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>
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly;
112 unsigned int sysctl_net_busy_poll __read_mostly;
115 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
116 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117 unsigned long nr_segs, loff_t pos);
118 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119 unsigned long nr_segs, loff_t pos);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
122 static int sock_close(struct inode *inode, struct file *file);
123 static unsigned int sock_poll(struct file *file,
124 struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
127 static long compat_sock_ioctl(struct file *file,
128 unsigned int cmd, unsigned long arg);
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_sendpage(struct file *file, struct page *page,
132 int offset, size_t size, loff_t *ppos, int more);
133 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134 struct pipe_inode_info *pipe, size_t len,
138 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
139 * in the operation structures but are done directly via the socketcall() multiplexor.
142 static const struct file_operations socket_file_ops = {
143 .owner = THIS_MODULE,
145 .aio_read = sock_aio_read,
146 .aio_write = sock_aio_write,
148 .unlocked_ioctl = sock_ioctl,
150 .compat_ioctl = compat_sock_ioctl,
153 .open = sock_no_open, /* special open code to disallow open via /proc */
154 .release = sock_close,
155 .fasync = sock_fasync,
156 .sendpage = sock_sendpage,
157 .splice_write = generic_splice_sendpage,
158 .splice_read = sock_splice_read,
162 * The protocol list. Each protocol is registered in here.
165 static DEFINE_SPINLOCK(net_family_lock);
166 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
169 * Statistics counters of the socket lists
172 static DEFINE_PER_CPU(int, sockets_in_use);
176 * Move socket addresses back and forth across the kernel/user
177 * divide and look after the messy bits.
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
191 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
193 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
197 if (copy_from_user(kaddr, uaddr, ulen))
199 return audit_sockaddr(ulen, kaddr);
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
219 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
220 void __user *uaddr, int __user *ulen)
225 BUG_ON(klen > sizeof(struct sockaddr_storage));
226 err = get_user(len, ulen);
234 if (audit_sockaddr(klen, kaddr))
236 if (copy_to_user(uaddr, kaddr, len))
240 * "fromlen shall refer to the value before truncation.."
243 return __put_user(klen, ulen);
246 static struct kmem_cache *sock_inode_cachep __read_mostly;
248 static struct inode *sock_alloc_inode(struct super_block *sb)
250 struct socket_alloc *ei;
251 struct socket_wq *wq;
253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
256 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
258 kmem_cache_free(sock_inode_cachep, ei);
261 init_waitqueue_head(&wq->wait);
262 wq->fasync_list = NULL;
263 RCU_INIT_POINTER(ei->socket.wq, wq);
265 ei->socket.state = SS_UNCONNECTED;
266 ei->socket.flags = 0;
267 ei->socket.ops = NULL;
268 ei->socket.sk = NULL;
269 ei->socket.file = NULL;
271 return &ei->vfs_inode;
274 static void sock_destroy_inode(struct inode *inode)
276 struct socket_alloc *ei;
277 struct socket_wq *wq;
279 ei = container_of(inode, struct socket_alloc, vfs_inode);
280 wq = rcu_dereference_protected(ei->socket.wq, 1);
282 kmem_cache_free(sock_inode_cachep, ei);
285 static void init_once(void *foo)
287 struct socket_alloc *ei = (struct socket_alloc *)foo;
289 inode_init_once(&ei->vfs_inode);
292 static int init_inodecache(void)
294 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
295 sizeof(struct socket_alloc),
297 (SLAB_HWCACHE_ALIGN |
298 SLAB_RECLAIM_ACCOUNT |
301 if (sock_inode_cachep == NULL)
306 static const struct super_operations sockfs_ops = {
307 .alloc_inode = sock_alloc_inode,
308 .destroy_inode = sock_destroy_inode,
309 .statfs = simple_statfs,
313 * sockfs_dname() is called from d_path().
315 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
317 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
318 dentry->d_inode->i_ino);
321 static const struct dentry_operations sockfs_dentry_operations = {
322 .d_dname = sockfs_dname,
325 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
326 int flags, const char *dev_name, void *data)
328 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
329 &sockfs_dentry_operations, SOCKFS_MAGIC);
332 static struct vfsmount *sock_mnt __read_mostly;
334 static struct file_system_type sock_fs_type = {
336 .mount = sockfs_mount,
337 .kill_sb = kill_anon_super,
341 * Obtains the first available file descriptor and sets it up for use.
343 * These functions create file structures and maps them to fd space
344 * of the current process. On success it returns file descriptor
345 * and file struct implicitly stored in sock->file.
346 * Note that another thread may close file descriptor before we return
347 * from this function. We use the fact that now we do not refer
348 * to socket after mapping. If one day we will need it, this
349 * function will increment ref. count on file by 1.
351 * In any case returned fd MAY BE not valid!
352 * This race condition is unavoidable
353 * with shared fd spaces, we cannot solve it inside kernel,
354 * but we take care of internal coherence yet.
357 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
359 struct qstr name = { .name = "" };
365 name.len = strlen(name.name);
366 } else if (sock->sk) {
367 name.name = sock->sk->sk_prot_creator->name;
368 name.len = strlen(name.name);
370 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
371 if (unlikely(!path.dentry))
372 return ERR_PTR(-ENOMEM);
373 path.mnt = mntget(sock_mnt);
375 d_instantiate(path.dentry, SOCK_INODE(sock));
376 SOCK_INODE(sock)->i_fop = &socket_file_ops;
378 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
380 if (unlikely(IS_ERR(file))) {
381 /* drop dentry, keep inode */
382 ihold(path.dentry->d_inode);
388 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
389 file->private_data = sock;
392 EXPORT_SYMBOL(sock_alloc_file);
394 static int sock_map_fd(struct socket *sock, int flags)
396 struct file *newfile;
397 int fd = get_unused_fd_flags(flags);
398 if (unlikely(fd < 0))
401 newfile = sock_alloc_file(sock, flags, NULL);
402 if (likely(!IS_ERR(newfile))) {
403 fd_install(fd, newfile);
408 return PTR_ERR(newfile);
411 struct socket *sock_from_file(struct file *file, int *err)
413 if (file->f_op == &socket_file_ops)
414 return file->private_data; /* set in sock_map_fd */
419 EXPORT_SYMBOL(sock_from_file);
422 * sockfd_lookup - Go from a file number to its socket slot
424 * @err: pointer to an error code return
426 * The file handle passed in is locked and the socket it is bound
427 * too is returned. If an error occurs the err pointer is overwritten
428 * with a negative errno code and NULL is returned. The function checks
429 * for both invalid handles and passing a handle which is not a socket.
431 * On a success the socket object pointer is returned.
434 struct socket *sockfd_lookup(int fd, int *err)
445 sock = sock_from_file(file, err);
450 EXPORT_SYMBOL(sockfd_lookup);
452 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
454 struct fd f = fdget(fd);
459 sock = sock_from_file(f.file, err);
461 *fput_needed = f.flags;
469 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
470 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
471 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
472 static ssize_t sockfs_getxattr(struct dentry *dentry,
473 const char *name, void *value, size_t size)
475 const char *proto_name;
480 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
481 proto_name = dentry->d_name.name;
482 proto_size = strlen(proto_name);
486 if (proto_size + 1 > size)
489 strncpy(value, proto_name, proto_size + 1);
491 error = proto_size + 1;
498 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
504 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
514 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
519 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
526 static const struct inode_operations sockfs_inode_ops = {
527 .getxattr = sockfs_getxattr,
528 .listxattr = sockfs_listxattr,
532 * sock_alloc - allocate a socket
534 * Allocate a new inode and socket object. The two are bound together
535 * and initialised. The socket is then returned. If we are out of inodes
539 static struct socket *sock_alloc(void)
544 inode = new_inode_pseudo(sock_mnt->mnt_sb);
548 sock = SOCKET_I(inode);
550 kmemcheck_annotate_bitfield(sock, type);
551 inode->i_ino = get_next_ino();
552 inode->i_mode = S_IFSOCK | S_IRWXUGO;
553 inode->i_uid = current_fsuid();
554 inode->i_gid = current_fsgid();
555 inode->i_op = &sockfs_inode_ops;
557 this_cpu_add(sockets_in_use, 1);
562 * In theory you can't get an open on this inode, but /proc provides
563 * a back door. Remember to keep it shut otherwise you'll let the
564 * creepy crawlies in.
