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>
92 #include <linux/nospec.h>
94 #include <asm/uaccess.h>
95 #include <asm/unistd.h>
97 #include <net/compat.h>
99 #include <net/cls_cgroup.h>
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <linux/atalk.h>
109 #include <net/busy_poll.h>
110 #include <linux/errqueue.h>
112 #ifdef CONFIG_NET_RX_BUSY_POLL
113 unsigned int sysctl_net_busy_read __read_mostly;
114 unsigned int sysctl_net_busy_poll __read_mostly;
117 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
118 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 static int sock_close(struct inode *inode, struct file *file);
122 static unsigned int sock_poll(struct file *file,
123 struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 static long compat_sock_ioctl(struct file *file,
127 unsigned int cmd, unsigned long arg);
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131 int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133 struct pipe_inode_info *pipe, size_t len,
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
141 static const struct file_operations socket_file_ops = {
142 .owner = THIS_MODULE,
144 .read_iter = sock_read_iter,
145 .write_iter = sock_write_iter,
147 .unlocked_ioctl = sock_ioctl,
149 .compat_ioctl = compat_sock_ioctl,
152 .release = sock_close,
153 .fasync = sock_fasync,
154 .sendpage = sock_sendpage,
155 .splice_write = generic_splice_sendpage,
156 .splice_read = sock_splice_read,
160 * The protocol list. Each protocol is registered in here.
163 static DEFINE_SPINLOCK(net_family_lock);
164 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
167 * Statistics counters of the socket lists
170 static DEFINE_PER_CPU(int, sockets_in_use);
174 * Move socket addresses back and forth across the kernel/user
175 * divide and look after the messy bits.
179 * move_addr_to_kernel - copy a socket address into kernel space
180 * @uaddr: Address in user space
181 * @kaddr: Address in kernel space
182 * @ulen: Length in user space
184 * The address is copied into kernel space. If the provided address is
185 * too long an error code of -EINVAL is returned. If the copy gives
186 * invalid addresses -EFAULT is returned. On a success 0 is returned.
189 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195 if (copy_from_user(kaddr, uaddr, ulen))
197 return audit_sockaddr(ulen, kaddr);
201 * move_addr_to_user - copy an address to user space
202 * @kaddr: kernel space address
203 * @klen: length of address in kernel
204 * @uaddr: user space address
205 * @ulen: pointer to user length field
207 * The value pointed to by ulen on entry is the buffer length available.
208 * This is overwritten with the buffer space used. -EINVAL is returned
209 * if an overlong buffer is specified or a negative buffer size. -EFAULT
210 * is returned if either the buffer or the length field are not
212 * After copying the data up to the limit the user specifies, the true
213 * length of the data is written over the length limit the user
214 * specified. Zero is returned for a success.
217 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
218 void __user *uaddr, int __user *ulen)
223 BUG_ON(klen > sizeof(struct sockaddr_storage));
224 err = get_user(len, ulen);
232 if (audit_sockaddr(klen, kaddr))
234 if (copy_to_user(uaddr, kaddr, len))
238 * "fromlen shall refer to the value before truncation.."
241 return __put_user(klen, ulen);
244 static struct kmem_cache *sock_inode_cachep __read_mostly;
246 static struct inode *sock_alloc_inode(struct super_block *sb)
248 struct socket_alloc *ei;
249 struct socket_wq *wq;
251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256 kmem_cache_free(sock_inode_cachep, ei);
259 init_waitqueue_head(&wq->wait);
260 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 d_inode(dentry)->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,
379 /* drop dentry, keep inode */
380 ihold(d_inode(path.dentry));
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 if (!strcmp(name, XATTR_NAME_SOCKPROTONAME)) {
475 if (dentry->d_name.len + 1 > size)
477 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
479 return dentry->d_name.len + 1;
484 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
490 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
500 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
505 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
512 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
514 int err = simple_setattr(dentry, iattr);
516 if (!err && (iattr->ia_valid & ATTR_UID)) {
517 struct socket *sock = SOCKET_I(d_inode(dentry));
519 sock->sk->sk_uid = iattr->ia_uid;
525 static const struct inode_operations sockfs_inode_ops = {
526 .getxattr = sockfs_getxattr,
527 .listxattr = sockfs_listxattr,
528 .setattr = sockfs_setattr,
532 * sock_alloc - allocate a socket
534 * Allocate a new inode and socket object. The two are bound together
535 * and initialised. The socket is then returned. If we are out of inodes
539 static struct socket *sock_alloc(void)
544 inode = new_inode_pseudo(sock_mnt->mnt_sb);
548 sock = SOCKET_I(inode);
550 kmemcheck_annotate_bitfield(sock, type);
551 inode->i_ino = get_next_ino();
552 inode->i_mode = S_IFSOCK | S_IRWXUGO;
553 inode->i_uid = current_fsuid();
554 inode->i_gid = current_fsgid();
555 inode->i_op = &sockfs_inode_ops;
557 this_cpu_add(sockets_in_use, 1);
562 * sock_release - close a socket
563 * @sock: socket to close
565 * The socket is released from the protocol stack if it has a release
566 * callback, and the inode is then released if the socket is bound to
567 * an inode not a file.
570 void sock_release(struct socket *sock)
573 struct module *owner = sock->ops->owner;
575 sock->ops->release(sock);
580 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
581 pr_err("%s: fasync list not empty!\n", __func__);
583 this_cpu_sub(sockets_in_use, 1);
585 iput(SOCK_INODE(sock));
590 EXPORT_SYMBOL(sock_release);
592 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
594 u8 flags = *tx_flags;
596 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
597 flags |= SKBTX_HW_TSTAMP;
599 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
600 flags |= SKBTX_SW_TSTAMP;
602 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
603 flags |= SKBTX_SCHED_TSTAMP;
605 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
606 flags |= SKBTX_ACK_TSTAMP;
610 EXPORT_SYMBOL(__sock_tx_timestamp);
612 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
614 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
615 BUG_ON(ret == -EIOCBQUEUED);
619 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
621 int err = security_socket_sendmsg(sock, msg,
624 return err ?: sock_sendmsg_nosec(sock, msg);
626 EXPORT_SYMBOL(sock_sendmsg);
628 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
629 struct kvec *vec, size_t num, size_t size)
631 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
632 return sock_sendmsg(sock, msg);
634 EXPORT_SYMBOL(kernel_sendmsg);
637 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
639 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
642 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
643 struct scm_timestamping tss;
645 struct skb_shared_hwtstamps *shhwtstamps =
648 /* Race occurred between timestamp enabling and packet
649 receiving. Fill in the current time for now. */
650 if (need_software_tstamp && skb->tstamp.tv64 == 0)
651 __net_timestamp(skb);
653 if (need_software_tstamp) {
654 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
656 skb_get_timestamp(skb, &tv);
657 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
661 skb_get_timestampns(skb, &ts);
662 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
667 memset(&tss, 0, sizeof(tss));
668 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
669 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
672 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
673 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
676 put_cmsg(msg, SOL_SOCKET,
677 SCM_TIMESTAMPING, sizeof(tss), &tss);
679 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
681 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
686 if (!sock_flag(sk, SOCK_WIFI_STATUS))
688 if (!skb->wifi_acked_valid)
691 ack = skb->wifi_acked;
693 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
695 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
697 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
700 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
701 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
702 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
705 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
708 sock_recv_timestamp(msg, sk, skb);
709 sock_recv_drops(msg, sk, skb);
711 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
713 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
714 size_t size, int flags)
716 return sock->ops->recvmsg(sock, msg, size, flags);
719 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
722 int err = security_socket_recvmsg(sock, msg, size, flags);
724 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
726 EXPORT_SYMBOL(sock_recvmsg);
729 * kernel_recvmsg - Receive a message from a socket (kernel space)
730 * @sock: The socket to receive the message from
731 * @msg: Received message
732 * @vec: Input s/g array for message data
733 * @num: Size of input s/g array
734 * @size: Number of bytes to read
735 * @flags: Message flags (MSG_DONTWAIT, etc...)
