknown try to hand it down
1033 * to the NIC driver.
1035 if (err == -ENOIOCTLCMD)
1036 err = dev_ioctl(net, cmd, argp);
1042 * With an ioctl, arg may well be a user mode pointer, but we don't know
1043 * what to do with it - that's up to the protocol still.
1046 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1048 struct socket *sock;
1050 void __user *argp = (void __user *)arg;
1054 sock = file->private_data;
1057 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1058 err = dev_ioctl(net, cmd, argp);
1060 #ifdef CONFIG_WEXT_CORE
1061 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1062 err = dev_ioctl(net, cmd, argp);
1069 if (get_user(pid, (int __user *)argp))
1071 err = f_setown(sock->file, pid, 1);
1075 err = put_user(f_getown(sock->file),
1076 (int __user *)argp);
1084 request_module("bridge");
1086 mutex_lock(&br_ioctl_mutex);
1088 err = br_ioctl_hook(net, cmd, argp);
1089 mutex_unlock(&br_ioctl_mutex);
1094 if (!vlan_ioctl_hook)
1095 request_module("8021q");
1097 mutex_lock(&vlan_ioctl_mutex);
1098 if (vlan_ioctl_hook)
1099 err = vlan_ioctl_hook(net, argp);
1100 mutex_unlock(&vlan_ioctl_mutex);
1105 if (!dlci_ioctl_hook)
1106 request_module("dlci");
1108 mutex_lock(&dlci_ioctl_mutex);
1109 if (dlci_ioctl_hook)
1110 err = dlci_ioctl_hook(cmd, argp);
1111 mutex_unlock(&dlci_ioctl_mutex);
1114 err = sock_do_ioctl(net, sock, cmd, arg);
1120 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1123 struct socket *sock = NULL;
1125 err = security_socket_create(family, type, protocol, 1);
1129 sock = sock_alloc();
1136 err = security_socket_post_create(sock, family, type, protocol, 1);
1148 EXPORT_SYMBOL(sock_create_lite);
1150 /* No kernel lock held - perfect */
1151 static unsigned int sock_poll(struct file *file, poll_table *wait)
1153 unsigned int busy_flag = 0;
1154 struct socket *sock;
1157 * We can't return errors to poll, so it's either yes or no.
1159 sock = file->private_data;
1161 if (sk_can_busy_loop(sock->sk)) {
1162 /* this socket can poll_ll so tell the system call */
1163 busy_flag = POLL_BUSY_LOOP;
1165 /* once, only if requested by syscall */
1166 if (wait && (wait->_key & POLL_BUSY_LOOP))
1167 sk_busy_loop(sock->sk, 1);
1170 return busy_flag | sock->ops->poll(file, sock, wait);
1173 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1175 struct socket *sock = file->private_data;
1177 return sock->ops->mmap(file, sock, vma);
1180 static int sock_close(struct inode *inode, struct file *filp)
1182 sock_release(SOCKET_I(inode));
1187 * Update the socket async list
1189 * Fasync_list locking strategy.
1191 * 1. fasync_list is modified only under process context socket lock
1192 * i.e. under semaphore.
1193 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1194 * or under socket lock
1197 static int sock_fasync(int fd, struct file *filp, int on)
1199 struct socket *sock = filp->private_data;
1200 struct sock *sk = sock->sk;
1201 struct socket_wq *wq;
1207 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1208 fasync_helper(fd, filp, on, &wq->fasync_list);
1210 if (!wq->fasync_list)
1211 sock_reset_flag(sk, SOCK_FASYNC);
1213 sock_set_flag(sk, SOCK_FASYNC);
1219 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1221 int sock_wake_async(struct socket *sock, int how, int band)
1223 struct socket_wq *wq;
1228 wq = rcu_dereference(sock->wq);
1229 if (!wq || !wq->fasync_list) {
1234 case SOCK_WAKE_WAITD:
1235 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1238 case SOCK_WAKE_SPACE:
1239 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1244 kill_fasync(&wq->fasync_list, SIGIO, band);
1247 kill_fasync(&wq->fasync_list, SIGURG, band);
1252 EXPORT_SYMBOL(sock_wake_async);
1254 int __sock_create(struct net *net, int family, int type, int protocol,
1255 struct socket **res, int kern)
1258 struct socket *sock;
1259 const struct net_proto_family *pf;
1262 * Check protocol is in range
1264 if (family < 0 || family >= NPROTO)
1265 return -EAFNOSUPPORT;
1266 if (type < 0 || type >= SOCK_MAX)
1271 This uglymoron is moved from INET layer to here to avoid
1272 deadlock in module load.
1274 if (family == PF_INET && type == SOCK_PACKET) {
1278 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1284 err = security_socket_create(family, type, protocol, kern);
1289 * Allocate the socket and allow the family to set things up. if
1290 * the protocol is 0, the family is instructed to select an appropriate
1293 sock = sock_alloc();
1295 net_warn_ratelimited("socket: no more sockets\n");
1296 return -ENFILE; /* Not exactly a match, but its the
1297 closest posix thing */
1302 #ifdef CONFIG_MODULES
1303 /* Attempt to load a protocol module if the find failed.
1305 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1306 * requested real, full-featured networking support upon configuration.
1307 * Otherwise module support will break!
