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/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.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>
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
108 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
109 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
113 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
115 static int sock_close(struct inode *inode, struct file *file);
116 static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
118 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
120 static long compat_sock_ioctl(struct file *file,
121 unsigned int cmd, unsigned long arg);
123 static int sock_fasync(int fd, struct file *filp, int on);
124 static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
126 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
127 struct pipe_inode_info *pipe, size_t len,
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
135 static const struct file_operations socket_file_ops = {
136 .owner = THIS_MODULE,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
141 .unlocked_ioctl = sock_ioctl,
143 .compat_ioctl = compat_sock_ioctl,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
151 .splice_read = sock_splice_read,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock);
159 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
162 * Statistics counters of the socket lists
165 static DEFINE_PER_CPU(int, sockets_in_use);
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
190 if (copy_from_user(kaddr, uaddr, ulen))
192 return audit_sockaddr(ulen, kaddr);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 static int move_addr_to_user(struct sockaddr *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
218 err = get_user(len, ulen);
223 if (len < 0 || len > sizeof(struct sockaddr_storage))
226 if (audit_sockaddr(klen, kaddr))
228 if (copy_to_user(uaddr, kaddr, len))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen, ulen);
238 static struct kmem_cache *sock_inode_cachep __read_mostly;
240 static struct inode *sock_alloc_inode(struct super_block *sb)
242 struct socket_alloc *ei;
243 struct socket_wq *wq;
245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
250 kmem_cache_free(sock_inode_cachep, ei);
253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 struct socket_alloc *ei;
269 struct socket_wq *wq;
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
272 wq = rcu_dereference_protected(ei->socket.wq, 1);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static int init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
293 if (sock_inode_cachep == NULL)
298 static const struct super_operations sockfs_ops = {
299 .alloc_inode = sock_alloc_inode,
300 .destroy_inode = sock_destroy_inode,
301 .statfs = simple_statfs,
305 * sockfs_dname() is called from d_path().
307 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
309 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310 dentry->d_inode->i_ino);
313 static const struct dentry_operations sockfs_dentry_operations = {
314 .d_dname = sockfs_dname,
317 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318 int flags, const char *dev_name, void *data)
320 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321 &sockfs_dentry_operations, SOCKFS_MAGIC);
324 static struct vfsmount *sock_mnt __read_mostly;
326 static struct file_system_type sock_fs_type = {
328 .mount = sockfs_mount,
329 .kill_sb = kill_anon_super,
333 * Obtains the first available file descriptor and sets it up for use.
335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
349 static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
351 struct qstr name = { .name = "" };
356 fd = get_unused_fd_flags(flags);
357 if (unlikely(fd < 0))
360 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
361 if (unlikely(!path.dentry)) {
365 path.mnt = mntget(sock_mnt);
367 d_instantiate(path.dentry, SOCK_INODE(sock));
368 SOCK_INODE(sock)->i_fop = &socket_file_ops;
370 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
372 if (unlikely(!file)) {
373 /* drop dentry, keep inode */
374 ihold(path.dentry->d_inode);
381 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
383 file->private_data = sock;
389 int sock_map_fd(struct socket *sock, int flags)
391 struct file *newfile;
392 int fd = sock_alloc_file(sock, &newfile, flags);
395 fd_install(fd, newfile);
399 EXPORT_SYMBOL(sock_map_fd);
401 static struct socket *sock_from_file(struct file *file, int *err)
403 if (file->f_op == &socket_file_ops)
404 return file->private_data; /* set in sock_map_fd */
411 * sockfd_lookup - Go from a file number to its socket slot
413 * @err: pointer to an error code return
415 * The file handle passed in is locked and the socket it is bound
416 * too is returned. If an error occurs the err pointer is overwritten
417 * with a negative errno code and NULL is returned. The function checks
418 * for both invalid handles and passing a handle which is not a socket.
420 * On a success the socket object pointer is returned.
423 struct socket *sockfd_lookup(int fd, int *err)
434 sock = sock_from_file(file, err);
439 EXPORT_SYMBOL(sockfd_lookup);
441 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
447 file = fget_light(fd, fput_needed);
449 sock = sock_from_file(file, err);
452 fput_light(file, *fput_needed);
458 * sock_alloc - allocate a socket
460 * Allocate a new inode and socket object. The two are bound together
461 * and initialised. The socket is then returned. If we are out of inodes
465 static struct socket *sock_alloc(void)
470 inode = new_inode_pseudo(sock_mnt->mnt_sb);
474 sock = SOCKET_I(inode);
476 kmemcheck_annotate_bitfield(sock, type);
477 inode->i_ino = get_next_ino();
478 inode->i_mode = S_IFSOCK | S_IRWXUGO;
479 inode->i_uid = current_fsuid();
480 inode->i_gid = current_fsgid();
482 percpu_add(sockets_in_use, 1);
487 * In theory you can't get an open on this inode, but /proc provides
488 * a back door. Remember to keep it shut otherwise you'll let the
489 * creepy crawlies in.
492 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
497 const struct file_operations bad_sock_fops = {
498 .owner = THIS_MODULE,
499 .open = sock_no_open,
500 .llseek = noop_llseek,
504 * sock_release - close a socket
505 * @sock: socket to close
507 * The socket is released from the protocol stack if it has a release
508 * callback, and the inode is then released if the socket is bound to
509 * an inode not a file.
