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 int sockfs_security_xattr_set(const struct xattr_handler *handler,
345 struct dentry *dentry, struct inode *inode,
346 const char *suffix, const void *value,
347 size_t size, int flags)
349 /* Handled by LSM. */
353 static const struct xattr_handler sockfs_security_xattr_handler = {
354 .prefix = XATTR_SECURITY_PREFIX,
355 .set = sockfs_security_xattr_set,
358 static const struct xattr_handler *sockfs_xattr_handlers[] = {
359 &sockfs_xattr_handler,
360 &sockfs_security_xattr_handler,
364 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
365 int flags, const char *dev_name, void *data)
367 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
368 sockfs_xattr_handlers,
369 &sockfs_dentry_operations, SOCKFS_MAGIC);
372 static struct vfsmount *sock_mnt __read_mostly;
374 static struct file_system_type sock_fs_type = {
376 .mount = sockfs_mount,
377 .kill_sb = kill_anon_super,
381 * Obtains the first available file descriptor and sets it up for use.
383 * These functions create file structures and maps them to fd space
384 * of the current process. On success it returns file descriptor
385 * and file struct implicitly stored in sock->file.
386 * Note that another thread may close file descriptor before we return
387 * from this function. We use the fact that now we do not refer
388 * to socket after mapping. If one day we will need it, this
389 * function will increment ref. count on file by 1.
391 * In any case returned fd MAY BE not valid!
392 * This race condition is unavoidable
393 * with shared fd spaces, we cannot solve it inside kernel,
394 * but we take care of internal coherence yet.
397 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
399 struct qstr name = { .name = "" };
405 name.len = strlen(name.name);
406 } else if (sock->sk) {
407 name.name = sock->sk->sk_prot_creator->name;
408 name.len = strlen(name.name);
410 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
411 if (unlikely(!path.dentry))
412 return ERR_PTR(-ENOMEM);
413 path.mnt = mntget(sock_mnt);
415 d_instantiate(path.dentry, SOCK_INODE(sock));
417 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
420 /* drop dentry, keep inode */
421 ihold(d_inode(path.dentry));
427 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
428 file->private_data = sock;
431 EXPORT_SYMBOL(sock_alloc_file);
433 static int sock_map_fd(struct socket *sock, int flags)
435 struct file *newfile;
436 int fd = get_unused_fd_flags(flags);
437 if (unlikely(fd < 0))
440 newfile = sock_alloc_file(sock, flags, NULL);
441 if (likely(!IS_ERR(newfile))) {
442 fd_install(fd, newfile);
447 return PTR_ERR(newfile);
450 struct socket *sock_from_file(struct file *file, int *err)
452 if (file->f_op == &socket_file_ops)
453 return file->private_data; /* set in sock_map_fd */
458 EXPORT_SYMBOL(sock_from_file);
461 * sockfd_lookup - Go from a file number to its socket slot
463 * @err: pointer to an error code return
465 * The file handle passed in is locked and the socket it is bound
466 * too is returned. If an error occurs the err pointer is overwritten
467 * with a negative errno code and NULL is returned. The function checks
468 * for both invalid handles and passing a handle which is not a socket.
470 * On a success the socket object pointer is returned.
473 struct socket *sockfd_lookup(int fd, int *err)
484 sock = sock_from_file(file, err);
489 EXPORT_SYMBOL(sockfd_lookup);
491 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
493 struct fd f = fdget(fd);
498 sock = sock_from_file(f.file, err);
500 *fput_needed = f.flags;
508 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
514 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
524 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
529 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
536 static const struct inode_operations sockfs_inode_ops = {
537 .listxattr = sockfs_listxattr,
541 * sock_alloc - allocate a socket
543 * Allocate a new inode and socket object. The two are bound together
544 * and initialised. The socket is then returned. If we are out of inodes
548 struct socket *sock_alloc(void)
553 inode = new_inode_pseudo(sock_mnt->mnt_sb);
557 sock = SOCKET_I(inode);
559 kmemcheck_annotate_bitfield(sock, type);
560 inode->i_ino = get_next_ino();
561 inode->i_mode = S_IFSOCK | S_IRWXUGO;
562 inode->i_uid = current_fsuid();
563 inode->i_gid = current_fsgid();
564 inode->i_op = &sockfs_inode_ops;
566 this_cpu_add(sockets_in_use, 1);
569 EXPORT_SYMBOL(sock_alloc);
572 * sock_release - close a socket
573 * @sock: socket to close
575 * The socket is released from the protocol stack if it has a release
576 * callback, and the inode is then released if the socket is bound to
577 * an inode not a file.
