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_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
140 static const struct file_operations socket_file_ops = {
141 .owner = THIS_MODULE,
143 .read_iter = sock_read_iter,
144 .write_iter = sock_write_iter,
146 .unlocked_ioctl = sock_ioctl,
148 .compat_ioctl = compat_sock_ioctl,
151 .release = sock_close,
152 .fasync = sock_fasync,
153 .sendpage = sock_sendpage,
154 .splice_write = generic_splice_sendpage,
155 .splice_read = sock_splice_read,
159 * The protocol list. Each protocol is registered in here.
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
166 * Statistics counters of the socket lists
169 static DEFINE_PER_CPU(int, sockets_in_use);
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
190 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
194 if (copy_from_user(kaddr, uaddr, ulen))
196 return audit_sockaddr(ulen, kaddr);
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217 void __user *uaddr, int __user *ulen)
222 BUG_ON(klen > sizeof(struct sockaddr_storage));
223 err = get_user(len, ulen);
231 if (audit_sockaddr(klen, kaddr))
233 if (copy_to_user(uaddr, kaddr, len))
237 * "fromlen shall refer to the value before truncation.."
240 return __put_user(klen, ulen);
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
245 static struct inode *sock_alloc_inode(struct super_block *sb)
247 struct socket_alloc *ei;
248 struct socket_wq *wq;
250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 kmem_cache_free(sock_inode_cachep, ei);
258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
269 return &ei->vfs_inode;
272 static void sock_destroy_inode(struct inode *inode)
274 struct socket_alloc *ei;
275 struct socket_wq *wq;
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
278 wq = rcu_dereference_protected(ei->socket.wq, 1);
280 kmem_cache_free(sock_inode_cachep, ei);
283 static void init_once(void *foo)
285 struct socket_alloc *ei = (struct socket_alloc *)foo;
287 inode_init_once(&ei->vfs_inode);
290 static int init_inodecache(void)
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293 sizeof(struct socket_alloc),
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
297 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
299 if (sock_inode_cachep == NULL)
304 static const struct super_operations sockfs_ops = {
305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
311 * sockfs_dname() is called from d_path().
313 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316 d_inode(dentry)->i_ino);
319 static const struct dentry_operations sockfs_dentry_operations = {
320 .d_dname = sockfs_dname,
323 static int sockfs_xattr_get(const struct xattr_handler *handler,
324 struct dentry *dentry, struct inode *inode,
325 const char *suffix, void *value, size_t size)
328 if (dentry->d_name.len + 1 > size)
330 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
332 return dentry->d_name.len + 1;
335 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
336 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
337 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
339 static const struct xattr_handler sockfs_xattr_handler = {
340 .name = XATTR_NAME_SOCKPROTONAME,
341 .get = sockfs_xattr_get,
344 static const struct xattr_handler *sockfs_xattr_handlers[] = {
345 &sockfs_xattr_handler,
349 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
350 int flags, const char *dev_name, void *data)
352 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
353 sockfs_xattr_handlers,
354 &sockfs_dentry_operations, SOCKFS_MAGIC);
357 static struct vfsmount *sock_mnt __read_mostly;
359 static struct file_system_type sock_fs_type = {
361 .mount = sockfs_mount,
362 .kill_sb = kill_anon_super,
366 * Obtains the first available file descriptor and sets it up for use.
368 * These functions create file structures and maps them to fd space
369 * of the current process. On success it returns file descriptor
370 * and file struct implicitly stored in sock->file.
371 * Note that another thread may close file descriptor before we return
372 * from this function. We use the fact that now we do not refer
373 * to socket after mapping. If one day we will need it, this
374 * function will increment ref. count on file by 1.
376 * In any case returned fd MAY BE not valid!
377 * This race condition is unavoidable
378 * with shared fd spaces, we cannot solve it inside kernel,
379 * but we take care of internal coherence yet.
382 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
384 struct qstr name = { .name = "" };
390 name.len = strlen(name.name);
391 } else if (sock->sk) {
392 name.name = sock->sk->sk_prot_creator->name;
393 name.len = strlen(name.name);
395 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
396 if (unlikely(!path.dentry))
397 return ERR_PTR(-ENOMEM);
398 path.mnt = mntget(sock_mnt);
400 d_instantiate(path.dentry, SOCK_INODE(sock));
402 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
405 /* drop dentry, keep inode */
406 ihold(d_inode(path.dentry));
412 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
413 file->private_data = sock;
416 EXPORT_SYMBOL(sock_alloc_file);
418 static int sock_map_fd(struct socket *sock, int flags)
420 struct file *newfile;
421 int fd = get_unused_fd_flags(flags);
422 if (unlikely(fd < 0))
425 newfile = sock_alloc_file(sock, flags, NULL);
426 if (likely(!IS_ERR(newfile))) {
427 fd_install(fd, newfile);
432 return PTR_ERR(newfile);
435 struct socket *sock_from_file(struct file *file, int *err)
437 if (file->f_op == &socket_file_ops)
438 return file->private_data; /* set in sock_map_fd */
443 EXPORT_SYMBOL(sock_from_file);
446 * sockfd_lookup - Go from a file number to its socket slot
448 * @err: pointer to an error code return
450 * The file handle passed in is locked and the socket it is bound
451 * too is returned. If an error occurs the err pointer is overwritten
452 * with a negative errno code and NULL is returned. The function checks
453 * for both invalid handles and passing a handle which is not a socket.
455 * On a success the socket object pointer is returned.
458 struct socket *sockfd_lookup(int fd, int *err)
469 sock = sock_from_file(file, err);
474 EXPORT_SYMBOL(sockfd_lookup);
476 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
478 struct fd f = fdget(fd);
483 sock = sock_from_file(f.file, err);
485 *fput_needed = f.flags;
493 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
499 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
509 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
514 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
521 static const struct inode_operations sockfs_inode_ops = {
522 .getxattr = generic_getxattr,
523 .listxattr = sockfs_listxattr,
527 * sock_alloc - allocate a socket
529 * Allocate a new inode and socket object. The two are bound together
530 * and initialised. The socket is then returned. If we are out of inodes
534 struct socket *sock_alloc(void)
539 inode = new_inode_pseudo(sock_mnt->mnt_sb);
543 sock = SOCKET_I(inode);
545 kmemcheck_annotate_bitfield(sock, type);
546 inode->i_ino = get_next_ino();
547 inode->i_mode = S_IFSOCK | S_IRWXUGO;
548 inode->i_uid = current_fsuid();
549 inode->i_gid = current_fsgid();
550 inode->i_op = &sockfs_inode_ops;
552 this_cpu_add(sockets_in_use, 1);
555 EXPORT_SYMBOL(sock_alloc);
558 * sock_release - close a socket
559 * @sock: socket to close
561 * The socket is released from the protocol stack if it has a release
562 * callback, and the inode is then released if the socket is bound to
563 * an inode not a file.