512 void sock_release(struct socket *sock)
515 struct module *owner = sock->ops->owner;
517 sock->ops->release(sock);
522 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
523 printk(KERN_ERR "sock_release: fasync list not empty!\n");
525 percpu_sub(sockets_in_use, 1);
527 iput(SOCK_INODE(sock));
532 EXPORT_SYMBOL(sock_release);
534 int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
537 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
538 *tx_flags |= SKBTX_HW_TSTAMP;
539 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
540 *tx_flags |= SKBTX_SW_TSTAMP;
543 EXPORT_SYMBOL(sock_tx_timestamp);
545 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
546 struct msghdr *msg, size_t size)
548 struct sock_iocb *si = kiocb_to_siocb(iocb);
550 sock_update_classid(sock->sk);
557 return sock->ops->sendmsg(iocb, sock, msg, size);
560 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
561 struct msghdr *msg, size_t size)
563 int err = security_socket_sendmsg(sock, msg, size);
565 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
568 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
571 struct sock_iocb siocb;
574 init_sync_kiocb(&iocb, NULL);
575 iocb.private = &siocb;
576 ret = __sock_sendmsg(&iocb, sock, msg, size);
577 if (-EIOCBQUEUED == ret)
578 ret = wait_on_sync_kiocb(&iocb);
581 EXPORT_SYMBOL(sock_sendmsg);
583 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
586 struct sock_iocb siocb;
589 init_sync_kiocb(&iocb, NULL);
590 iocb.private = &siocb;
591 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
592 if (-EIOCBQUEUED == ret)
593 ret = wait_on_sync_kiocb(&iocb);
597 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
598 struct kvec *vec, size_t num, size_t size)
600 mm_segment_t oldfs = get_fs();
605 * the following is safe, since for compiler definitions of kvec and
606 * iovec are identical, yielding the same in-core layout and alignment
608 msg->msg_iov = (struct iovec *)vec;
609 msg->msg_iovlen = num;
610 result = sock_sendmsg(sock, msg, size);
614 EXPORT_SYMBOL(kernel_sendmsg);
616 static int ktime2ts(ktime_t kt, struct timespec *ts)
619 *ts = ktime_to_timespec(kt);
627 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
629 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
632 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
633 struct timespec ts[3];
635 struct skb_shared_hwtstamps *shhwtstamps =
638 /* Race occurred between timestamp enabling and packet
639 receiving. Fill in the current time for now. */
640 if (need_software_tstamp && skb->tstamp.tv64 == 0)
641 __net_timestamp(skb);
643 if (need_software_tstamp) {
644 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
646 skb_get_timestamp(skb, &tv);
647 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
650 skb_get_timestampns(skb, &ts[0]);
651 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
652 sizeof(ts[0]), &ts[0]);
657 memset(ts, 0, sizeof(ts));
658 if (skb->tstamp.tv64 &&
659 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
660 skb_get_timestampns(skb, ts + 0);
664 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
665 ktime2ts(shhwtstamps->syststamp, ts + 1))
667 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
668 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
672 put_cmsg(msg, SOL_SOCKET,
673 SCM_TIMESTAMPING, sizeof(ts), &ts);
675 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
677 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
680 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
681 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
682 sizeof(__u32), &skb->dropcount);
685 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
688 sock_recv_timestamp(msg, sk, skb);
689 sock_recv_drops(msg, sk, skb);
691 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
693 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
694 struct msghdr *msg, size_t size, int flags)
696 struct sock_iocb *si = kiocb_to_siocb(iocb);
698 sock_update_classid(sock->sk);
706 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
709 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
710 struct msghdr *msg, size_t size, int flags)
712 int err = security_socket_recvmsg(sock, msg, size, flags);
714 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
717 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
718 size_t size, int flags)
721 struct sock_iocb siocb;
724 init_sync_kiocb(&iocb, NULL);
725 iocb.private = &siocb;
726 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
727 if (-EIOCBQUEUED == ret)
728 ret = wait_on_sync_kiocb(&iocb);
731 EXPORT_SYMBOL(sock_recvmsg);
733 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
734 size_t size, int flags)
737 struct sock_iocb siocb;
740 init_sync_kiocb(&iocb, NULL);
741 iocb.private = &siocb;
742 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
743 if (-EIOCBQUEUED == ret)
744 ret = wait_on_sync_kiocb(&iocb);
749 * kernel_recvmsg - Receive a message from a socket (kernel space)
750 * @sock: The socket to receive the message from
751 * @msg: Received message
752 * @vec: Input s/g array for message data
753 * @num: Size of input s/g array
754 * @size: Number of bytes to read
755 * @flags: Message flags (MSG_DONTWAIT, etc...)
757 * On return the msg structure contains the scatter/gather array passed in the
758 * vec argument. The array is modified so that it consists of the unfilled
759 * portion of the original array.
761 * The returned value is the total number of bytes received, or an error.
763 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
764 struct kvec *vec, size_t num, size_t size, int flags)
766 mm_segment_t oldfs = get_fs();
771 * the following is safe, since for compiler definitions of kvec and
772 * iovec are identical, yielding the same in-core layout and alignment
774 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
775 result = sock_recvmsg(sock, msg, size, flags);
779 EXPORT_SYMBOL(kernel_recvmsg);
781 static void sock_aio_dtor(struct kiocb *iocb)
783 kfree(iocb->private);
786 static ssize_t sock_sendpage(struct file *file, struct page *page,
787 int offset, size_t size, loff_t *ppos, int more)
792 sock = file->private_data;
794 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
795 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
798 return kernel_sendpage(sock, page, offset, size, flags);
801 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
802 struct pipe_inode_info *pipe, size_t len,
805 struct socket *sock = file->private_data;
807 if (unlikely(!sock->ops->splice_read))
810 sock_update_classid(sock->sk);
812 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
815 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
816 struct sock_iocb *siocb)
818 if (!is_sync_kiocb(iocb)) {
819 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
822 iocb->ki_dtor = sock_aio_dtor;
826 iocb->private = siocb;
830 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
831 struct file *file, const struct iovec *iov,
832 unsigned long nr_segs)
834 struct socket *sock = file->private_data;
838 for (i = 0; i < nr_segs; i++)
839 size += iov[i].iov_len;
841 msg->msg_name = NULL;
842 msg->msg_namelen = 0;
843 msg->msg_control = NULL;
844 msg->msg_controllen = 0;
845 msg->msg_iov = (struct iovec *)iov;
846 msg->msg_iovlen = nr_segs;
847 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
849 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
852 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
853 unsigned long nr_segs, loff_t pos)
855 struct sock_iocb siocb, *x;
860 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
864 x = alloc_sock_iocb(iocb, &siocb);
867 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
870 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
871 struct file *file, const struct iovec *iov,
872 unsigned long nr_segs)
874 struct socket *sock = file->private_data;
878 for (i = 0; i < nr_segs; i++)
879 size += iov[i].iov_len;
881 msg->msg_name = NULL;
882 msg->msg_namelen = 0;
883 msg->msg_control = NULL;
884 msg->msg_controllen = 0;
885 msg->msg_iov = (struct iovec *)iov;
886 msg->msg_iovlen = nr_segs;
887 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
888 if (sock->type == SOCK_SEQPACKET)
889 msg->msg_flags |= MSG_EOR;
891 return __sock_sendmsg(iocb, sock, msg, size);
894 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
895 unsigned long nr_segs, loff_t pos)
897 struct sock_iocb siocb, *x;
902 x = alloc_sock_iocb(iocb, &siocb);
906 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
910 * Atomic setting of ioctl hooks to avoid race
911 * with module unload.