580 void sock_release(struct socket *sock)
583 struct module *owner = sock->ops->owner;
585 sock->ops->release(sock);
590 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
591 pr_err("%s: fasync list not empty!\n", __func__);
593 this_cpu_sub(sockets_in_use, 1);
595 iput(SOCK_INODE(sock));
600 EXPORT_SYMBOL(sock_release);
602 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
604 u8 flags = *tx_flags;
606 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
607 flags |= SKBTX_HW_TSTAMP;
609 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
610 flags |= SKBTX_SW_TSTAMP;
612 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
613 flags |= SKBTX_SCHED_TSTAMP;
617 EXPORT_SYMBOL(__sock_tx_timestamp);
619 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
621 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
622 BUG_ON(ret == -EIOCBQUEUED);
626 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
628 int err = security_socket_sendmsg(sock, msg,
631 return err ?: sock_sendmsg_nosec(sock, msg);
633 EXPORT_SYMBOL(sock_sendmsg);
635 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
636 struct kvec *vec, size_t num, size_t size)
638 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
639 return sock_sendmsg(sock, msg);
641 EXPORT_SYMBOL(kernel_sendmsg);
644 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
646 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
649 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
650 struct scm_timestamping tss;
652 struct skb_shared_hwtstamps *shhwtstamps =
655 /* Race occurred between timestamp enabling and packet
656 receiving. Fill in the current time for now. */
657 if (need_software_tstamp && skb->tstamp.tv64 == 0)
658 __net_timestamp(skb);
660 if (need_software_tstamp) {
661 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
663 skb_get_timestamp(skb, &tv);
664 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
668 skb_get_timestampns(skb, &ts);
669 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
674 memset(&tss, 0, sizeof(tss));
675 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
676 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
679 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
680 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
683 put_cmsg(msg, SOL_SOCKET,
684 SCM_TIMESTAMPING, sizeof(tss), &tss);
686 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
688 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
693 if (!sock_flag(sk, SOCK_WIFI_STATUS))
695 if (!skb->wifi_acked_valid)
698 ack = skb->wifi_acked;
700 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
702 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
704 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
707 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
708 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
709 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
712 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
715 sock_recv_timestamp(msg, sk, skb);
716 sock_recv_drops(msg, sk, skb);
718 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
720 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
723 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
726 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
728 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
730 return err ?: sock_recvmsg_nosec(sock, msg, flags);
732 EXPORT_SYMBOL(sock_recvmsg);
735 * kernel_recvmsg - Receive a message from a socket (kernel space)
736 * @sock: The socket to receive the message from
737 * @msg: Received message
738 * @vec: Input s/g array for message data
739 * @num: Size of input s/g array
740 * @size: Number of bytes to read
741 * @flags: Message flags (MSG_DONTWAIT, etc...)
743 * On return the msg structure contains the scatter/gather array passed in the
744 * vec argument. The array is modified so that it consists of the unfilled
745 * portion of the original array.
747 * The returned value is the total number of bytes received, or an error.