737 * On return the msg structure contains the scatter/gather array passed in the
738 * vec argument. The array is modified so that it consists of the unfilled
739 * portion of the original array.
741 * The returned value is the total number of bytes received, or an error.
743 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
744 struct kvec *vec, size_t num, size_t size, int flags)
746 mm_segment_t oldfs = get_fs();
749 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
751 result = sock_recvmsg(sock, msg, size, flags);
755 EXPORT_SYMBOL(kernel_recvmsg);
757 static ssize_t sock_sendpage(struct file *file, struct page *page,
758 int offset, size_t size, loff_t *ppos, int more)
763 sock = file->private_data;
765 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
766 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
769 return kernel_sendpage(sock, page, offset, size, flags);
772 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
773 struct pipe_inode_info *pipe, size_t len,
776 struct socket *sock = file->private_data;
778 if (unlikely(!sock->ops->splice_read))
781 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
784 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
786 struct file *file = iocb->ki_filp;
787 struct socket *sock = file->private_data;
788 struct msghdr msg = {.msg_iter = *to,
792 if (file->f_flags & O_NONBLOCK)
793 msg.msg_flags = MSG_DONTWAIT;
795 if (iocb->ki_pos != 0)
798 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
801 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
806 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
808 struct file *file = iocb->ki_filp;
809 struct socket *sock = file->private_data;
810 struct msghdr msg = {.msg_iter = *from,
814 if (iocb->ki_pos != 0)
817 if (file->f_flags & O_NONBLOCK)
818 msg.msg_flags = MSG_DONTWAIT;
820 if (sock->type == SOCK_SEQPACKET)
821 msg.msg_flags |= MSG_EOR;
823 res = sock_sendmsg(sock, &msg);
824 *from = msg.msg_iter;
829 * Atomic setting of ioctl hooks to avoid race
830 * with module unload.
833 static DEFINE_MUTEX(br_ioctl_mutex);
834 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
836 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
838 mutex_lock(&br_ioctl_mutex);
839 br_ioctl_hook = hook;
840 mutex_unlock(&br_ioctl_mutex);
842 EXPORT_SYMBOL(brioctl_set);
844 static DEFINE_MUTEX(vlan_ioctl_mutex);
845 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
847 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
849 mutex_lock(&vlan_ioctl_mutex);
850 vlan_ioctl_hook = hook;
851 mutex_unlock(&vlan_ioctl_mutex);
853 EXPORT_SYMBOL(vlan_ioctl_set);
855 static DEFINE_MUTEX(dlci_ioctl_mutex);
856 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
858 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
860 mutex_lock(&dlci_ioctl_mutex);
861 dlci_ioctl_hook = hook;
862 mutex_unlock(&dlci_ioctl_mutex);
864 EXPORT_SYMBOL(dlci_ioctl_set);
866 static long sock_do_ioctl(struct net *net, struct socket *sock,
867 unsigned int cmd, unsigned long arg)
870 void __user *argp = (void __user *)arg;
872 err = sock->ops->ioctl(sock, cmd, arg);
875 * If this ioctl is unknown try to hand it down
878 if (err == -ENOIOCTLCMD)
879 err = dev_ioctl(net, cmd, argp);
885 * With an ioctl, arg may well be a user mode pointer, but we don't know
886 * what to do with it - that's up to the protocol still.
889 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
893 void __user *argp = (void __user *)arg;
897 sock = file->private_data;
900 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
901 err = dev_ioctl(net, cmd, argp);
903 #ifdef CONFIG_WEXT_CORE
904 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
905 err = dev_ioctl(net, cmd, argp);
912 if (get_user(pid, (int __user *)argp))
914 f_setown(sock->file, pid, 1);
919 err = put_user(f_getown(sock->file),
928 request_module("bridge");
930 mutex_lock(&br_ioctl_mutex);
932 err = br_ioctl_hook(net, cmd, argp);
933 mutex_unlock(&br_ioctl_mutex);
938 if (!vlan_ioctl_hook)
939 request_module("8021q");
941 mutex_lock(&vlan_ioctl_mutex);
943 err = vlan_ioctl_hook(net, argp);
944 mutex_unlock(&vlan_ioctl_mutex);
949 if (!dlci_ioctl_hook)
950 request_module("dlci");
952 mutex_lock(&dlci_ioctl_mutex);
954 err = dlci_ioctl_hook(cmd, argp);
955 mutex_unlock(&dlci_ioctl_mutex);
958 err = sock_do_ioctl(net, sock, cmd, arg);
964 int sock_create_lite(int family, int type, int protocol, struct socket **res)
967 struct socket *sock = NULL;
969 err = security_socket_create(family, type, protocol, 1);
980 err = security_socket_post_create(sock, family, type, protocol, 1);
992 EXPORT_SYMBOL(sock_create_lite);
994 /* No kernel lock held - perfect */
995 static unsigned int sock_poll(struct file *file, poll_table *wait)
997 unsigned int busy_flag = 0;
1001 * We can't return errors to poll, so it's either yes or no.
1003 sock = file->private_data;
1005 if (sk_can_busy_loop(sock->sk)) {
1006 /* this socket can poll_ll so tell the system call */
1007 busy_flag = POLL_BUSY_LOOP;
1009 /* once, only if requested by syscall */
1010 if (wait && (wait->_key & POLL_BUSY_LOOP))
1011 sk_busy_loop(sock->sk, 1);
1014 return busy_flag | sock->ops->poll(file, sock, wait);
1017 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1019 struct socket *sock = file->private_data;
1021 return sock->ops->mmap(file, sock, vma);
1024 static int sock_close(struct inode *inode, struct file *filp)
1026 sock_release(SOCKET_I(inode));
1031 * Update the socket async list
1033 * Fasync_list locking strategy.