566 void sock_release(struct socket *sock)
569 struct module *owner = sock->ops->owner;
571 sock->ops->release(sock);
576 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
577 pr_err("%s: fasync list not empty!\n", __func__);
579 this_cpu_sub(sockets_in_use, 1);
581 iput(SOCK_INODE(sock));
586 EXPORT_SYMBOL(sock_release);
588 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
590 u8 flags = *tx_flags;
592 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
593 flags |= SKBTX_HW_TSTAMP;
595 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
596 flags |= SKBTX_SW_TSTAMP;
598 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
599 flags |= SKBTX_SCHED_TSTAMP;
603 EXPORT_SYMBOL(__sock_tx_timestamp);
605 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
607 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
608 BUG_ON(ret == -EIOCBQUEUED);
612 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
614 int err = security_socket_sendmsg(sock, msg,
617 return err ?: sock_sendmsg_nosec(sock, msg);
619 EXPORT_SYMBOL(sock_sendmsg);
621 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
622 struct kvec *vec, size_t num, size_t size)
624 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
625 return sock_sendmsg(sock, msg);
627 EXPORT_SYMBOL(kernel_sendmsg);
630 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
632 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
635 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
636 struct scm_timestamping tss;
638 struct skb_shared_hwtstamps *shhwtstamps =
641 /* Race occurred between timestamp enabling and packet
642 receiving. Fill in the current time for now. */
643 if (need_software_tstamp && skb->tstamp.tv64 == 0)
644 __net_timestamp(skb);
646 if (need_software_tstamp) {
647 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
649 skb_get_timestamp(skb, &tv);
650 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
654 skb_get_timestampns(skb, &ts);
655 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
660 memset(&tss, 0, sizeof(tss));
661 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
662 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
665 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
666 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
669 put_cmsg(msg, SOL_SOCKET,
670 SCM_TIMESTAMPING, sizeof(tss), &tss);
672 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
674 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
679 if (!sock_flag(sk, SOCK_WIFI_STATUS))
681 if (!skb->wifi_acked_valid)
684 ack = skb->wifi_acked;
686 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
688 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
690 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
693 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
694 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
695 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
698 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
701 sock_recv_timestamp(msg, sk, skb);
702 sock_recv_drops(msg, sk, skb);
704 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
706 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
709 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
712 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
714 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
716 return err ?: sock_recvmsg_nosec(sock, msg, flags);
718 EXPORT_SYMBOL(sock_recvmsg);
721 * kernel_recvmsg - Receive a message from a socket (kernel space)
722 * @sock: The socket to receive the message from
723 * @msg: Received message
724 * @vec: Input s/g array for message data
725 * @num: Size of input s/g array
726 * @size: Number of bytes to read
727 * @flags: Message flags (MSG_DONTWAIT, etc...)
729 * On return the msg structure contains the scatter/gather array passed in the
730 * vec argument. The array is modified so that it consists of the unfilled
731 * portion of the original array.
733 * The returned value is the total number of bytes received, or an error.
735 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
736 struct kvec *vec, size_t num, size_t size, int flags)
738 mm_segment_t oldfs = get_fs();
741 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
743 result = sock_recvmsg(sock, msg, flags);
747 EXPORT_SYMBOL(kernel_recvmsg);
749 static ssize_t sock_sendpage(struct file *file, struct page *page,
750 int offset, size_t size, loff_t *ppos, int more)
755 sock = file->private_data;
757 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
758 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
761 return kernel_sendpage(sock, page, offset, size, flags);
764 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
765 struct pipe_inode_info *pipe, size_t len,
768 struct socket *sock = file->private_data;
770 if (unlikely(!sock->ops->splice_read))
773 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
776 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
778 struct file *file = iocb->ki_filp;
779 struct socket *sock = file->private_data;
780 struct msghdr msg = {.msg_iter = *to,
784 if (file->f_flags & O_NONBLOCK)
785 msg.msg_flags = MSG_DONTWAIT;
787 if (iocb->ki_pos != 0)
790 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
793 res = sock_recvmsg(sock, &msg, msg.msg_flags);
798 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
800 struct file *file = iocb->ki_filp;
801 struct socket *sock = file->private_data;
802 struct msghdr msg = {.msg_iter = *from,
806 if (iocb->ki_pos != 0)
809 if (file->f_flags & O_NONBLOCK)
810 msg.msg_flags = MSG_DONTWAIT;
812 if (sock->type == SOCK_SEQPACKET)
813 msg.msg_flags |= MSG_EOR;
815 res = sock_sendmsg(sock, &msg);
816 *from = msg.msg_iter;
821 * Atomic setting of ioctl hooks to avoid race
822 * with module unload.
825 static DEFINE_MUTEX(br_ioctl_mutex);
826 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
828 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
830 mutex_lock(&br_ioctl_mutex);
831 br_ioctl_hook = hook;
832 mutex_unlock(&br_ioctl_mutex);
834 EXPORT_SYMBOL(brioctl_set);
836 static DEFINE_MUTEX(vlan_ioctl_mutex);
837 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
839 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
841 mutex_lock(&vlan_ioctl_mutex);
842 vlan_ioctl_hook = hook;
843 mutex_unlock(&vlan_ioctl_mutex);
845 EXPORT_SYMBOL(vlan_ioctl_set);
847 static DEFINE_MUTEX(dlci_ioctl_mutex);
848 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
850 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
852 mutex_lock(&dlci_ioctl_mutex);
853 dlci_ioctl_hook = hook;
854 mutex_unlock(&dlci_ioctl_mutex);
856 EXPORT_SYMBOL(dlci_ioctl_set);
858 static long sock_do_ioctl(struct net *net, struct socket *sock,
859 unsigned int cmd, unsigned long arg)
862 void __user *argp = (void __user *)arg;
864 err = sock->ops->ioctl(sock, cmd, arg);
867 * If this ioctl is unknown try to hand it down
870 if (err == -ENOIOCTLCMD)
871 err = dev_ioctl(net, cmd, argp);
877 * With an ioctl, arg may well be a user mode pointer, but we don't know
878 * what to do with it - that's up to the protocol still.
881 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
885 void __user *argp = (void __user *)arg;
889 sock = file->private_data;
892 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
893 err = dev_ioctl(net, cmd, argp);
895 #ifdef CONFIG_WEXT_CORE
896 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
897 err = dev_ioctl(net, cmd, argp);
904 if (get_user(pid, (int __user *)argp))
906 f_setown(sock->file, pid, 1);
911 err = put_user(f_getown(sock->file),
920 request_module("bridge");
922 mutex_lock(&br_ioctl_mutex);
924 err = br_ioctl_hook(net, cmd, argp);
925 mutex_unlock(&br_ioctl_mutex);
930 if (!vlan_ioctl_hook)
931 request_module("8021q");
933 mutex_lock(&vlan_ioctl_mutex);
935 err = vlan_ioctl_hook(net, argp);
936 mutex_unlock(&vlan_ioctl_mutex);
941 if (!dlci_ioctl_hook)
942 request_module("dlci");
944 mutex_lock(&dlci_ioctl_mutex);
946 err = dlci_ioctl_hook(cmd, argp);
947 mutex_unlock(&dlci_ioctl_mutex);
950 err = sock_do_ioctl(net, sock, cmd, arg);
956 int sock_create_lite(int family, int type, int protocol, struct socket **res)
959 struct socket *sock = NULL;
961 err = security_socket_create(family, type, protocol, 1);
972 err = security_socket_post_create(sock, family, type, protocol, 1);
984 EXPORT_SYMBOL(sock_create_lite);
986 /* No kernel lock held - perfect */
987 static unsigned int sock_poll(struct file *file, poll_table *wait)
989 unsigned int busy_flag = 0;
993 * We can't return errors to poll, so it's either yes or no.
