2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117 unsigned long nr_segs, loff_t pos);
118 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119 unsigned long nr_segs, loff_t pos);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
122 static int sock_close(struct inode *inode, struct file *file);
123 static unsigned int sock_poll(struct file *file,
124 struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
127 static long compat_sock_ioctl(struct file *file,
128 unsigned int cmd, unsigned long arg);
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_sendpage(struct file *file, struct page *page,
132 int offset, size_t size, loff_t *ppos, int more);
133 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134 struct pipe_inode_info *pipe, size_t len,
138 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
139 * in the operation structures but are done directly via the socketcall() multiplexor.
142 static const struct file_operations socket_file_ops = {
143 .owner = THIS_MODULE,
145 .aio_read = sock_aio_read,
146 .aio_write = sock_aio_write,
148 .unlocked_ioctl = sock_ioctl,
150 .compat_ioctl = compat_sock_ioctl,
153 .release = sock_close,
154 .fasync = sock_fasync,
155 .sendpage = sock_sendpage,
156 .splice_write = generic_splice_sendpage,
157 .splice_read = sock_splice_read,
161 * The protocol list. Each protocol is registered in here.
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
168 * Statistics counters of the socket lists
171 static DEFINE_PER_CPU(int, sockets_in_use);
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
192 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
196 if (copy_from_user(kaddr, uaddr, ulen))
198 return audit_sockaddr(ulen, kaddr);
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219 void __user *uaddr, int __user *ulen)
224 BUG_ON(klen > sizeof(struct sockaddr_storage));
225 err = get_user(len, ulen);
233 if (audit_sockaddr(klen, kaddr))
235 if (copy_to_user(uaddr, kaddr, len))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen, ulen);
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
247 static struct inode *sock_alloc_inode(struct super_block *sb)
249 struct socket_alloc *ei;
250 struct socket_wq *wq;
252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
257 kmem_cache_free(sock_inode_cachep, ei);
260 init_waitqueue_head(&wq->wait);
261 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
270 return &ei->vfs_inode;
273 static void sock_destroy_inode(struct inode *inode)
275 struct socket_alloc *ei;
276 struct socket_wq *wq;
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
279 wq = rcu_dereference_protected(ei->socket.wq, 1);
281 kmem_cache_free(sock_inode_cachep, ei);
284 static void init_once(void *foo)
286 struct socket_alloc *ei = (struct socket_alloc *)foo;
288 inode_init_once(&ei->vfs_inode);
291 static int init_inodecache(void)
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294 sizeof(struct socket_alloc),
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
300 if (sock_inode_cachep == NULL)
305 static const struct super_operations sockfs_ops = {
306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
312 * sockfs_dname() is called from d_path().
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 dentry->d_inode->i_ino);
320 static const struct dentry_operations sockfs_dentry_operations = {
321 .d_dname = sockfs_dname,
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
331 static struct vfsmount *sock_mnt __read_mostly;
333 static struct file_system_type sock_fs_type = {
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
340 * Obtains the first available file descriptor and sets it up for use.
342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
358 struct qstr name = { .name = "" };
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
372 path.mnt = mntget(sock_mnt);
374 d_instantiate(path.dentry, SOCK_INODE(sock));
376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 if (unlikely(IS_ERR(file))) {
379 /* drop dentry, keep inode */
380 ihold(path.dentry->d_inode);
386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387 file->private_data = sock;
390 EXPORT_SYMBOL(sock_alloc_file);
392 static int sock_map_fd(struct socket *sock, int flags)
394 struct file *newfile;
395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
399 newfile = sock_alloc_file(sock, flags, NULL);
400 if (likely(!IS_ERR(newfile))) {
401 fd_install(fd, newfile);
406 return PTR_ERR(newfile);
409 struct socket *sock_from_file(struct file *file, int *err)
411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
417 EXPORT_SYMBOL(sock_from_file);
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket *sockfd_lookup(int fd, int *err)
443 sock = sock_from_file(file, err);
448 EXPORT_SYMBOL(sockfd_lookup);
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452 struct fd f = fdget(fd);
457 sock = sock_from_file(f.file, err);
459 *fput_needed = f.flags;
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
473 const char *proto_name;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
484 if (proto_size + 1 > size)
487 strncpy(value, proto_name, proto_size + 1);
489 error = proto_size + 1;
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
524 static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
530 * sock_alloc - allocate a socket
532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
537 static struct socket *sock_alloc(void)
542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
546 sock = SOCKET_I(inode);
548 kmemcheck_annotate_bitfield(sock, type);
549 inode->i_ino = get_next_ino();
550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
553 inode->i_op = &sockfs_inode_ops;
555 this_cpu_add(sockets_in_use, 1);
560 * sock_release - close a socket
561 * @sock: socket to close
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
568 void sock_release(struct socket *sock)
571 struct module *owner = sock->ops->owner;
573 sock->ops->release(sock);
578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579 pr_err("%s: fasync list not empty!\n", __func__);
581 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
584 this_cpu_sub(sockets_in_use, 1);
586 iput(SOCK_INODE(sock));
591 EXPORT_SYMBOL(sock_release);
593 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
595 u8 flags = *tx_flags;
597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
598 flags |= SKBTX_HW_TSTAMP;
600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
601 flags |= SKBTX_SW_TSTAMP;
603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
604 flags |= SKBTX_SCHED_TSTAMP;
606 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
607 flags |= SKBTX_ACK_TSTAMP;
611 EXPORT_SYMBOL(__sock_tx_timestamp);
613 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
614 struct msghdr *msg, size_t size)
616 struct sock_iocb *si = kiocb_to_siocb(iocb);
623 return sock->ops->sendmsg(iocb, sock, msg, size);
626 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
627 struct msghdr *msg, size_t size)
629 int err = security_socket_sendmsg(sock, msg, size);
631 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
634 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
635 size_t size, bool nosec)
638 struct sock_iocb siocb;
641 init_sync_kiocb(&iocb, NULL);
642 iocb.private = &siocb;
643 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
644 __sock_sendmsg(&iocb, sock, msg, size);
645 if (-EIOCBQUEUED == ret)
646 ret = wait_on_sync_kiocb(&iocb);
650 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
652 return do_sock_sendmsg(sock, msg, size, false);
654 EXPORT_SYMBOL(sock_sendmsg);
656 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
658 return do_sock_sendmsg(sock, msg, size, true);
661 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
662 struct kvec *vec, size_t num, size_t size)
664 mm_segment_t oldfs = get_fs();