1309 if (rcu_access_pointer(net_families[family]) == NULL)
1310 request_module("net-pf-%d", family);
1314 pf = rcu_dereference(net_families[family]);
1315 err = -EAFNOSUPPORT;
1320 * We will call the ->create function, that possibly is in a loadable
1321 * module, so we have to bump that loadable module refcnt first.
1323 if (!try_module_get(pf->owner))
1326 /* Now protected by module ref count */
1329 err = pf->create(net, sock, protocol, kern);
1331 goto out_module_put;
1334 * Now to bump the refcnt of the [loadable] module that owns this
1335 * socket at sock_release time we decrement its refcnt.
1337 if (!try_module_get(sock->ops->owner))
1338 goto out_module_busy;
1341 * Now that we're done with the ->create function, the [loadable]
1342 * module can have its refcnt decremented
1344 module_put(pf->owner);
1345 err = security_socket_post_create(sock, family, type, protocol, kern);
1347 goto out_sock_release;
1353 err = -EAFNOSUPPORT;
1356 module_put(pf->owner);
1363 goto out_sock_release;
1365 EXPORT_SYMBOL(__sock_create);
1367 int sock_create(int family, int type, int protocol, struct socket **res)
1369 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1371 EXPORT_SYMBOL(sock_create);
1373 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1375 return __sock_create(&init_net, family, type, protocol, res, 1);
1377 EXPORT_SYMBOL(sock_create_kern);
1379 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1382 struct socket *sock;
1385 /* Check the SOCK_* constants for consistency. */
1386 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1387 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1388 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1389 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1391 flags = type & ~SOCK_TYPE_MASK;
1392 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1394 type &= SOCK_TYPE_MASK;
1396 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1397 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1399 retval = sock_create(family, type, protocol, &sock);
1403 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1408 /* It may be already another descriptor 8) Not kernel problem. */
1417 * Create a pair of connected sockets.
1420 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1421 int __user *, usockvec)
1423 struct socket *sock1, *sock2;
1425 struct file *newfile1, *newfile2;
1428 flags = type & ~SOCK_TYPE_MASK;
1429 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1431 type &= SOCK_TYPE_MASK;
1433 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1434 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1437 * Obtain the first socket and check if the underlying protocol
1438 * supports the socketpair call.
1441 err = sock_create(family, type, protocol, &sock1);
1445 err = sock_create(family, type, protocol, &sock2);
1449 err = sock1->ops->socketpair(sock1, sock2);
1451 goto out_release_both;
1453 fd1 = get_unused_fd_flags(flags);
1454 if (unlikely(fd1 < 0)) {
1456 goto out_release_both;
1459 fd2 = get_unused_fd_flags(flags);
1460 if (unlikely(fd2 < 0)) {
1462 goto out_put_unused_1;
1465 newfile1 = sock_alloc_file(sock1, flags, NULL);
1466 if (unlikely(IS_ERR(newfile1))) {
1467 err = PTR_ERR(newfile1);
1468 goto out_put_unused_both;
1471 newfile2 = sock_alloc_file(sock2, flags, NULL);
1472 if (IS_ERR(newfile2)) {
1473 err = PTR_ERR(newfile2);
1477 err = put_user(fd1, &usockvec[0]);
1481 err = put_user(fd2, &usockvec[1]);
1485 audit_fd_pair(fd1, fd2);
1487 fd_install(fd1, newfile1);
1488 fd_install(fd2, newfile2);
1489 /* fd1 and fd2 may be already another descriptors.
1490 * Not kernel problem.
1506 sock_release(sock2);
1509 out_put_unused_both:
1514 sock_release(sock2);
1516 sock_release(sock1);
1522 * Bind a name to a socket. Nothing much to do here since it's
1523 * the protocol's responsibility to handle the local address.
1525 * We move the socket address to kernel space before we call
1526 * the protocol layer (having also checked the address is ok).
1529 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1531 struct socket *sock;
1532 struct sockaddr_storage address;
1533 int err, fput_needed;
1535 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1537 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1539 err = security_socket_bind(sock,
1540 (struct sockaddr *)&address,
1543 err = sock->ops->bind(sock,
1547 fput_light(sock->file, fput_needed);
1553 * Perform a listen. Basically, we allow the protocol to do anything
1554 * necessary for a listen, and if that works, we mark the socket as
1555 * ready for listening.
1558 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1560 struct socket *sock;
1561 int err, fput_needed;
1564 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1566 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1567 if ((unsigned int)backlog > somaxconn)
1568 backlog = somaxconn;
1570 err = security_socket_listen(sock, backlog);
1572 err = sock->ops->listen(sock, backlog);
1574 fput_light(sock->file, fput_needed);
1580 * For accept, we attempt to create a new socket, set up the link
1581 * with the client, wake up the client, then return the new
1582 * connected fd. We collect the address of the connector in kernel
1583 * space and move it to user at the very end. This is unclean because
1584 * we open the socket then return an error.
1586 * 1003.1g adds the ability to recvmsg() to query connection pending
1587 * status to recvmsg. We need to add that support in a way thats
1588 * clean when we restucture accept also.
1591 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1592 int __user *, upeer_addrlen, int, flags)
1594 struct socket *sock, *newsock;
1595 struct file *newfile;
1596 int err, len, newfd, fput_needed;
1597 struct sockaddr_storage address;
1599 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1602 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1603 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1605 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1610 newsock = sock_alloc();
1614 newsock->type = sock->type;
1615 newsock->ops = sock->ops;
1618 * We don't need try_module_get here, as the listening socket (sock)
1619 * has the protocol module (sock->ops->owner) held.