914 static DEFINE_MUTEX(br_ioctl_mutex);
915 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
917 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
919 mutex_lock(&br_ioctl_mutex);
920 br_ioctl_hook = hook;
921 mutex_unlock(&br_ioctl_mutex);
923 EXPORT_SYMBOL(brioctl_set);
925 static DEFINE_MUTEX(vlan_ioctl_mutex);
926 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
928 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
930 mutex_lock(&vlan_ioctl_mutex);
931 vlan_ioctl_hook = hook;
932 mutex_unlock(&vlan_ioctl_mutex);
934 EXPORT_SYMBOL(vlan_ioctl_set);
936 static DEFINE_MUTEX(dlci_ioctl_mutex);
937 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
939 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
941 mutex_lock(&dlci_ioctl_mutex);
942 dlci_ioctl_hook = hook;
943 mutex_unlock(&dlci_ioctl_mutex);
945 EXPORT_SYMBOL(dlci_ioctl_set);
947 static long sock_do_ioctl(struct net *net, struct socket *sock,
948 unsigned int cmd, unsigned long arg)
951 void __user *argp = (void __user *)arg;
953 err = sock->ops->ioctl(sock, cmd, arg);
956 * If this ioctl is unknown try to hand it down
959 if (err == -ENOIOCTLCMD)
960 err = dev_ioctl(net, cmd, argp);
966 * With an ioctl, arg may well be a user mode pointer, but we don't know
967 * what to do with it - that's up to the protocol still.
970 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
974 void __user *argp = (void __user *)arg;
978 sock = file->private_data;
981 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
982 err = dev_ioctl(net, cmd, argp);
984 #ifdef CONFIG_WEXT_CORE
985 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
986 err = dev_ioctl(net, cmd, argp);
993 if (get_user(pid, (int __user *)argp))
995 err = f_setown(sock->file, pid, 1);
999 err = put_user(f_getown(sock->file),
1000 (int __user *)argp);
1008 request_module("bridge");
1010 mutex_lock(&br_ioctl_mutex);
1012 err = br_ioctl_hook(net, cmd, argp);
1013 mutex_unlock(&br_ioctl_mutex);
1018 if (!vlan_ioctl_hook)
1019 request_module("8021q");
1021 mutex_lock(&vlan_ioctl_mutex);
1022 if (vlan_ioctl_hook)
1023 err = vlan_ioctl_hook(net, argp);
1024 mutex_unlock(&vlan_ioctl_mutex);
1029 if (!dlci_ioctl_hook)
1030 request_module("dlci");
1032 mutex_lock(&dlci_ioctl_mutex);
1033 if (dlci_ioctl_hook)
1034 err = dlci_ioctl_hook(cmd, argp);
1035 mutex_unlock(&dlci_ioctl_mutex);
1038 err = sock_do_ioctl(net, sock, cmd, arg);
1044 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1047 struct socket *sock = NULL;
1049 err = security_socket_create(family, type, protocol, 1);
1053 sock = sock_alloc();
1060 err = security_socket_post_create(sock, family, type, protocol, 1);
1072 EXPORT_SYMBOL(sock_create_lite);
1074 /* No kernel lock held - perfect */
1075 static unsigned int sock_poll(struct file *file, poll_table *wait)
1077 struct socket *sock;
1080 * We can't return errors to poll, so it's either yes or no.
1082 sock = file->private_data;
1083 return sock->ops->poll(file, sock, wait);
1086 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1088 struct socket *sock = file->private_data;
1090 return sock->ops->mmap(file, sock, vma);
1093 static int sock_close(struct inode *inode, struct file *filp)
1096 * It was possible the inode is NULL we were
1097 * closing an unfinished socket.
1101 printk(KERN_DEBUG "sock_close: NULL inode\n");
1104 sock_release(SOCKET_I(inode));
1109 * Update the socket async list
1111 * Fasync_list locking strategy.
1113 * 1. fasync_list is modified only under process context socket lock
1114 * i.e. under semaphore.
1115 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1116 * or under socket lock
1119 static int sock_fasync(int fd, struct file *filp, int on)
1121 struct socket *sock = filp->private_data;
1122 struct sock *sk = sock->sk;
1123 struct socket_wq *wq;
1129 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1130 fasync_helper(fd, filp, on, &wq->fasync_list);
1132 if (!wq->fasync_list)
1133 sock_reset_flag(sk, SOCK_FASYNC);
1135 sock_set_flag(sk, SOCK_FASYNC);
1141 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1143 int sock_wake_async(struct socket *sock, int how, int band)
1145 struct socket_wq *wq;
1150 wq = rcu_dereference(sock->wq);
1151 if (!wq || !wq->fasync_list) {
1156 case SOCK_WAKE_WAITD:
1157 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1160 case SOCK_WAKE_SPACE:
1161 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1166 kill_fasync(&wq->fasync_list, SIGIO, band);
1169 kill_fasync(&wq->fasync_list, SIGURG, band);
1174 EXPORT_SYMBOL(sock_wake_async);
1176 int __sock_create(struct net *net, int family, int type, int protocol,
1177 struct socket **res, int kern)
1180 struct socket *sock;
1181 const struct net_proto_family *pf;
1184 * Check protocol is in range
1186 if (family < 0 || family >= NPROTO)
1187 return -EAFNOSUPPORT;
1188 if (type < 0 || type >= SOCK_MAX)
1193 This uglymoron is moved from INET layer to here to avoid
1194 deadlock in module load.
1196 if (family == PF_INET && type == SOCK_PACKET) {
1200 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1206 err = security_socket_create(family, type, protocol, kern);
1211 * Allocate the socket and allow the family to set things up. if
1212 * the protocol is 0, the family is instructed to select an appropriate
1215 sock = sock_alloc();
1217 if (net_ratelimit())
1218 printk(KERN_WARNING "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 = sock_alloc_file(sock1, &newfile1, flags);
1377 if (unlikely(fd1 < 0)) {
1379 goto out_release_both;
1382 fd2 = sock_alloc_file(sock2, &newfile2, flags);
1383 if (unlikely(fd2 < 0)) {
1387 sock_release(sock2);
1391 audit_fd_pair(fd1, fd2);
1392 fd_install(fd1, newfile1);
1393 fd_install(fd2, newfile2);
1394 /* fd1 and fd2 may be already another descriptors.
1395 * Not kernel problem.
1398 err = put_user(fd1, &usockvec[0]);
1400 err = put_user(fd2, &usockvec[1]);
1409 sock_release(sock2);
1411 sock_release(sock1);
1417 * Bind a name to a socket. Nothing much to do here since it's
1418 * the protocol's responsibility to handle the local address.
1420 * We move the socket address to kernel space before we call
1421 * the protocol layer (having also checked the address is ok).
1424 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1426 struct socket *sock;
1427 struct sockaddr_storage address;
1428 int err, fput_needed;
1430 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1432 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1434 err = security_socket_bind(sock,
1435 (struct sockaddr *)&address,
1438 err = sock->ops->bind(sock,
1442 fput_light(sock->file, fput_needed);
1448 * Perform a listen. Basically, we allow the protocol to do anything
1449 * necessary for a listen, and if that works, we mark the socket as
1450 * ready for listening.