1035 * 1. fasync_list is modified only under process context socket lock
1036 * i.e. under semaphore.
1037 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1038 * or under socket lock
1041 static int sock_fasync(int fd, struct file *filp, int on)
1043 struct socket *sock = filp->private_data;
1044 struct sock *sk = sock->sk;
1045 struct socket_wq *wq;
1051 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1052 fasync_helper(fd, filp, on, &wq->fasync_list);
1054 if (!wq->fasync_list)
1055 sock_reset_flag(sk, SOCK_FASYNC);
1057 sock_set_flag(sk, SOCK_FASYNC);
1063 /* This function may be called only under rcu_lock */
1065 int sock_wake_async(struct socket_wq *wq, int how, int band)
1067 if (!wq || !wq->fasync_list)
1071 case SOCK_WAKE_WAITD:
1072 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1075 case SOCK_WAKE_SPACE:
1076 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1081 kill_fasync(&wq->fasync_list, SIGIO, band);
1084 kill_fasync(&wq->fasync_list, SIGURG, band);
1089 EXPORT_SYMBOL(sock_wake_async);
1091 int __sock_create(struct net *net, int family, int type, int protocol,
1092 struct socket **res, int kern)
1095 struct socket *sock;
1096 const struct net_proto_family *pf;
1099 * Check protocol is in range
1101 if (family < 0 || family >= NPROTO)
1102 return -EAFNOSUPPORT;
1103 if (type < 0 || type >= SOCK_MAX)
1108 This uglymoron is moved from INET layer to here to avoid
1109 deadlock in module load.
1111 if (family == PF_INET && type == SOCK_PACKET) {
1115 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1121 err = security_socket_create(family, type, protocol, kern);
1126 * Allocate the socket and allow the family to set things up. if
1127 * the protocol is 0, the family is instructed to select an appropriate
1130 sock = sock_alloc();
1132 net_warn_ratelimited("socket: no more sockets\n");
1133 return -ENFILE; /* Not exactly a match, but its the
1134 closest posix thing */
1139 #ifdef CONFIG_MODULES
1140 /* Attempt to load a protocol module if the find failed.
1142 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1143 * requested real, full-featured networking support upon configuration.
1144 * Otherwise module support will break!
1146 if (rcu_access_pointer(net_families[family]) == NULL)
1147 request_module("net-pf-%d", family);
1151 pf = rcu_dereference(net_families[family]);
1152 err = -EAFNOSUPPORT;
1157 * We will call the ->create function, that possibly is in a loadable
1158 * module, so we have to bump that loadable module refcnt first.
1160 if (!try_module_get(pf->owner))
1163 /* Now protected by module ref count */
1166 err = pf->create(net, sock, protocol, kern);
1168 goto out_module_put;
1171 * Now to bump the refcnt of the [loadable] module that owns this
1172 * socket at sock_release time we decrement its refcnt.
1174 if (!try_module_get(sock->ops->owner))
1175 goto out_module_busy;
1178 * Now that we're done with the ->create function, the [loadable]
1179 * module can have its refcnt decremented
1181 module_put(pf->owner);
1182 err = security_socket_post_create(sock, family, type, protocol, kern);
1184 goto out_sock_release;
1190 err = -EAFNOSUPPORT;
1193 module_put(pf->owner);
1200 goto out_sock_release;
1202 EXPORT_SYMBOL(__sock_create);
1204 int sock_create(int family, int type, int protocol, struct socket **res)
1206 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1208 EXPORT_SYMBOL(sock_create);
1210 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1212 return __sock_create(net, family, type, protocol, res, 1);
1214 EXPORT_SYMBOL(sock_create_kern);
1216 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1219 struct socket *sock;
1222 /* Check the SOCK_* constants for consistency. */
1223 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1224 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1225 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1226 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1228 flags = type & ~SOCK_TYPE_MASK;
1229 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1231 type &= SOCK_TYPE_MASK;
1233 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1234 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1236 retval = sock_create(family, type, protocol, &sock);
1240 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1245 /* It may be already another descriptor 8) Not kernel problem. */
1254 * Create a pair of connected sockets.
1257 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1258 int __user *, usockvec)
1260 struct socket *sock1, *sock2;
1262 struct file *newfile1, *newfile2;
1265 flags = type & ~SOCK_TYPE_MASK;
1266 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1268 type &= SOCK_TYPE_MASK;
1270 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1271 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1274 * Obtain the first socket and check if the underlying protocol
1275 * supports the socketpair call.
1278 err = sock_create(family, type, protocol, &sock1);
1282 err = sock_create(family, type, protocol, &sock2);
1286 err = sock1->ops->socketpair(sock1, sock2);
1288 goto out_release_both;
1290 fd1 = get_unused_fd_flags(flags);
1291 if (unlikely(fd1 < 0)) {
1293 goto out_release_both;
1296 fd2 = get_unused_fd_flags(flags);
1297 if (unlikely(fd2 < 0)) {
1299 goto out_put_unused_1;
1302 newfile1 = sock_alloc_file(sock1, flags, NULL);
1303 if (IS_ERR(newfile1)) {
1304 err = PTR_ERR(newfile1);
1305 goto out_put_unused_both;
1308 newfile2 = sock_alloc_file(sock2, flags, NULL);
1309 if (IS_ERR(newfile2)) {
1310 err = PTR_ERR(newfile2);
1314 err = put_user(fd1, &usockvec[0]);
1318 err = put_user(fd2, &usockvec[1]);
1322 audit_fd_pair(fd1, fd2);
1324 fd_install(fd1, newfile1);
1325 fd_install(fd2, newfile2);
1326 /* fd1 and fd2 may be already another descriptors.
1327 * Not kernel problem.
1343 sock_release(sock2);
1346 out_put_unused_both:
1351 sock_release(sock2);
1353 sock_release(sock1);
1359 * Bind a name to a socket. Nothing much to do here since it's
1360 * the protocol's responsibility to handle the local address.
1362 * We move the socket address to kernel space before we call
1363 * the protocol layer (having also checked the address is ok).
1366 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1368 struct socket *sock;
1369 struct sockaddr_storage address;
1370 int err, fput_needed;
1372 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1374 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1376 err = security_socket_bind(sock,
1377 (struct sockaddr *)&address,
1380 err = sock->ops->bind(sock,
1384 fput_light(sock->file, fput_needed);
1390 * Perform a listen. Basically, we allow the protocol to do anything
1391 * necessary for a listen, and if that works, we mark the socket as
1392 * ready for listening.