567 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
572 const struct file_operations bad_sock_fops = {
573 .owner = THIS_MODULE,
574 .open = sock_no_open,
575 .llseek = noop_llseek,
579 * sock_release - close a socket
580 * @sock: socket to close
582 * The socket is released from the protocol stack if it has a release
583 * callback, and the inode is then released if the socket is bound to
584 * an inode not a file.
587 void sock_release(struct socket *sock)
590 struct module *owner = sock->ops->owner;
592 sock->ops->release(sock);
597 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
598 pr_err("%s: fasync list not empty!\n", __func__);
600 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
603 this_cpu_sub(sockets_in_use, 1);
605 iput(SOCK_INODE(sock));
610 EXPORT_SYMBOL(sock_release);
612 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
615 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
616 *tx_flags |= SKBTX_HW_TSTAMP;
617 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
618 *tx_flags |= SKBTX_SW_TSTAMP;
619 if (sock_flag(sk, SOCK_WIFI_STATUS))
620 *tx_flags |= SKBTX_WIFI_STATUS;
622 EXPORT_SYMBOL(sock_tx_timestamp);
624 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
625 struct msghdr *msg, size_t size)
627 struct sock_iocb *si = kiocb_to_siocb(iocb);
634 return sock->ops->sendmsg(iocb, sock, msg, size);
637 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
638 struct msghdr *msg, size_t size)
640 int err = security_socket_sendmsg(sock, msg, size);
642 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
645 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
648 struct sock_iocb siocb;
651 init_sync_kiocb(&iocb, NULL);
652 iocb.private = &siocb;
653 ret = __sock_sendmsg(&iocb, sock, msg, size);
654 if (-EIOCBQUEUED == ret)
655 ret = wait_on_sync_kiocb(&iocb);
658 EXPORT_SYMBOL(sock_sendmsg);
660 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
663 struct sock_iocb siocb;
666 init_sync_kiocb(&iocb, NULL);
667 iocb.private = &siocb;
668 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
669 if (-EIOCBQUEUED == ret)
670 ret = wait_on_sync_kiocb(&iocb);
674 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
675 struct kvec *vec, size_t num, size_t size)
677 mm_segment_t oldfs = get_fs();
682 * the following is safe, since for compiler definitions of kvec and
683 * iovec are identical, yielding the same in-core layout and alignment
685 msg->msg_iov = (struct iovec *)vec;
686 msg->msg_iovlen = num;
687 result = sock_sendmsg(sock, msg, size);
691 EXPORT_SYMBOL(kernel_sendmsg);
694 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
696 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
699 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
700 struct timespec ts[3];
702 struct skb_shared_hwtstamps *shhwtstamps =
705 /* Race occurred between timestamp enabling and packet
706 receiving. Fill in the current time for now. */
707 if (need_software_tstamp && skb->tstamp.tv64 == 0)
708 __net_timestamp(skb);
710 if (need_software_tstamp) {
711 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
713 skb_get_timestamp(skb, &tv);
714 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
717 skb_get_timestampns(skb, &ts[0]);
718 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
719 sizeof(ts[0]), &ts[0]);
724 memset(ts, 0, sizeof(ts));
725 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
726 ktime_to_timespec_cond(skb->tstamp, ts + 0))
729 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
730 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
733 put_cmsg(msg, SOL_SOCKET,
734 SCM_TIMESTAMPING, sizeof(ts), &ts);
736 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
738 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
743 if (!sock_flag(sk, SOCK_WIFI_STATUS))
745 if (!skb->wifi_acked_valid)
748 ack = skb->wifi_acked;
750 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
752 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
754 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
757 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
758 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
759 sizeof(__u32), &skb->dropcount);
762 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
765 sock_recv_timestamp(msg, sk, skb);
766 sock_recv_drops(msg, sk, skb);
768 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
770 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
771 struct msghdr *msg, size_t size, int flags)
773 struct sock_iocb *si = kiocb_to_siocb(iocb);
781 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
784 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
785 struct msghdr *msg, size_t size, int flags)
787 int err = security_socket_recvmsg(sock, msg, size, flags);
789 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
792 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
793 size_t size, int flags)
796 struct sock_iocb siocb;
799 init_sync_kiocb(&iocb, NULL);
800 iocb.private = &siocb;
801 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
802 if (-EIOCBQUEUED == ret)
803 ret = wait_on_sync_kiocb(&iocb);
806 EXPORT_SYMBOL(sock_recvmsg);
808 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
809 size_t size, int flags)
812 struct sock_iocb siocb;
815 init_sync_kiocb(&iocb, NULL);
816 iocb.private = &siocb;
817 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
818 if (-EIOCBQUEUED == ret)
819 ret = wait_on_sync_kiocb(&iocb);
824 * kernel_recvmsg - Receive a message from a socket (kernel space)
825 * @sock: The socket to receive the message from
826 * @msg: Received message
827 * @vec: Input s/g array for message data
828 * @num: Size of input s/g array
829 * @size: Number of bytes to read
830 * @flags: Message flags (MSG_DONTWAIT, etc...)
832 * On return the msg structure contains the scatter/gather array passed in the
833 * vec argument. The array is modified so that it consists of the unfilled
834 * portion of the original array.
836 * The returned value is the total number of bytes received, or an error.
838 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
839 struct kvec *vec, size_t num, size_t size, int flags)
841 mm_segment_t oldfs = get_fs();
846 * the following is safe, since for compiler definitions of kvec and
847 * iovec are identical, yielding the same in-core layout and alignment
849 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
850 result = sock_recvmsg(sock, msg, size, flags);
854 EXPORT_SYMBOL(kernel_recvmsg);
856 static ssize_t sock_sendpage(struct file *file, struct page *page,
857 int offset, size_t size, loff_t *ppos, int more)
862 sock = file->private_data;
864 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
865 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
868 return kernel_sendpage(sock, page, offset, size, flags);
871 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
872 struct pipe_inode_info *pipe, size_t len,
875 struct socket *sock = file->private_data;
877 if (unlikely(!sock->ops->splice_read))
880 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
883 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
884 struct sock_iocb *siocb)
886 if (!is_sync_kiocb(iocb))
890 iocb->private = siocb;
894 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
895 struct file *file, const struct iovec *iov,
896 unsigned long nr_segs)
898 struct socket *sock = file->private_data;
902 for (i = 0; i < nr_segs; i++)
903 size += iov[i].iov_len;
905 msg->msg_name = NULL;
906 msg->msg_namelen = 0;
907 msg->msg_control = NULL;
908 msg->msg_controllen = 0;
909 msg->msg_iov = (struct iovec *)iov;
910 msg->msg_iovlen = nr_segs;
911 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
913 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
916 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
917 unsigned long nr_segs, loff_t pos)
919 struct sock_iocb siocb, *x;
924 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
928 x = alloc_sock_iocb(iocb, &siocb);
931 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
934 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
935 struct file *file, const struct iovec *iov,
936 unsigned long nr_segs)
938 struct socket *sock = file->private_data;
942 for (i = 0; i < nr_segs; i++)
943 size += iov[i].iov_len;
945 msg->msg_name = NULL;
946 msg->msg_namelen = 0;
947 msg->msg_control = NULL;
948 msg->msg_controllen = 0;
949 msg->msg_iov = (struct iovec *)iov;
950 msg->msg_iovlen = nr_segs;
951 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
952 if (sock->type == SOCK_SEQPACKET)
953 msg->msg_flags |= MSG_EOR;
955 return __sock_sendmsg(iocb, sock, msg, size);
958 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
959 unsigned long nr_segs, loff_t pos)
961 struct sock_iocb siocb, *x;
966 x = alloc_sock_iocb(iocb, &siocb);
970 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
974 * Atomic setting of ioctl hooks to avoid race
975 * with module unload.