669 * the following is safe, since for compiler definitions of kvec and
670 * iovec are identical, yielding the same in-core layout and alignment
672 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
673 result = sock_sendmsg(sock, msg, size);
677 EXPORT_SYMBOL(kernel_sendmsg);
680 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
682 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
685 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
686 struct scm_timestamping tss;
688 struct skb_shared_hwtstamps *shhwtstamps =
691 /* Race occurred between timestamp enabling and packet
692 receiving. Fill in the current time for now. */
693 if (need_software_tstamp && skb->tstamp.tv64 == 0)
694 __net_timestamp(skb);
696 if (need_software_tstamp) {
697 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
699 skb_get_timestamp(skb, &tv);
700 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
704 skb_get_timestampns(skb, &ts);
705 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
710 memset(&tss, 0, sizeof(tss));
711 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
712 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
715 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
716 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
719 put_cmsg(msg, SOL_SOCKET,
720 SCM_TIMESTAMPING, sizeof(tss), &tss);
722 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
724 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
729 if (!sock_flag(sk, SOCK_WIFI_STATUS))
731 if (!skb->wifi_acked_valid)
734 ack = skb->wifi_acked;
736 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
738 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
740 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
743 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
744 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
745 sizeof(__u32), &skb->dropcount);
748 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
751 sock_recv_timestamp(msg, sk, skb);
752 sock_recv_drops(msg, sk, skb);
754 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
756 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
757 struct msghdr *msg, size_t size, int flags)
759 struct sock_iocb *si = kiocb_to_siocb(iocb);
767 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
770 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
771 struct msghdr *msg, size_t size, int flags)
773 int err = security_socket_recvmsg(sock, msg, size, flags);
775 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
778 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
779 size_t size, int flags)
782 struct sock_iocb siocb;
785 init_sync_kiocb(&iocb, NULL);
786 iocb.private = &siocb;
787 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
788 if (-EIOCBQUEUED == ret)
789 ret = wait_on_sync_kiocb(&iocb);
792 EXPORT_SYMBOL(sock_recvmsg);
794 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
795 size_t size, int flags)
798 struct sock_iocb siocb;
801 init_sync_kiocb(&iocb, NULL);
802 iocb.private = &siocb;
803 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
804 if (-EIOCBQUEUED == ret)
805 ret = wait_on_sync_kiocb(&iocb);
810 * kernel_recvmsg - Receive a message from a socket (kernel space)
811 * @sock: The socket to receive the message from
812 * @msg: Received message
813 * @vec: Input s/g array for message data
814 * @num: Size of input s/g array
815 * @size: Number of bytes to read
816 * @flags: Message flags (MSG_DONTWAIT, etc...)
818 * On return the msg structure contains the scatter/gather array passed in the
819 * vec argument. The array is modified so that it consists of the unfilled
820 * portion of the original array.
822 * The returned value is the total number of bytes received, or an error.
824 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
825 struct kvec *vec, size_t num, size_t size, int flags)
827 mm_segment_t oldfs = get_fs();
832 * the following is safe, since for compiler definitions of kvec and
833 * iovec are identical, yielding the same in-core layout and alignment
835 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
836 result = sock_recvmsg(sock, msg, size, flags);
840 EXPORT_SYMBOL(kernel_recvmsg);
842 static ssize_t sock_sendpage(struct file *file, struct page *page,
843 int offset, size_t size, loff_t *ppos, int more)
848 sock = file->private_data;
850 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
851 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
854 return kernel_sendpage(sock, page, offset, size, flags);
857 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
858 struct pipe_inode_info *pipe, size_t len,
861 struct socket *sock = file->private_data;
863 if (unlikely(!sock->ops->splice_read))
866 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
869 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
870 struct sock_iocb *siocb)
873 iocb->private = siocb;
877 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
878 struct file *file, const struct iovec *iov,
879 unsigned long nr_segs)
881 struct socket *sock = file->private_data;
883 msg->msg_name = NULL;
884 msg->msg_namelen = 0;
885 msg->msg_control = NULL;
886 msg->msg_controllen = 0;
887 iov_iter_init(&msg->msg_iter, READ, iov, nr_segs, iocb->ki_nbytes);
888 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
890 return __sock_recvmsg(iocb, sock, msg, iocb->ki_nbytes, msg->msg_flags);
893 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
894 unsigned long nr_segs, loff_t pos)
896 struct sock_iocb siocb, *x;
901 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
905 x = alloc_sock_iocb(iocb, &siocb);
908 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
911 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
912 struct file *file, const struct iovec *iov,
913 unsigned long nr_segs)
915 struct socket *sock = file->private_data;
917 msg->msg_name = NULL;
918 msg->msg_namelen = 0;
919 msg->msg_control = NULL;
920 msg->msg_controllen = 0;
921 iov_iter_init(&msg->msg_iter, WRITE, iov, nr_segs, iocb->ki_nbytes);
922 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
923 if (sock->type == SOCK_SEQPACKET)
924 msg->msg_flags |= MSG_EOR;
926 return __sock_sendmsg(iocb, sock, msg, iocb->ki_nbytes);
929 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
930 unsigned long nr_segs, loff_t pos)
932 struct sock_iocb siocb, *x;
937 x = alloc_sock_iocb(iocb, &siocb);
941 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
945 * Atomic setting of ioctl hooks to avoid race
946 * with module unload.
949 static DEFINE_MUTEX(br_ioctl_mutex);
950 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
952 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
954 mutex_lock(&br_ioctl_mutex);
955 br_ioctl_hook = hook;
956 mutex_unlock(&br_ioctl_mutex);
958 EXPORT_SYMBOL(brioctl_set);
960 static DEFINE_MUTEX(vlan_ioctl_mutex);
961 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
963 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
965 mutex_lock(&vlan_ioctl_mutex);
966 vlan_ioctl_hook = hook;
967 mutex_unlock(&vlan_ioctl_mutex);
969 EXPORT_SYMBOL(vlan_ioctl_set);
971 static DEFINE_MUTEX(dlci_ioctl_mutex);
972 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
974 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
976 mutex_lock(&dlci_ioctl_mutex);
977 dlci_ioctl_hook = hook;
978 mutex_unlock(&dlci_ioctl_mutex);
980 EXPORT_SYMBOL(dlci_ioctl_set);
982 static long sock_do_ioctl(struct net *net, struct socket *sock,
983 unsigned int cmd, unsigned long arg)
986 void __user *argp = (void __user *)arg;
988 err = sock->ops->ioctl(sock, cmd, arg);
991 * If this ioctl is unknown try to hand it down
994 if (err == -ENOIOCTLCMD)
995 err = dev_ioctl(net, cmd, argp);
1001 * With an ioctl, arg may well be a user mode pointer, but we don't know
1002 * what to do with it - that's up to the protocol still.