995 sock = file->private_data;
997 if (sk_can_busy_loop(sock->sk)) {
998 /* this socket can poll_ll so tell the system call */
999 busy_flag = POLL_BUSY_LOOP;
1001 /* once, only if requested by syscall */
1002 if (wait && (wait->_key & POLL_BUSY_LOOP))
1003 sk_busy_loop(sock->sk, 1);
1006 return busy_flag | sock->ops->poll(file, sock, wait);
1009 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1011 struct socket *sock = file->private_data;
1013 return sock->ops->mmap(file, sock, vma);
1016 static int sock_close(struct inode *inode, struct file *filp)
1018 sock_release(SOCKET_I(inode));
1023 * Update the socket async list
1025 * Fasync_list locking strategy.
1027 * 1. fasync_list is modified only under process context socket lock
1028 * i.e. under semaphore.
1029 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1030 * or under socket lock
1033 static int sock_fasync(int fd, struct file *filp, int on)
1035 struct socket *sock = filp->private_data;
1036 struct sock *sk = sock->sk;
1037 struct socket_wq *wq;
1043 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1044 fasync_helper(fd, filp, on, &wq->fasync_list);
1046 if (!wq->fasync_list)
1047 sock_reset_flag(sk, SOCK_FASYNC);
1049 sock_set_flag(sk, SOCK_FASYNC);
1055 /* This function may be called only under rcu_lock */
1057 int sock_wake_async(struct socket_wq *wq, int how, int band)
1059 if (!wq || !wq->fasync_list)
1063 case SOCK_WAKE_WAITD:
1064 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1067 case SOCK_WAKE_SPACE:
1068 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1073 kill_fasync(&wq->fasync_list, SIGIO, band);
1076 kill_fasync(&wq->fasync_list, SIGURG, band);
1081 EXPORT_SYMBOL(sock_wake_async);
1083 int __sock_create(struct net *net, int family, int type, int protocol,
1084 struct socket **res, int kern)
1087 struct socket *sock;
1088 const struct net_proto_family *pf;
1091 * Check protocol is in range
1093 if (family < 0 || family >= NPROTO)
1094 return -EAFNOSUPPORT;
1095 if (type < 0 || type >= SOCK_MAX)
1100 This uglymoron is moved from INET layer to here to avoid
1101 deadlock in module load.
1103 if (family == PF_INET && type == SOCK_PACKET) {
1104 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1109 err = security_socket_create(family, type, protocol, kern);
1114 * Allocate the socket and allow the family to set things up. if
1115 * the protocol is 0, the family is instructed to select an appropriate
1118 sock = sock_alloc();
1120 net_warn_ratelimited("socket: no more sockets\n");
1121 return -ENFILE; /* Not exactly a match, but its the
1122 closest posix thing */
1127 #ifdef CONFIG_MODULES
1128 /* Attempt to load a protocol module if the find failed.
1130 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1131 * requested real, full-featured networking support upon configuration.
1132 * Otherwise module support will break!
1134 if (rcu_access_pointer(net_families[family]) == NULL)
1135 request_module("net-pf-%d", family);
1139 pf = rcu_dereference(net_families[family]);
1140 err = -EAFNOSUPPORT;
1145 * We will call the ->create function, that possibly is in a loadable
1146 * module, so we have to bump that loadable module refcnt first.
1148 if (!try_module_get(pf->owner))
1151 /* Now protected by module ref count */
1154 err = pf->create(net, sock, protocol, kern);
1156 goto out_module_put;
1159 * Now to bump the refcnt of the [loadable] module that owns this
1160 * socket at sock_release time we decrement its refcnt.
1162 if (!try_module_get(sock->ops->owner))
1163 goto out_module_busy;
1166 * Now that we're done with the ->create function, the [loadable]
1167 * module can have its refcnt decremented
1169 module_put(pf->owner);
1170 err = security_socket_post_create(sock, family, type, protocol, kern);
1172 goto out_sock_release;
1178 err = -EAFNOSUPPORT;
1181 module_put(pf->owner);
1188 goto out_sock_release;
1190 EXPORT_SYMBOL(__sock_create);
1192 int sock_create(int family, int type, int protocol, struct socket **res)
1194 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1196 EXPORT_SYMBOL(sock_create);
1198 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1200 return __sock_create(net, family, type, protocol, res, 1);
1202 EXPORT_SYMBOL(sock_create_kern);
1204 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1207 struct socket *sock;
1210 /* Check the SOCK_* constants for consistency. */
1211 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1212 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1213 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1214 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1216 flags = type & ~SOCK_TYPE_MASK;
1217 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1219 type &= SOCK_TYPE_MASK;
1221 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1222 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1224 retval = sock_create(family, type, protocol, &sock);
1228 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1233 /* It may be already another descriptor 8) Not kernel problem. */
1242 * Create a pair of connected sockets.
1245 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1246 int __user *, usockvec)
1248 struct socket *sock1, *sock2;
1250 struct file *newfile1, *newfile2;
1253 flags = type & ~SOCK_TYPE_MASK;
1254 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1256 type &= SOCK_TYPE_MASK;
1258 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1259 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1262 * Obtain the first socket and check if the underlying protocol
1263 * supports the socketpair call.
1266 err = sock_create(family, type, protocol, &sock1);
1270 err = sock_create(family, type, protocol, &sock2);
1274 err = sock1->ops->socketpair(sock1, sock2);
1276 goto out_release_both;
1278 fd1 = get_unused_fd_flags(flags);
1279 if (unlikely(fd1 < 0)) {
1281 goto out_release_both;
1284 fd2 = get_unused_fd_flags(flags);
1285 if (unlikely(fd2 < 0)) {
1287 goto out_put_unused_1;
1290 newfile1 = sock_alloc_file(sock1, flags, NULL);
1291 if (IS_ERR(newfile1)) {
1292 err = PTR_ERR(newfile1);
1293 goto out_put_unused_both;
1296 newfile2 = sock_alloc_file(sock2, flags, NULL);
1297 if (IS_ERR(newfile2)) {
1298 err = PTR_ERR(newfile2);
1302 err = put_user(fd1, &usockvec[0]);
1306 err = put_user(fd2, &usockvec[1]);
1310 audit_fd_pair(fd1, fd2);
1312 fd_install(fd1, newfile1);
1313 fd_install(fd2, newfile2);
1314 /* fd1 and fd2 may be already another descriptors.