1395 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1397 struct socket *sock;
1398 int err, fput_needed;
1401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1403 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1404 if ((unsigned int)backlog > somaxconn)
1405 backlog = somaxconn;
1407 err = security_socket_listen(sock, backlog);
1409 err = sock->ops->listen(sock, backlog);
1411 fput_light(sock->file, fput_needed);
1417 * For accept, we attempt to create a new socket, set up the link
1418 * with the client, wake up the client, then return the new
1419 * connected fd. We collect the address of the connector in kernel
1420 * space and move it to user at the very end. This is unclean because
1421 * we open the socket then return an error.
1423 * 1003.1g adds the ability to recvmsg() to query connection pending
1424 * status to recvmsg. We need to add that support in a way thats
1425 * clean when we restucture accept also.
1428 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1429 int __user *, upeer_addrlen, int, flags)
1431 struct socket *sock, *newsock;
1432 struct file *newfile;
1433 int err, len, newfd, fput_needed;
1434 struct sockaddr_storage address;
1436 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1439 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1440 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1442 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1447 newsock = sock_alloc();
1451 newsock->type = sock->type;
1452 newsock->ops = sock->ops;
1455 * We don't need try_module_get here, as the listening socket (sock)
1456 * has the protocol module (sock->ops->owner) held.
1458 __module_get(newsock->ops->owner);
1460 newfd = get_unused_fd_flags(flags);
1461 if (unlikely(newfd < 0)) {
1463 sock_release(newsock);
1466 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1467 if (IS_ERR(newfile)) {
1468 err = PTR_ERR(newfile);
1469 put_unused_fd(newfd);
1470 sock_release(newsock);
1474 err = security_socket_accept(sock, newsock);
1478 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1482 if (upeer_sockaddr) {
1483 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1485 err = -ECONNABORTED;
1488 err = move_addr_to_user(&address,
1489 len, upeer_sockaddr, upeer_addrlen);
1494 /* File flags are not inherited via accept() unlike another OSes. */
1496 fd_install(newfd, newfile);
1500 fput_light(sock->file, fput_needed);
1505 put_unused_fd(newfd);
1509 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1510 int __user *, upeer_addrlen)
1512 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1516 * Attempt to connect to a socket with the server address. The address
1517 * is in user space so we verify it is OK and move it to kernel space.
1519 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1522 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1523 * other SEQPACKET protocols that take time to connect() as it doesn't
1524 * include the -EINPROGRESS status for such sockets.
1527 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1530 struct socket *sock;
1531 struct sockaddr_storage address;
1532 int err, fput_needed;
1534 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1537 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1542 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1546 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1547 sock->file->f_flags);
1549 fput_light(sock->file, fput_needed);
1555 * Get the local address ('name') of a socket object. Move the obtained
1556 * name to user space.
1559 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1560 int __user *, usockaddr_len)
1562 struct socket *sock;
1563 struct sockaddr_storage address;
1564 int len, err, fput_needed;
1566 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1570 err = security_socket_getsockname(sock);
1574 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1577 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1580 fput_light(sock->file, fput_needed);
1586 * Get the remote address ('name') of a socket object. Move the obtained
1587 * name to user space.
1590 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1591 int __user *, usockaddr_len)
1593 struct socket *sock;
1594 struct sockaddr_storage address;
1595 int len, err, fput_needed;
1597 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1599 err = security_socket_getpeername(sock);
1601 fput_light(sock->file, fput_needed);
1606 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1609 err = move_addr_to_user(&address, len, usockaddr,
1611 fput_light(sock->file, fput_needed);
1617 * Send a datagram to a given address. We move the address into kernel
1618 * space and check the user space data area is readable before invoking
1622 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1623 unsigned int, flags, struct sockaddr __user *, addr,
1626 struct socket *sock;
1627 struct sockaddr_storage address;
1633 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1636 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640 msg.msg_name = NULL;
1641 msg.msg_control = NULL;
1642 msg.msg_controllen = 0;
1643 msg.msg_namelen = 0;
1645 err = move_addr_to_kernel(addr, addr_len, &address);
1648 msg.msg_name = (struct sockaddr *)&address;
1649 msg.msg_namelen = addr_len;
1651 if (sock->file->f_flags & O_NONBLOCK)
1652 flags |= MSG_DONTWAIT;
1653 msg.msg_flags = flags;
1654 err = sock_sendmsg(sock, &msg);
1657 fput_light(sock->file, fput_needed);
1663 * Send a datagram down a socket.
1666 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1667 unsigned int, flags)
1669 return sys_sendto(fd, buff, len, flags, NULL, 0);
1673 * Receive a frame from the socket and optionally record the address of the
1674 * sender. We verify the buffers are writable and if needed move the
1675 * sender address from kernel to user space.
1678 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1679 unsigned int, flags, struct sockaddr __user *, addr,
1680 int __user *, addr_len)
1682 struct socket *sock;
1685 struct sockaddr_storage address;
1689 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1692 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1696 msg.msg_control = NULL;
1697 msg.msg_controllen = 0;
1698 /* Save some cycles and don't copy the address if not needed */
1699 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1700 /* We assume all kernel code knows the size of sockaddr_storage */
1701 msg.msg_namelen = 0;
1702 msg.msg_iocb = NULL;
1704 if (sock->file->f_flags & O_NONBLOCK)
1705 flags |= MSG_DONTWAIT;
1706 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1708 if (err >= 0 && addr != NULL) {
1709 err2 = move_addr_to_user(&address,
1710 msg.msg_namelen, addr, addr_len);
1715 fput_light(sock->file, fput_needed);
1721 * Receive a datagram from a socket.
1724 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1725 unsigned int, flags)
1727 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1731 * Set a socket option. Because we don't know the option lengths we have
1732 * to pass the user mode parameter for the protocols to sort out.
1735 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1736 char __user *, optval, int, optlen)
1738 int err, fput_needed;
1739 struct socket *sock;
1744 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1746 err = security_socket_setsockopt(sock, level, optname);
1750 if (level == SOL_SOCKET)
1752 sock_setsockopt(sock, level, optname, optval,
1756 sock->ops->setsockopt(sock, level, optname, optval,
1759 fput_light(sock->file, fput_needed);
1765 * Get a socket option. Because we don't know the option lengths we have
1766 * to pass a user mode parameter for the protocols to sort out.
1769 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1770 char __user *, optval, int __user *, optlen)
1772 int err, fput_needed;
1773 struct socket *sock;
1775 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1777 err = security_socket_getsockopt(sock, level, optname);
1781 if (level == SOL_SOCKET)
1783 sock_getsockopt(sock, level, optname, optval,
1787 sock->ops->getsockopt(sock, level, optname, optval,
1790 fput_light(sock->file, fput_needed);
1796 * Shutdown a socket.