978 static DEFINE_MUTEX(br_ioctl_mutex);
979 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
981 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
983 mutex_lock(&br_ioctl_mutex);
984 br_ioctl_hook = hook;
985 mutex_unlock(&br_ioctl_mutex);
987 EXPORT_SYMBOL(brioctl_set);
989 static DEFINE_MUTEX(vlan_ioctl_mutex);
990 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
992 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
994 mutex_lock(&vlan_ioctl_mutex);
995 vlan_ioctl_hook = hook;
996 mutex_unlock(&vlan_ioctl_mutex);
998 EXPORT_SYMBOL(vlan_ioctl_set);
1000 static DEFINE_MUTEX(dlci_ioctl_mutex);
1001 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1003 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1005 mutex_lock(&dlci_ioctl_mutex);
1006 dlci_ioctl_hook = hook;
1007 mutex_unlock(&dlci_ioctl_mutex);
1009 EXPORT_SYMBOL(dlci_ioctl_set);
1011 static long sock_do_ioctl(struct net *net, struct socket *sock,
1012 unsigned int cmd, unsigned long arg)
1015 void __user *argp = (void __user *)arg;
1017 err = sock->ops->ioctl(sock, cmd, arg);
1020 * If this ioctl is unknown try to hand it down
1021 * to the NIC driver.
1023 if (err == -ENOIOCTLCMD)
1024 err = dev_ioctl(net, cmd, argp);
1030 * With an ioctl, arg may well be a user mode pointer, but we don't know
1031 * what to do with it - that's up to the protocol still.
1034 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1036 struct socket *sock;
1038 void __user *argp = (void __user *)arg;
1042 sock = file->private_data;
1045 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1046 err = dev_ioctl(net, cmd, argp);
1048 #ifdef CONFIG_WEXT_CORE
1049 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1050 err = dev_ioctl(net, cmd, argp);
1057 if (get_user(pid, (int __user *)argp))
1059 err = f_setown(sock->file, pid, 1);
1063 err = put_user(f_getown(sock->file),
1064 (int __user *)argp);
1072 request_module("bridge");
1074 mutex_lock(&br_ioctl_mutex);
1076 err = br_ioctl_hook(net, cmd, argp);
1077 mutex_unlock(&br_ioctl_mutex);
1082 if (!vlan_ioctl_hook)
1083 request_module("8021q");
1085 mutex_lock(&vlan_ioctl_mutex);
1086 if (vlan_ioctl_hook)
1087 err = vlan_ioctl_hook(net, argp);
1088 mutex_unlock(&vlan_ioctl_mutex);
1093 if (!dlci_ioctl_hook)
1094 request_module("dlci");
1096 mutex_lock(&dlci_ioctl_mutex);
1097 if (dlci_ioctl_hook)
1098 err = dlci_ioctl_hook(cmd, argp);
1099 mutex_unlock(&dlci_ioctl_mutex);
1102 err = sock_do_ioctl(net, sock, cmd, arg);
1108 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1111 struct socket *sock = NULL;
1113 err = security_socket_create(family, type, protocol, 1);
1117 sock = sock_alloc();
1124 err = security_socket_post_create(sock, family, type, protocol, 1);
1136 EXPORT_SYMBOL(sock_create_lite);
1138 /* No kernel lock held - perfect */
1139 static unsigned int sock_poll(struct file *file, poll_table *wait)
1141 unsigned int busy_flag = 0;
1142 struct socket *sock;
1145 * We can't return errors to poll, so it's either yes or no.
1147 sock = file->private_data;
1149 if (sk_can_busy_loop(sock->sk)) {
1150 /* this socket can poll_ll so tell the system call */
1151 busy_flag = POLL_BUSY_LOOP;
1153 /* once, only if requested by syscall */
1154 if (wait && (wait->_key & POLL_BUSY_LOOP))
1155 sk_busy_loop(sock->sk, 1);
1158 return busy_flag | sock->ops->poll(file, sock, wait);
1161 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1163 struct socket *sock = file->private_data;
1165 return sock->ops->mmap(file, sock, vma);
1168 static int sock_close(struct inode *inode, struct file *filp)
1170 sock_release(SOCKET_I(inode));
1175 * Update the socket async list
1177 * Fasync_list locking strategy.
1179 * 1. fasync_list is modified only under process context socket lock
1180 * i.e. under semaphore.
1181 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1182 * or under socket lock
1185 static int sock_fasync(int fd, struct file *filp, int on)
1187 struct socket *sock = filp->private_data;
1188 struct sock *sk = sock->sk;
1189 struct socket_wq *wq;
1195 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1196 fasync_helper(fd, filp, on, &wq->fasync_list);
1198 if (!wq->fasync_list)
1199 sock_reset_flag(sk, SOCK_FASYNC);
1201 sock_set_flag(sk, SOCK_FASYNC);
1207 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1209 int sock_wake_async(struct socket *sock, int how, int band)
1211 struct socket_wq *wq;
1216 wq = rcu_dereference(sock->wq);
1217 if (!wq || !wq->fasync_list) {
1222 case SOCK_WAKE_WAITD:
1223 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1226 case SOCK_WAKE_SPACE:
1227 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1232 kill_fasync(&wq->fasync_list, SIGIO, band);
1235 kill_fasync(&wq->fasync_list, SIGURG, band);
1240 EXPORT_SYMBOL(sock_wake_async);
1242 int __sock_create(struct net *net, int family, int type, int protocol,
1243 struct socket **res, int kern)
1246 struct socket *sock;
1247 const struct net_proto_family *pf;
1250 * Check protocol is in range
1252 if (family < 0 || family >= NPROTO)
1253 return -EAFNOSUPPORT;
1254 if (type < 0 || type >= SOCK_MAX)
1259 This uglymoron is moved from INET layer to here to avoid
1260 deadlock in module load.
1262 if (family == PF_INET && type == SOCK_PACKET) {
1266 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1272 err = security_socket_create(family, type, protocol, kern);
1277 * Allocate the socket and allow the family to set things up. if
1278 * the protocol is 0, the family is instructed to select an appropriate
1281 sock = sock_alloc();
1283 net_warn_ratelimited("socket: no more sockets\n");
1284 return -ENFILE; /* Not exactly a match, but its the
1285 closest posix thing */
1290 #ifdef CONFIG_MODULES
1291 /* Attempt to load a protocol module if the find failed.
1293 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1294 * requested real, full-featured networking support upon configuration.