1005 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1007 struct socket *sock;
1009 void __user *argp = (void __user *)arg;
1013 sock = file->private_data;
1016 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1017 err = dev_ioctl(net, cmd, argp);
1019 #ifdef CONFIG_WEXT_CORE
1020 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1021 err = dev_ioctl(net, cmd, argp);
1028 if (get_user(pid, (int __user *)argp))
1030 f_setown(sock->file, pid, 1);
1035 err = put_user(f_getown(sock->file),
1036 (int __user *)argp);
1044 request_module("bridge");
1046 mutex_lock(&br_ioctl_mutex);
1048 err = br_ioctl_hook(net, cmd, argp);
1049 mutex_unlock(&br_ioctl_mutex);
1054 if (!vlan_ioctl_hook)
1055 request_module("8021q");
1057 mutex_lock(&vlan_ioctl_mutex);
1058 if (vlan_ioctl_hook)
1059 err = vlan_ioctl_hook(net, argp);
1060 mutex_unlock(&vlan_ioctl_mutex);
1065 if (!dlci_ioctl_hook)
1066 request_module("dlci");
1068 mutex_lock(&dlci_ioctl_mutex);
1069 if (dlci_ioctl_hook)
1070 err = dlci_ioctl_hook(cmd, argp);
1071 mutex_unlock(&dlci_ioctl_mutex);
1074 err = sock_do_ioctl(net, sock, cmd, arg);
1080 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1083 struct socket *sock = NULL;
1085 err = security_socket_create(family, type, protocol, 1);
1089 sock = sock_alloc();
1096 err = security_socket_post_create(sock, family, type, protocol, 1);
1108 EXPORT_SYMBOL(sock_create_lite);
1110 /* No kernel lock held - perfect */
1111 static unsigned int sock_poll(struct file *file, poll_table *wait)
1113 unsigned int busy_flag = 0;
1114 struct socket *sock;
1117 * We can't return errors to poll, so it's either yes or no.
1119 sock = file->private_data;
1121 if (sk_can_busy_loop(sock->sk)) {
1122 /* this socket can poll_ll so tell the system call */
1123 busy_flag = POLL_BUSY_LOOP;
1125 /* once, only if requested by syscall */
1126 if (wait && (wait->_key & POLL_BUSY_LOOP))
1127 sk_busy_loop(sock->sk, 1);
1130 return busy_flag | sock->ops->poll(file, sock, wait);
1133 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1135 struct socket *sock = file->private_data;
1137 return sock->ops->mmap(file, sock, vma);
1140 static int sock_close(struct inode *inode, struct file *filp)
1142 sock_release(SOCKET_I(inode));
1147 * Update the socket async list
1149 * Fasync_list locking strategy.
1151 * 1. fasync_list is modified only under process context socket lock
1152 * i.e. under semaphore.
1153 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1154 * or under socket lock
1157 static int sock_fasync(int fd, struct file *filp, int on)
1159 struct socket *sock = filp->private_data;
1160 struct sock *sk = sock->sk;
1161 struct socket_wq *wq;
1167 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1168 fasync_helper(fd, filp, on, &wq->fasync_list);
1170 if (!wq->fasync_list)
1171 sock_reset_flag(sk, SOCK_FASYNC);
1173 sock_set_flag(sk, SOCK_FASYNC);
1179 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1181 int sock_wake_async(struct socket *sock, int how, int band)
1183 struct socket_wq *wq;
1188 wq = rcu_dereference(sock->wq);
1189 if (!wq || !wq->fasync_list) {
1194 case SOCK_WAKE_WAITD:
1195 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1198 case SOCK_WAKE_SPACE:
1199 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1204 kill_fasync(&wq->fasync_list, SIGIO, band);
1207 kill_fasync(&wq->fasync_list, SIGURG, band);
1212 EXPORT_SYMBOL(sock_wake_async);
1214 int __sock_create(struct net *net, int family, int type, int protocol,
1215 struct socket **res, int kern)
1218 struct socket *sock;
1219 const struct net_proto_family *pf;
1222 * Check protocol is in range
1224 if (family < 0 || family >= NPROTO)
1225 return -EAFNOSUPPORT;
1226 if (type < 0 || type >= SOCK_MAX)
1231 This uglymoron is moved from INET layer to here to avoid
1232 deadlock in module load.
1234 if (family == PF_INET && type == SOCK_PACKET) {
1238 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1244 err = security_socket_create(family, type, protocol, kern);
1249 * Allocate the socket and allow the family to set things up. if
1250 * the protocol is 0, the family is instructed to select an appropriate
1253 sock = sock_alloc();
1255 net_warn_ratelimited("socket: no more sockets\n");
1256 return -ENFILE; /* Not exactly a match, but its the
1257 closest posix thing */
1262 #ifdef CONFIG_MODULES
1263 /* Attempt to load a protocol module if the find failed.
1265 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1266 * requested real, full-featured networking support upon configuration.
1267 * Otherwise module support will break!
1269 if (rcu_access_pointer(net_families[family]) == NULL)
1270 request_module("net-pf-%d", family);
1274 pf = rcu_dereference(net_families[family]);
1275 err = -EAFNOSUPPORT;
1280 * We will call the ->create function, that possibly is in a loadable
1281 * module, so we have to bump that loadable module refcnt first.
1283 if (!try_module_get(pf->owner))
1286 /* Now protected by module ref count */
1289 err = pf->create(net, sock, protocol, kern);
1291 goto out_module_put;
1294 * Now to bump the refcnt of the [loadable] module that owns this
1295 * socket at sock_release time we decrement its refcnt.
1297 if (!try_module_get(sock->ops->owner))
1298 goto out_module_busy;
1301 * Now that we're done with the ->create function, the [loadable]
1302 * module can have its refcnt decremented
1304 module_put(pf->owner);
1305 err = security_socket_post_create(sock, family, type, protocol, kern);
1307 goto out_sock_release;
1313 err = -EAFNOSUPPORT;
1316 module_put(pf->owner);
1323 goto out_sock_release;
1325 EXPORT_SYMBOL(__sock_create);
1327 int sock_create(int family, int type, int protocol, struct socket **res)
1329 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1331 EXPORT_SYMBOL(sock_create);
1333 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1335 return __sock_create(&init_net, family, type, protocol, res, 1);
1337 EXPORT_SYMBOL(sock_create_kern);
1339 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1342 struct socket *sock;
1345 /* Check the SOCK_* constants for consistency. */
1346 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1347 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1348 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1349 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
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;
1359 retval = sock_create(family, type, protocol, &sock);
1363 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1368 /* It may be already another descriptor 8) Not kernel problem. */
1377 * Create a pair of connected sockets.
1380 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1381 int __user *, usockvec)
1383 struct socket *sock1, *sock2;
1385 struct file *newfile1, *newfile2;
1388 flags = type & ~SOCK_TYPE_MASK;
1389 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1391 type &= SOCK_TYPE_MASK;
1393 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1394 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1397 * Obtain the first socket and check if the underlying protocol
1398 * supports the socketpair call.