1453 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1455 struct socket *sock;
1456 int err, fput_needed;
1459 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1461 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1462 if ((unsigned)backlog > somaxconn)
1463 backlog = somaxconn;
1465 err = security_socket_listen(sock, backlog);
1467 err = sock->ops->listen(sock, backlog);
1469 fput_light(sock->file, fput_needed);
1475 * For accept, we attempt to create a new socket, set up the link
1476 * with the client, wake up the client, then return the new
1477 * connected fd. We collect the address of the connector in kernel
1478 * space and move it to user at the very end. This is unclean because
1479 * we open the socket then return an error.
1481 * 1003.1g adds the ability to recvmsg() to query connection pending
1482 * status to recvmsg. We need to add that support in a way thats
1483 * clean when we restucture accept also.
1486 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1487 int __user *, upeer_addrlen, int, flags)
1489 struct socket *sock, *newsock;
1490 struct file *newfile;
1491 int err, len, newfd, fput_needed;
1492 struct sockaddr_storage address;
1494 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1497 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1498 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1500 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1505 newsock = sock_alloc();
1509 newsock->type = sock->type;
1510 newsock->ops = sock->ops;
1513 * We don't need try_module_get here, as the listening socket (sock)
1514 * has the protocol module (sock->ops->owner) held.
1516 __module_get(newsock->ops->owner);
1518 newfd = sock_alloc_file(newsock, &newfile, flags);
1519 if (unlikely(newfd < 0)) {
1521 sock_release(newsock);
1525 err = security_socket_accept(sock, newsock);
1529 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1533 if (upeer_sockaddr) {
1534 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1536 err = -ECONNABORTED;
1539 err = move_addr_to_user((struct sockaddr *)&address,
1540 len, upeer_sockaddr, upeer_addrlen);
1545 /* File flags are not inherited via accept() unlike another OSes. */
1547 fd_install(newfd, newfile);
1551 fput_light(sock->file, fput_needed);
1556 put_unused_fd(newfd);
1560 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1561 int __user *, upeer_addrlen)
1563 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1567 * Attempt to connect to a socket with the server address. The address
1568 * is in user space so we verify it is OK and move it to kernel space.
1570 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1573 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1574 * other SEQPACKET protocols that take time to connect() as it doesn't
1575 * include the -EINPROGRESS status for such sockets.
1578 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1581 struct socket *sock;
1582 struct sockaddr_storage address;
1583 int err, fput_needed;
1585 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1588 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1593 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1597 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1598 sock->file->f_flags);
1600 fput_light(sock->file, fput_needed);
1606 * Get the local address ('name') of a socket object. Move the obtained
1607 * name to user space.
1610 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1611 int __user *, usockaddr_len)
1613 struct socket *sock;
1614 struct sockaddr_storage address;
1615 int len, err, fput_needed;
1617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1621 err = security_socket_getsockname(sock);
1625 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1628 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1631 fput_light(sock->file, fput_needed);
1637 * Get the remote address ('name') of a socket object. Move the obtained
1638 * name to user space.
1641 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1642 int __user *, usockaddr_len)
1644 struct socket *sock;
1645 struct sockaddr_storage address;
1646 int len, err, fput_needed;
1648 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1650 err = security_socket_getpeername(sock);
1652 fput_light(sock->file, fput_needed);
1657 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1660 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1662 fput_light(sock->file, fput_needed);
1668 * Send a datagram to a given address. We move the address into kernel
1669 * space and check the user space data area is readable before invoking
1673 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1674 unsigned, flags, struct sockaddr __user *, addr,
1677 struct socket *sock;
1678 struct sockaddr_storage address;
1686 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690 iov.iov_base = buff;
1692 msg.msg_name = NULL;
1695 msg.msg_control = NULL;
1696 msg.msg_controllen = 0;
1697 msg.msg_namelen = 0;
1699 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1702 msg.msg_name = (struct sockaddr *)&address;
1703 msg.msg_namelen = addr_len;
1705 if (sock->file->f_flags & O_NONBLOCK)
1706 flags |= MSG_DONTWAIT;
1707 msg.msg_flags = flags;
1708 err = sock_sendmsg(sock, &msg, len);
1711 fput_light(sock->file, fput_needed);
1717 * Send a datagram down a socket.
1720 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1723 return sys_sendto(fd, buff, len, flags, NULL, 0);
1727 * Receive a frame from the socket and optionally record the address of the
1728 * sender. We verify the buffers are writable and if needed move the
1729 * sender address from kernel to user space.
1732 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1733 unsigned, flags, struct sockaddr __user *, addr,
1734 int __user *, addr_len)
1736 struct socket *sock;
1739 struct sockaddr_storage address;
1745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1749 msg.msg_control = NULL;
1750 msg.msg_controllen = 0;
1754 iov.iov_base = ubuf;
1755 msg.msg_name = (struct sockaddr *)&address;
1756 msg.msg_namelen = sizeof(address);
1757 if (sock->file->f_flags & O_NONBLOCK)
1758 flags |= MSG_DONTWAIT;
1759 err = sock_recvmsg(sock, &msg, size, flags);
1761 if (err >= 0 && addr != NULL) {
1762 err2 = move_addr_to_user((struct sockaddr *)&address,
1763 msg.msg_namelen, addr, addr_len);
1768 fput_light(sock->file, fput_needed);
1774 * Receive a datagram from a socket.
1777 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1780 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1784 * Set a socket option. Because we don't know the option lengths we have
1785 * to pass the user mode parameter for the protocols to sort out.
1788 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1789 char __user *, optval, int, optlen)
1791 int err, fput_needed;
1792 struct socket *sock;
1797 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1799 err = security_socket_setsockopt(sock, level, optname);
1803 if (level == SOL_SOCKET)
1805 sock_setsockopt(sock, level, optname, optval,
1809 sock->ops->setsockopt(sock, level, optname, optval,
1812 fput_light(sock->file, fput_needed);
1818 * Get a socket option. Because we don't know the option lengths we have
1819 * to pass a user mode parameter for the protocols to sort out.
1822 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1823 char __user *, optval, int __user *, optlen)
1825 int err, fput_needed;
1826 struct socket *sock;
1828 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1830 err = security_socket_getsockopt(sock, level, optname);
1834 if (level == SOL_SOCKET)
1836 sock_getsockopt(sock, level, optname, optval,
1840 sock->ops->getsockopt(sock, level, optname, optval,
1843 fput_light(sock->file, fput_needed);
1849 * Shutdown a socket.
1852 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1854 int err, fput_needed;
1855 struct socket *sock;
1857 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1859 err = security_socket_shutdown(sock, how);
1861 err = sock->ops->shutdown(sock, how);
1862 fput_light(sock->file, fput_needed);
1867 /* A couple of helpful macros for getting the address of the 32/64 bit
1868 * fields which are the same type (int / unsigned) on our platforms.