1295 * Otherwise module support will break!
1297 if (rcu_access_pointer(net_families[family]) == NULL)
1298 request_module("net-pf-%d", family);
1302 pf = rcu_dereference(net_families[family]);
1303 err = -EAFNOSUPPORT;
1308 * We will call the ->create function, that possibly is in a loadable
1309 * module, so we have to bump that loadable module refcnt first.
1311 if (!try_module_get(pf->owner))
1314 /* Now protected by module ref count */
1317 err = pf->create(net, sock, protocol, kern);
1319 goto out_module_put;
1322 * Now to bump the refcnt of the [loadable] module that owns this
1323 * socket at sock_release time we decrement its refcnt.
1325 if (!try_module_get(sock->ops->owner))
1326 goto out_module_busy;
1329 * Now that we're done with the ->create function, the [loadable]
1330 * module can have its refcnt decremented
1332 module_put(pf->owner);
1333 err = security_socket_post_create(sock, family, type, protocol, kern);
1335 goto out_sock_release;
1341 err = -EAFNOSUPPORT;
1344 module_put(pf->owner);
1351 goto out_sock_release;
1353 EXPORT_SYMBOL(__sock_create);
1355 int sock_create(int family, int type, int protocol, struct socket **res)
1357 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1359 EXPORT_SYMBOL(sock_create);
1361 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1363 return __sock_create(&init_net, family, type, protocol, res, 1);
1365 EXPORT_SYMBOL(sock_create_kern);
1367 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1370 struct socket *sock;
1373 /* Check the SOCK_* constants for consistency. */
1374 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1375 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1376 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1377 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1379 flags = type & ~SOCK_TYPE_MASK;
1380 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1382 type &= SOCK_TYPE_MASK;
1384 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1385 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1387 retval = sock_create(family, type, protocol, &sock);
1391 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1396 /* It may be already another descriptor 8) Not kernel problem. */
1405 * Create a pair of connected sockets.
1408 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1409 int __user *, usockvec)
1411 struct socket *sock1, *sock2;
1413 struct file *newfile1, *newfile2;
1416 flags = type & ~SOCK_TYPE_MASK;
1417 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1419 type &= SOCK_TYPE_MASK;
1421 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1422 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1425 * Obtain the first socket and check if the underlying protocol
1426 * supports the socketpair call.
1429 err = sock_create(family, type, protocol, &sock1);
1433 err = sock_create(family, type, protocol, &sock2);
1437 err = sock1->ops->socketpair(sock1, sock2);
1439 goto out_release_both;
1441 fd1 = get_unused_fd_flags(flags);
1442 if (unlikely(fd1 < 0)) {
1444 goto out_release_both;
1447 fd2 = get_unused_fd_flags(flags);
1448 if (unlikely(fd2 < 0)) {
1450 goto out_put_unused_1;
1453 newfile1 = sock_alloc_file(sock1, flags, NULL);
1454 if (unlikely(IS_ERR(newfile1))) {
1455 err = PTR_ERR(newfile1);
1456 goto out_put_unused_both;
1459 newfile2 = sock_alloc_file(sock2, flags, NULL);
1460 if (IS_ERR(newfile2)) {
1461 err = PTR_ERR(newfile2);
1465 err = put_user(fd1, &usockvec[0]);
1469 err = put_user(fd2, &usockvec[1]);
1473 audit_fd_pair(fd1, fd2);
1475 fd_install(fd1, newfile1);
1476 fd_install(fd2, newfile2);
1477 /* fd1 and fd2 may be already another descriptors.
1478 * Not kernel problem.
1494 sock_release(sock2);
1497 out_put_unused_both:
1502 sock_release(sock2);
1504 sock_release(sock1);
1510 * Bind a name to a socket. Nothing much to do here since it's
1511 * the protocol's responsibility to handle the local address.
1513 * We move the socket address to kernel space before we call
1514 * the protocol layer (having also checked the address is ok).
1517 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1519 struct socket *sock;
1520 struct sockaddr_storage address;
1521 int err, fput_needed;
1523 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1525 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1527 err = security_socket_bind(sock,
1528 (struct sockaddr *)&address,
1531 err = sock->ops->bind(sock,
1535 fput_light(sock->file, fput_needed);
1541 * Perform a listen. Basically, we allow the protocol to do anything
1542 * necessary for a listen, and if that works, we mark the socket as
1543 * ready for listening.
1546 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1548 struct socket *sock;
1549 int err, fput_needed;
1552 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1554 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1555 if ((unsigned int)backlog > somaxconn)
1556 backlog = somaxconn;
1558 err = security_socket_listen(sock, backlog);
1560 err = sock->ops->listen(sock, backlog);
1562 fput_light(sock->file, fput_needed);
1568 * For accept, we attempt to create a new socket, set up the link
1569 * with the client, wake up the client, then return the new
1570 * connected fd. We collect the address of the connector in kernel
1571 * space and move it to user at the very end. This is unclean because
1572 * we open the socket then return an error.
1574 * 1003.1g adds the ability to recvmsg() to query connection pending
1575 * status to recvmsg. We need to add that support in a way thats
1576 * clean when we restucture accept also.
1579 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1580 int __user *, upeer_addrlen, int, flags)
1582 struct socket *sock, *newsock;
1583 struct file *newfile;
1584 int err, len, newfd, fput_needed;
1585 struct sockaddr_storage address;
1587 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1590 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1591 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1598 newsock = sock_alloc();
1602 newsock->type = sock->type;
1603 newsock->ops = sock->ops;
1606 * We don't need try_module_get here, as the listening socket (sock)
1607 * has the protocol module (sock->ops->owner) held.
1609 __module_get(newsock->ops->owner);
1611 newfd = get_unused_fd_flags(flags);
1612 if (unlikely(newfd < 0)) {
1614 sock_release(newsock);
1617 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1618 if (unlikely(IS_ERR(newfile))) {
1619 err = PTR_ERR(newfile);
1620 put_unused_fd(newfd);
1621 sock_release(newsock);
1625 err = security_socket_accept(sock, newsock);
1629 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1633 if (upeer_sockaddr) {
1634 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1636 err = -ECONNABORTED;
1639 err = move_addr_to_user(&address,
1640 len, upeer_sockaddr, upeer_addrlen);
1645 /* File flags are not inherited via accept() unlike another OSes. */
1647 fd_install(newfd, newfile);
1651 fput_light(sock->file, fput_needed);
1656 put_unused_fd(newfd);
1660 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1661 int __user *, upeer_addrlen)
1663 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1667 * Attempt to connect to a socket with the server address. The address
1668 * is in user space so we verify it is OK and move it to kernel space.
1670 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1673 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1674 * other SEQPACKET protocols that take time to connect() as it doesn't
1675 * include the -EINPROGRESS status for such sockets.
1678 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1681 struct socket *sock;
1682 struct sockaddr_storage address;
1683 int err, fput_needed;
1685 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1688 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1693 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1697 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1698 sock->file->f_flags);
1700 fput_light(sock->file, fput_needed);
1706 * Get the local address ('name') of a socket object. Move the obtained
1707 * name to user space.
1710 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1711 int __user *, usockaddr_len)
1713 struct socket *sock;
1714 struct sockaddr_storage address;
1715 int len, err, fput_needed;
1717 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1721 err = security_socket_getsockname(sock);
1725 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1728 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1731 fput_light(sock->file, fput_needed);
1737 * Get the remote address ('name') of a socket object. Move the obtained
1738 * name to user space.