1315 * Not kernel problem.
1331 sock_release(sock2);
1334 out_put_unused_both:
1339 sock_release(sock2);
1341 sock_release(sock1);
1347 * Bind a name to a socket. Nothing much to do here since it's
1348 * the protocol's responsibility to handle the local address.
1350 * We move the socket address to kernel space before we call
1351 * the protocol layer (having also checked the address is ok).
1354 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1356 struct socket *sock;
1357 struct sockaddr_storage address;
1358 int err, fput_needed;
1360 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1362 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1364 err = security_socket_bind(sock,
1365 (struct sockaddr *)&address,
1368 err = sock->ops->bind(sock,
1372 fput_light(sock->file, fput_needed);
1378 * Perform a listen. Basically, we allow the protocol to do anything
1379 * necessary for a listen, and if that works, we mark the socket as
1380 * ready for listening.
1383 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1385 struct socket *sock;
1386 int err, fput_needed;
1389 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1391 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1392 if ((unsigned int)backlog > somaxconn)
1393 backlog = somaxconn;
1395 err = security_socket_listen(sock, backlog);
1397 err = sock->ops->listen(sock, backlog);
1399 fput_light(sock->file, fput_needed);
1405 * For accept, we attempt to create a new socket, set up the link
1406 * with the client, wake up the client, then return the new
1407 * connected fd. We collect the address of the connector in kernel
1408 * space and move it to user at the very end. This is unclean because
1409 * we open the socket then return an error.
1411 * 1003.1g adds the ability to recvmsg() to query connection pending
1412 * status to recvmsg. We need to add that support in a way thats
1413 * clean when we restucture accept also.
1416 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1417 int __user *, upeer_addrlen, int, flags)
1419 struct socket *sock, *newsock;
1420 struct file *newfile;
1421 int err, len, newfd, fput_needed;
1422 struct sockaddr_storage address;
1424 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1427 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1428 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1430 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1435 newsock = sock_alloc();
1439 newsock->type = sock->type;
1440 newsock->ops = sock->ops;
1443 * We don't need try_module_get here, as the listening socket (sock)
1444 * has the protocol module (sock->ops->owner) held.
1446 __module_get(newsock->ops->owner);
1448 newfd = get_unused_fd_flags(flags);
1449 if (unlikely(newfd < 0)) {
1451 sock_release(newsock);
1454 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1455 if (IS_ERR(newfile)) {
1456 err = PTR_ERR(newfile);
1457 put_unused_fd(newfd);
1458 sock_release(newsock);
1462 err = security_socket_accept(sock, newsock);
1466 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1470 if (upeer_sockaddr) {
1471 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1473 err = -ECONNABORTED;
1476 err = move_addr_to_user(&address,
1477 len, upeer_sockaddr, upeer_addrlen);
1482 /* File flags are not inherited via accept() unlike another OSes. */
1484 fd_install(newfd, newfile);
1488 fput_light(sock->file, fput_needed);
1493 put_unused_fd(newfd);
1497 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1498 int __user *, upeer_addrlen)
1500 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1504 * Attempt to connect to a socket with the server address. The address
1505 * is in user space so we verify it is OK and move it to kernel space.
1507 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1510 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1511 * other SEQPACKET protocols that take time to connect() as it doesn't
1512 * include the -EINPROGRESS status for such sockets.
1515 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1518 struct socket *sock;
1519 struct sockaddr_storage address;
1520 int err, fput_needed;
1522 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1525 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1530 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1534 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1535 sock->file->f_flags);
1537 fput_light(sock->file, fput_needed);
1543 * Get the local address ('name') of a socket object. Move the obtained
1544 * name to user space.
1547 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1548 int __user *, usockaddr_len)
1550 struct socket *sock;
1551 struct sockaddr_storage address;
1552 int len, err, fput_needed;
1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1558 err = security_socket_getsockname(sock);
1562 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1565 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1568 fput_light(sock->file, fput_needed);
1574 * Get the remote address ('name') of a socket object. Move the obtained
1575 * name to user space.
1578 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1579 int __user *, usockaddr_len)
1581 struct socket *sock;
1582 struct sockaddr_storage address;
1583 int len, err, fput_needed;
1585 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1587 err = security_socket_getpeername(sock);
1589 fput_light(sock->file, fput_needed);
1594 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1597 err = move_addr_to_user(&address, len, usockaddr,
1599 fput_light(sock->file, fput_needed);
1605 * Send a datagram to a given address. We move the address into kernel
1606 * space and check the user space data area is readable before invoking
1610 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1611 unsigned int, flags, struct sockaddr __user *, addr,
1614 struct socket *sock;
1615 struct sockaddr_storage address;
1621 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1624 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1628 msg.msg_name = NULL;
1629 msg.msg_control = NULL;
1630 msg.msg_controllen = 0;
1631 msg.msg_namelen = 0;
1633 err = move_addr_to_kernel(addr, addr_len, &address);
1636 msg.msg_name = (struct sockaddr *)&address;
1637 msg.msg_namelen = addr_len;
1639 if (sock->file->f_flags & O_NONBLOCK)
1640 flags |= MSG_DONTWAIT;
1641 msg.msg_flags = flags;
1642 err = sock_sendmsg(sock, &msg);
1645 fput_light(sock->file, fput_needed);
1651 * Send a datagram down a socket.
1654 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1655 unsigned int, flags)
1657 return sys_sendto(fd, buff, len, flags, NULL, 0);
1661 * Receive a frame from the socket and optionally record the address of the
1662 * sender. We verify the buffers are writable and if needed move the
1663 * sender address from kernel to user space.
1666 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1667 unsigned int, flags, struct sockaddr __user *, addr,
1668 int __user *, addr_len)
1670 struct socket *sock;
1673 struct sockaddr_storage address;
1677 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1680 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1684 msg.msg_control = NULL;
1685 msg.msg_controllen = 0;
1686 /* Save some cycles and don't copy the address if not needed */
1687 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1688 /* We assume all kernel code knows the size of sockaddr_storage */
1689 msg.msg_namelen = 0;
1690 msg.msg_iocb = NULL;
1691 if (sock->file->f_flags & O_NONBLOCK)
1692 flags |= MSG_DONTWAIT;
1693 err = sock_recvmsg(sock, &msg, flags);
1695 if (err >= 0 && addr != NULL) {
1696 err2 = move_addr_to_user(&address,
1697 msg.msg_namelen, addr, addr_len);
1702 fput_light(sock->file, fput_needed);
1708 * Receive a datagram from a socket.
1711 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1712 unsigned int, flags)
1714 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1718 * Set a socket option. Because we don't know the option lengths we have
1719 * to pass the user mode parameter for the protocols to sort out.