1870 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1871 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1872 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1874 struct used_address {
1875 struct sockaddr_storage name;
1876 unsigned int name_len;
1879 static int copy_msghdr_from_user(struct msghdr *kmsg,
1880 struct msghdr __user *umsg)
1882 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1884 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1889 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1890 struct msghdr *msg_sys, unsigned flags,
1891 struct used_address *used_address)
1893 struct compat_msghdr __user *msg_compat =
1894 (struct compat_msghdr __user *)msg;
1895 struct sockaddr_storage address;
1896 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1897 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1898 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1899 /* 20 is size of ipv6_pktinfo */
1900 unsigned char *ctl_buf = ctl;
1901 int err, ctl_len, iov_size, total_len;
1904 if (MSG_CMSG_COMPAT & flags) {
1905 if (get_compat_msghdr(msg_sys, msg_compat))
1908 err = copy_msghdr_from_user(msg_sys, msg);
1913 /* do not move before msg_sys is valid */
1915 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1918 /* Check whether to allocate the iovec area */
1920 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1921 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1922 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1927 /* This will also move the address data into kernel space */
1928 if (MSG_CMSG_COMPAT & flags) {
1929 err = verify_compat_iovec(msg_sys, iov,
1930 (struct sockaddr *)&address,
1933 err = verify_iovec(msg_sys, iov,
1934 (struct sockaddr *)&address,
1942 if (msg_sys->msg_controllen > INT_MAX)
1944 ctl_len = msg_sys->msg_controllen;
1945 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1947 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1951 ctl_buf = msg_sys->msg_control;
1952 ctl_len = msg_sys->msg_controllen;
1953 } else if (ctl_len) {
1954 if (ctl_len > sizeof(ctl)) {
1955 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1956 if (ctl_buf == NULL)
1961 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1962 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1963 * checking falls down on this.
1965 if (copy_from_user(ctl_buf,
1966 (void __user __force *)msg_sys->msg_control,
1969 msg_sys->msg_control = ctl_buf;
1971 msg_sys->msg_flags = flags;
1973 if (sock->file->f_flags & O_NONBLOCK)
1974 msg_sys->msg_flags |= MSG_DONTWAIT;
1976 * If this is sendmmsg() and current destination address is same as
1977 * previously succeeded address, omit asking LSM's decision.
1978 * used_address->name_len is initialized to UINT_MAX so that the first
1979 * destination address never matches.
1981 if (used_address && msg_sys->msg_name &&
1982 used_address->name_len == msg_sys->msg_namelen &&
1983 !memcmp(&used_address->name, msg_sys->msg_name,
1984 used_address->name_len)) {
1985 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1988 err = sock_sendmsg(sock, msg_sys, total_len);
1990 * If this is sendmmsg() and sending to current destination address was
1991 * successful, remember it.
1993 if (used_address && err >= 0) {
1994 used_address->name_len = msg_sys->msg_namelen;
1995 if (msg_sys->msg_name)
1996 memcpy(&used_address->name, msg_sys->msg_name,
1997 used_address->name_len);
2002 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2004 if (iov != iovstack)
2005 sock_kfree_s(sock->sk, iov, iov_size);
2011 * BSD sendmsg interface
2014 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2016 int fput_needed, err;
2017 struct msghdr msg_sys;
2018 struct socket *sock;
2020 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2024 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2026 fput_light(sock->file, fput_needed);
2031 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2033 if (flags & MSG_CMSG_COMPAT)
2035 return __sys_sendmsg(fd, msg, flags);
2039 * Linux sendmmsg interface
2042 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2045 int fput_needed, err, datagrams;
2046 struct socket *sock;
2047 struct mmsghdr __user *entry;
2048 struct compat_mmsghdr __user *compat_entry;
2049 struct msghdr msg_sys;
2050 struct used_address used_address;
2052 if (vlen > UIO_MAXIOV)
2057 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2061 used_address.name_len = UINT_MAX;
2063 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2066 while (datagrams < vlen) {
2067 if (MSG_CMSG_COMPAT & flags) {
2068 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2069 &msg_sys, flags, &used_address);
2072 err = __put_user(err, &compat_entry->msg_len);
2075 err = ___sys_sendmsg(sock,
2076 (struct msghdr __user *)entry,
2077 &msg_sys, flags, &used_address);
2080 err = put_user(err, &entry->msg_len);
2089 fput_light(sock->file, fput_needed);
2091 /* We only return an error if no datagrams were able to be sent */
2098 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2099 unsigned int, vlen, unsigned int, flags)
2101 if (flags & MSG_CMSG_COMPAT)
2103 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2106 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2107 struct msghdr *msg_sys, unsigned flags, int nosec)
2109 struct compat_msghdr __user *msg_compat =
2110 (struct compat_msghdr __user *)msg;
2111 struct iovec iovstack[UIO_FASTIOV];
2112 struct iovec *iov = iovstack;
2113 unsigned long cmsg_ptr;
2114 int err, iov_size, total_len, len;
2116 /* kernel mode address */
2117 struct sockaddr_storage addr;
2119 /* user mode address pointers */
2120 struct sockaddr __user *uaddr;
2121 int __user *uaddr_len;
2123 if (MSG_CMSG_COMPAT & flags) {
2124 if (get_compat_msghdr(msg_sys, msg_compat))
2127 err = copy_msghdr_from_user(msg_sys, msg);
2133 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2136 /* Check whether to allocate the iovec area */
2138 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2139 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2140 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2146 * Save the user-mode address (verify_iovec will change the
2147 * kernel msghdr to use the kernel address space)
2150 uaddr = (__force void __user *)msg_sys->msg_name;
2151 uaddr_len = COMPAT_NAMELEN(msg);
2152 if (MSG_CMSG_COMPAT & flags) {
2153 err = verify_compat_iovec(msg_sys, iov,
2154 (struct sockaddr *)&addr,
2157 err = verify_iovec(msg_sys, iov,
2158 (struct sockaddr *)&addr,
2164 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2165 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2167 if (sock->file->f_flags & O_NONBLOCK)
2168 flags |= MSG_DONTWAIT;
2169 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2175 if (uaddr != NULL) {
2176 err = move_addr_to_user((struct sockaddr *)&addr,
2177 msg_sys->msg_namelen, uaddr,
2182 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2186 if (MSG_CMSG_COMPAT & flags)
2187 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2188 &msg_compat->msg_controllen);
2190 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2191 &msg->msg_controllen);
2197 if (iov != iovstack)
2198 sock_kfree_s(sock->sk, iov, iov_size);
2204 * BSD recvmsg interface
2207 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2209 int fput_needed, err;
2210 struct msghdr msg_sys;
2211 struct socket *sock;
2213 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2217 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2219 fput_light(sock->file, fput_needed);
2224 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2225 unsigned int, flags)
2227 if (flags & MSG_CMSG_COMPAT)
2229 return __sys_recvmsg(fd, msg, flags);
2233 * Linux recvmmsg interface
2236 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2237 unsigned int flags, struct timespec *timeout)
2239 int fput_needed, err, datagrams;
2240 struct socket *sock;
2241 struct mmsghdr __user *entry;
2242 struct compat_mmsghdr __user *compat_entry;
2243 struct msghdr msg_sys;
2244 struct timespec end_time;
2247 poll_select_set_timeout(&end_time, timeout->tv_sec,
2253 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2257 err = sock_error(sock->sk);
2262 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2264 while (datagrams < vlen) {
2266 * No need to ask LSM for more than the first datagram.