1741 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1742 int __user *, usockaddr_len)
1744 struct socket *sock;
1745 struct sockaddr_storage address;
1746 int len, err, fput_needed;
1748 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1750 err = security_socket_getpeername(sock);
1752 fput_light(sock->file, fput_needed);
1757 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1760 err = move_addr_to_user(&address, len, usockaddr,
1762 fput_light(sock->file, fput_needed);
1768 * Send a datagram to a given address. We move the address into kernel
1769 * space and check the user space data area is readable before invoking
1773 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1774 unsigned int, flags, struct sockaddr __user *, addr,
1777 struct socket *sock;
1778 struct sockaddr_storage address;
1786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1790 iov.iov_base = buff;
1792 msg.msg_name = NULL;
1795 msg.msg_control = NULL;
1796 msg.msg_controllen = 0;
1797 msg.msg_namelen = 0;
1799 err = move_addr_to_kernel(addr, addr_len, &address);
1802 msg.msg_name = (struct sockaddr *)&address;
1803 msg.msg_namelen = addr_len;
1805 if (sock->file->f_flags & O_NONBLOCK)
1806 flags |= MSG_DONTWAIT;
1807 msg.msg_flags = flags;
1808 err = sock_sendmsg(sock, &msg, len);
1811 fput_light(sock->file, fput_needed);
1817 * Send a datagram down a socket.
1820 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1821 unsigned int, flags)
1823 return sys_sendto(fd, buff, len, flags, NULL, 0);
1827 * Receive a frame from the socket and optionally record the address of the
1828 * sender. We verify the buffers are writable and if needed move the
1829 * sender address from kernel to user space.
1832 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1833 unsigned int, flags, struct sockaddr __user *, addr,
1834 int __user *, addr_len)
1836 struct socket *sock;
1839 struct sockaddr_storage address;
1845 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1849 msg.msg_control = NULL;
1850 msg.msg_controllen = 0;
1854 iov.iov_base = ubuf;
1855 /* Save some cycles and don't copy the address if not needed */
1856 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1857 /* We assume all kernel code knows the size of sockaddr_storage */
1858 msg.msg_namelen = 0;
1859 if (sock->file->f_flags & O_NONBLOCK)
1860 flags |= MSG_DONTWAIT;
1861 err = sock_recvmsg(sock, &msg, size, flags);
1863 if (err >= 0 && addr != NULL) {
1864 err2 = move_addr_to_user(&address,
1865 msg.msg_namelen, addr, addr_len);
1870 fput_light(sock->file, fput_needed);
1876 * Receive a datagram from a socket.
1879 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1880 unsigned int, flags)
1882 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1886 * Set a socket option. Because we don't know the option lengths we have
1887 * to pass the user mode parameter for the protocols to sort out.
1890 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1891 char __user *, optval, int, optlen)
1893 int err, fput_needed;
1894 struct socket *sock;
1899 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1901 err = security_socket_setsockopt(sock, level, optname);
1905 if (level == SOL_SOCKET)
1907 sock_setsockopt(sock, level, optname, optval,
1911 sock->ops->setsockopt(sock, level, optname, optval,
1914 fput_light(sock->file, fput_needed);
1920 * Get a socket option. Because we don't know the option lengths we have
1921 * to pass a user mode parameter for the protocols to sort out.
1924 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1925 char __user *, optval, int __user *, optlen)
1927 int err, fput_needed;
1928 struct socket *sock;
1930 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1932 err = security_socket_getsockopt(sock, level, optname);
1936 if (level == SOL_SOCKET)
1938 sock_getsockopt(sock, level, optname, optval,
1942 sock->ops->getsockopt(sock, level, optname, optval,
1945 fput_light(sock->file, fput_needed);
1951 * Shutdown a socket.
1954 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1956 int err, fput_needed;
1957 struct socket *sock;
1959 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1961 err = security_socket_shutdown(sock, how);
1963 err = sock->ops->shutdown(sock, how);
1964 fput_light(sock->file, fput_needed);
1969 /* A couple of helpful macros for getting the address of the 32/64 bit
1970 * fields which are the same type (int / unsigned) on our platforms.
1972 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1973 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1974 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1976 struct used_address {
1977 struct sockaddr_storage name;
1978 unsigned int name_len;
1981 static int copy_msghdr_from_user(struct msghdr *kmsg,
1982 struct msghdr __user *umsg)
1984 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1987 if (kmsg->msg_namelen < 0)
1990 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1991 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1995 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1996 struct msghdr *msg_sys, unsigned int flags,
1997 struct used_address *used_address)
1999 struct compat_msghdr __user *msg_compat =
2000 (struct compat_msghdr __user *)msg;
2001 struct sockaddr_storage address;
2002 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2003 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2004 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2005 /* 20 is size of ipv6_pktinfo */
2006 unsigned char *ctl_buf = ctl;
2007 int err, ctl_len, total_len;
2010 if (MSG_CMSG_COMPAT & flags) {
2011 if (get_compat_msghdr(msg_sys, msg_compat))
2014 err = copy_msghdr_from_user(msg_sys, msg);
2019 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2021 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2024 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2030 /* This will also move the address data into kernel space */
2031 if (MSG_CMSG_COMPAT & flags) {
2032 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2034 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2041 if (msg_sys->msg_controllen > INT_MAX)
2043 ctl_len = msg_sys->msg_controllen;
2044 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2046 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2050 ctl_buf = msg_sys->msg_control;
2051 ctl_len = msg_sys->msg_controllen;
2052 } else if (ctl_len) {
2053 if (ctl_len > sizeof(ctl)) {
2054 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2055 if (ctl_buf == NULL)
2060 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2061 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2062 * checking falls down on this.
2064 if (copy_from_user(ctl_buf,
2065 (void __user __force *)msg_sys->msg_control,
2068 msg_sys->msg_control = ctl_buf;
2070 msg_sys->msg_flags = flags;
2072 if (sock->file->f_flags & O_NONBLOCK)
2073 msg_sys->msg_flags |= MSG_DONTWAIT;
2075 * If this is sendmmsg() and current destination address is same as
2076 * previously succeeded address, omit asking LSM's decision.
2077 * used_address->name_len is initialized to UINT_MAX so that the first
2078 * destination address never matches.
2080 if (used_address && msg_sys->msg_name &&
2081 used_address->name_len == msg_sys->msg_namelen &&
2082 !memcmp(&used_address->name, msg_sys->msg_name,
2083 used_address->name_len)) {
2084 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2087 err = sock_sendmsg(sock, msg_sys, total_len);
2089 * If this is sendmmsg() and sending to current destination address was
2090 * successful, remember it.