1722 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1723 char __user *, optval, int, optlen)
1725 int err, fput_needed;
1726 struct socket *sock;
1731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733 err = security_socket_setsockopt(sock, level, optname);
1737 if (level == SOL_SOCKET)
1739 sock_setsockopt(sock, level, optname, optval,
1743 sock->ops->setsockopt(sock, level, optname, optval,
1746 fput_light(sock->file, fput_needed);
1752 * Get a socket option. Because we don't know the option lengths we have
1753 * to pass a user mode parameter for the protocols to sort out.
1756 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1757 char __user *, optval, int __user *, optlen)
1759 int err, fput_needed;
1760 struct socket *sock;
1762 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1764 err = security_socket_getsockopt(sock, level, optname);
1768 if (level == SOL_SOCKET)
1770 sock_getsockopt(sock, level, optname, optval,
1774 sock->ops->getsockopt(sock, level, optname, optval,
1777 fput_light(sock->file, fput_needed);
1783 * Shutdown a socket.
1786 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1788 int err, fput_needed;
1789 struct socket *sock;
1791 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1793 err = security_socket_shutdown(sock, how);
1795 err = sock->ops->shutdown(sock, how);
1796 fput_light(sock->file, fput_needed);
1801 /* A couple of helpful macros for getting the address of the 32/64 bit
1802 * fields which are the same type (int / unsigned) on our platforms.
1804 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1805 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1806 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1808 struct used_address {
1809 struct sockaddr_storage name;
1810 unsigned int name_len;
1813 static int copy_msghdr_from_user(struct msghdr *kmsg,
1814 struct user_msghdr __user *umsg,
1815 struct sockaddr __user **save_addr,
1818 struct sockaddr __user *uaddr;
1819 struct iovec __user *uiov;
1823 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1824 __get_user(uaddr, &umsg->msg_name) ||
1825 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1826 __get_user(uiov, &umsg->msg_iov) ||
1827 __get_user(nr_segs, &umsg->msg_iovlen) ||
1828 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1829 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1830 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1834 kmsg->msg_namelen = 0;
1836 if (kmsg->msg_namelen < 0)
1839 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1840 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1845 if (uaddr && kmsg->msg_namelen) {
1847 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1853 kmsg->msg_name = NULL;
1854 kmsg->msg_namelen = 0;
1857 if (nr_segs > UIO_MAXIOV)
1860 kmsg->msg_iocb = NULL;
1862 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1863 UIO_FASTIOV, iov, &kmsg->msg_iter);
1866 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1867 struct msghdr *msg_sys, unsigned int flags,
1868 struct used_address *used_address,
1869 unsigned int allowed_msghdr_flags)
1871 struct compat_msghdr __user *msg_compat =
1872 (struct compat_msghdr __user *)msg;
1873 struct sockaddr_storage address;
1874 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1875 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1876 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1877 /* 20 is size of ipv6_pktinfo */
1878 unsigned char *ctl_buf = ctl;
1882 msg_sys->msg_name = &address;
1884 if (MSG_CMSG_COMPAT & flags)
1885 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1887 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1893 if (msg_sys->msg_controllen > INT_MAX)
1895 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1896 ctl_len = msg_sys->msg_controllen;
1897 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1899 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1903 ctl_buf = msg_sys->msg_control;
1904 ctl_len = msg_sys->msg_controllen;
1905 } else if (ctl_len) {
1906 if (ctl_len > sizeof(ctl)) {
1907 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1908 if (ctl_buf == NULL)
1913 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1914 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1915 * checking falls down on this.
1917 if (copy_from_user(ctl_buf,
1918 (void __user __force *)msg_sys->msg_control,
1921 msg_sys->msg_control = ctl_buf;
1923 msg_sys->msg_flags = flags;
1925 if (sock->file->f_flags & O_NONBLOCK)
1926 msg_sys->msg_flags |= MSG_DONTWAIT;
1928 * If this is sendmmsg() and current destination address is same as
1929 * previously succeeded address, omit asking LSM's decision.
1930 * used_address->name_len is initialized to UINT_MAX so that the first
1931 * destination address never matches.
1933 if (used_address && msg_sys->msg_name &&
1934 used_address->name_len == msg_sys->msg_namelen &&
1935 !memcmp(&used_address->name, msg_sys->msg_name,
1936 used_address->name_len)) {
1937 err = sock_sendmsg_nosec(sock, msg_sys);
1940 err = sock_sendmsg(sock, msg_sys);
1942 * If this is sendmmsg() and sending to current destination address was
1943 * successful, remember it.
1945 if (used_address && err >= 0) {
1946 used_address->name_len = msg_sys->msg_namelen;
1947 if (msg_sys->msg_name)
1948 memcpy(&used_address->name, msg_sys->msg_name,
1949 used_address->name_len);
1954 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1961 * BSD sendmsg interface
1964 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1966 int fput_needed, err;
1967 struct msghdr msg_sys;
1968 struct socket *sock;
1970 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1974 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
1976 fput_light(sock->file, fput_needed);
1981 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1983 if (flags & MSG_CMSG_COMPAT)
1985 return __sys_sendmsg(fd, msg, flags);
1989 * Linux sendmmsg interface
1992 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1995 int fput_needed, err, datagrams;
1996 struct socket *sock;
1997 struct mmsghdr __user *entry;
1998 struct compat_mmsghdr __user *compat_entry;
1999 struct msghdr msg_sys;
2000 struct used_address used_address;
2001 unsigned int oflags = flags;
2003 if (vlen > UIO_MAXIOV)
2008 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2012 used_address.name_len = UINT_MAX;
2014 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2018 while (datagrams < vlen) {
2019 if (datagrams == vlen - 1)
2022 if (MSG_CMSG_COMPAT & flags) {
2023 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2024 &msg_sys, flags, &used_address, MSG_EOR);
2027 err = __put_user(err, &compat_entry->msg_len);
2030 err = ___sys_sendmsg(sock,
2031 (struct user_msghdr __user *)entry,
2032 &msg_sys, flags, &used_address, MSG_EOR);
2035 err = put_user(err, &entry->msg_len);
2045 fput_light(sock->file, fput_needed);
2047 /* We only return an error if no datagrams were able to be sent */
2054 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2055 unsigned int, vlen, unsigned int, flags)
2057 if (flags & MSG_CMSG_COMPAT)
2059 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2062 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2063 struct msghdr *msg_sys, unsigned int flags, int nosec)
2065 struct compat_msghdr __user *msg_compat =
2066 (struct compat_msghdr __user *)msg;
2067 struct iovec iovstack[UIO_FASTIOV];
2068 struct iovec *iov = iovstack;
2069 unsigned long cmsg_ptr;
2073 /* kernel mode address */
2074 struct sockaddr_storage addr;
2076 /* user mode address pointers */
2077 struct sockaddr __user *uaddr;
2078 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2080 msg_sys->msg_name = &addr;
2082 if (MSG_CMSG_COMPAT & flags)
2083 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2085 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2089 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2090 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2092 /* We assume all kernel code knows the size of sockaddr_storage */
2093 msg_sys->msg_namelen = 0;
2095 if (sock->file->f_flags & O_NONBLOCK)
2096 flags |= MSG_DONTWAIT;
2097 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2102 if (uaddr != NULL) {
2103 err = move_addr_to_user(&addr,
2104 msg_sys->msg_namelen, uaddr,
2109 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2113 if (MSG_CMSG_COMPAT & flags)
2114 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2115 &msg_compat->msg_controllen);
2117 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2118 &msg->msg_controllen);
2129 * BSD recvmsg interface
2132 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2134 int fput_needed, err;
2135 struct msghdr msg_sys;
2136 struct socket *sock;
2138 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2142 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2144 fput_light(sock->file, fput_needed);
2149 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2150 unsigned int, flags)
2152 if (flags & MSG_CMSG_COMPAT)
2154 return __sys_recvmsg(fd, msg, flags);
2158 * Linux recvmmsg interface
2161 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2162 unsigned int flags, struct timespec *timeout)
2164 int fput_needed, err, datagrams;
2165 struct socket *sock;
2166 struct mmsghdr __user *entry;
2167 struct compat_mmsghdr __user *compat_entry;
2168 struct msghdr msg_sys;
2169 struct timespec64 end_time;
2170 struct timespec64 timeout64;
2173 poll_select_set_timeout(&end_time, timeout->tv_sec,
2179 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2183 err = sock_error(sock->sk);
2188 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2190 while (datagrams < vlen) {
2192 * No need to ask LSM for more than the first datagram.