1401 err = sock_create(family, type, protocol, &sock1);
1405 err = sock_create(family, type, protocol, &sock2);
1409 err = sock1->ops->socketpair(sock1, sock2);
1411 goto out_release_both;
1413 fd1 = get_unused_fd_flags(flags);
1414 if (unlikely(fd1 < 0)) {
1416 goto out_release_both;
1419 fd2 = get_unused_fd_flags(flags);
1420 if (unlikely(fd2 < 0)) {
1422 goto out_put_unused_1;
1425 newfile1 = sock_alloc_file(sock1, flags, NULL);
1426 if (unlikely(IS_ERR(newfile1))) {
1427 err = PTR_ERR(newfile1);
1428 goto out_put_unused_both;
1431 newfile2 = sock_alloc_file(sock2, flags, NULL);
1432 if (IS_ERR(newfile2)) {
1433 err = PTR_ERR(newfile2);
1437 err = put_user(fd1, &usockvec[0]);
1441 err = put_user(fd2, &usockvec[1]);
1445 audit_fd_pair(fd1, fd2);
1447 fd_install(fd1, newfile1);
1448 fd_install(fd2, newfile2);
1449 /* fd1 and fd2 may be already another descriptors.
1450 * Not kernel problem.
1466 sock_release(sock2);
1469 out_put_unused_both:
1474 sock_release(sock2);
1476 sock_release(sock1);
1482 * Bind a name to a socket. Nothing much to do here since it's
1483 * the protocol's responsibility to handle the local address.
1485 * We move the socket address to kernel space before we call
1486 * the protocol layer (having also checked the address is ok).
1489 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1491 struct socket *sock;
1492 struct sockaddr_storage address;
1493 int err, fput_needed;
1495 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1497 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1499 err = security_socket_bind(sock,
1500 (struct sockaddr *)&address,
1503 err = sock->ops->bind(sock,
1507 fput_light(sock->file, fput_needed);
1513 * Perform a listen. Basically, we allow the protocol to do anything
1514 * necessary for a listen, and if that works, we mark the socket as
1515 * ready for listening.
1518 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1520 struct socket *sock;
1521 int err, fput_needed;
1524 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1526 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1527 if ((unsigned int)backlog > somaxconn)
1528 backlog = somaxconn;
1530 err = security_socket_listen(sock, backlog);
1532 err = sock->ops->listen(sock, backlog);
1534 fput_light(sock->file, fput_needed);
1540 * For accept, we attempt to create a new socket, set up the link
1541 * with the client, wake up the client, then return the new
1542 * connected fd. We collect the address of the connector in kernel
1543 * space and move it to user at the very end. This is unclean because
1544 * we open the socket then return an error.
1546 * 1003.1g adds the ability to recvmsg() to query connection pending
1547 * status to recvmsg. We need to add that support in a way thats
1548 * clean when we restucture accept also.
1551 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1552 int __user *, upeer_addrlen, int, flags)
1554 struct socket *sock, *newsock;
1555 struct file *newfile;
1556 int err, len, newfd, fput_needed;
1557 struct sockaddr_storage address;
1559 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1562 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1563 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1565 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1570 newsock = sock_alloc();
1574 newsock->type = sock->type;
1575 newsock->ops = sock->ops;
1578 * We don't need try_module_get here, as the listening socket (sock)
1579 * has the protocol module (sock->ops->owner) held.
1581 __module_get(newsock->ops->owner);
1583 newfd = get_unused_fd_flags(flags);
1584 if (unlikely(newfd < 0)) {
1586 sock_release(newsock);
1589 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1590 if (unlikely(IS_ERR(newfile))) {
1591 err = PTR_ERR(newfile);
1592 put_unused_fd(newfd);
1593 sock_release(newsock);
1597 err = security_socket_accept(sock, newsock);
1601 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1605 if (upeer_sockaddr) {
1606 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1608 err = -ECONNABORTED;
1611 err = move_addr_to_user(&address,
1612 len, upeer_sockaddr, upeer_addrlen);
1617 /* File flags are not inherited via accept() unlike another OSes. */
1619 fd_install(newfd, newfile);
1623 fput_light(sock->file, fput_needed);
1628 put_unused_fd(newfd);
1632 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1633 int __user *, upeer_addrlen)
1635 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1639 * Attempt to connect to a socket with the server address. The address
1640 * is in user space so we verify it is OK and move it to kernel space.
1642 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1645 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1646 * other SEQPACKET protocols that take time to connect() as it doesn't
1647 * include the -EINPROGRESS status for such sockets.
1650 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1653 struct socket *sock;
1654 struct sockaddr_storage address;
1655 int err, fput_needed;
1657 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1660 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1665 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1669 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1670 sock->file->f_flags);
1672 fput_light(sock->file, fput_needed);
1678 * Get the local address ('name') of a socket object. Move the obtained
1679 * name to user space.
1682 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1683 int __user *, usockaddr_len)
1685 struct socket *sock;
1686 struct sockaddr_storage address;
1687 int len, err, fput_needed;
1689 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1693 err = security_socket_getsockname(sock);
1697 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1700 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1703 fput_light(sock->file, fput_needed);
1709 * Get the remote address ('name') of a socket object. Move the obtained
1710 * name to user space.
1713 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1714 int __user *, usockaddr_len)
1716 struct socket *sock;
1717 struct sockaddr_storage address;
1718 int len, err, fput_needed;
1720 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1722 err = security_socket_getpeername(sock);
1724 fput_light(sock->file, fput_needed);
1729 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1732 err = move_addr_to_user(&address, len, usockaddr,
1734 fput_light(sock->file, fput_needed);
1740 * Send a datagram to a given address. We move the address into kernel
1741 * space and check the user space data area is readable before invoking
1745 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1746 unsigned int, flags, struct sockaddr __user *, addr,
1749 struct socket *sock;
1750 struct sockaddr_storage address;
1758 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1762 iov.iov_base = buff;
1764 msg.msg_name = NULL;
1765 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1766 msg.msg_control = NULL;
1767 msg.msg_controllen = 0;
1768 msg.msg_namelen = 0;
1770 err = move_addr_to_kernel(addr, addr_len, &address);
1773 msg.msg_name = (struct sockaddr *)&address;
1774 msg.msg_namelen = addr_len;
1776 if (sock->file->f_flags & O_NONBLOCK)
1777 flags |= MSG_DONTWAIT;
1778 msg.msg_flags = flags;
1779 err = sock_sendmsg(sock, &msg, len);
1782 fput_light(sock->file, fput_needed);
1788 * Send a datagram down a socket.
1791 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1792 unsigned int, flags)
1794 return sys_sendto(fd, buff, len, flags, NULL, 0);
1798 * Receive a frame from the socket and optionally record the address of the
1799 * sender. We verify the buffers are writable and if needed move the
1800 * sender address from kernel to user space.
1803 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1804 unsigned int, flags, struct sockaddr __user *, addr,
1805 int __user *, addr_len)
1807 struct socket *sock;
1810 struct sockaddr_storage address;
1816 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1820 msg.msg_control = NULL;
1821 msg.msg_controllen = 0;
1823 iov.iov_base = ubuf;
1824 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1825 /* Save some cycles and don't copy the address if not needed */
1826 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1827 /* We assume all kernel code knows the size of sockaddr_storage */
1828 msg.msg_namelen = 0;
1829 if (sock->file->f_flags & O_NONBLOCK)
1830 flags |= MSG_DONTWAIT;
1831 err = sock_recvmsg(sock, &msg, size, flags);
1833 if (err >= 0 && addr != NULL) {
1834 err2 = move_addr_to_user(&address,
1835 msg.msg_namelen, addr, addr_len);
1840 fput_light(sock->file, fput_needed);
1846 * Receive a datagram from a socket.