2092 if (used_address && err >= 0) {
2093 used_address->name_len = msg_sys->msg_namelen;
2094 if (msg_sys->msg_name)
2095 memcpy(&used_address->name, msg_sys->msg_name,
2096 used_address->name_len);
2101 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2103 if (iov != iovstack)
2110 * BSD sendmsg interface
2113 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2115 int fput_needed, err;
2116 struct msghdr msg_sys;
2117 struct socket *sock;
2119 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2123 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2125 fput_light(sock->file, fput_needed);
2130 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2132 if (flags & MSG_CMSG_COMPAT)
2134 return __sys_sendmsg(fd, msg, flags);
2138 * Linux sendmmsg interface
2141 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2144 int fput_needed, err, datagrams;
2145 struct socket *sock;
2146 struct mmsghdr __user *entry;
2147 struct compat_mmsghdr __user *compat_entry;
2148 struct msghdr msg_sys;
2149 struct used_address used_address;
2151 if (vlen > UIO_MAXIOV)
2156 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2160 used_address.name_len = UINT_MAX;
2162 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2165 while (datagrams < vlen) {
2166 if (MSG_CMSG_COMPAT & flags) {
2167 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2168 &msg_sys, flags, &used_address);
2171 err = __put_user(err, &compat_entry->msg_len);
2174 err = ___sys_sendmsg(sock,
2175 (struct msghdr __user *)entry,
2176 &msg_sys, flags, &used_address);
2179 err = put_user(err, &entry->msg_len);
2188 fput_light(sock->file, fput_needed);
2190 /* We only return an error if no datagrams were able to be sent */
2197 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2198 unsigned int, vlen, unsigned int, flags)
2200 if (flags & MSG_CMSG_COMPAT)
2202 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2205 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2206 struct msghdr *msg_sys, unsigned int flags, int nosec)
2208 struct compat_msghdr __user *msg_compat =
2209 (struct compat_msghdr __user *)msg;
2210 struct iovec iovstack[UIO_FASTIOV];
2211 struct iovec *iov = iovstack;
2212 unsigned long cmsg_ptr;
2213 int err, total_len, len;
2215 /* kernel mode address */
2216 struct sockaddr_storage addr;
2218 /* user mode address pointers */
2219 struct sockaddr __user *uaddr;
2220 int __user *uaddr_len;
2222 if (MSG_CMSG_COMPAT & flags) {
2223 if (get_compat_msghdr(msg_sys, msg_compat))
2226 err = copy_msghdr_from_user(msg_sys, msg);
2231 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2233 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2236 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2242 /* Save the user-mode address (verify_iovec will change the
2243 * kernel msghdr to use the kernel address space)
2245 uaddr = (__force void __user *)msg_sys->msg_name;
2246 uaddr_len = COMPAT_NAMELEN(msg);
2247 if (MSG_CMSG_COMPAT & flags)
2248 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2250 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2255 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2256 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2258 /* We assume all kernel code knows the size of sockaddr_storage */
2259 msg_sys->msg_namelen = 0;
2261 if (sock->file->f_flags & O_NONBLOCK)
2262 flags |= MSG_DONTWAIT;
2263 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2269 if (uaddr != NULL) {
2270 err = move_addr_to_user(&addr,
2271 msg_sys->msg_namelen, uaddr,
2276 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2280 if (MSG_CMSG_COMPAT & flags)
2281 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2282 &msg_compat->msg_controllen);
2284 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2285 &msg->msg_controllen);
2291 if (iov != iovstack)
2298 * BSD recvmsg interface
2301 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2303 int fput_needed, err;
2304 struct msghdr msg_sys;
2305 struct socket *sock;
2307 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2311 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2313 fput_light(sock->file, fput_needed);
2318 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2319 unsigned int, flags)
2321 if (flags & MSG_CMSG_COMPAT)
2323 return __sys_recvmsg(fd, msg, flags);
2327 * Linux recvmmsg interface
2330 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2331 unsigned int flags, struct timespec *timeout)
2333 int fput_needed, err, datagrams;
2334 struct socket *sock;
2335 struct mmsghdr __user *entry;
2336 struct compat_mmsghdr __user *compat_entry;
2337 struct msghdr msg_sys;
2338 struct timespec end_time;
2341 poll_select_set_timeout(&end_time, timeout->tv_sec,
2347 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2351 err = sock_error(sock->sk);
2356 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2358 while (datagrams < vlen) {
2360 * No need to ask LSM for more than the first datagram.
2362 if (MSG_CMSG_COMPAT & flags) {
2363 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2364 &msg_sys, flags & ~MSG_WAITFORONE,
2368 err = __put_user(err, &compat_entry->msg_len);
2371 err = ___sys_recvmsg(sock,
2372 (struct msghdr __user *)entry,
2373 &msg_sys, flags & ~MSG_WAITFORONE,
2377 err = put_user(err, &entry->msg_len);
2385 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2386 if (flags & MSG_WAITFORONE)
2387 flags |= MSG_DONTWAIT;
2390 ktime_get_ts(timeout);
2391 *timeout = timespec_sub(end_time, *timeout);
2392 if (timeout->tv_sec < 0) {
2393 timeout->tv_sec = timeout->tv_nsec = 0;
2397 /* Timeout, return less than vlen datagrams */
2398 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2402 /* Out of band data, return right away */
2403 if (msg_sys.msg_flags & MSG_OOB)
2408 fput_light(sock->file, fput_needed);
2413 if (datagrams != 0) {
2415 * We may return less entries than requested (vlen) if the
2416 * sock is non block and there aren't enough datagrams...
2418 if (err != -EAGAIN) {
2420 * ... or if recvmsg returns an error after we
2421 * received some datagrams, where we record the
2422 * error to return on the next call or if the
2423 * app asks about it using getsockopt(SO_ERROR).
2425 sock->sk->sk_err = -err;
2434 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2435 unsigned int, vlen, unsigned int, flags,
2436 struct timespec __user *, timeout)
2439 struct timespec timeout_sys;
2441 if (flags & MSG_CMSG_COMPAT)
2445 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2447 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2450 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2452 if (datagrams > 0 &&
2453 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2454 datagrams = -EFAULT;
2459 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2460 /* Argument list sizes for sys_socketcall */
2461 #define AL(x) ((x) * sizeof(unsigned long))
2462 static const unsigned char nargs[21] = {
2463 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2464 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2465 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2472 * System call vectors.
2474 * Argument checking cleaned up. Saved 20% in size.
2475 * This function doesn't need to set the kernel lock because
2476 * it is set by the callees.
2479 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2481 unsigned long a[AUDITSC_ARGS];
2482 unsigned long a0, a1;
2486 if (call < 1 || call > SYS_SENDMMSG)
2490 if (len > sizeof(a))
2493 /* copy_from_user should be SMP safe. */
2494 if (copy_from_user(a, args, len))
2497 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2506 err = sys_socket(a0, a1, a[2]);
2509 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2512 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2515 err = sys_listen(a0, a1);
2518 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2519 (int __user *)a[2], 0);
2521 case SYS_GETSOCKNAME:
2523 sys_getsockname(a0, (struct sockaddr __user *)a1,
2524 (int __user *)a[2]);
2526 case SYS_GETPEERNAME:
2528 sys_getpeername(a0, (struct sockaddr __user *)a1,
2529 (int __user *)a[2]);
2531 case SYS_SOCKETPAIR:
2532 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2535 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2538 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2539 (struct sockaddr __user *)a[4], a[5]);
2542 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2545 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2546 (struct sockaddr __user *)a[4],
2547 (int __user *)a[5]);
2550 err = sys_shutdown(a0, a1);
2552 case SYS_SETSOCKOPT:
2553 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2555 case SYS_GETSOCKOPT:
2557 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2558 (int __user *)a[4]);
2561 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2564 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2567 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2570 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2571 (struct timespec __user *)a[4]);
2574 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2575 (int __user *)a[2], a[3]);
2584 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2587 * sock_register - add a socket protocol handler
2588 * @ops: description of protocol
2590 * This function is called by a protocol handler that wants to
2591 * advertise its address family, and have it linked into the
2592 * socket interface. The value ops->family coresponds to the
2593 * socket system call protocol family.