2268 if (MSG_CMSG_COMPAT & flags) {
2269 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2270 &msg_sys, flags & ~MSG_WAITFORONE,
2274 err = __put_user(err, &compat_entry->msg_len);
2277 err = ___sys_recvmsg(sock,
2278 (struct msghdr __user *)entry,
2279 &msg_sys, flags & ~MSG_WAITFORONE,
2283 err = put_user(err, &entry->msg_len);
2291 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2292 if (flags & MSG_WAITFORONE)
2293 flags |= MSG_DONTWAIT;
2296 ktime_get_ts(timeout);
2297 *timeout = timespec_sub(end_time, *timeout);
2298 if (timeout->tv_sec < 0) {
2299 timeout->tv_sec = timeout->tv_nsec = 0;
2303 /* Timeout, return less than vlen datagrams */
2304 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2308 /* Out of band data, return right away */
2309 if (msg_sys.msg_flags & MSG_OOB)
2314 fput_light(sock->file, fput_needed);
2319 if (datagrams != 0) {
2321 * We may return less entries than requested (vlen) if the
2322 * sock is non block and there aren't enough datagrams...
2324 if (err != -EAGAIN) {
2326 * ... or if recvmsg returns an error after we
2327 * received some datagrams, where we record the
2328 * error to return on the next call or if the
2329 * app asks about it using getsockopt(SO_ERROR).
2331 sock->sk->sk_err = -err;
2340 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2341 unsigned int, vlen, unsigned int, flags,
2342 struct timespec __user *, timeout)
2345 struct timespec timeout_sys;
2347 if (flags & MSG_CMSG_COMPAT)
2351 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2353 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2356 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2358 if (datagrams > 0 &&
2359 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2360 datagrams = -EFAULT;
2365 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2366 /* Argument list sizes for sys_socketcall */
2367 #define AL(x) ((x) * sizeof(unsigned long))
2368 static const unsigned char nargs[21] = {
2369 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2370 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2371 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2378 * System call vectors.
2380 * Argument checking cleaned up. Saved 20% in size.
2381 * This function doesn't need to set the kernel lock because
2382 * it is set by the callees.
2385 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2388 unsigned long a0, a1;
2392 if (call < 1 || call > SYS_SENDMMSG)
2396 if (len > sizeof(a))
2399 /* copy_from_user should be SMP safe. */
2400 if (copy_from_user(a, args, len))
2403 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2410 err = sys_socket(a0, a1, a[2]);
2413 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2416 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2419 err = sys_listen(a0, a1);
2422 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2423 (int __user *)a[2], 0);
2425 case SYS_GETSOCKNAME:
2427 sys_getsockname(a0, (struct sockaddr __user *)a1,
2428 (int __user *)a[2]);
2430 case SYS_GETPEERNAME:
2432 sys_getpeername(a0, (struct sockaddr __user *)a1,
2433 (int __user *)a[2]);
2435 case SYS_SOCKETPAIR:
2436 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2439 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2442 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2443 (struct sockaddr __user *)a[4], a[5]);
2446 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2449 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2450 (struct sockaddr __user *)a[4],
2451 (int __user *)a[5]);
2454 err = sys_shutdown(a0, a1);
2456 case SYS_SETSOCKOPT:
2457 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2459 case SYS_GETSOCKOPT:
2461 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2462 (int __user *)a[4]);
2465 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2468 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2471 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2474 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2475 (struct timespec __user *)a[4]);
2478 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2479 (int __user *)a[2], a[3]);
2488 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2491 * sock_register - add a socket protocol handler
2492 * @ops: description of protocol
2494 * This function is called by a protocol handler that wants to
2495 * advertise its address family, and have it linked into the
2496 * socket interface. The value ops->family coresponds to the
2497 * socket system call protocol family.
2499 int sock_register(const struct net_proto_family *ops)
2503 if (ops->family >= NPROTO) {
2504 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2509 spin_lock(&net_family_lock);
2510 if (rcu_dereference_protected(net_families[ops->family],
2511 lockdep_is_held(&net_family_lock)))
2514 rcu_assign_pointer(net_families[ops->family], ops);
2517 spin_unlock(&net_family_lock);
2519 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2522 EXPORT_SYMBOL(sock_register);
2525 * sock_unregister - remove a protocol handler
2526 * @family: protocol family to remove
2528 * This function is called by a protocol handler that wants to
2529 * remove its address family, and have it unlinked from the
2530 * new socket creation.
2532 * If protocol handler is a module, then it can use module reference
2533 * counts to protect against new references. If protocol handler is not
2534 * a module then it needs to provide its own protection in
2535 * the ops->create routine.
2537 void sock_unregister(int family)
2539 BUG_ON(family < 0 || family >= NPROTO);
2541 spin_lock(&net_family_lock);
2542 RCU_INIT_POINTER(net_families[family], NULL);
2543 spin_unlock(&net_family_lock);
2547 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2549 EXPORT_SYMBOL(sock_unregister);
2551 static int __init sock_init(void)
2556 * Initialize sock SLAB cache.
2562 * Initialize skbuff SLAB cache
2567 * Initialize the protocols module.
2572 err = register_filesystem(&sock_fs_type);
2575 sock_mnt = kern_mount(&sock_fs_type);
2576 if (IS_ERR(sock_mnt)) {
2577 err = PTR_ERR(sock_mnt);
2581 /* The real protocol initialization is performed in later initcalls.