1621 __module_get(newsock->ops->owner);
1623 newfd = get_unused_fd_flags(flags);
1624 if (unlikely(newfd < 0)) {
1626 sock_release(newsock);
1629 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1630 if (unlikely(IS_ERR(newfile))) {
1631 err = PTR_ERR(newfile);
1632 put_unused_fd(newfd);
1633 sock_release(newsock);
1637 err = security_socket_accept(sock, newsock);
1641 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1645 if (upeer_sockaddr) {
1646 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1648 err = -ECONNABORTED;
1651 err = move_addr_to_user(&address,
1652 len, upeer_sockaddr, upeer_addrlen);
1657 /* File flags are not inherited via accept() unlike another OSes. */
1659 fd_install(newfd, newfile);
1663 fput_light(sock->file, fput_needed);
1668 put_unused_fd(newfd);
1672 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1673 int __user *, upeer_addrlen)
1675 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1679 * Attempt to connect to a socket with the server address. The address
1680 * is in user space so we verify it is OK and move it to kernel space.
1682 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1685 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1686 * other SEQPACKET protocols that take time to connect() as it doesn't
1687 * include the -EINPROGRESS status for such sockets.
1690 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1693 struct socket *sock;
1694 struct sockaddr_storage address;
1695 int err, fput_needed;
1697 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1700 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1705 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1709 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1710 sock->file->f_flags);
1712 fput_light(sock->file, fput_needed);
1718 * Get the local address ('name') of a socket object. Move the obtained
1719 * name to user space.
1722 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1723 int __user *, usockaddr_len)
1725 struct socket *sock;
1726 struct sockaddr_storage address;
1727 int len, err, fput_needed;
1729 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733 err = security_socket_getsockname(sock);
1737 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1740 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1743 fput_light(sock->file, fput_needed);
1749 * Get the remote address ('name') of a socket object. Move the obtained
1750 * name to user space.
1753 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1754 int __user *, usockaddr_len)
1756 struct socket *sock;
1757 struct sockaddr_storage address;
1758 int len, err, fput_needed;
1760 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1762 err = security_socket_getpeername(sock);
1764 fput_light(sock->file, fput_needed);
1769 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1772 err = move_addr_to_user(&address, len, usockaddr,
1774 fput_light(sock->file, fput_needed);
1780 * Send a datagram to a given address. We move the address into kernel
1781 * space and check the user space data area is readable before invoking
1785 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1786 unsigned int, flags, struct sockaddr __user *, addr,
1789 struct socket *sock;
1790 struct sockaddr_storage address;
1798 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1802 iov.iov_base = buff;
1804 msg.msg_name = NULL;
1807 msg.msg_control = NULL;
1808 msg.msg_controllen = 0;
1809 msg.msg_namelen = 0;
1811 err = move_addr_to_kernel(addr, addr_len, &address);
1814 msg.msg_name = (struct sockaddr *)&address;
1815 msg.msg_namelen = addr_len;
1817 if (sock->file->f_flags & O_NONBLOCK)
1818 flags |= MSG_DONTWAIT;
1819 msg.msg_flags = flags;
1820 err = sock_sendmsg(sock, &msg, len);
1823 fput_light(sock->file, fput_needed);
1829 * Send a datagram down a socket.
1832 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1833 unsigned int, flags)
1835 return sys_sendto(fd, buff, len, flags, NULL, 0);
1839 * Receive a frame from the socket and optionally record the address of the
1840 * sender. We verify the buffers are writable and if needed move the
1841 * sender address from kernel to user space.
1844 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1845 unsigned int, flags, struct sockaddr __user *, addr,
1846 int __user *, addr_len)
1848 struct socket *sock;
1851 struct sockaddr_storage address;
1857 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1861 msg.msg_control = NULL;
1862 msg.msg_controllen = 0;
1866 iov.iov_base = ubuf;
1867 /* Save some cycles and don't copy the address if not needed */
1868 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1869 /* We assume all kernel code knows the size of sockaddr_storage */
1870 msg.msg_namelen = 0;
1871 if (sock->file->f_flags & O_NONBLOCK)
1872 flags |= MSG_DONTWAIT;
1873 err = sock_recvmsg(sock, &msg, size, flags);
1875 if (err >= 0 && addr != NULL) {
1876 err2 = move_addr_to_user(&address,
1877 msg.msg_namelen, addr, addr_len);
1882 fput_light(sock->file, fput_needed);
1888 * Receive a datagram from a socket.
1891 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1892 unsigned int, flags)
1894 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1898 * Set a socket option. Because we don't know the option lengths we have
1899 * to pass the user mode parameter for the protocols to sort out.
1902 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1903 char __user *, optval, int, optlen)
1905 int err, fput_needed;
1906 struct socket *sock;
1911 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1913 err = security_socket_setsockopt(sock, level, optname);
1917 if (level == SOL_SOCKET)
1919 sock_setsockopt(sock, level, optname, optval,
1923 sock->ops->setsockopt(sock, level, optname, optval,
1926 fput_light(sock->file, fput_needed);
1932 * Get a socket option. Because we don't know the option lengths we have
1933 * to pass a user mode parameter for the protocols to sort out.