749 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
750 struct kvec *vec, size_t num, size_t size, int flags)
752 mm_segment_t oldfs = get_fs();
755 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
757 result = sock_recvmsg(sock, msg, flags);
761 EXPORT_SYMBOL(kernel_recvmsg);
763 static ssize_t sock_sendpage(struct file *file, struct page *page,
764 int offset, size_t size, loff_t *ppos, int more)
769 sock = file->private_data;
771 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
772 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
775 return kernel_sendpage(sock, page, offset, size, flags);
778 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
779 struct pipe_inode_info *pipe, size_t len,
782 struct socket *sock = file->private_data;
784 if (unlikely(!sock->ops->splice_read))
787 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
790 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
792 struct file *file = iocb->ki_filp;
793 struct socket *sock = file->private_data;
794 struct msghdr msg = {.msg_iter = *to,
798 if (file->f_flags & O_NONBLOCK)
799 msg.msg_flags = MSG_DONTWAIT;
801 if (iocb->ki_pos != 0)
804 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
807 res = sock_recvmsg(sock, &msg, msg.msg_flags);
812 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
814 struct file *file = iocb->ki_filp;
815 struct socket *sock = file->private_data;
816 struct msghdr msg = {.msg_iter = *from,
820 if (iocb->ki_pos != 0)
823 if (file->f_flags & O_NONBLOCK)
824 msg.msg_flags = MSG_DONTWAIT;
826 if (sock->type == SOCK_SEQPACKET)
827 msg.msg_flags |= MSG_EOR;
829 res = sock_sendmsg(sock, &msg);
830 *from = msg.msg_iter;
835 * Atomic setting of ioctl hooks to avoid race
836 * with module unload.
839 static DEFINE_MUTEX(br_ioctl_mutex);
840 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
842 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
844 mutex_lock(&br_ioctl_mutex);
845 br_ioctl_hook = hook;
846 mutex_unlock(&br_ioctl_mutex);
848 EXPORT_SYMBOL(brioctl_set);
850 static DEFINE_MUTEX(vlan_ioctl_mutex);
851 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
853 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
855 mutex_lock(&vlan_ioctl_mutex);
856 vlan_ioctl_hook = hook;
857 mutex_unlock(&vlan_ioctl_mutex);
859 EXPORT_SYMBOL(vlan_ioctl_set);
861 static DEFINE_MUTEX(dlci_ioctl_mutex);
862 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
864 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
866 mutex_lock(&dlci_ioctl_mutex);
867 dlci_ioctl_hook = hook;
868 mutex_unlock(&dlci_ioctl_mutex);
870 EXPORT_SYMBOL(dlci_ioctl_set);
872 static long sock_do_ioctl(struct net *net, struct socket *sock,
873 unsigned int cmd, unsigned long arg)
876 void __user *argp = (void __user *)arg;
878 err = sock->ops->ioctl(sock, cmd, arg);
881 * If this ioctl is unknown try to hand it down
884 if (err == -ENOIOCTLCMD)
885 err = dev_ioctl(net, cmd, argp);
891 * With an ioctl, arg may well be a user mode pointer, but we don't know
892 * what to do with it - that's up to the protocol still.
895 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
899 void __user *argp = (void __user *)arg;
903 sock = file->private_data;
906 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
907 err = dev_ioctl(net, cmd, argp);
909 #ifdef CONFIG_WEXT_CORE
910 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
911 err = dev_ioctl(net, cmd, argp);
918 if (get_user(pid, (int __user *)argp))
920 f_setown(sock->file, pid, 1);
925 err = put_user(f_getown(sock->file),
934 request_module("bridge");
936 mutex_lock(&br_ioctl_mutex);
938 err = br_ioctl_hook(net, cmd, argp);
939 mutex_unlock(&br_ioctl_mutex);
944 if (!vlan_ioctl_hook)
945 request_module("8021q");
947 mutex_lock(&vlan_ioctl_mutex);
949 err = vlan_ioctl_hook(net, argp);
950 mutex_unlock(&vlan_ioctl_mutex);
955 if (!dlci_ioctl_hook)
956 request_module("dlci");
958 mutex_lock(&dlci_ioctl_mutex);
960 err = dlci_ioctl_hook(cmd, argp);
961 mutex_unlock(&dlci_ioctl_mutex);
964 err = sock_do_ioctl(net, sock, cmd, arg);
970 int sock_create_lite(int family, int type, int protocol, struct socket **res)
973 struct socket *sock = NULL;
975 err = security_socket_create(family, type, protocol, 1);
986 err = security_socket_post_create(sock, family, type, protocol, 1);
998 EXPORT_SYMBOL(sock_create_lite);
1000 /* No kernel lock held - perfect */
1001 static unsigned int sock_poll(struct file *file, poll_table *wait)
1003 unsigned int busy_flag = 0;
1004 struct socket *sock;
1007 * We can't return errors to poll, so it's either yes or no.