1849 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1850 unsigned int, flags)
1852 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1856 * Set a socket option. Because we don't know the option lengths we have
1857 * to pass the user mode parameter for the protocols to sort out.
1860 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1861 char __user *, optval, int, optlen)
1863 int err, fput_needed;
1864 struct socket *sock;
1869 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1871 err = security_socket_setsockopt(sock, level, optname);
1875 if (level == SOL_SOCKET)
1877 sock_setsockopt(sock, level, optname, optval,
1881 sock->ops->setsockopt(sock, level, optname, optval,
1884 fput_light(sock->file, fput_needed);
1890 * Get a socket option. Because we don't know the option lengths we have
1891 * to pass a user mode parameter for the protocols to sort out.
1894 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1895 char __user *, optval, int __user *, optlen)
1897 int err, fput_needed;
1898 struct socket *sock;
1900 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1902 err = security_socket_getsockopt(sock, level, optname);
1906 if (level == SOL_SOCKET)
1908 sock_getsockopt(sock, level, optname, optval,
1912 sock->ops->getsockopt(sock, level, optname, optval,
1915 fput_light(sock->file, fput_needed);
1921 * Shutdown a socket.
1924 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1926 int err, fput_needed;
1927 struct socket *sock;
1929 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1931 err = security_socket_shutdown(sock, how);
1933 err = sock->ops->shutdown(sock, how);
1934 fput_light(sock->file, fput_needed);
1939 /* A couple of helpful macros for getting the address of the 32/64 bit
1940 * fields which are the same type (int / unsigned) on our platforms.
1942 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1943 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1944 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1946 struct used_address {
1947 struct sockaddr_storage name;
1948 unsigned int name_len;
1951 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1952 struct user_msghdr __user *umsg,
1953 struct sockaddr __user **save_addr,
1956 struct sockaddr __user *uaddr;
1957 struct iovec __user *uiov;
1961 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1962 __get_user(uaddr, &umsg->msg_name) ||
1963 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1964 __get_user(uiov, &umsg->msg_iov) ||
1965 __get_user(nr_segs, &umsg->msg_iovlen) ||
1966 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1967 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1968 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1972 kmsg->msg_namelen = 0;
1974 if (kmsg->msg_namelen < 0)
1977 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1978 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1983 if (uaddr && kmsg->msg_namelen) {
1985 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1991 kmsg->msg_name = NULL;
1992 kmsg->msg_namelen = 0;
1995 if (nr_segs > UIO_MAXIOV)
1998 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
2000 UIO_FASTIOV, *iov, iov);
2002 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
2003 *iov, nr_segs, err);
2007 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2008 struct msghdr *msg_sys, unsigned int flags,
2009 struct used_address *used_address)
2011 struct compat_msghdr __user *msg_compat =
2012 (struct compat_msghdr __user *)msg;
2013 struct sockaddr_storage address;
2014 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2015 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2016 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2017 /* 20 is size of ipv6_pktinfo */
2018 unsigned char *ctl_buf = ctl;
2019 int ctl_len, total_len;
2022 msg_sys->msg_name = &address;
2024 if (MSG_CMSG_COMPAT & flags)
2025 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2027 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2034 if (msg_sys->msg_controllen > INT_MAX)
2036 ctl_len = msg_sys->msg_controllen;
2037 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2039 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2043 ctl_buf = msg_sys->msg_control;
2044 ctl_len = msg_sys->msg_controllen;
2045 } else if (ctl_len) {
2046 if (ctl_len > sizeof(ctl)) {
2047 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2048 if (ctl_buf == NULL)
2053 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2054 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2055 * checking falls down on this.
2057 if (copy_from_user(ctl_buf,
2058 (void __user __force *)msg_sys->msg_control,
2061 msg_sys->msg_control = ctl_buf;
2063 msg_sys->msg_flags = flags;
2065 if (sock->file->f_flags & O_NONBLOCK)
2066 msg_sys->msg_flags |= MSG_DONTWAIT;
2068 * If this is sendmmsg() and current destination address is same as
2069 * previously succeeded address, omit asking LSM's decision.
2070 * used_address->name_len is initialized to UINT_MAX so that the first
2071 * destination address never matches.
2073 if (used_address && msg_sys->msg_name &&
2074 used_address->name_len == msg_sys->msg_namelen &&
2075 !memcmp(&used_address->name, msg_sys->msg_name,
2076 used_address->name_len)) {
2077 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2080 err = sock_sendmsg(sock, msg_sys, total_len);
2082 * If this is sendmmsg() and sending to current destination address was
2083 * successful, remember it.
2085 if (used_address && err >= 0) {
2086 used_address->name_len = msg_sys->msg_namelen;
2087 if (msg_sys->msg_name)
2088 memcpy(&used_address->name, msg_sys->msg_name,
2089 used_address->name_len);
2094 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2096 if (iov != iovstack)
2102 * BSD sendmsg interface
2105 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2107 int fput_needed, err;
2108 struct msghdr msg_sys;
2109 struct socket *sock;
2111 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2115 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2117 fput_light(sock->file, fput_needed);
2122 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2124 if (flags & MSG_CMSG_COMPAT)
2126 return __sys_sendmsg(fd, msg, flags);
2130 * Linux sendmmsg interface
2133 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2136 int fput_needed, err, datagrams;
2137 struct socket *sock;
2138 struct mmsghdr __user *entry;
2139 struct compat_mmsghdr __user *compat_entry;
2140 struct msghdr msg_sys;
2141 struct used_address used_address;
2143 if (vlen > UIO_MAXIOV)
2148 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2152 used_address.name_len = UINT_MAX;
2154 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2157 while (datagrams < vlen) {
2158 if (MSG_CMSG_COMPAT & flags) {
2159 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2160 &msg_sys, flags, &used_address);
2163 err = __put_user(err, &compat_entry->msg_len);
2166 err = ___sys_sendmsg(sock,
2167 (struct user_msghdr __user *)entry,
2168 &msg_sys, flags, &used_address);
2171 err = put_user(err, &entry->msg_len);
2180 fput_light(sock->file, fput_needed);
2182 /* We only return an error if no datagrams were able to be sent */
2189 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2190 unsigned int, vlen, unsigned int, flags)
2192 if (flags & MSG_CMSG_COMPAT)
2194 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2197 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2198 struct msghdr *msg_sys, unsigned int flags, int nosec)
2200 struct compat_msghdr __user *msg_compat =
2201 (struct compat_msghdr __user *)msg;
2202 struct iovec iovstack[UIO_FASTIOV];
2203 struct iovec *iov = iovstack;
2204 unsigned long cmsg_ptr;
2208 /* kernel mode address */
2209 struct sockaddr_storage addr;
2211 /* user mode address pointers */
2212 struct sockaddr __user *uaddr;
2213 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2215 msg_sys->msg_name = &addr;
2217 if (MSG_CMSG_COMPAT & flags)
2218 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2220 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2225 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2226 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2228 /* We assume all kernel code knows the size of sockaddr_storage */
2229 msg_sys->msg_namelen = 0;
2231 if (sock->file->f_flags & O_NONBLOCK)
2232 flags |= MSG_DONTWAIT;
2233 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2239 if (uaddr != NULL) {
2240 err = move_addr_to_user(&addr,
2241 msg_sys->msg_namelen, uaddr,
2246 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2250 if (MSG_CMSG_COMPAT & flags)
2251 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2252 &msg_compat->msg_controllen);
2254 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2255 &msg->msg_controllen);
2261 if (iov != iovstack)
2267 * BSD recvmsg interface
2270 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2272 int fput_needed, err;
2273 struct msghdr msg_sys;
2274 struct socket *sock;
2276 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2280 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2282 fput_light(sock->file, fput_needed);
2287 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2288 unsigned int, flags)
2290 if (flags & MSG_CMSG_COMPAT)
2292 return __sys_recvmsg(fd, msg, flags);
2296 * Linux recvmmsg interface
2299 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2300 unsigned int flags, struct timespec *timeout)
2302 int fput_needed, err, datagrams;
2303 struct socket *sock;
2304 struct mmsghdr __user *entry;
2305 struct compat_mmsghdr __user *compat_entry;
2306 struct msghdr msg_sys;
2307 struct timespec end_time;
2310 poll_select_set_timeout(&end_time, timeout->tv_sec,
2316 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2320 err = sock_error(sock->sk);
2325 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2327 while (datagrams < vlen) {
2329 * No need to ask LSM for more than the first datagram.