2595 int sock_register(const struct net_proto_family *ops)
2599 if (ops->family >= NPROTO) {
2600 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2604 spin_lock(&net_family_lock);
2605 if (rcu_dereference_protected(net_families[ops->family],
2606 lockdep_is_held(&net_family_lock)))
2609 rcu_assign_pointer(net_families[ops->family], ops);
2612 spin_unlock(&net_family_lock);
2614 pr_info("NET: Registered protocol family %d\n", ops->family);
2617 EXPORT_SYMBOL(sock_register);
2620 * sock_unregister - remove a protocol handler
2621 * @family: protocol family to remove
2623 * This function is called by a protocol handler that wants to
2624 * remove its address family, and have it unlinked from the
2625 * new socket creation.
2627 * If protocol handler is a module, then it can use module reference
2628 * counts to protect against new references. If protocol handler is not
2629 * a module then it needs to provide its own protection in
2630 * the ops->create routine.
2632 void sock_unregister(int family)
2634 BUG_ON(family < 0 || family >= NPROTO);
2636 spin_lock(&net_family_lock);
2637 RCU_INIT_POINTER(net_families[family], NULL);
2638 spin_unlock(&net_family_lock);
2642 pr_info("NET: Unregistered protocol family %d\n", family);
2644 EXPORT_SYMBOL(sock_unregister);
2646 static int __init sock_init(void)
2650 * Initialize the network sysctl infrastructure.
2652 err = net_sysctl_init();
2657 * Initialize skbuff SLAB cache
2662 * Initialize the protocols module.
2667 err = register_filesystem(&sock_fs_type);
2670 sock_mnt = kern_mount(&sock_fs_type);
2671 if (IS_ERR(sock_mnt)) {
2672 err = PTR_ERR(sock_mnt);
2676 /* The real protocol initialization is performed in later initcalls.
2679 #ifdef CONFIG_NETFILTER
2680 err = netfilter_init();
2685 ptp_classifier_init();
2691 unregister_filesystem(&sock_fs_type);
2696 core_initcall(sock_init); /* early initcall */
2698 #ifdef CONFIG_PROC_FS
2699 void socket_seq_show(struct seq_file *seq)
2704 for_each_possible_cpu(cpu)
2705 counter += per_cpu(sockets_in_use, cpu);
2707 /* It can be negative, by the way. 8) */
2711 seq_printf(seq, "sockets: used %d\n", counter);
2713 #endif /* CONFIG_PROC_FS */
2715 #ifdef CONFIG_COMPAT
2716 static int do_siocgstamp(struct net *net, struct socket *sock,
2717 unsigned int cmd, void __user *up)
2719 mm_segment_t old_fs = get_fs();
2724 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2727 err = compat_put_timeval(&ktv, up);
2732 static int do_siocgstampns(struct net *net, struct socket *sock,
2733 unsigned int cmd, void __user *up)
2735 mm_segment_t old_fs = get_fs();
2736 struct timespec kts;
2740 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2743 err = compat_put_timespec(&kts, up);
2748 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2750 struct ifreq __user *uifr;
2753 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2754 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2757 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2761 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2767 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2769 struct compat_ifconf ifc32;
2771 struct ifconf __user *uifc;
2772 struct compat_ifreq __user *ifr32;
2773 struct ifreq __user *ifr;
2777 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2780 memset(&ifc, 0, sizeof(ifc));
2781 if (ifc32.ifcbuf == 0) {
2785 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2787 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2788 sizeof(struct ifreq);
2789 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2791 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2792 ifr32 = compat_ptr(ifc32.ifcbuf);
2793 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2794 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2800 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2803 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2807 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2811 ifr32 = compat_ptr(ifc32.ifcbuf);
2813 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2814 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2815 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2821 if (ifc32.ifcbuf == 0) {
2822 /* Translate from 64-bit structure multiple to
2826 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2831 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2837 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2839 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2840 bool convert_in = false, convert_out = false;
2841 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2842 struct ethtool_rxnfc __user *rxnfc;
2843 struct ifreq __user *ifr;
2844 u32 rule_cnt = 0, actual_rule_cnt;
2849 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2852 compat_rxnfc = compat_ptr(data);
2854 if (get_user(ethcmd, &compat_rxnfc->cmd))
2857 /* Most ethtool structures are defined without padding.
2858 * Unfortunately struct ethtool_rxnfc is an exception.
2863 case ETHTOOL_GRXCLSRLALL:
2864 /* Buffer size is variable */
2865 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2867 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2869 buf_size += rule_cnt * sizeof(u32);
2871 case ETHTOOL_GRXRINGS:
2872 case ETHTOOL_GRXCLSRLCNT:
2873 case ETHTOOL_GRXCLSRULE:
2874 case ETHTOOL_SRXCLSRLINS:
2877 case ETHTOOL_SRXCLSRLDEL:
2878 buf_size += sizeof(struct ethtool_rxnfc);
2883 ifr = compat_alloc_user_space(buf_size);
2884 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2886 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2889 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2890 &ifr->ifr_ifru.ifru_data))
2894 /* We expect there to be holes between fs.m_ext and
2895 * fs.ring_cookie and at the end of fs, but nowhere else.
2897 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2898 sizeof(compat_rxnfc->fs.m_ext) !=
2899 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2900 sizeof(rxnfc->fs.m_ext));
2902 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2903 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2904 offsetof(struct ethtool_rxnfc, fs.location) -
2905 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2907 if (copy_in_user(rxnfc, compat_rxnfc,
2908 (void __user *)(&rxnfc->fs.m_ext + 1) -
2909 (void __user *)rxnfc) ||
2910 copy_in_user(&rxnfc->fs.ring_cookie,
2911 &compat_rxnfc->fs.ring_cookie,
2912 (void __user *)(&rxnfc->fs.location + 1) -
2913 (void __user *)&rxnfc->fs.ring_cookie) ||
2914 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2915 sizeof(rxnfc->rule_cnt)))
2919 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2924 if (copy_in_user(compat_rxnfc, rxnfc,
2925 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2926 (const void __user *)rxnfc) ||
2927 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2928 &rxnfc->fs.ring_cookie,
2929 (const void __user *)(&rxnfc->fs.location + 1) -
2930 (const void __user *)&rxnfc->fs.ring_cookie) ||
2931 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2932 sizeof(rxnfc->rule_cnt)))
2935 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2936 /* As an optimisation, we only copy the actual
2937 * number of rules that the underlying
2938 * function returned. Since Mallory might
2939 * change the rule count in user memory, we
2940 * check that it is less than the rule count
2941 * originally given (as the user buffer size),
2942 * which has been range-checked.