1799 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1801 int err, fput_needed;
1802 struct socket *sock;
1804 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1806 err = security_socket_shutdown(sock, how);
1808 err = sock->ops->shutdown(sock, how);
1809 fput_light(sock->file, fput_needed);
1814 /* A couple of helpful macros for getting the address of the 32/64 bit
1815 * fields which are the same type (int / unsigned) on our platforms.
1817 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1818 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1819 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1821 struct used_address {
1822 struct sockaddr_storage name;
1823 unsigned int name_len;
1826 static int copy_msghdr_from_user(struct msghdr *kmsg,
1827 struct user_msghdr __user *umsg,
1828 struct sockaddr __user **save_addr,
1831 struct sockaddr __user *uaddr;
1832 struct iovec __user *uiov;
1836 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1837 __get_user(uaddr, &umsg->msg_name) ||
1838 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1839 __get_user(uiov, &umsg->msg_iov) ||
1840 __get_user(nr_segs, &umsg->msg_iovlen) ||
1841 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1842 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1843 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1847 kmsg->msg_namelen = 0;
1849 if (kmsg->msg_namelen < 0)
1852 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1853 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1858 if (uaddr && kmsg->msg_namelen) {
1860 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1866 kmsg->msg_name = NULL;
1867 kmsg->msg_namelen = 0;
1870 if (nr_segs > UIO_MAXIOV)
1873 kmsg->msg_iocb = NULL;
1875 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1876 UIO_FASTIOV, iov, &kmsg->msg_iter);
1879 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1880 struct msghdr *msg_sys, unsigned int flags,
1881 struct used_address *used_address)
1883 struct compat_msghdr __user *msg_compat =
1884 (struct compat_msghdr __user *)msg;
1885 struct sockaddr_storage address;
1886 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1887 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1888 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1889 /* 20 is size of ipv6_pktinfo */
1890 unsigned char *ctl_buf = ctl;
1894 msg_sys->msg_name = &address;
1896 if (MSG_CMSG_COMPAT & flags)
1897 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1899 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1905 if (msg_sys->msg_controllen > INT_MAX)
1907 ctl_len = msg_sys->msg_controllen;
1908 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1910 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1914 ctl_buf = msg_sys->msg_control;
1915 ctl_len = msg_sys->msg_controllen;
1916 } else if (ctl_len) {
1917 if (ctl_len > sizeof(ctl)) {
1918 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1919 if (ctl_buf == NULL)
1924 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1925 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1926 * checking falls down on this.
1928 if (copy_from_user(ctl_buf,
1929 (void __user __force *)msg_sys->msg_control,
1932 msg_sys->msg_control = ctl_buf;
1934 msg_sys->msg_flags = flags;
1936 if (sock->file->f_flags & O_NONBLOCK)
1937 msg_sys->msg_flags |= MSG_DONTWAIT;
1939 * If this is sendmmsg() and current destination address is same as
1940 * previously succeeded address, omit asking LSM's decision.
1941 * used_address->name_len is initialized to UINT_MAX so that the first
1942 * destination address never matches.
1944 if (used_address && msg_sys->msg_name &&
1945 used_address->name_len == msg_sys->msg_namelen &&
1946 !memcmp(&used_address->name, msg_sys->msg_name,
1947 used_address->name_len)) {
1948 err = sock_sendmsg_nosec(sock, msg_sys);
1951 err = sock_sendmsg(sock, msg_sys);
1953 * If this is sendmmsg() and sending to current destination address was
1954 * successful, remember it.
1956 if (used_address && err >= 0) {
1957 used_address->name_len = msg_sys->msg_namelen;
1958 if (msg_sys->msg_name)
1959 memcpy(&used_address->name, msg_sys->msg_name,
1960 used_address->name_len);
1965 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1972 * BSD sendmsg interface
1975 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1977 int fput_needed, err;
1978 struct msghdr msg_sys;
1979 struct socket *sock;
1981 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1985 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1987 fput_light(sock->file, fput_needed);
1992 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1994 if (flags & MSG_CMSG_COMPAT)
1996 return __sys_sendmsg(fd, msg, flags);
2000 * Linux sendmmsg interface
2003 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2006 int fput_needed, err, datagrams;
2007 struct socket *sock;
2008 struct mmsghdr __user *entry;
2009 struct compat_mmsghdr __user *compat_entry;
2010 struct msghdr msg_sys;
2011 struct used_address used_address;
2013 if (vlen > UIO_MAXIOV)
2018 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022 used_address.name_len = UINT_MAX;
2024 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2027 while (datagrams < vlen) {
2028 if (MSG_CMSG_COMPAT & flags) {
2029 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2030 &msg_sys, flags, &used_address);
2033 err = __put_user(err, &compat_entry->msg_len);
2036 err = ___sys_sendmsg(sock,
2037 (struct user_msghdr __user *)entry,
2038 &msg_sys, flags, &used_address);
2041 err = put_user(err, &entry->msg_len);
2048 if (msg_data_left(&msg_sys))
2052 fput_light(sock->file, fput_needed);
2054 /* We only return an error if no datagrams were able to be sent */
2061 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2062 unsigned int, vlen, unsigned int, flags)
2064 if (flags & MSG_CMSG_COMPAT)
2066 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2069 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2070 struct msghdr *msg_sys, unsigned int flags, int nosec)
2072 struct compat_msghdr __user *msg_compat =
2073 (struct compat_msghdr __user *)msg;
2074 struct iovec iovstack[UIO_FASTIOV];
2075 struct iovec *iov = iovstack;
2076 unsigned long cmsg_ptr;
2080 /* kernel mode address */
2081 struct sockaddr_storage addr;
2083 /* user mode address pointers */
2084 struct sockaddr __user *uaddr;
2085 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2087 msg_sys->msg_name = &addr;
2089 if (MSG_CMSG_COMPAT & flags)
2090 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2092 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2095 total_len = iov_iter_count(&msg_sys->msg_iter);
2097 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2098 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2100 /* We assume all kernel code knows the size of sockaddr_storage */
2101 msg_sys->msg_namelen = 0;
2103 if (sock->file->f_flags & O_NONBLOCK)
2104 flags |= MSG_DONTWAIT;
2105 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2111 if (uaddr != NULL) {
2112 err = move_addr_to_user(&addr,
2113 msg_sys->msg_namelen, uaddr,
2118 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2122 if (MSG_CMSG_COMPAT & flags)
2123 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2124 &msg_compat->msg_controllen);
2126 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2127 &msg->msg_controllen);
2138 * BSD recvmsg interface
2141 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2143 int fput_needed, err;
2144 struct msghdr msg_sys;
2145 struct socket *sock;
2147 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2151 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2153 fput_light(sock->file, fput_needed);
2158 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2159 unsigned int, flags)
2161 if (flags & MSG_CMSG_COMPAT)
2163 return __sys_recvmsg(fd, msg, flags);
2167 * Linux recvmmsg interface
2170 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2171 unsigned int flags, struct timespec *timeout)
2173 int fput_needed, err, datagrams;
2174 struct socket *sock;
2175 struct mmsghdr __user *entry;
2176 struct compat_mmsghdr __user *compat_entry;
2177 struct msghdr msg_sys;
2178 struct timespec end_time;
2181 poll_select_set_timeout(&end_time, timeout->tv_sec,
2187 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2191 err = sock_error(sock->sk);
2198 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2200 while (datagrams < vlen) {
2202 * No need to ask LSM for more than the first datagram.