2194 if (MSG_CMSG_COMPAT & flags) {
2195 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2196 &msg_sys, flags & ~MSG_WAITFORONE,
2200 err = __put_user(err, &compat_entry->msg_len);
2203 err = ___sys_recvmsg(sock,
2204 (struct user_msghdr __user *)entry,
2205 &msg_sys, flags & ~MSG_WAITFORONE,
2209 err = put_user(err, &entry->msg_len);
2217 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2218 if (flags & MSG_WAITFORONE)
2219 flags |= MSG_DONTWAIT;
2222 ktime_get_ts64(&timeout64);
2223 *timeout = timespec64_to_timespec(
2224 timespec64_sub(end_time, timeout64));
2225 if (timeout->tv_sec < 0) {
2226 timeout->tv_sec = timeout->tv_nsec = 0;
2230 /* Timeout, return less than vlen datagrams */
2231 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2235 /* Out of band data, return right away */
2236 if (msg_sys.msg_flags & MSG_OOB)
2244 if (datagrams == 0) {
2250 * We may return less entries than requested (vlen) if the
2251 * sock is non block and there aren't enough datagrams...
2253 if (err != -EAGAIN) {
2255 * ... or if recvmsg returns an error after we
2256 * received some datagrams, where we record the
2257 * error to return on the next call or if the
2258 * app asks about it using getsockopt(SO_ERROR).
2260 sock->sk->sk_err = -err;
2263 fput_light(sock->file, fput_needed);
2268 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2269 unsigned int, vlen, unsigned int, flags,
2270 struct timespec __user *, timeout)
2273 struct timespec timeout_sys;
2275 if (flags & MSG_CMSG_COMPAT)
2279 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2281 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2284 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2286 if (datagrams > 0 &&
2287 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2288 datagrams = -EFAULT;
2293 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2294 /* Argument list sizes for sys_socketcall */
2295 #define AL(x) ((x) * sizeof(unsigned long))
2296 static const unsigned char nargs[21] = {
2297 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2298 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2299 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2306 * System call vectors.
2308 * Argument checking cleaned up. Saved 20% in size.
2309 * This function doesn't need to set the kernel lock because
2310 * it is set by the callees.
2313 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2315 unsigned long a[AUDITSC_ARGS];
2316 unsigned long a0, a1;
2320 if (call < 1 || call > SYS_SENDMMSG)
2324 if (len > sizeof(a))
2327 /* copy_from_user should be SMP safe. */
2328 if (copy_from_user(a, args, len))
2331 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2340 err = sys_socket(a0, a1, a[2]);
2343 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2346 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2349 err = sys_listen(a0, a1);
2352 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2353 (int __user *)a[2], 0);
2355 case SYS_GETSOCKNAME:
2357 sys_getsockname(a0, (struct sockaddr __user *)a1,
2358 (int __user *)a[2]);
2360 case SYS_GETPEERNAME:
2362 sys_getpeername(a0, (struct sockaddr __user *)a1,
2363 (int __user *)a[2]);
2365 case SYS_SOCKETPAIR:
2366 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2369 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2372 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2373 (struct sockaddr __user *)a[4], a[5]);
2376 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2379 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2380 (struct sockaddr __user *)a[4],
2381 (int __user *)a[5]);
2384 err = sys_shutdown(a0, a1);
2386 case SYS_SETSOCKOPT:
2387 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2389 case SYS_GETSOCKOPT:
2391 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2392 (int __user *)a[4]);
2395 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2398 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2401 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2404 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2405 (struct timespec __user *)a[4]);
2408 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2409 (int __user *)a[2], a[3]);
2418 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2421 * sock_register - add a socket protocol handler
2422 * @ops: description of protocol
2424 * This function is called by a protocol handler that wants to
2425 * advertise its address family, and have it linked into the
2426 * socket interface. The value ops->family corresponds to the
2427 * socket system call protocol family.
2429 int sock_register(const struct net_proto_family *ops)
2433 if (ops->family >= NPROTO) {
2434 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2438 spin_lock(&net_family_lock);
2439 if (rcu_dereference_protected(net_families[ops->family],
2440 lockdep_is_held(&net_family_lock)))
2443 rcu_assign_pointer(net_families[ops->family], ops);
2446 spin_unlock(&net_family_lock);
2448 pr_info("NET: Registered protocol family %d\n", ops->family);
2451 EXPORT_SYMBOL(sock_register);
2454 * sock_unregister - remove a protocol handler
2455 * @family: protocol family to remove
2457 * This function is called by a protocol handler that wants to
2458 * remove its address family, and have it unlinked from the
2459 * new socket creation.
2461 * If protocol handler is a module, then it can use module reference
2462 * counts to protect against new references. If protocol handler is not
2463 * a module then it needs to provide its own protection in
2464 * the ops->create routine.
2466 void sock_unregister(int family)
2468 BUG_ON(family < 0 || family >= NPROTO);
2470 spin_lock(&net_family_lock);
2471 RCU_INIT_POINTER(net_families[family], NULL);
2472 spin_unlock(&net_family_lock);
2476 pr_info("NET: Unregistered protocol family %d\n", family);
2478 EXPORT_SYMBOL(sock_unregister);
2480 static int __init sock_init(void)
2484 * Initialize the network sysctl infrastructure.