1936 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1937 char __user *, optval, int __user *, optlen)
1939 int err, fput_needed;
1940 struct socket *sock;
1942 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1944 err = security_socket_getsockopt(sock, level, optname);
1948 if (level == SOL_SOCKET)
1950 sock_getsockopt(sock, level, optname, optval,
1954 sock->ops->getsockopt(sock, level, optname, optval,
1957 fput_light(sock->file, fput_needed);
1963 * Shutdown a socket.
1966 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1968 int err, fput_needed;
1969 struct socket *sock;
1971 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1973 err = security_socket_shutdown(sock, how);
1975 err = sock->ops->shutdown(sock, how);
1976 fput_light(sock->file, fput_needed);
1981 /* A couple of helpful macros for getting the address of the 32/64 bit
1982 * fields which are the same type (int / unsigned) on our platforms.
1984 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1985 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1986 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1988 struct used_address {
1989 struct sockaddr_storage name;
1990 unsigned int name_len;
1993 static int copy_msghdr_from_user(struct msghdr *kmsg,
1994 struct msghdr __user *umsg)
1996 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1999 if (kmsg->msg_name == NULL)
2000 kmsg->msg_namelen = 0;
2002 if (kmsg->msg_namelen < 0)
2005 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2006 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2010 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
2011 struct msghdr *msg_sys, unsigned int flags,
2012 struct used_address *used_address)
2014 struct compat_msghdr __user *msg_compat =
2015 (struct compat_msghdr __user *)msg;
2016 struct sockaddr_storage address;
2017 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2018 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2019 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2020 /* 20 is size of ipv6_pktinfo */
2021 unsigned char *ctl_buf = ctl;
2022 int err, ctl_len, total_len;
2025 if (MSG_CMSG_COMPAT & flags) {
2026 if (get_compat_msghdr(msg_sys, msg_compat))
2029 err = copy_msghdr_from_user(msg_sys, msg);
2034 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2036 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2039 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2045 /* This will also move the address data into kernel space */
2046 if (MSG_CMSG_COMPAT & flags) {
2047 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2049 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2056 if (msg_sys->msg_controllen > INT_MAX)
2058 ctl_len = msg_sys->msg_controllen;
2059 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2061 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2065 ctl_buf = msg_sys->msg_control;
2066 ctl_len = msg_sys->msg_controllen;
2067 } else if (ctl_len) {
2068 if (ctl_len > sizeof(ctl)) {
2069 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2070 if (ctl_buf == NULL)
2075 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2076 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2077 * checking falls down on this.
2079 if (copy_from_user(ctl_buf,
2080 (void __user __force *)msg_sys->msg_control,
2083 msg_sys->msg_control = ctl_buf;
2085 msg_sys->msg_flags = flags;
2087 if (sock->file->f_flags & O_NONBLOCK)
2088 msg_sys->msg_flags |= MSG_DONTWAIT;
2090 * If this is sendmmsg() and current destination address is same as
2091 * previously succeeded address, omit asking LSM's decision.
2092 * used_address->name_len is initialized to UINT_MAX so that the first
2093 * destination address never matches.
2095 if (used_address && msg_sys->msg_name &&
2096 used_address->name_len == msg_sys->msg_namelen &&
2097 !memcmp(&used_address->name, msg_sys->msg_name,
2098 used_address->name_len)) {
2099 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2102 err = sock_sendmsg(sock, msg_sys, total_len);
2104 * If this is sendmmsg() and sending to current destination address was
2105 * successful, remember it.
2107 if (used_address && err >= 0) {
2108 used_address->name_len = msg_sys->msg_namelen;
2109 if (msg_sys->msg_name)
2110 memcpy(&used_address->name, msg_sys->msg_name,
2111 used_address->name_len);
2116 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2118 if (iov != iovstack)
2125 * BSD sendmsg interface
2128 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2130 int fput_needed, err;
2131 struct msghdr msg_sys;
2132 struct socket *sock;
2134 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2138 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2140 fput_light(sock->file, fput_needed);
2145 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2147 if (flags & MSG_CMSG_COMPAT)
2149 return __sys_sendmsg(fd, msg, flags);
2153 * Linux sendmmsg interface
2156 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2159 int fput_needed, err, datagrams;
2160 struct socket *sock;
2161 struct mmsghdr __user *entry;
2162 struct compat_mmsghdr __user *compat_entry;
2163 struct msghdr msg_sys;
2164 struct used_address used_address;
2166 if (vlen > UIO_MAXIOV)
2171 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2175 used_address.name_len = UINT_MAX;
2177 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2180 while (datagrams < vlen) {
2181 if (MSG_CMSG_COMPAT & flags) {
2182 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2183 &msg_sys, flags, &used_address);
2186 err = __put_user(err, &compat_entry->msg_len);
2189 err = ___sys_sendmsg(sock,
2190 (struct msghdr __user *)entry,
2191 &msg_sys, flags, &used_address);
2194 err = put_user(err, &entry->msg_len);
2203 fput_light(sock->file, fput_needed);
2205 /* We only return an error if no datagrams were able to be sent */
2212 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2213 unsigned int, vlen, unsigned int, flags)
2215 if (flags & MSG_CMSG_COMPAT)
2217 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2220 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2221 struct msghdr *msg_sys, unsigned int flags, int nosec)
2223 struct compat_msghdr __user *msg_compat =
2224 (struct compat_msghdr __user *)msg;
2225 struct iovec iovstack[UIO_FASTIOV];
2226 struct iovec *iov = iovstack;
2227 unsigned long cmsg_ptr;
2228 int err, total_len, len;
2230 /* kernel mode address */
2231 struct sockaddr_storage addr;
2233 /* user mode address pointers */
2234 struct sockaddr __user *uaddr;
2235 int __user *uaddr_len;
2237 if (MSG_CMSG_COMPAT & flags) {
2238 if (get_compat_msghdr(msg_sys, msg_compat))
2241 err = copy_msghdr_from_user(msg_sys, msg);
2246 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2248 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2251 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2257 /* Save the user-mode address (verify_iovec will change the