1009 sock = file->private_data;
1011 if (sk_can_busy_loop(sock->sk)) {
1012 /* this socket can poll_ll so tell the system call */
1013 busy_flag = POLL_BUSY_LOOP;
1015 /* once, only if requested by syscall */
1016 if (wait && (wait->_key & POLL_BUSY_LOOP))
1017 sk_busy_loop(sock->sk, 1);
1020 return busy_flag | sock->ops->poll(file, sock, wait);
1023 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1025 struct socket *sock = file->private_data;
1027 return sock->ops->mmap(file, sock, vma);
1030 static int sock_close(struct inode *inode, struct file *filp)
1032 sock_release(SOCKET_I(inode));
1037 * Update the socket async list
1039 * Fasync_list locking strategy.
1041 * 1. fasync_list is modified only under process context socket lock
1042 * i.e. under semaphore.
1043 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1044 * or under socket lock
1047 static int sock_fasync(int fd, struct file *filp, int on)
1049 struct socket *sock = filp->private_data;
1050 struct sock *sk = sock->sk;
1051 struct socket_wq *wq;
1057 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1058 fasync_helper(fd, filp, on, &wq->fasync_list);
1060 if (!wq->fasync_list)
1061 sock_reset_flag(sk, SOCK_FASYNC);
1063 sock_set_flag(sk, SOCK_FASYNC);
1069 /* This function may be called only under rcu_lock */
1071 int sock_wake_async(struct socket_wq *wq, int how, int band)
1073 if (!wq || !wq->fasync_list)
1077 case SOCK_WAKE_WAITD:
1078 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1081 case SOCK_WAKE_SPACE:
1082 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1087 kill_fasync(&wq->fasync_list, SIGIO, band);
1090 kill_fasync(&wq->fasync_list, SIGURG, band);
1095 EXPORT_SYMBOL(sock_wake_async);
1097 int __sock_create(struct net *net, int family, int type, int protocol,
1098 struct socket **res, int kern)
1101 struct socket *sock;
1102 const struct net_proto_family *pf;
1105 * Check protocol is in range
1107 if (family < 0 || family >= NPROTO)
1108 return -EAFNOSUPPORT;
1109 if (type < 0 || type >= SOCK_MAX)
1114 This uglymoron is moved from INET layer to here to avoid
1115 deadlock in module load.
1117 if (family == PF_INET && type == SOCK_PACKET) {
1118 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1123 err = security_socket_create(family, type, protocol, kern);
1128 * Allocate the socket and allow the family to set things up. if
1129 * the protocol is 0, the family is instructed to select an appropriate
1132 sock = sock_alloc();
1134 net_warn_ratelimited("socket: no more sockets\n");
1135 return -ENFILE; /* Not exactly a match, but its the
1136 closest posix thing */
1141 #ifdef CONFIG_MODULES
1142 /* Attempt to load a protocol module if the find failed.
1144 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1145 * requested real, full-featured networking support upon configuration.
1146 * Otherwise module support will break!
1148 if (rcu_access_pointer(net_families[family]) == NULL)
1149 request_module("net-pf-%d", family);
1153 pf = rcu_dereference(net_families[family]);
1154 err = -EAFNOSUPPORT;
1159 * We will call the ->create function, that possibly is in a loadable
1160 * module, so we have to bump that loadable module refcnt first.
1162 if (!try_module_get(pf->owner))
1165 /* Now protected by module ref count */
1168 err = pf->create(net, sock, protocol, kern);
1170 goto out_module_put;
1173 * Now to bump the refcnt of the [loadable] module that owns this
1174 * socket at sock_release time we decrement its refcnt.