2204 if (MSG_CMSG_COMPAT & flags) {
2205 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2206 &msg_sys, flags & ~MSG_WAITFORONE,
2210 err = __put_user(err, &compat_entry->msg_len);
2213 err = ___sys_recvmsg(sock,
2214 (struct user_msghdr __user *)entry,
2215 &msg_sys, flags & ~MSG_WAITFORONE,
2219 err = put_user(err, &entry->msg_len);
2227 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2228 if (flags & MSG_WAITFORONE)
2229 flags |= MSG_DONTWAIT;
2232 ktime_get_ts(timeout);
2233 *timeout = timespec_sub(end_time, *timeout);
2234 if (timeout->tv_sec < 0) {
2235 timeout->tv_sec = timeout->tv_nsec = 0;
2239 /* Timeout, return less than vlen datagrams */
2240 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2244 /* Out of band data, return right away */
2245 if (msg_sys.msg_flags & MSG_OOB)
2252 if (datagrams == 0) {
2258 * We may return less entries than requested (vlen) if the
2259 * sock is non block and there aren't enough datagrams...
2261 if (err != -EAGAIN) {
2263 * ... or if recvmsg returns an error after we
2264 * received some datagrams, where we record the
2265 * error to return on the next call or if the
2266 * app asks about it using getsockopt(SO_ERROR).
2268 sock->sk->sk_err = -err;
2271 fput_light(sock->file, fput_needed);
2276 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2277 unsigned int, vlen, unsigned int, flags,
2278 struct timespec __user *, timeout)
2281 struct timespec timeout_sys;
2283 if (flags & MSG_CMSG_COMPAT)
2287 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2289 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2292 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2294 if (datagrams > 0 &&
2295 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2296 datagrams = -EFAULT;
2301 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2302 /* Argument list sizes for sys_socketcall */
2303 #define AL(x) ((x) * sizeof(unsigned long))
2304 static const unsigned char nargs[21] = {
2305 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2306 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2307 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2314 * System call vectors.
2316 * Argument checking cleaned up. Saved 20% in size.
2317 * This function doesn't need to set the kernel lock because
2318 * it is set by the callees.
2321 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2323 unsigned long a[AUDITSC_ARGS];
2324 unsigned long a0, a1;
2328 if (call < 1 || call > SYS_SENDMMSG)
2330 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2333 if (len > sizeof(a))
2336 /* copy_from_user should be SMP safe. */
2337 if (copy_from_user(a, args, len))
2340 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2349 err = sys_socket(a0, a1, a[2]);
2352 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2355 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2358 err = sys_listen(a0, a1);
2361 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2362 (int __user *)a[2], 0);
2364 case SYS_GETSOCKNAME:
2366 sys_getsockname(a0, (struct sockaddr __user *)a1,
2367 (int __user *)a[2]);
2369 case SYS_GETPEERNAME:
2371 sys_getpeername(a0, (struct sockaddr __user *)a1,
2372 (int __user *)a[2]);
2374 case SYS_SOCKETPAIR:
2375 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2378 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2381 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2382 (struct sockaddr __user *)a[4], a[5]);
2385 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2388 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2389 (struct sockaddr __user *)a[4],
2390 (int __user *)a[5]);
2393 err = sys_shutdown(a0, a1);
2395 case SYS_SETSOCKOPT:
2396 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2398 case SYS_GETSOCKOPT:
2400 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2401 (int __user *)a[4]);
2404 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2407 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2410 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2413 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2414 (struct timespec __user *)a[4]);
2417 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2418 (int __user *)a[2], a[3]);
2427 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2430 * sock_register - add a socket protocol handler
2431 * @ops: description of protocol
2433 * This function is called by a protocol handler that wants to
2434 * advertise its address family, and have it linked into the
2435 * socket interface. The value ops->family corresponds to the
2436 * socket system call protocol family.
2438 int sock_register(const struct net_proto_family *ops)
2442 if (ops->family >= NPROTO) {
2443 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2447 spin_lock(&net_family_lock);
2448 if (rcu_dereference_protected(net_families[ops->family],
2449 lockdep_is_held(&net_family_lock)))
2452 rcu_assign_pointer(net_families[ops->family], ops);
2455 spin_unlock(&net_family_lock);
2457 pr_info("NET: Registered protocol family %d\n", ops->family);
2460 EXPORT_SYMBOL(sock_register);
2463 * sock_unregister - remove a protocol handler
2464 * @family: protocol family to remove
2466 * This function is called by a protocol handler that wants to
2467 * remove its address family, and have it unlinked from the
2468 * new socket creation.
2470 * If protocol handler is a module, then it can use module reference
2471 * counts to protect against new references. If protocol handler is not
2472 * a module then it needs to provide its own protection in
2473 * the ops->create routine.
2475 void sock_unregister(int family)
2477 BUG_ON(family < 0 || family >= NPROTO);
2479 spin_lock(&net_family_lock);
2480 RCU_INIT_POINTER(net_families[family], NULL);
2481 spin_unlock(&net_family_lock);
2485 pr_info("NET: Unregistered protocol family %d\n", family);
2487 EXPORT_SYMBOL(sock_unregister);
2489 static int __init sock_init(void)
2493 * Initialize the network sysctl infrastructure.
2495 err = net_sysctl_init();
2500 * Initialize skbuff SLAB cache
2505 * Initialize the protocols module.
2510 err = register_filesystem(&sock_fs_type);
2513 sock_mnt = kern_mount(&sock_fs_type);
2514 if (IS_ERR(sock_mnt)) {
2515 err = PTR_ERR(sock_mnt);
2519 /* The real protocol initialization is performed in later initcalls.