2258 * kernel msghdr to use the kernel address space)
2260 uaddr = (__force void __user *)msg_sys->msg_name;
2261 uaddr_len = COMPAT_NAMELEN(msg);
2262 if (MSG_CMSG_COMPAT & flags)
2263 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2265 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2270 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2271 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2273 /* We assume all kernel code knows the size of sockaddr_storage */
2274 msg_sys->msg_namelen = 0;
2276 if (sock->file->f_flags & O_NONBLOCK)
2277 flags |= MSG_DONTWAIT;
2278 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2284 if (uaddr != NULL) {
2285 err = move_addr_to_user(&addr,
2286 msg_sys->msg_namelen, uaddr,
2291 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2295 if (MSG_CMSG_COMPAT & flags)
2296 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2297 &msg_compat->msg_controllen);
2299 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2300 &msg->msg_controllen);
2306 if (iov != iovstack)
2313 * BSD recvmsg interface
2316 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2318 int fput_needed, err;
2319 struct msghdr msg_sys;
2320 struct socket *sock;
2322 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2326 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2328 fput_light(sock->file, fput_needed);
2333 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2334 unsigned int, flags)
2336 if (flags & MSG_CMSG_COMPAT)
2338 return __sys_recvmsg(fd, msg, flags);
2342 * Linux recvmmsg interface
2345 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2346 unsigned int flags, struct timespec *timeout)
2348 int fput_needed, err, datagrams;
2349 struct socket *sock;
2350 struct mmsghdr __user *entry;
2351 struct compat_mmsghdr __user *compat_entry;
2352 struct msghdr msg_sys;
2353 struct timespec end_time;
2356 poll_select_set_timeout(&end_time, timeout->tv_sec,
2362 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2366 err = sock_error(sock->sk);
2371 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2373 while (datagrams < vlen) {
2375 * No need to ask LSM for more than the first datagram.
2377 if (MSG_CMSG_COMPAT & flags) {
2378 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2379 &msg_sys, flags & ~MSG_WAITFORONE,
2383 err = __put_user(err, &compat_entry->msg_len);
2386 err = ___sys_recvmsg(sock,
2387 (struct msghdr __user *)entry,
2388 &msg_sys, flags & ~MSG_WAITFORONE,
2392 err = put_user(err, &entry->msg_len);
2400 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2401 if (flags & MSG_WAITFORONE)
2402 flags |= MSG_DONTWAIT;
2405 ktime_get_ts(timeout);
2406 *timeout = timespec_sub(end_time, *timeout);
2407 if (timeout->tv_sec < 0) {
2408 timeout->tv_sec = timeout->tv_nsec = 0;
2412 /* Timeout, return less than vlen datagrams */
2413 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2417 /* Out of band data, return right away */
2418 if (msg_sys.msg_flags & MSG_OOB)
2423 fput_light(sock->file, fput_needed);
2428 if (datagrams != 0) {
2430 * We may return less entries than requested (vlen) if the
2431 * sock is non block and there aren't enough datagrams...
2433 if (err != -EAGAIN) {
2435 * ... or if recvmsg returns an error after we
2436 * received some datagrams, where we record the
2437 * error to return on the next call or if the
2438 * app asks about it using getsockopt(SO_ERROR).
2440 sock->sk->sk_err = -err;
2449 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2450 unsigned int, vlen, unsigned int, flags,
2451 struct timespec __user *, timeout)
2454 struct timespec timeout_sys;
2456 if (flags & MSG_CMSG_COMPAT)
2460 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2462 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2465 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2467 if (datagrams > 0 &&
2468 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2469 datagrams = -EFAULT;
2474 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2475 /* Argument list sizes for sys_socketcall */
2476 #define AL(x) ((x) * sizeof(unsigned long))
2477 static const unsigned char nargs[21] = {
2478 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2479 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2480 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2487 * System call vectors.
2489 * Argument checking cleaned up. Saved 20% in size.
2490 * This function doesn't need to set the kernel lock because
2491 * it is set by the callees.
2494 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2496 unsigned long a[AUDITSC_ARGS];
2497 unsigned long a0, a1;
2501 if (call < 1 || call > SYS_SENDMMSG)
2505 if (len > sizeof(a))
2508 /* copy_from_user should be SMP safe. */
2509 if (copy_from_user(a, args, len))
2512 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2521 err = sys_socket(a0, a1, a[2]);
2524 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2527 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2530 err = sys_listen(a0, a1);
2533 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2534 (int __user *)a[2], 0);
2536 case SYS_GETSOCKNAME:
2538 sys_getsockname(a0, (struct sockaddr __user *)a1,
2539 (int __user *)a[2]);
2541 case SYS_GETPEERNAME:
2543 sys_getpeername(a0, (struct sockaddr __user *)a1,
2544 (int __user *)a[2]);
2546 case SYS_SOCKETPAIR:
2547 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2550 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2553 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2554 (struct sockaddr __user *)a[4], a[5]);
2557 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2560 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2561 (struct sockaddr __user *)a[4],
2562 (int __user *)a[5]);
2565 err = sys_shutdown(a0, a1);
2567 case SYS_SETSOCKOPT:
2568 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2570 case SYS_GETSOCKOPT:
2572 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2573 (int __user *)a[4]);
2576 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2579 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2582 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2585 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2586 (struct timespec __user *)a[4]);
2589 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2590 (int __user *)a[2], a[3]);
2599 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2602 * sock_register - add a socket protocol handler
2603 * @ops: description of protocol
2605 * This function is called by a protocol handler that wants to
2606 * advertise its address family, and have it linked into the
2607 * socket interface. The value ops->family corresponds to the
2608 * socket system call protocol family.