1176 if (!try_module_get(sock->ops->owner))
1177 goto out_module_busy;
1180 * Now that we're done with the ->create function, the [loadable]
1181 * module can have its refcnt decremented
1183 module_put(pf->owner);
1184 err = security_socket_post_create(sock, family, type, protocol, kern);
1186 goto out_sock_release;
1192 err = -EAFNOSUPPORT;
1195 module_put(pf->owner);
1202 goto out_sock_release;
1204 EXPORT_SYMBOL(__sock_create);
1206 int sock_create(int family, int type, int protocol, struct socket **res)
1208 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1210 EXPORT_SYMBOL(sock_create);
1212 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1214 return __sock_create(net, family, type, protocol, res, 1);
1216 EXPORT_SYMBOL(sock_create_kern);
1218 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1221 struct socket *sock;
1224 /* Check the SOCK_* constants for consistency. */
1225 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1226 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1227 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1228 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1230 flags = type & ~SOCK_TYPE_MASK;
1231 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1233 type &= SOCK_TYPE_MASK;
1235 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1236 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1238 retval = sock_create(family, type, protocol, &sock);
1242 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1247 /* It may be already another descriptor 8) Not kernel problem. */
1256 * Create a pair of connected sockets.
1259 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1260 int __user *, usockvec)
1262 struct socket *sock1, *sock2;
1264 struct file *newfile1, *newfile2;
1267 flags = type & ~SOCK_TYPE_MASK;
1268 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1270 type &= SOCK_TYPE_MASK;
1272 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1273 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1276 * Obtain the first socket and check if the underlying protocol
1277 * supports the socketpair call.
1280 err = sock_create(family, type, protocol, &sock1);
1284 err = sock_create(family, type, protocol, &sock2);
1288 err = sock1->ops->socketpair(sock1, sock2);
1290 goto out_release_both;
1292 fd1 = get_unused_fd_flags(flags);
1293 if (unlikely(fd1 < 0)) {
1295 goto out_release_both;
1298 fd2 = get_unused_fd_flags(flags);
1299 if (unlikely(fd2 < 0)) {
1301 goto out_put_unused_1;
1304 newfile1 = sock_alloc_file(sock1, flags, NULL);
1305 if (IS_ERR(newfile1)) {
1306 err = PTR_ERR(newfile1);
1307 goto out_put_unused_both;
1310 newfile2 = sock_alloc_file(sock2, flags, NULL);
1311 if (IS_ERR(newfile2)) {
1312 err = PTR_ERR(newfile2);
1316 err = put_user(fd1, &usockvec[0]);
1320 err = put_user(fd2, &usockvec[1]);
1324 audit_fd_pair(fd1, fd2);
1326 fd_install(fd1, newfile1);
1327 fd_install(fd2, newfile2);
1328 /* fd1 and fd2 may be already another descriptors.
1329 * Not kernel problem.
1345 sock_release(sock2);
1348 out_put_unused_both:
1353 sock_release(sock2);
1355 sock_release(sock1);
1361 * Bind a name to a socket. Nothing much to do here since it's
1362 * the protocol's responsibility to handle the local address.
1364 * We move the socket address to kernel space before we call
1365 * the protocol layer (having also checked the address is ok).
1368 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1370 struct socket *sock;
1371 struct sockaddr_storage address;
1372 int err, fput_needed;
1374 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1376 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1378 err = security_socket_bind(sock,
1379 (struct sockaddr *)&address,
1382 err = sock->ops->bind(sock,
1386 fput_light(sock->file, fput_needed);
1392 * Perform a listen. Basically, we allow the protocol to do anything
1393 * necessary for a listen, and if that works, we mark the socket as
1394 * ready for listening.