2584 #ifdef CONFIG_NETFILTER
2588 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2589 skb_timestamping_init();
2596 unregister_filesystem(&sock_fs_type);
2601 core_initcall(sock_init); /* early initcall */
2603 #ifdef CONFIG_PROC_FS
2604 void socket_seq_show(struct seq_file *seq)
2609 for_each_possible_cpu(cpu)
2610 counter += per_cpu(sockets_in_use, cpu);
2612 /* It can be negative, by the way. 8) */
2616 seq_printf(seq, "sockets: used %d\n", counter);
2618 #endif /* CONFIG_PROC_FS */
2620 #ifdef CONFIG_COMPAT
2621 static int do_siocgstamp(struct net *net, struct socket *sock,
2622 unsigned int cmd, struct compat_timeval __user *up)
2624 mm_segment_t old_fs = get_fs();
2629 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2632 err = put_user(ktv.tv_sec, &up->tv_sec);
2633 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2638 static int do_siocgstampns(struct net *net, struct socket *sock,
2639 unsigned int cmd, struct compat_timespec __user *up)
2641 mm_segment_t old_fs = get_fs();
2642 struct timespec kts;
2646 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2649 err = put_user(kts.tv_sec, &up->tv_sec);
2650 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2655 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2657 struct ifreq __user *uifr;
2660 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2661 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2664 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2668 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2674 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2676 struct compat_ifconf ifc32;
2678 struct ifconf __user *uifc;
2679 struct compat_ifreq __user *ifr32;
2680 struct ifreq __user *ifr;
2684 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2687 memset(&ifc, 0, sizeof(ifc));
2688 if (ifc32.ifcbuf == 0) {
2692 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2694 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2695 sizeof(struct ifreq);
2696 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2698 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2699 ifr32 = compat_ptr(ifc32.ifcbuf);
2700 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2701 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2707 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2710 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2714 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2718 ifr32 = compat_ptr(ifc32.ifcbuf);
2720 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2721 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2722 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2728 if (ifc32.ifcbuf == 0) {
2729 /* Translate from 64-bit structure multiple to
2733 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2738 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2744 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2746 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2747 bool convert_in = false, convert_out = false;
2748 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2749 struct ethtool_rxnfc __user *rxnfc;
2750 struct ifreq __user *ifr;
2751 u32 rule_cnt = 0, actual_rule_cnt;
2756 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2759 compat_rxnfc = compat_ptr(data);
2761 if (get_user(ethcmd, &compat_rxnfc->cmd))
2764 /* Most ethtool structures are defined without padding.
2765 * Unfortunately struct ethtool_rxnfc is an exception.
2770 case ETHTOOL_GRXCLSRLALL:
2771 /* Buffer size is variable */
2772 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2774 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2776 buf_size += rule_cnt * sizeof(u32);
2778 case ETHTOOL_GRXRINGS:
2779 case ETHTOOL_GRXCLSRLCNT:
2780 case ETHTOOL_GRXCLSRULE:
2783 case ETHTOOL_SRXCLSRLDEL:
2784 case ETHTOOL_SRXCLSRLINS:
2785 buf_size += sizeof(struct ethtool_rxnfc);
2790 ifr = compat_alloc_user_space(buf_size);
2791 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2793 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2796 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2797 &ifr->ifr_ifru.ifru_data))
2801 /* We expect there to be holes between fs.m_ext and
2802 * fs.ring_cookie and at the end of fs, but nowhere else.
2804 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2805 sizeof(compat_rxnfc->fs.m_ext) !=
2806 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2807 sizeof(rxnfc->fs.m_ext));
2809 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2810 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2811 offsetof(struct ethtool_rxnfc, fs.location) -
2812 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2814 if (copy_in_user(rxnfc, compat_rxnfc,
2815 (void *)(&rxnfc->fs.m_ext + 1) -
2817 copy_in_user(&rxnfc->fs.ring_cookie,
2818 &compat_rxnfc->fs.ring_cookie,
2819 (void *)(&rxnfc->fs.location + 1) -
2820 (void *)&rxnfc->fs.ring_cookie) ||
2821 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2822 sizeof(rxnfc->rule_cnt)))
2826 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2831 if (copy_in_user(compat_rxnfc, rxnfc,
2832 (const void *)(&rxnfc->fs.m_ext + 1) -
2833 (const void *)rxnfc) ||
2834 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2835 &rxnfc->fs.ring_cookie,
2836 (const void *)(&rxnfc->fs.location + 1) -
2837 (const void *)&rxnfc->fs.ring_cookie) ||
2838 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2839 sizeof(rxnfc->rule_cnt)))
2842 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2843 /* As an optimisation, we only copy the actual
2844 * number of rules that the underlying
2845 * function returned. Since Mallory might
2846 * change the rule count in user memory, we
2847 * check that it is less than the rule count
2848 * originally given (as the user buffer size),
2849 * which has been range-checked.
2851 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2853 if (actual_rule_cnt < rule_cnt)
2854 rule_cnt = actual_rule_cnt;
2855 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2856 &rxnfc->rule_locs[0],
2857 rule_cnt * sizeof(u32)))
2865 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2868 compat_uptr_t uptr32;
2869 struct ifreq __user *uifr;
2871 uifr = compat_alloc_user_space(sizeof(*uifr));
2872 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2875 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2878 uptr = compat_ptr(uptr32);
2880 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2883 return dev_ioctl(net, SIOCWANDEV, uifr);
2886 static int bond_ioctl(struct net *net, unsigned int cmd,
2887 struct compat_ifreq __user *ifr32)
2890 struct ifreq __user *uifr;
2891 mm_segment_t old_fs;
2897 case SIOCBONDENSLAVE:
2898 case SIOCBONDRELEASE:
2899 case SIOCBONDSETHWADDR:
2900 case SIOCBONDCHANGEACTIVE:
2901 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2906 err = dev_ioctl(net, cmd,
2907 (struct ifreq __user __force *) &kifr);
2911 case SIOCBONDSLAVEINFOQUERY:
2912 case SIOCBONDINFOQUERY:
2913 uifr = compat_alloc_user_space(sizeof(*uifr));
2914 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2917 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2920 datap = compat_ptr(data);
2921 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2924 return dev_ioctl(net, cmd, uifr);
2930 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2931 struct compat_ifreq __user *u_ifreq32)
2933 struct ifreq __user *u_ifreq64;
2934 char tmp_buf[IFNAMSIZ];
2935 void __user *data64;
2938 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2941 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2943 data64 = compat_ptr(data32);
2945 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2947 /* Don't check these user accesses, just let that get trapped
2948 * in the ioctl handler instead.