2486 err = net_sysctl_init();
2491 * Initialize skbuff SLAB cache
2496 * Initialize the protocols module.
2501 err = register_filesystem(&sock_fs_type);
2504 sock_mnt = kern_mount(&sock_fs_type);
2505 if (IS_ERR(sock_mnt)) {
2506 err = PTR_ERR(sock_mnt);
2510 /* The real protocol initialization is performed in later initcalls.
2513 #ifdef CONFIG_NETFILTER
2514 err = netfilter_init();
2519 ptp_classifier_init();
2525 unregister_filesystem(&sock_fs_type);
2530 core_initcall(sock_init); /* early initcall */
2532 #ifdef CONFIG_PROC_FS
2533 void socket_seq_show(struct seq_file *seq)
2538 for_each_possible_cpu(cpu)
2539 counter += per_cpu(sockets_in_use, cpu);
2541 /* It can be negative, by the way. 8) */
2545 seq_printf(seq, "sockets: used %d\n", counter);
2547 #endif /* CONFIG_PROC_FS */
2549 #ifdef CONFIG_COMPAT
2550 static int do_siocgstamp(struct net *net, struct socket *sock,
2551 unsigned int cmd, void __user *up)
2553 mm_segment_t old_fs = get_fs();
2558 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2561 err = compat_put_timeval(&ktv, up);
2566 static int do_siocgstampns(struct net *net, struct socket *sock,
2567 unsigned int cmd, void __user *up)
2569 mm_segment_t old_fs = get_fs();
2570 struct timespec kts;
2574 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2577 err = compat_put_timespec(&kts, up);
2582 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2584 struct ifreq __user *uifr;
2587 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2588 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2591 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2595 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2601 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2603 struct compat_ifconf ifc32;
2605 struct ifconf __user *uifc;
2606 struct compat_ifreq __user *ifr32;
2607 struct ifreq __user *ifr;
2611 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2614 memset(&ifc, 0, sizeof(ifc));
2615 if (ifc32.ifcbuf == 0) {
2619 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2621 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2622 sizeof(struct ifreq);
2623 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2625 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2626 ifr32 = compat_ptr(ifc32.ifcbuf);
2627 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2628 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2634 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2637 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2641 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2645 ifr32 = compat_ptr(ifc32.ifcbuf);
2647 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2648 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2649 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2655 if (ifc32.ifcbuf == 0) {
2656 /* Translate from 64-bit structure multiple to
2660 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2665 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2671 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2673 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2674 bool convert_in = false, convert_out = false;
2675 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2676 struct ethtool_rxnfc __user *rxnfc;
2677 struct ifreq __user *ifr;
2678 u32 rule_cnt = 0, actual_rule_cnt;
2683 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2686 compat_rxnfc = compat_ptr(data);
2688 if (get_user(ethcmd, &compat_rxnfc->cmd))
2691 /* Most ethtool structures are defined without padding.
2692 * Unfortunately struct ethtool_rxnfc is an exception.
2697 case ETHTOOL_GRXCLSRLALL:
2698 /* Buffer size is variable */
2699 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2701 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2703 buf_size += rule_cnt * sizeof(u32);
2705 case ETHTOOL_GRXRINGS:
2706 case ETHTOOL_GRXCLSRLCNT:
2707 case ETHTOOL_GRXCLSRULE:
2708 case ETHTOOL_SRXCLSRLINS:
2711 case ETHTOOL_SRXCLSRLDEL:
2712 buf_size += sizeof(struct ethtool_rxnfc);
2717 ifr = compat_alloc_user_space(buf_size);
2718 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2720 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2723 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2724 &ifr->ifr_ifru.ifru_data))
2728 /* We expect there to be holes between fs.m_ext and
2729 * fs.ring_cookie and at the end of fs, but nowhere else.
2731 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2732 sizeof(compat_rxnfc->fs.m_ext) !=
2733 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2734 sizeof(rxnfc->fs.m_ext));
2736 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2737 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2738 offsetof(struct ethtool_rxnfc, fs.location) -
2739 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2741 if (copy_in_user(rxnfc, compat_rxnfc,
2742 (void __user *)(&rxnfc->fs.m_ext + 1) -
2743 (void __user *)rxnfc) ||
2744 copy_in_user(&rxnfc->fs.ring_cookie,
2745 &compat_rxnfc->fs.ring_cookie,
2746 (void __user *)(&rxnfc->fs.location + 1) -
2747 (void __user *)&rxnfc->fs.ring_cookie) ||
2748 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2749 sizeof(rxnfc->rule_cnt)))
2753 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2758 if (copy_in_user(compat_rxnfc, rxnfc,
2759 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2760 (const void __user *)rxnfc) ||
2761 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2762 &rxnfc->fs.ring_cookie,
2763 (const void __user *)(&rxnfc->fs.location + 1) -
2764 (const void __user *)&rxnfc->fs.ring_cookie) ||
2765 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2766 sizeof(rxnfc->rule_cnt)))
2769 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2770 /* As an optimisation, we only copy the actual
2771 * number of rules that the underlying
2772 * function returned. Since Mallory might
2773 * change the rule count in user memory, we
2774 * check that it is less than the rule count
2775 * originally given (as the user buffer size),
2776 * which has been range-checked.