2331 if (MSG_CMSG_COMPAT & flags) {
2332 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2333 &msg_sys, flags & ~MSG_WAITFORONE,
2337 err = __put_user(err, &compat_entry->msg_len);
2340 err = ___sys_recvmsg(sock,
2341 (struct user_msghdr __user *)entry,
2342 &msg_sys, flags & ~MSG_WAITFORONE,
2346 err = put_user(err, &entry->msg_len);
2354 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2355 if (flags & MSG_WAITFORONE)
2356 flags |= MSG_DONTWAIT;
2359 ktime_get_ts(timeout);
2360 *timeout = timespec_sub(end_time, *timeout);
2361 if (timeout->tv_sec < 0) {
2362 timeout->tv_sec = timeout->tv_nsec = 0;
2366 /* Timeout, return less than vlen datagrams */
2367 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2371 /* Out of band data, return right away */
2372 if (msg_sys.msg_flags & MSG_OOB)
2377 fput_light(sock->file, fput_needed);
2382 if (datagrams != 0) {
2384 * We may return less entries than requested (vlen) if the
2385 * sock is non block and there aren't enough datagrams...
2387 if (err != -EAGAIN) {
2389 * ... or if recvmsg returns an error after we
2390 * received some datagrams, where we record the
2391 * error to return on the next call or if the
2392 * app asks about it using getsockopt(SO_ERROR).
2394 sock->sk->sk_err = -err;
2403 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2404 unsigned int, vlen, unsigned int, flags,
2405 struct timespec __user *, timeout)
2408 struct timespec timeout_sys;
2410 if (flags & MSG_CMSG_COMPAT)
2414 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2416 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2419 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2421 if (datagrams > 0 &&
2422 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2423 datagrams = -EFAULT;
2428 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2429 /* Argument list sizes for sys_socketcall */
2430 #define AL(x) ((x) * sizeof(unsigned long))
2431 static const unsigned char nargs[21] = {
2432 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2433 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2434 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2441 * System call vectors.
2443 * Argument checking cleaned up. Saved 20% in size.
2444 * This function doesn't need to set the kernel lock because
2445 * it is set by the callees.
2448 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2450 unsigned long a[AUDITSC_ARGS];
2451 unsigned long a0, a1;
2455 if (call < 1 || call > SYS_SENDMMSG)
2459 if (len > sizeof(a))
2462 /* copy_from_user should be SMP safe. */
2463 if (copy_from_user(a, args, len))
2466 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2475 err = sys_socket(a0, a1, a[2]);
2478 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2481 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2484 err = sys_listen(a0, a1);
2487 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2488 (int __user *)a[2], 0);
2490 case SYS_GETSOCKNAME:
2492 sys_getsockname(a0, (struct sockaddr __user *)a1,
2493 (int __user *)a[2]);
2495 case SYS_GETPEERNAME:
2497 sys_getpeername(a0, (struct sockaddr __user *)a1,
2498 (int __user *)a[2]);
2500 case SYS_SOCKETPAIR:
2501 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2504 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2507 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2508 (struct sockaddr __user *)a[4], a[5]);
2511 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2514 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2515 (struct sockaddr __user *)a[4],
2516 (int __user *)a[5]);
2519 err = sys_shutdown(a0, a1);
2521 case SYS_SETSOCKOPT:
2522 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2524 case SYS_GETSOCKOPT:
2526 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2527 (int __user *)a[4]);
2530 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2533 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2536 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2539 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2540 (struct timespec __user *)a[4]);
2543 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2544 (int __user *)a[2], a[3]);
2553 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2556 * sock_register - add a socket protocol handler
2557 * @ops: description of protocol
2559 * This function is called by a protocol handler that wants to
2560 * advertise its address family, and have it linked into the
2561 * socket interface. The value ops->family corresponds to the
2562 * socket system call protocol family.
2564 int sock_register(const struct net_proto_family *ops)
2568 if (ops->family >= NPROTO) {
2569 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2573 spin_lock(&net_family_lock);
2574 if (rcu_dereference_protected(net_families[ops->family],
2575 lockdep_is_held(&net_family_lock)))
2578 rcu_assign_pointer(net_families[ops->family], ops);
2581 spin_unlock(&net_family_lock);
2583 pr_info("NET: Registered protocol family %d\n", ops->family);
2586 EXPORT_SYMBOL(sock_register);
2589 * sock_unregister - remove a protocol handler
2590 * @family: protocol family to remove
2592 * This function is called by a protocol handler that wants to
2593 * remove its address family, and have it unlinked from the
2594 * new socket creation.
2596 * If protocol handler is a module, then it can use module reference
2597 * counts to protect against new references. If protocol handler is not
2598 * a module then it needs to provide its own protection in
2599 * the ops->create routine.
2601 void sock_unregister(int family)
2603 BUG_ON(family < 0 || family >= NPROTO);
2605 spin_lock(&net_family_lock);
2606 RCU_INIT_POINTER(net_families[family], NULL);
2607 spin_unlock(&net_family_lock);
2611 pr_info("NET: Unregistered protocol family %d\n", family);
2613 EXPORT_SYMBOL(sock_unregister);
2615 static int __init sock_init(void)
2619 * Initialize the network sysctl infrastructure.