2950 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2953 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2956 return dev_ioctl(net, cmd, u_ifreq64);
2959 static int dev_ifsioc(struct net *net, struct socket *sock,
2960 unsigned int cmd, struct compat_ifreq __user *uifr32)
2962 struct ifreq __user *uifr;
2965 uifr = compat_alloc_user_space(sizeof(*uifr));
2966 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2969 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2980 case SIOCGIFBRDADDR:
2981 case SIOCGIFDSTADDR:
2982 case SIOCGIFNETMASK:
2987 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2995 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2996 struct compat_ifreq __user *uifr32)
2999 struct compat_ifmap __user *uifmap32;
3000 mm_segment_t old_fs;
3003 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3004 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3005 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3006 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3007 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3008 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3009 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3010 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3016 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3019 if (cmd == SIOCGIFMAP && !err) {
3020 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3021 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3022 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3023 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3024 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3025 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3026 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3033 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3036 compat_uptr_t uptr32;
3037 struct ifreq __user *uifr;
3039 uifr = compat_alloc_user_space(sizeof(*uifr));
3040 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3043 if (get_user(uptr32, &uifr32->ifr_data))
3046 uptr = compat_ptr(uptr32);
3048 if (put_user(uptr, &uifr->ifr_data))
3051 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3056 struct sockaddr rt_dst; /* target address */
3057 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3058 struct sockaddr rt_genmask; /* target network mask (IP) */
3059 unsigned short rt_flags;
3062 unsigned char rt_tos;
3063 unsigned char rt_class;
3065 short rt_metric; /* +1 for binary compatibility! */
3066 /* char * */ u32 rt_dev; /* forcing the device at add */
3067 u32 rt_mtu; /* per route MTU/Window */
3068 u32 rt_window; /* Window clamping */
3069 unsigned short rt_irtt; /* Initial RTT */
3072 struct in6_rtmsg32 {
3073 struct in6_addr rtmsg_dst;
3074 struct in6_addr rtmsg_src;
3075 struct in6_addr rtmsg_gateway;
3085 static int routing_ioctl(struct net *net, struct socket *sock,
3086 unsigned int cmd, void __user *argp)
3090 struct in6_rtmsg r6;
3094 mm_segment_t old_fs = get_fs();
3096 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3097 struct in6_rtmsg32 __user *ur6 = argp;
3098 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3099 3 * sizeof(struct in6_addr));
3100 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3101 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3102 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3103 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3104 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3105 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3106 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3110 struct rtentry32 __user *ur4 = argp;
3111 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3112 3 * sizeof(struct sockaddr));
3113 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3114 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3115 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3116 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3117 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3118 ret |= __get_user(rtdev, &(ur4->rt_dev));
3120 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3121 r4.rt_dev = (char __user __force *)devname;
3135 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3142 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3143 * for some operations; this forces use of the newer bridge-utils that
3144 * use compatible ioctls
3146 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3150 if (get_user(tmp, argp))
3152 if (tmp == BRCTL_GET_VERSION)
3153 return BRCTL_VERSION + 1;
3157 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3158 unsigned int cmd, unsigned long arg)
3160 void __user *argp = compat_ptr(arg);
3161 struct sock *sk = sock->sk;
3162 struct net *net = sock_net(sk);
3164 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3165 return siocdevprivate_ioctl(net, cmd, argp);
3170 return old_bridge_ioctl(argp);
3172 return dev_ifname32(net, argp);
3174 return dev_ifconf(net, argp);
3176 return ethtool_ioctl(net, argp);
3178 return compat_siocwandev(net, argp);
3181 return compat_sioc_ifmap(net, cmd, argp);
3182 case SIOCBONDENSLAVE:
3183 case SIOCBONDRELEASE:
3184 case SIOCBONDSETHWADDR:
3185 case SIOCBONDSLAVEINFOQUERY:
3186 case SIOCBONDINFOQUERY:
3187 case SIOCBONDCHANGEACTIVE:
3188 return bond_ioctl(net, cmd, argp);
3191 return routing_ioctl(net, sock, cmd, argp);
3193 return do_siocgstamp(net, sock, cmd, argp);
3195 return do_siocgstampns(net, sock, cmd, argp);
3197 return compat_siocshwtstamp(net, argp);
3209 return sock_ioctl(file, cmd, arg);
3226 case SIOCSIFHWBROADCAST:
3228 case SIOCGIFBRDADDR:
3229 case SIOCSIFBRDADDR:
3230 case SIOCGIFDSTADDR:
3231 case SIOCSIFDSTADDR:
3232 case SIOCGIFNETMASK:
3233 case SIOCSIFNETMASK:
3244 return dev_ifsioc(net, sock, cmd, argp);
3250 return sock_do_ioctl(net, sock, cmd, arg);
3253 /* Prevent warning from compat_sys_ioctl, these always
3254 * result in -EINVAL in the native case anyway. */
3267 return -ENOIOCTLCMD;
3270 static long compat_sock_ioctl(struct file *file, unsigned cmd,
3273 struct socket *sock = file->private_data;
3274 int ret = -ENOIOCTLCMD;
3281 if (sock->ops->compat_ioctl)
3282 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3284 if (ret == -ENOIOCTLCMD &&
3285 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3286 ret = compat_wext_handle_ioctl(net, cmd, arg);
3288 if (ret == -ENOIOCTLCMD)
3289 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3295 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3297 return sock->ops->bind(sock, addr, addrlen);
3299 EXPORT_SYMBOL(kernel_bind);
3301 int kernel_listen(struct socket *sock, int backlog)
3303 return sock->ops->listen(sock, backlog);
3305 EXPORT_SYMBOL(kernel_listen);
3307 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3309 struct sock *sk = sock->sk;
3312 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3317 err = sock->ops->accept(sock, *newsock, flags);
3319 sock_release(*newsock);
3324 (*newsock)->ops = sock->ops;
3325 __module_get((*newsock)->ops->owner);
3330 EXPORT_SYMBOL(kernel_accept);
3332 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3335 return sock->ops->connect(sock, addr, addrlen, flags);
3337 EXPORT_SYMBOL(kernel_connect);
3339 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3342 return sock->ops->getname(sock, addr, addrlen, 0);
3344 EXPORT_SYMBOL(kernel_getsockname);
3346 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3349 return sock->ops->getname(sock, addr, addrlen, 1);
3351 EXPORT_SYMBOL(kernel_getpeername);
3353 int kernel_getsockopt(struct socket *sock, int level, int optname,
3354 char *optval, int *optlen)
3356 mm_segment_t oldfs = get_fs();
3357 char __user *uoptval;
3358 int __user *uoptlen;
3361 uoptval = (char __user __force *) optval;
3362 uoptlen = (int __user __force *) optlen;
3365 if (level == SOL_SOCKET)
3366 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3368 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3373 EXPORT_SYMBOL(kernel_getsockopt);
3375 int kernel_setsockopt(struct socket *sock, int level, int optname,
3376 char *optval, unsigned int optlen)
3378 mm_segment_t oldfs = get_fs();
3379 char __user *uoptval;
3382 uoptval = (char __user __force *) optval;
3385 if (level == SOL_SOCKET)
3386 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3388 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3393 EXPORT_SYMBOL(kernel_setsockopt);
3395 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3396 size_t size, int flags)
3398 sock_update_classid(sock->sk);
3400 if (sock->ops->sendpage)
3401 return sock->ops->sendpage(sock, page, offset, size, flags);
3403 return sock_no_sendpage(sock, page, offset, size, flags);
3405 EXPORT_SYMBOL(kernel_sendpage);
3407 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3409 mm_segment_t oldfs = get_fs();
3413 err = sock->ops->ioctl(sock, cmd, arg);
3418 EXPORT_SYMBOL(kernel_sock_ioctl);
3420 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3422 return sock->ops->shutdown(sock, how);
3424 EXPORT_SYMBOL(kernel_sock_shutdown);