2778 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2780 if (actual_rule_cnt < rule_cnt)
2781 rule_cnt = actual_rule_cnt;
2782 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2783 &rxnfc->rule_locs[0],
2784 rule_cnt * sizeof(u32)))
2792 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2795 compat_uptr_t uptr32;
2796 struct ifreq __user *uifr;
2798 uifr = compat_alloc_user_space(sizeof(*uifr));
2799 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2802 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2805 uptr = compat_ptr(uptr32);
2807 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2810 return dev_ioctl(net, SIOCWANDEV, uifr);
2813 static int bond_ioctl(struct net *net, unsigned int cmd,
2814 struct compat_ifreq __user *ifr32)
2817 mm_segment_t old_fs;
2821 case SIOCBONDENSLAVE:
2822 case SIOCBONDRELEASE:
2823 case SIOCBONDSETHWADDR:
2824 case SIOCBONDCHANGEACTIVE:
2825 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2830 err = dev_ioctl(net, cmd,
2831 (struct ifreq __user __force *) &kifr);
2836 return -ENOIOCTLCMD;
2840 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2841 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2842 struct compat_ifreq __user *u_ifreq32)
2844 struct ifreq __user *u_ifreq64;
2845 char tmp_buf[IFNAMSIZ];
2846 void __user *data64;
2849 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2852 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2854 data64 = compat_ptr(data32);
2856 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2858 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2861 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2864 return dev_ioctl(net, cmd, u_ifreq64);
2867 static int dev_ifsioc(struct net *net, struct socket *sock,
2868 unsigned int cmd, struct compat_ifreq __user *uifr32)
2870 struct ifreq __user *uifr;
2873 uifr = compat_alloc_user_space(sizeof(*uifr));
2874 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2877 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2888 case SIOCGIFBRDADDR:
2889 case SIOCGIFDSTADDR:
2890 case SIOCGIFNETMASK:
2895 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2903 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2904 struct compat_ifreq __user *uifr32)
2907 struct compat_ifmap __user *uifmap32;
2908 mm_segment_t old_fs;
2911 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2912 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2913 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2914 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2915 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2916 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2917 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2918 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2924 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2927 if (cmd == SIOCGIFMAP && !err) {
2928 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2929 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2930 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2931 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2932 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2933 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2934 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2943 struct sockaddr rt_dst; /* target address */
2944 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2945 struct sockaddr rt_genmask; /* target network mask (IP) */
2946 unsigned short rt_flags;
2949 unsigned char rt_tos;
2950 unsigned char rt_class;
2952 short rt_metric; /* +1 for binary compatibility! */
2953 /* char * */ u32 rt_dev; /* forcing the device at add */
2954 u32 rt_mtu; /* per route MTU/Window */
2955 u32 rt_window; /* Window clamping */
2956 unsigned short rt_irtt; /* Initial RTT */
2959 struct in6_rtmsg32 {
2960 struct in6_addr rtmsg_dst;
2961 struct in6_addr rtmsg_src;
2962 struct in6_addr rtmsg_gateway;
2972 static int routing_ioctl(struct net *net, struct socket *sock,
2973 unsigned int cmd, void __user *argp)
2977 struct in6_rtmsg r6;
2981 mm_segment_t old_fs = get_fs();
2983 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2984 struct in6_rtmsg32 __user *ur6 = argp;
2985 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2986 3 * sizeof(struct in6_addr));
2987 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2988 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2989 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2990 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2991 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2992 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2993 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2997 struct rtentry32 __user *ur4 = argp;
2998 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2999 3 * sizeof(struct sockaddr));
3000 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3001 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3002 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3003 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3004 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3005 ret |= get_user(rtdev, &(ur4->rt_dev));
3007 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3008 r4.rt_dev = (char __user __force *)devname;
3022 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3029 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3030 * for some operations; this forces use of the newer bridge-utils that
3031 * use compatible ioctls
3033 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3037 if (get_user(tmp, argp))
3039 if (tmp == BRCTL_GET_VERSION)
3040 return BRCTL_VERSION + 1;
3044 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3045 unsigned int cmd, unsigned long arg)
3047 void __user *argp = compat_ptr(arg);
3048 struct sock *sk = sock->sk;
3049 struct net *net = sock_net(sk);
3051 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3052 return compat_ifr_data_ioctl(net, cmd, argp);
3057 return old_bridge_ioctl(argp);
3059 return dev_ifname32(net, argp);
3061 return dev_ifconf(net, argp);
3063 return ethtool_ioctl(net, argp);
3065 return compat_siocwandev(net, argp);
3068 return compat_sioc_ifmap(net, cmd, argp);
3069 case SIOCBONDENSLAVE:
3070 case SIOCBONDRELEASE:
3071 case SIOCBONDSETHWADDR:
3072 case SIOCBONDCHANGEACTIVE:
3073 return bond_ioctl(net, cmd, argp);
3076 return routing_ioctl(net, sock, cmd, argp);
3078 return do_siocgstamp(net, sock, cmd, argp);
3080 return do_siocgstampns(net, sock, cmd, argp);
3081 case SIOCBONDSLAVEINFOQUERY:
3082 case SIOCBONDINFOQUERY:
3085 return compat_ifr_data_ioctl(net, cmd, argp);
3097 return sock_ioctl(file, cmd, arg);
3114 case SIOCSIFHWBROADCAST:
3116 case SIOCGIFBRDADDR:
3117 case SIOCSIFBRDADDR:
3118 case SIOCGIFDSTADDR:
3119 case SIOCSIFDSTADDR:
3120 case SIOCGIFNETMASK:
3121 case SIOCSIFNETMASK:
3132 return dev_ifsioc(net, sock, cmd, argp);
3138 return sock_do_ioctl(net, sock, cmd, arg);
3141 return -ENOIOCTLCMD;
3144 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3147 struct socket *sock = file->private_data;
3148 int ret = -ENOIOCTLCMD;
3155 if (sock->ops->compat_ioctl)
3156 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3158 if (ret == -ENOIOCTLCMD &&
3159 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3160 ret = compat_wext_handle_ioctl(net, cmd, arg);
3162 if (ret == -ENOIOCTLCMD)
3163 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3169 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3171 return sock->ops->bind(sock, addr, addrlen);
3173 EXPORT_SYMBOL(kernel_bind);
3175 int kernel_listen(struct socket *sock, int backlog)
3177 return sock->ops->listen(sock, backlog);
3179 EXPORT_SYMBOL(kernel_listen);
3181 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3183 struct sock *sk = sock->sk;
3186 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3191 err = sock->ops->accept(sock, *newsock, flags);
3193 sock_release(*newsock);
3198 (*newsock)->ops = sock->ops;
3199 __module_get((*newsock)->ops->owner);
3204 EXPORT_SYMBOL(kernel_accept);
3206 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3209 return sock->ops->connect(sock, addr, addrlen, flags);
3211 EXPORT_SYMBOL(kernel_connect);
3213 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3216 return sock->ops->getname(sock, addr, addrlen, 0);
3218 EXPORT_SYMBOL(kernel_getsockname);
3220 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3223 return sock->ops->getname(sock, addr, addrlen, 1);
3225 EXPORT_SYMBOL(kernel_getpeername);
3227 int kernel_getsockopt(struct socket *sock, int level, int optname,
3228 char *optval, int *optlen)
3230 mm_segment_t oldfs = get_fs();
3231 char __user *uoptval;
3232 int __user *uoptlen;
3235 uoptval = (char __user __force *) optval;
3236 uoptlen = (int __user __force *) optlen;
3239 if (level == SOL_SOCKET)
3240 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3242 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3247 EXPORT_SYMBOL(kernel_getsockopt);
3249 int kernel_setsockopt(struct socket *sock, int level, int optname,
3250 char *optval, unsigned int optlen)
3252 mm_segment_t oldfs = get_fs();
3253 char __user *uoptval;
3256 uoptval = (char __user __force *) optval;
3259 if (level == SOL_SOCKET)
3260 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3262 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3267 EXPORT_SYMBOL(kernel_setsockopt);
3269 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3270 size_t size, int flags)
3272 if (sock->ops->sendpage)
3273 return sock->ops->sendpage(sock, page, offset, size, flags);
3275 return sock_no_sendpage(sock, page, offset, size, flags);
3277 EXPORT_SYMBOL(kernel_sendpage);
3279 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3281 mm_segment_t oldfs = get_fs();
3285 err = sock->ops->ioctl(sock, cmd, arg);
3290 EXPORT_SYMBOL(kernel_sock_ioctl);
3292 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3294 return sock->ops->shutdown(sock, how);
3296 EXPORT_SYMBOL(kernel_sock_shutdown);