2522 #ifdef CONFIG_NETFILTER
2523 err = netfilter_init();
2528 ptp_classifier_init();
2534 unregister_filesystem(&sock_fs_type);
2539 core_initcall(sock_init); /* early initcall */
2541 static int __init jit_init(void)
2543 #ifdef CONFIG_BPF_JIT_ALWAYS_ON
2548 pure_initcall(jit_init);
2550 #ifdef CONFIG_PROC_FS
2551 void socket_seq_show(struct seq_file *seq)
2556 for_each_possible_cpu(cpu)
2557 counter += per_cpu(sockets_in_use, cpu);
2559 /* It can be negative, by the way. 8) */
2563 seq_printf(seq, "sockets: used %d\n", counter);
2565 #endif /* CONFIG_PROC_FS */
2567 #ifdef CONFIG_COMPAT
2568 static int do_siocgstamp(struct net *net, struct socket *sock,
2569 unsigned int cmd, void __user *up)
2571 mm_segment_t old_fs = get_fs();
2576 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2579 err = compat_put_timeval(&ktv, up);
2584 static int do_siocgstampns(struct net *net, struct socket *sock,
2585 unsigned int cmd, void __user *up)
2587 mm_segment_t old_fs = get_fs();
2588 struct timespec kts;
2592 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2595 err = compat_put_timespec(&kts, up);
2600 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2602 struct ifreq __user *uifr;
2605 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2606 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2609 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2613 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2619 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2621 struct compat_ifconf ifc32;
2623 struct ifconf __user *uifc;
2624 struct compat_ifreq __user *ifr32;
2625 struct ifreq __user *ifr;
2629 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2632 memset(&ifc, 0, sizeof(ifc));
2633 if (ifc32.ifcbuf == 0) {
2637 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2639 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2640 sizeof(struct ifreq);
2641 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2643 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2644 ifr32 = compat_ptr(ifc32.ifcbuf);
2645 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2646 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2652 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2655 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2659 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2663 ifr32 = compat_ptr(ifc32.ifcbuf);
2665 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2666 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2667 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2673 if (ifc32.ifcbuf == 0) {
2674 /* Translate from 64-bit structure multiple to
2678 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2683 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2689 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2691 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2692 bool convert_in = false, convert_out = false;
2693 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2694 struct ethtool_rxnfc __user *rxnfc;
2695 struct ifreq __user *ifr;
2696 u32 rule_cnt = 0, actual_rule_cnt;
2701 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2704 compat_rxnfc = compat_ptr(data);
2706 if (get_user(ethcmd, &compat_rxnfc->cmd))
2709 /* Most ethtool structures are defined without padding.
2710 * Unfortunately struct ethtool_rxnfc is an exception.
2715 case ETHTOOL_GRXCLSRLALL:
2716 /* Buffer size is variable */
2717 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2719 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2721 buf_size += rule_cnt * sizeof(u32);
2723 case ETHTOOL_GRXRINGS:
2724 case ETHTOOL_GRXCLSRLCNT:
2725 case ETHTOOL_GRXCLSRULE:
2726 case ETHTOOL_SRXCLSRLINS:
2729 case ETHTOOL_SRXCLSRLDEL:
2730 buf_size += sizeof(struct ethtool_rxnfc);
2735 ifr = compat_alloc_user_space(buf_size);
2736 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2738 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2741 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2742 &ifr->ifr_ifru.ifru_data))
2746 /* We expect there to be holes between fs.m_ext and
2747 * fs.ring_cookie and at the end of fs, but nowhere else.
2749 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2750 sizeof(compat_rxnfc->fs.m_ext) !=
2751 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2752 sizeof(rxnfc->fs.m_ext));
2754 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2755 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2756 offsetof(struct ethtool_rxnfc, fs.location) -
2757 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2759 if (copy_in_user(rxnfc, compat_rxnfc,
2760 (void __user *)(&rxnfc->fs.m_ext + 1) -
2761 (void __user *)rxnfc) ||
2762 copy_in_user(&rxnfc->fs.ring_cookie,
2763 &compat_rxnfc->fs.ring_cookie,
2764 (void __user *)(&rxnfc->fs.location + 1) -
2765 (void __user *)&rxnfc->fs.ring_cookie))
2767 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2768 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2770 } else if (copy_in_user(&rxnfc->rule_cnt,
2771 &compat_rxnfc->rule_cnt,
2772 sizeof(rxnfc->rule_cnt)))
2776 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2781 if (copy_in_user(compat_rxnfc, rxnfc,
2782 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2783 (const void __user *)rxnfc) ||
2784 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2785 &rxnfc->fs.ring_cookie,
2786 (const void __user *)(&rxnfc->fs.location + 1) -
2787 (const void __user *)&rxnfc->fs.ring_cookie) ||
2788 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2789 sizeof(rxnfc->rule_cnt)))
2792 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2793 /* As an optimisation, we only copy the actual
2794 * number of rules that the underlying
2795 * function returned. Since Mallory might
2796 * change the rule count in user memory, we
2797 * check that it is less than the rule count
2798 * originally given (as the user buffer size),
2799 * which has been range-checked.