2621 err = net_sysctl_init();
2626 * Initialize skbuff SLAB cache
2631 * Initialize the protocols module.
2636 err = register_filesystem(&sock_fs_type);
2639 sock_mnt = kern_mount(&sock_fs_type);
2640 if (IS_ERR(sock_mnt)) {
2641 err = PTR_ERR(sock_mnt);
2645 /* The real protocol initialization is performed in later initcalls.
2648 #ifdef CONFIG_NETFILTER
2649 err = netfilter_init();
2654 ptp_classifier_init();
2660 unregister_filesystem(&sock_fs_type);
2665 core_initcall(sock_init); /* early initcall */
2667 #ifdef CONFIG_PROC_FS
2668 void socket_seq_show(struct seq_file *seq)
2673 for_each_possible_cpu(cpu)
2674 counter += per_cpu(sockets_in_use, cpu);
2676 /* It can be negative, by the way. 8) */
2680 seq_printf(seq, "sockets: used %d\n", counter);
2682 #endif /* CONFIG_PROC_FS */
2684 #ifdef CONFIG_COMPAT
2685 static int do_siocgstamp(struct net *net, struct socket *sock,
2686 unsigned int cmd, void __user *up)
2688 mm_segment_t old_fs = get_fs();
2693 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2696 err = compat_put_timeval(&ktv, up);
2701 static int do_siocgstampns(struct net *net, struct socket *sock,
2702 unsigned int cmd, void __user *up)
2704 mm_segment_t old_fs = get_fs();
2705 struct timespec kts;
2709 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2712 err = compat_put_timespec(&kts, up);
2717 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2719 struct ifreq __user *uifr;
2722 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2723 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2726 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2730 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2736 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2738 struct compat_ifconf ifc32;
2740 struct ifconf __user *uifc;
2741 struct compat_ifreq __user *ifr32;
2742 struct ifreq __user *ifr;
2746 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2749 memset(&ifc, 0, sizeof(ifc));
2750 if (ifc32.ifcbuf == 0) {
2754 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2756 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2757 sizeof(struct ifreq);
2758 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2760 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2761 ifr32 = compat_ptr(ifc32.ifcbuf);
2762 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2763 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2769 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2772 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2776 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2780 ifr32 = compat_ptr(ifc32.ifcbuf);
2782 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2783 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2784 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2790 if (ifc32.ifcbuf == 0) {
2791 /* Translate from 64-bit structure multiple to
2795 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2800 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2806 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2808 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2809 bool convert_in = false, convert_out = false;
2810 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2811 struct ethtool_rxnfc __user *rxnfc;
2812 struct ifreq __user *ifr;
2813 u32 rule_cnt = 0, actual_rule_cnt;
2818 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2821 compat_rxnfc = compat_ptr(data);
2823 if (get_user(ethcmd, &compat_rxnfc->cmd))
2826 /* Most ethtool structures are defined without padding.
2827 * Unfortunately struct ethtool_rxnfc is an exception.
2832 case ETHTOOL_GRXCLSRLALL:
2833 /* Buffer size is variable */
2834 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2836 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2838 buf_size += rule_cnt * sizeof(u32);
2840 case ETHTOOL_GRXRINGS:
2841 case ETHTOOL_GRXCLSRLCNT:
2842 case ETHTOOL_GRXCLSRULE:
2843 case ETHTOOL_SRXCLSRLINS:
2846 case ETHTOOL_SRXCLSRLDEL:
2847 buf_size += sizeof(struct ethtool_rxnfc);
2852 ifr = compat_alloc_user_space(buf_size);
2853 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2855 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2858 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2859 &ifr->ifr_ifru.ifru_data))
2863 /* We expect there to be holes between fs.m_ext and
2864 * fs.ring_cookie and at the end of fs, but nowhere else.
2866 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2867 sizeof(compat_rxnfc->fs.m_ext) !=
2868 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2869 sizeof(rxnfc->fs.m_ext));
2871 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2872 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2873 offsetof(struct ethtool_rxnfc, fs.location) -
2874 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2876 if (copy_in_user(rxnfc, compat_rxnfc,
2877 (void __user *)(&rxnfc->fs.m_ext + 1) -
2878 (void __user *)rxnfc) ||
2879 copy_in_user(&rxnfc->fs.ring_cookie,
2880 &compat_rxnfc->fs.ring_cookie,
2881 (void __user *)(&rxnfc->fs.location + 1) -
2882 (void __user *)&rxnfc->fs.ring_cookie) ||
2883 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2884 sizeof(rxnfc->rule_cnt)))
2888 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2893 if (copy_in_user(compat_rxnfc, rxnfc,
2894 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2895 (const void __user *)rxnfc) ||
2896 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2897 &rxnfc->fs.ring_cookie,
2898 (const void __user *)(&rxnfc->fs.location + 1) -
2899 (const void __user *)&rxnfc->fs.ring_cookie) ||
2900 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2901 sizeof(rxnfc->rule_cnt)))
2904 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2905 /* As an optimisation, we only copy the actual
2906 * number of rules that the underlying
2907 * function returned. Since Mallory might
2908 * change the rule count in user memory, we
2909 * check that it is less than the rule count
2910 * originally given (as the user buffer size),
2911 * which has been range-checked.