2944 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2946 if (actual_rule_cnt < rule_cnt)
2947 rule_cnt = actual_rule_cnt;
2948 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2949 &rxnfc->rule_locs[0],
2950 rule_cnt * sizeof(u32)))
2958 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2961 compat_uptr_t uptr32;
2962 struct ifreq __user *uifr;
2964 uifr = compat_alloc_user_space(sizeof(*uifr));
2965 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2968 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2971 uptr = compat_ptr(uptr32);
2973 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2976 return dev_ioctl(net, SIOCWANDEV, uifr);
2979 static int bond_ioctl(struct net *net, unsigned int cmd,
2980 struct compat_ifreq __user *ifr32)
2983 mm_segment_t old_fs;
2987 case SIOCBONDENSLAVE:
2988 case SIOCBONDRELEASE:
2989 case SIOCBONDSETHWADDR:
2990 case SIOCBONDCHANGEACTIVE:
2991 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2996 err = dev_ioctl(net, cmd,
2997 (struct ifreq __user __force *) &kifr);
3002 return -ENOIOCTLCMD;
3006 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3007 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3008 struct compat_ifreq __user *u_ifreq32)
3010 struct ifreq __user *u_ifreq64;
3011 char tmp_buf[IFNAMSIZ];
3012 void __user *data64;
3015 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3018 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3020 data64 = compat_ptr(data32);
3022 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3024 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3027 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3030 return dev_ioctl(net, cmd, u_ifreq64);
3033 static int dev_ifsioc(struct net *net, struct socket *sock,
3034 unsigned int cmd, struct compat_ifreq __user *uifr32)
3036 struct ifreq __user *uifr;
3039 uifr = compat_alloc_user_space(sizeof(*uifr));
3040 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3043 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3054 case SIOCGIFBRDADDR:
3055 case SIOCGIFDSTADDR:
3056 case SIOCGIFNETMASK:
3061 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3069 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3070 struct compat_ifreq __user *uifr32)
3073 struct compat_ifmap __user *uifmap32;
3074 mm_segment_t old_fs;
3077 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3078 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3079 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3080 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3081 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3082 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3083 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3084 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3090 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3093 if (cmd == SIOCGIFMAP && !err) {
3094 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3095 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3096 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3097 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3098 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3099 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3100 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3109 struct sockaddr rt_dst; /* target address */
3110 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3111 struct sockaddr rt_genmask; /* target network mask (IP) */
3112 unsigned short rt_flags;
3115 unsigned char rt_tos;
3116 unsigned char rt_class;
3118 short rt_metric; /* +1 for binary compatibility! */
3119 /* char * */ u32 rt_dev; /* forcing the device at add */
3120 u32 rt_mtu; /* per route MTU/Window */
3121 u32 rt_window; /* Window clamping */
3122 unsigned short rt_irtt; /* Initial RTT */
3125 struct in6_rtmsg32 {
3126 struct in6_addr rtmsg_dst;
3127 struct in6_addr rtmsg_src;
3128 struct in6_addr rtmsg_gateway;
3138 static int routing_ioctl(struct net *net, struct socket *sock,
3139 unsigned int cmd, void __user *argp)
3143 struct in6_rtmsg r6;
3147 mm_segment_t old_fs = get_fs();
3149 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3150 struct in6_rtmsg32 __user *ur6 = argp;
3151 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3152 3 * sizeof(struct in6_addr));
3153 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3154 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3155 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3156 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3157 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3158 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3159 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3163 struct rtentry32 __user *ur4 = argp;
3164 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3165 3 * sizeof(struct sockaddr));
3166 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3167 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3168 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3169 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3170 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3171 ret |= get_user(rtdev, &(ur4->rt_dev));
3173 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3174 r4.rt_dev = (char __user __force *)devname;
3188 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3195 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3196 * for some operations; this forces use of the newer bridge-utils that
3197 * use compatible ioctls
3199 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3203 if (get_user(tmp, argp))
3205 if (tmp == BRCTL_GET_VERSION)
3206 return BRCTL_VERSION + 1;
3210 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3211 unsigned int cmd, unsigned long arg)
3213 void __user *argp = compat_ptr(arg);
3214 struct sock *sk = sock->sk;
3215 struct net *net = sock_net(sk);
3217 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3218 return compat_ifr_data_ioctl(net, cmd, argp);
3223 return old_bridge_ioctl(argp);
3225 return dev_ifname32(net, argp);
3227 return dev_ifconf(net, argp);
3229 return ethtool_ioctl(net, argp);
3231 return compat_siocwandev(net, argp);
3234 return compat_sioc_ifmap(net, cmd, argp);
3235 case SIOCBONDENSLAVE:
3236 case SIOCBONDRELEASE:
3237 case SIOCBONDSETHWADDR:
3238 case SIOCBONDCHANGEACTIVE:
3239 return bond_ioctl(net, cmd, argp);
3242 return routing_ioctl(net, sock, cmd, argp);
3244 return do_siocgstamp(net, sock, cmd, argp);
3246 return do_siocgstampns(net, sock, cmd, argp);
3247 case SIOCBONDSLAVEINFOQUERY:
3248 case SIOCBONDINFOQUERY:
3251 return compat_ifr_data_ioctl(net, cmd, argp);
3263 return sock_ioctl(file, cmd, arg);
3280 case SIOCSIFHWBROADCAST:
3282 case SIOCGIFBRDADDR:
3283 case SIOCSIFBRDADDR:
3284 case SIOCGIFDSTADDR:
3285 case SIOCSIFDSTADDR:
3286 case SIOCGIFNETMASK:
3287 case SIOCSIFNETMASK:
3298 return dev_ifsioc(net, sock, cmd, argp);
3304 return sock_do_ioctl(net, sock, cmd, arg);
3307 return -ENOIOCTLCMD;
3310 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3313 struct socket *sock = file->private_data;
3314 int ret = -ENOIOCTLCMD;
3321 if (sock->ops->compat_ioctl)
3322 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3324 if (ret == -ENOIOCTLCMD &&
3325 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3326 ret = compat_wext_handle_ioctl(net, cmd, arg);
3328 if (ret == -ENOIOCTLCMD)
3329 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3335 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3337 return sock->ops->bind(sock, addr, addrlen);
3339 EXPORT_SYMBOL(kernel_bind);
3341 int kernel_listen(struct socket *sock, int backlog)
3343 return sock->ops->listen(sock, backlog);
3345 EXPORT_SYMBOL(kernel_listen);
3347 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3349 struct sock *sk = sock->sk;
3352 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3357 err = sock->ops->accept(sock, *newsock, flags);
3359 sock_release(*newsock);
3364 (*newsock)->ops = sock->ops;
3365 __module_get((*newsock)->ops->owner);
3370 EXPORT_SYMBOL(kernel_accept);
3372 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3375 return sock->ops->connect(sock, addr, addrlen, flags);
3377 EXPORT_SYMBOL(kernel_connect);
3379 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3382 return sock->ops->getname(sock, addr, addrlen, 0);
3384 EXPORT_SYMBOL(kernel_getsockname);
3386 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3389 return sock->ops->getname(sock, addr, addrlen, 1);
3391 EXPORT_SYMBOL(kernel_getpeername);
3393 int kernel_getsockopt(struct socket *sock, int level, int optname,
3394 char *optval, int *optlen)
3396 mm_segment_t oldfs = get_fs();
3397 char __user *uoptval;
3398 int __user *uoptlen;
3401 uoptval = (char __user __force *) optval;
3402 uoptlen = (int __user __force *) optlen;
3405 if (level == SOL_SOCKET)
3406 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3408 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3413 EXPORT_SYMBOL(kernel_getsockopt);
3415 int kernel_setsockopt(struct socket *sock, int level, int optname,
3416 char *optval, unsigned int optlen)
3418 mm_segment_t oldfs = get_fs();
3419 char __user *uoptval;
3422 uoptval = (char __user __force *) optval;
3425 if (level == SOL_SOCKET)
3426 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3428 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3433 EXPORT_SYMBOL(kernel_setsockopt);
3435 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3436 size_t size, int flags)
3438 if (sock->ops->sendpage)
3439 return sock->ops->sendpage(sock, page, offset, size, flags);
3441 return sock_no_sendpage(sock, page, offset, size, flags);
3443 EXPORT_SYMBOL(kernel_sendpage);
3445 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3447 mm_segment_t oldfs = get_fs();
3451 err = sock->ops->ioctl(sock, cmd, arg);
3456 EXPORT_SYMBOL(kernel_sock_ioctl);
3458 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3460 return sock->ops->shutdown(sock, how);
3462 EXPORT_SYMBOL(kernel_sock_shutdown);