2610 int sock_register(const struct net_proto_family *ops)
2614 if (ops->family >= NPROTO) {
2615 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2619 spin_lock(&net_family_lock);
2620 if (rcu_dereference_protected(net_families[ops->family],
2621 lockdep_is_held(&net_family_lock)))
2624 rcu_assign_pointer(net_families[ops->family], ops);
2627 spin_unlock(&net_family_lock);
2629 pr_info("NET: Registered protocol family %d\n", ops->family);
2632 EXPORT_SYMBOL(sock_register);
2635 * sock_unregister - remove a protocol handler
2636 * @family: protocol family to remove
2638 * This function is called by a protocol handler that wants to
2639 * remove its address family, and have it unlinked from the
2640 * new socket creation.
2642 * If protocol handler is a module, then it can use module reference
2643 * counts to protect against new references. If protocol handler is not
2644 * a module then it needs to provide its own protection in
2645 * the ops->create routine.
2647 void sock_unregister(int family)
2649 BUG_ON(family < 0 || family >= NPROTO);
2651 spin_lock(&net_family_lock);
2652 RCU_INIT_POINTER(net_families[family], NULL);
2653 spin_unlock(&net_family_lock);
2657 pr_info("NET: Unregistered protocol family %d\n", family);
2659 EXPORT_SYMBOL(sock_unregister);
2661 static int __init sock_init(void)
2665 * Initialize the network sysctl infrastructure.
2667 err = net_sysctl_init();
2672 * Initialize skbuff SLAB cache
2677 * Initialize the protocols module.
2682 err = register_filesystem(&sock_fs_type);
2685 sock_mnt = kern_mount(&sock_fs_type);
2686 if (IS_ERR(sock_mnt)) {
2687 err = PTR_ERR(sock_mnt);
2691 /* The real protocol initialization is performed in later initcalls.
2694 #ifdef CONFIG_NETFILTER
2695 err = netfilter_init();
2700 ptp_classifier_init();
2706 unregister_filesystem(&sock_fs_type);
2711 core_initcall(sock_init); /* early initcall */
2713 #ifdef CONFIG_PROC_FS
2714 void socket_seq_show(struct seq_file *seq)
2719 for_each_possible_cpu(cpu)
2720 counter += per_cpu(sockets_in_use, cpu);
2722 /* It can be negative, by the way. 8) */
2726 seq_printf(seq, "sockets: used %d\n", counter);
2728 #endif /* CONFIG_PROC_FS */
2730 #ifdef CONFIG_COMPAT
2731 static int do_siocgstamp(struct net *net, struct socket *sock,
2732 unsigned int cmd, void __user *up)
2734 mm_segment_t old_fs = get_fs();
2739 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2742 err = compat_put_timeval(&ktv, up);
2747 static int do_siocgstampns(struct net *net, struct socket *sock,
2748 unsigned int cmd, void __user *up)
2750 mm_segment_t old_fs = get_fs();
2751 struct timespec kts;
2755 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2758 err = compat_put_timespec(&kts, up);
2763 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2765 struct ifreq __user *uifr;
2768 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2769 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2772 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2776 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2782 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2784 struct compat_ifconf ifc32;
2786 struct ifconf __user *uifc;
2787 struct compat_ifreq __user *ifr32;
2788 struct ifreq __user *ifr;
2792 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2795 memset(&ifc, 0, sizeof(ifc));
2796 if (ifc32.ifcbuf == 0) {
2800 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2802 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2803 sizeof(struct ifreq);
2804 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2806 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2807 ifr32 = compat_ptr(ifc32.ifcbuf);
2808 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2809 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2815 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2818 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2822 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2826 ifr32 = compat_ptr(ifc32.ifcbuf);
2828 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2829 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2830 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2836 if (ifc32.ifcbuf == 0) {
2837 /* Translate from 64-bit structure multiple to
2841 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2846 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2852 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2854 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2855 bool convert_in = false, convert_out = false;
2856 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2857 struct ethtool_rxnfc __user *rxnfc;
2858 struct ifreq __user *ifr;
2859 u32 rule_cnt = 0, actual_rule_cnt;
2864 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2867 compat_rxnfc = compat_ptr(data);
2869 if (get_user(ethcmd, &compat_rxnfc->cmd))
2872 /* Most ethtool structures are defined without padding.
2873 * Unfortunately struct ethtool_rxnfc is an exception.