2801 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2803 if (actual_rule_cnt < rule_cnt)
2804 rule_cnt = actual_rule_cnt;
2805 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2806 &rxnfc->rule_locs[0],
2807 rule_cnt * sizeof(u32)))
2815 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2818 compat_uptr_t uptr32;
2819 struct ifreq __user *uifr;
2821 uifr = compat_alloc_user_space(sizeof(*uifr));
2822 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2825 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2828 uptr = compat_ptr(uptr32);
2830 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2833 return dev_ioctl(net, SIOCWANDEV, uifr);
2836 static int bond_ioctl(struct net *net, unsigned int cmd,
2837 struct compat_ifreq __user *ifr32)
2840 mm_segment_t old_fs;
2844 case SIOCBONDENSLAVE:
2845 case SIOCBONDRELEASE:
2846 case SIOCBONDSETHWADDR:
2847 case SIOCBONDCHANGEACTIVE:
2848 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2853 err = dev_ioctl(net, cmd,
2854 (struct ifreq __user __force *) &kifr);
2859 return -ENOIOCTLCMD;
2863 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2864 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2865 struct compat_ifreq __user *u_ifreq32)
2867 struct ifreq __user *u_ifreq64;
2868 char tmp_buf[IFNAMSIZ];
2869 void __user *data64;
2872 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2875 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2877 data64 = compat_ptr(data32);
2879 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2881 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2884 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2887 return dev_ioctl(net, cmd, u_ifreq64);
2890 static int dev_ifsioc(struct net *net, struct socket *sock,
2891 unsigned int cmd, struct compat_ifreq __user *uifr32)
2893 struct ifreq __user *uifr;
2896 uifr = compat_alloc_user_space(sizeof(*uifr));
2897 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2900 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2911 case SIOCGIFBRDADDR:
2912 case SIOCGIFDSTADDR:
2913 case SIOCGIFNETMASK:
2918 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2926 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2927 struct compat_ifreq __user *uifr32)
2930 struct compat_ifmap __user *uifmap32;
2931 mm_segment_t old_fs;
2934 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2935 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2936 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2937 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2938 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2939 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2940 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2941 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2947 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2950 if (cmd == SIOCGIFMAP && !err) {
2951 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2952 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2953 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2954 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2955 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2956 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2957 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2966 struct sockaddr rt_dst; /* target address */
2967 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2968 struct sockaddr rt_genmask; /* target network mask (IP) */
2969 unsigned short rt_flags;
2972 unsigned char rt_tos;
2973 unsigned char rt_class;
2975 short rt_metric; /* +1 for binary compatibility! */
2976 /* char * */ u32 rt_dev; /* forcing the device at add */
2977 u32 rt_mtu; /* per route MTU/Window */
2978 u32 rt_window; /* Window clamping */
2979 unsigned short rt_irtt; /* Initial RTT */
2982 struct in6_rtmsg32 {
2983 struct in6_addr rtmsg_dst;
2984 struct in6_addr rtmsg_src;
2985 struct in6_addr rtmsg_gateway;
2995 static int routing_ioctl(struct net *net, struct socket *sock,
2996 unsigned int cmd, void __user *argp)
3000 struct in6_rtmsg r6;
3004 mm_segment_t old_fs = get_fs();
3006 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3007 struct in6_rtmsg32 __user *ur6 = argp;
3008 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3009 3 * sizeof(struct in6_addr));
3010 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3011 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3012 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3013 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3014 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3015 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3016 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3020 struct rtentry32 __user *ur4 = argp;
3021 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3022 3 * sizeof(struct sockaddr));
3023 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3024 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3025 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3026 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3027 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3028 ret |= get_user(rtdev, &(ur4->rt_dev));
3030 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3031 r4.rt_dev = (char __user __force *)devname;
3045 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3052 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3053 * for some operations; this forces use of the newer bridge-utils that
3054 * use compatible ioctls
3056 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3060 if (get_user(tmp, argp))
3062 if (tmp == BRCTL_GET_VERSION)
3063 return BRCTL_VERSION + 1;
3067 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3068 unsigned int cmd, unsigned long arg)
3070 void __user *argp = compat_ptr(arg);
3071 struct sock *sk = sock->sk;
3072 struct net *net = sock_net(sk);
3074 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3075 return compat_ifr_data_ioctl(net, cmd, argp);
3080 return old_bridge_ioctl(argp);
3082 return dev_ifname32(net, argp);
3084 return dev_ifconf(net, argp);
3086 return ethtool_ioctl(net, argp);
3088 return compat_siocwandev(net, argp);
3091 return compat_sioc_ifmap(net, cmd, argp);
3092 case SIOCBONDENSLAVE:
3093 case SIOCBONDRELEASE:
3094 case SIOCBONDSETHWADDR:
3095 case SIOCBONDCHANGEACTIVE:
3096 return bond_ioctl(net, cmd, argp);
3099 return routing_ioctl(net, sock, cmd, argp);
3101 return do_siocgstamp(net, sock, cmd, argp);
3103 return do_siocgstampns(net, sock, cmd, argp);
3104 case SIOCBONDSLAVEINFOQUERY:
3105 case SIOCBONDINFOQUERY:
3108 return compat_ifr_data_ioctl(net, cmd, argp);
3120 return sock_ioctl(file, cmd, arg);
3137 case SIOCSIFHWBROADCAST:
3139 case SIOCGIFBRDADDR:
3140 case SIOCSIFBRDADDR:
3141 case SIOCGIFDSTADDR:
3142 case SIOCSIFDSTADDR:
3143 case SIOCGIFNETMASK:
3144 case SIOCSIFNETMASK:
3155 return dev_ifsioc(net, sock, cmd, argp);
3161 return sock_do_ioctl(net, sock, cmd, arg);
3164 return -ENOIOCTLCMD;
3167 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3170 struct socket *sock = file->private_data;
3171 int ret = -ENOIOCTLCMD;
3178 if (sock->ops->compat_ioctl)
3179 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3181 if (ret == -ENOIOCTLCMD &&
3182 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3183 ret = compat_wext_handle_ioctl(net, cmd, arg);
3185 if (ret == -ENOIOCTLCMD)
3186 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3192 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3194 return sock->ops->bind(sock, addr, addrlen);
3196 EXPORT_SYMBOL(kernel_bind);
3198 int kernel_listen(struct socket *sock, int backlog)
3200 return sock->ops->listen(sock, backlog);
3202 EXPORT_SYMBOL(kernel_listen);
3204 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3206 struct sock *sk = sock->sk;
3209 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3214 err = sock->ops->accept(sock, *newsock, flags);
3216 sock_release(*newsock);
3221 (*newsock)->ops = sock->ops;
3222 __module_get((*newsock)->ops->owner);
3227 EXPORT_SYMBOL(kernel_accept);
3229 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3232 return sock->ops->connect(sock, addr, addrlen, flags);
3234 EXPORT_SYMBOL(kernel_connect);
3236 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3239 return sock->ops->getname(sock, addr, addrlen, 0);
3241 EXPORT_SYMBOL(kernel_getsockname);
3243 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3246 return sock->ops->getname(sock, addr, addrlen, 1);
3248 EXPORT_SYMBOL(kernel_getpeername);
3250 int kernel_getsockopt(struct socket *sock, int level, int optname,
3251 char *optval, int *optlen)
3253 mm_segment_t oldfs = get_fs();
3254 char __user *uoptval;
3255 int __user *uoptlen;
3258 uoptval = (char __user __force *) optval;
3259 uoptlen = (int __user __force *) optlen;
3262 if (level == SOL_SOCKET)
3263 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3265 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3270 EXPORT_SYMBOL(kernel_getsockopt);
3272 int kernel_setsockopt(struct socket *sock, int level, int optname,
3273 char *optval, unsigned int optlen)
3275 mm_segment_t oldfs = get_fs();
3276 char __user *uoptval;
3279 uoptval = (char __user __force *) optval;
3282 if (level == SOL_SOCKET)
3283 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3285 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3290 EXPORT_SYMBOL(kernel_setsockopt);
3292 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3293 size_t size, int flags)
3295 if (sock->ops->sendpage)
3296 return sock->ops->sendpage(sock, page, offset, size, flags);
3298 return sock_no_sendpage(sock, page, offset, size, flags);
3300 EXPORT_SYMBOL(kernel_sendpage);
3302 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3304 mm_segment_t oldfs = get_fs();
3308 err = sock->ops->ioctl(sock, cmd, arg);
3313 EXPORT_SYMBOL(kernel_sock_ioctl);
3315 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3317 return sock->ops->shutdown(sock, how);
3319 EXPORT_SYMBOL(kernel_sock_shutdown);