2913 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2915 if (actual_rule_cnt < rule_cnt)
2916 rule_cnt = actual_rule_cnt;
2917 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2918 &rxnfc->rule_locs[0],
2919 rule_cnt * sizeof(u32)))
2927 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2930 compat_uptr_t uptr32;
2931 struct ifreq __user *uifr;
2933 uifr = compat_alloc_user_space(sizeof(*uifr));
2934 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2937 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2940 uptr = compat_ptr(uptr32);
2942 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2945 return dev_ioctl(net, SIOCWANDEV, uifr);
2948 static int bond_ioctl(struct net *net, unsigned int cmd,
2949 struct compat_ifreq __user *ifr32)
2952 mm_segment_t old_fs;
2956 case SIOCBONDENSLAVE:
2957 case SIOCBONDRELEASE:
2958 case SIOCBONDSETHWADDR:
2959 case SIOCBONDCHANGEACTIVE:
2960 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2965 err = dev_ioctl(net, cmd,
2966 (struct ifreq __user __force *) &kifr);
2971 return -ENOIOCTLCMD;
2975 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2976 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2977 struct compat_ifreq __user *u_ifreq32)
2979 struct ifreq __user *u_ifreq64;
2980 char tmp_buf[IFNAMSIZ];
2981 void __user *data64;
2984 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2987 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2989 data64 = compat_ptr(data32);
2991 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2993 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2996 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2999 return dev_ioctl(net, cmd, u_ifreq64);
3002 static int dev_ifsioc(struct net *net, struct socket *sock,
3003 unsigned int cmd, struct compat_ifreq __user *uifr32)
3005 struct ifreq __user *uifr;
3008 uifr = compat_alloc_user_space(sizeof(*uifr));
3009 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3012 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3023 case SIOCGIFBRDADDR:
3024 case SIOCGIFDSTADDR:
3025 case SIOCGIFNETMASK:
3030 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3038 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3039 struct compat_ifreq __user *uifr32)
3042 struct compat_ifmap __user *uifmap32;
3043 mm_segment_t old_fs;
3046 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3047 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3048 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3049 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3050 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3051 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3052 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3053 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3059 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3062 if (cmd == SIOCGIFMAP && !err) {
3063 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3064 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3065 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3066 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3067 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3068 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3069 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3078 struct sockaddr rt_dst; /* target address */
3079 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3080 struct sockaddr rt_genmask; /* target network mask (IP) */
3081 unsigned short rt_flags;
3084 unsigned char rt_tos;
3085 unsigned char rt_class;
3087 short rt_metric; /* +1 for binary compatibility! */
3088 /* char * */ u32 rt_dev; /* forcing the device at add */
3089 u32 rt_mtu; /* per route MTU/Window */
3090 u32 rt_window; /* Window clamping */
3091 unsigned short rt_irtt; /* Initial RTT */
3094 struct in6_rtmsg32 {
3095 struct in6_addr rtmsg_dst;
3096 struct in6_addr rtmsg_src;
3097 struct in6_addr rtmsg_gateway;
3107 static int routing_ioctl(struct net *net, struct socket *sock,
3108 unsigned int cmd, void __user *argp)
3112 struct in6_rtmsg r6;
3116 mm_segment_t old_fs = get_fs();
3118 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3119 struct in6_rtmsg32 __user *ur6 = argp;
3120 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3121 3 * sizeof(struct in6_addr));
3122 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3123 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3124 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3125 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3126 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3127 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3128 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3132 struct rtentry32 __user *ur4 = argp;
3133 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3134 3 * sizeof(struct sockaddr));
3135 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3136 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3137 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3138 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3139 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3140 ret |= get_user(rtdev, &(ur4->rt_dev));
3142 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3143 r4.rt_dev = (char __user __force *)devname;
3157 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3164 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3165 * for some operations; this forces use of the newer bridge-utils that
3166 * use compatible ioctls
3168 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3172 if (get_user(tmp, argp))
3174 if (tmp == BRCTL_GET_VERSION)
3175 return BRCTL_VERSION + 1;
3179 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3180 unsigned int cmd, unsigned long arg)
3182 void __user *argp = compat_ptr(arg);
3183 struct sock *sk = sock->sk;
3184 struct net *net = sock_net(sk);
3186 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3187 return compat_ifr_data_ioctl(net, cmd, argp);
3192 return old_bridge_ioctl(argp);
3194 return dev_ifname32(net, argp);
3196 return dev_ifconf(net, argp);
3198 return ethtool_ioctl(net, argp);
3200 return compat_siocwandev(net, argp);
3203 return compat_sioc_ifmap(net, cmd, argp);
3204 case SIOCBONDENSLAVE:
3205 case SIOCBONDRELEASE:
3206 case SIOCBONDSETHWADDR:
3207 case SIOCBONDCHANGEACTIVE:
3208 return bond_ioctl(net, cmd, argp);
3211 return routing_ioctl(net, sock, cmd, argp);
3213 return do_siocgstamp(net, sock, cmd, argp);
3215 return do_siocgstampns(net, sock, cmd, argp);
3216 case SIOCBONDSLAVEINFOQUERY:
3217 case SIOCBONDINFOQUERY:
3220 return compat_ifr_data_ioctl(net, cmd, argp);
3232 return sock_ioctl(file, cmd, arg);
3249 case SIOCSIFHWBROADCAST:
3251 case SIOCGIFBRDADDR:
3252 case SIOCSIFBRDADDR:
3253 case SIOCGIFDSTADDR:
3254 case SIOCSIFDSTADDR:
3255 case SIOCGIFNETMASK:
3256 case SIOCSIFNETMASK:
3267 return dev_ifsioc(net, sock, cmd, argp);
3273 return sock_do_ioctl(net, sock, cmd, arg);
3276 return -ENOIOCTLCMD;
3279 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3282 struct socket *sock = file->private_data;
3283 int ret = -ENOIOCTLCMD;
3290 if (sock->ops->compat_ioctl)
3291 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3293 if (ret == -ENOIOCTLCMD &&
3294 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3295 ret = compat_wext_handle_ioctl(net, cmd, arg);
3297 if (ret == -ENOIOCTLCMD)
3298 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3304 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3306 return sock->ops->bind(sock, addr, addrlen);
3308 EXPORT_SYMBOL(kernel_bind);
3310 int kernel_listen(struct socket *sock, int backlog)
3312 return sock->ops->listen(sock, backlog);
3314 EXPORT_SYMBOL(kernel_listen);
3316 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3318 struct sock *sk = sock->sk;
3321 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3326 err = sock->ops->accept(sock, *newsock, flags);
3328 sock_release(*newsock);
3333 (*newsock)->ops = sock->ops;
3334 __module_get((*newsock)->ops->owner);
3339 EXPORT_SYMBOL(kernel_accept);
3341 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3344 return sock->ops->connect(sock, addr, addrlen, flags);
3346 EXPORT_SYMBOL(kernel_connect);
3348 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3351 return sock->ops->getname(sock, addr, addrlen, 0);
3353 EXPORT_SYMBOL(kernel_getsockname);
3355 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3358 return sock->ops->getname(sock, addr, addrlen, 1);
3360 EXPORT_SYMBOL(kernel_getpeername);
3362 int kernel_getsockopt(struct socket *sock, int level, int optname,
3363 char *optval, int *optlen)
3365 mm_segment_t oldfs = get_fs();
3366 char __user *uoptval;
3367 int __user *uoptlen;
3370 uoptval = (char __user __force *) optval;
3371 uoptlen = (int __user __force *) optlen;
3374 if (level == SOL_SOCKET)
3375 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3377 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3382 EXPORT_SYMBOL(kernel_getsockopt);
3384 int kernel_setsockopt(struct socket *sock, int level, int optname,
3385 char *optval, unsigned int optlen)
3387 mm_segment_t oldfs = get_fs();
3388 char __user *uoptval;
3391 uoptval = (char __user __force *) optval;
3394 if (level == SOL_SOCKET)
3395 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3397 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3402 EXPORT_SYMBOL(kernel_setsockopt);
3404 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3405 size_t size, int flags)
3407 if (sock->ops->sendpage)
3408 return sock->ops->sendpage(sock, page, offset, size, flags);
3410 return sock_no_sendpage(sock, page, offset, size, flags);
3412 EXPORT_SYMBOL(kernel_sendpage);
3414 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3416 mm_segment_t oldfs = get_fs();
3420 err = sock->ops->ioctl(sock, cmd, arg);
3425 EXPORT_SYMBOL(kernel_sock_ioctl);
3427 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3429 return sock->ops->shutdown(sock, how);
3431 EXPORT_SYMBOL(kernel_sock_shutdown);