2878 case ETHTOOL_GRXCLSRLALL:
2879 /* Buffer size is variable */
2880 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2882 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2884 buf_size += rule_cnt * sizeof(u32);
2886 case ETHTOOL_GRXRINGS:
2887 case ETHTOOL_GRXCLSRLCNT:
2888 case ETHTOOL_GRXCLSRULE:
2889 case ETHTOOL_SRXCLSRLINS:
2892 case ETHTOOL_SRXCLSRLDEL:
2893 buf_size += sizeof(struct ethtool_rxnfc);
2898 ifr = compat_alloc_user_space(buf_size);
2899 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2901 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2904 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2905 &ifr->ifr_ifru.ifru_data))
2909 /* We expect there to be holes between fs.m_ext and
2910 * fs.ring_cookie and at the end of fs, but nowhere else.
2912 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2913 sizeof(compat_rxnfc->fs.m_ext) !=
2914 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2915 sizeof(rxnfc->fs.m_ext));
2917 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2918 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2919 offsetof(struct ethtool_rxnfc, fs.location) -
2920 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2922 if (copy_in_user(rxnfc, compat_rxnfc,
2923 (void __user *)(&rxnfc->fs.m_ext + 1) -
2924 (void __user *)rxnfc) ||
2925 copy_in_user(&rxnfc->fs.ring_cookie,
2926 &compat_rxnfc->fs.ring_cookie,
2927 (void __user *)(&rxnfc->fs.location + 1) -
2928 (void __user *)&rxnfc->fs.ring_cookie) ||
2929 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2930 sizeof(rxnfc->rule_cnt)))
2934 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2939 if (copy_in_user(compat_rxnfc, rxnfc,
2940 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2941 (const void __user *)rxnfc) ||
2942 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2943 &rxnfc->fs.ring_cookie,
2944 (const void __user *)(&rxnfc->fs.location + 1) -
2945 (const void __user *)&rxnfc->fs.ring_cookie) ||
2946 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2947 sizeof(rxnfc->rule_cnt)))
2950 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2951 /* As an optimisation, we only copy the actual
2952 * number of rules that the underlying
2953 * function returned. Since Mallory might
2954 * change the rule count in user memory, we
2955 * check that it is less than the rule count
2956 * originally given (as the user buffer size),
2957 * which has been range-checked.
2959 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2961 if (actual_rule_cnt < rule_cnt)
2962 rule_cnt = actual_rule_cnt;
2963 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2964 &rxnfc->rule_locs[0],
2965 rule_cnt * sizeof(u32)))
2973 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2976 compat_uptr_t uptr32;
2977 struct ifreq __user *uifr;
2979 uifr = compat_alloc_user_space(sizeof(*uifr));
2980 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2983 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2986 uptr = compat_ptr(uptr32);
2988 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2991 return dev_ioctl(net, SIOCWANDEV, uifr);
2994 static int bond_ioctl(struct net *net, unsigned int cmd,
2995 struct compat_ifreq __user *ifr32)
2998 mm_segment_t old_fs;
3002 case SIOCBONDENSLAVE:
3003 case SIOCBONDRELEASE:
3004 case SIOCBONDSETHWADDR:
3005 case SIOCBONDCHANGEACTIVE:
3006 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
3011 err = dev_ioctl(net, cmd,
3012 (struct ifreq __user __force *) &kifr);
3017 return -ENOIOCTLCMD;
3021 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3022 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3023 struct compat_ifreq __user *u_ifreq32)
3025 struct ifreq __user *u_ifreq64;
3026 char tmp_buf[IFNAMSIZ];
3027 void __user *data64;
3030 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3033 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3035 data64 = compat_ptr(data32);
3037 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3039 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3042 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3045 return dev_ioctl(net, cmd, u_ifreq64);
3048 static int dev_ifsioc(struct net *net, struct socket *sock,
3049 unsigned int cmd, struct compat_ifreq __user *uifr32)
3051 struct ifreq __user *uifr;
3054 uifr = compat_alloc_user_space(sizeof(*uifr));
3055 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3058 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3069 case SIOCGIFBRDADDR:
3070 case SIOCGIFDSTADDR:
3071 case SIOCGIFNETMASK:
3076 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3084 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3085 struct compat_ifreq __user *uifr32)
3088 struct compat_ifmap __user *uifmap32;
3089 mm_segment_t old_fs;
3092 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3093 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3094 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3095 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3096 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3097 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3098 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3099 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3105 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3108 if (cmd == SIOCGIFMAP && !err) {
3109 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3110 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3111 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3112 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3113 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3114 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3115 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3124 struct sockaddr rt_dst; /* target address */
3125 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3126 struct sockaddr rt_genmask; /* target network mask (IP) */
3127 unsigned short rt_flags;
3130 unsigned char rt_tos;
3131 unsigned char rt_class;
3133 short rt_metric; /* +1 for binary compatibility! */
3134 /* char * */ u32 rt_dev; /* forcing the device at add */
3135 u32 rt_mtu; /* per route MTU/Window */
3136 u32 rt_window; /* Window clamping */
3137 unsigned short rt_irtt; /* Initial RTT */
3140 struct in6_rtmsg32 {
3141 struct in6_addr rtmsg_dst;
3142 struct in6_addr rtmsg_src;
3143 struct in6_addr rtmsg_gateway;
3153 static int routing_ioctl(struct net *net, struct socket *sock,
3154 unsigned int cmd, void __user *argp)
3158 struct in6_rtmsg r6;
3162 mm_segment_t old_fs = get_fs();
3164 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3165 struct in6_rtmsg32 __user *ur6 = argp;
3166 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3167 3 * sizeof(struct in6_addr));
3168 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3169 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3170 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3171 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3172 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3173 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3174 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3178 struct rtentry32 __user *ur4 = argp;
3179 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3180 3 * sizeof(struct sockaddr));
3181 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3182 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3183 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3184 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3185 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3186 ret |= get_user(rtdev, &(ur4->rt_dev));
3188 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3189 r4.rt_dev = (char __user __force *)devname;
3203 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3210 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3211 * for some operations; this forces use of the newer bridge-utils that
3212 * use compatible ioctls
3214 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3218 if (get_user(tmp, argp))
3220 if (tmp == BRCTL_GET_VERSION)
3221 return BRCTL_VERSION + 1;
3225 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3226 unsigned int cmd, unsigned long arg)
3228 void __user *argp = compat_ptr(arg);
3229 struct sock *sk = sock->sk;
3230 struct net *net = sock_net(sk);
3232 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3233 return compat_ifr_data_ioctl(net, cmd, argp);
3238 return old_bridge_ioctl(argp);
3240 return dev_ifname32(net, argp);
3242 return dev_ifconf(net, argp);
3244 return ethtool_ioctl(net, argp);
3246 return compat_siocwandev(net, argp);
3249 return compat_sioc_ifmap(net, cmd, argp);
3250 case SIOCBONDENSLAVE:
3251 case SIOCBONDRELEASE:
3252 case SIOCBONDSETHWADDR:
3253 case SIOCBONDCHANGEACTIVE:
3254 return bond_ioctl(net, cmd, argp);
3257 return routing_ioctl(net, sock, cmd, argp);
3259 return do_siocgstamp(net, sock, cmd, argp);
3261 return do_siocgstampns(net, sock, cmd, argp);
3262 case SIOCBONDSLAVEINFOQUERY:
3263 case SIOCBONDINFOQUERY:
3266 return compat_ifr_data_ioctl(net, cmd, argp);
3278 return sock_ioctl(file, cmd, arg);
3295 case SIOCSIFHWBROADCAST:
3297 case SIOCGIFBRDADDR:
3298 case SIOCSIFBRDADDR:
3299 case SIOCGIFDSTADDR:
3300 case SIOCSIFDSTADDR:
3301 case SIOCGIFNETMASK:
3302 case SIOCSIFNETMASK:
3313 return dev_ifsioc(net, sock, cmd, argp);
3319 return sock_do_ioctl(net, sock, cmd, arg);
3322 return -ENOIOCTLCMD;
3325 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3328 struct socket *sock = file->private_data;
3329 int ret = -ENOIOCTLCMD;
3336 if (sock->ops->compat_ioctl)
3337 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3339 if (ret == -ENOIOCTLCMD &&
3340 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3341 ret = compat_wext_handle_ioctl(net, cmd, arg);
3343 if (ret == -ENOIOCTLCMD)
3344 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3350 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3352 return sock->ops->bind(sock, addr, addrlen);
3354 EXPORT_SYMBOL(kernel_bind);
3356 int kernel_listen(struct socket *sock, int backlog)
3358 return sock->ops->listen(sock, backlog);
3360 EXPORT_SYMBOL(kernel_listen);
3362 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3364 struct sock *sk = sock->sk;
3367 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3372 err = sock->ops->accept(sock, *newsock, flags);
3374 sock_release(*newsock);
3379 (*newsock)->ops = sock->ops;
3380 __module_get((*newsock)->ops->owner);
3385 EXPORT_SYMBOL(kernel_accept);
3387 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3390 return sock->ops->connect(sock, addr, addrlen, flags);
3392 EXPORT_SYMBOL(kernel_connect);
3394 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3397 return sock->ops->getname(sock, addr, addrlen, 0);
3399 EXPORT_SYMBOL(kernel_getsockname);
3401 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3404 return sock->ops->getname(sock, addr, addrlen, 1);
3406 EXPORT_SYMBOL(kernel_getpeername);
3408 int kernel_getsockopt(struct socket *sock, int level, int optname,
3409 char *optval, int *optlen)
3411 mm_segment_t oldfs = get_fs();
3412 char __user *uoptval;
3413 int __user *uoptlen;
3416 uoptval = (char __user __force *) optval;
3417 uoptlen = (int __user __force *) optlen;
3420 if (level == SOL_SOCKET)
3421 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3423 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3428 EXPORT_SYMBOL(kernel_getsockopt);
3430 int kernel_setsockopt(struct socket *sock, int level, int optname,
3431 char *optval, unsigned int optlen)
3433 mm_segment_t oldfs = get_fs();
3434 char __user *uoptval;
3437 uoptval = (char __user __force *) optval;
3440 if (level == SOL_SOCKET)
3441 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3443 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3448 EXPORT_SYMBOL(kernel_setsockopt);
3450 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3451 size_t size, int flags)
3453 if (sock->ops->sendpage)
3454 return sock->ops->sendpage(sock, page, offset, size, flags);
3456 return sock_no_sendpage(sock, page, offset, size, flags);
3458 EXPORT_SYMBOL(kernel_sendpage);
3460 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3462 mm_segment_t oldfs = get_fs();
3466 err = sock->ops->ioctl(sock, cmd, arg);
3471 EXPORT_SYMBOL(kernel_sock_ioctl);
3473 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3475 return sock->ops->shutdown(sock, how);
3477 EXPORT_SYMBOL(kernel_sock_shutdown);