1397 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1399 struct socket *sock;
1400 int err, fput_needed;
1403 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1405 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1406 if ((unsigned int)backlog > somaxconn)
1407 backlog = somaxconn;
1409 err = security_socket_listen(sock, backlog);
1411 err = sock->ops->listen(sock, backlog);
1413 fput_light(sock->file, fput_needed);
1419 * For accept, we attempt to create a new socket, set up the link
1420 * with the client, wake up the client, then return the new
1421 * connected fd. We collect the address of the connector in kernel
1422 * space and move it to user at the very end. This is unclean because
1423 * we open the socket then return an error.
1425 * 1003.1g adds the ability to recvmsg() to query connection pending
1426 * status to recvmsg. We need to add that support in a way thats
1427 * clean when we restucture accept also.
1430 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1431 int __user *, upeer_addrlen, int, flags)
1433 struct socket *sock, *newsock;
1434 struct file *newfile;
1435 int err, len, newfd, fput_needed;
1436 struct sockaddr_storage address;
1438 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1441 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1442 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1444 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1449 newsock = sock_alloc();
1453 newsock->type = sock->type;
1454 newsock->ops = sock->ops;
1457 * We don't need try_module_get here, as the listening socket (sock)
1458 * has the protocol module (sock->ops->owner) held.
1460 __module_get(newsock->ops->owner);
1462 newfd = get_unused_fd_flags(flags);
1463 if (unlikely(newfd < 0)) {
1465 sock_release(newsock);
1468 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1469 if (IS_ERR(newfile)) {
1470 err = PTR_ERR(newfile);
1471 put_unused_fd(newfd);
1472 sock_release(newsock);
1476 err = security_socket_accept(sock, newsock);
1480 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1484 if (upeer_sockaddr) {
1485 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1487 err = -ECONNABORTED;
1490 err = move_addr_to_user(&address,
1491 len, upeer_sockaddr, upeer_addrlen);
1496 /* File flags are not inherited via accept() unlike another OSes. */
1498 fd_install(newfd, newfile);
1502 fput_light(sock->file, fput_needed);
1507 put_unused_fd(newfd);
1511 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1512 int __user *, upeer_addrlen)
1514 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1518 * Attempt to connect to a socket with the server address. The address
1519 * is in user space so we verify it is OK and move it to kernel space.
1521 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1524 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1525 * other SEQPACKET protocols that take time to connect() as it doesn't
1526 * include the -EINPROGRESS status for such sockets.
1529 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1532 struct socket *sock;
1533 struct sockaddr_storage address;
1534 int err, fput_needed;
1536 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1539 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1544 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1548 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1549 sock->file->f_flags);
1551 fput_light(sock->file, fput_needed);
1557 * Get the local address ('name') of a socket object. Move the obtained
1558 * name to user space.
1561 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1562 int __user *, usockaddr_len)
1564 struct socket *sock;
1565 struct sockaddr_storage address;
1566 int len, err, fput_needed;
1568 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1572 err = security_socket_getsockname(sock);
1576 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1579 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1582 fput_light(sock->file, fput_needed);
1588 * Get the remote address ('name') of a socket object. Move the obtained
1589 * name to user space.
1592 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1593 int __user *, usockaddr_len)
1595 struct socket *sock;
1596 struct sockaddr_storage address;
1597 int len, err, fput_needed;
1599 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1601 err = security_socket_getpeername(sock);
1603 fput_light(sock->file, fput_needed);
1608 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1611 err = move_addr_to_user(&address, len, usockaddr,
1613 fput_light(sock->file, fput_needed);
1619 * Send a datagram to a given address. We move the address into kernel
1620 * space and check the user space data area is readable before invoking
1624 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1625 unsigned int, flags, struct sockaddr __user *, addr,
1628 struct socket *sock;
1629 struct sockaddr_storage address;
1635 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1638 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1642 msg.msg_name = NULL;
1643 msg.msg_control = NULL;
1644 msg.msg_controllen = 0;
1645 msg.msg_namelen = 0;
1647 err = move_addr_to_kernel(addr, addr_len, &address);
1650 msg.msg_name = (struct sockaddr *)&address;
1651 msg.msg_namelen = addr_len;
1653 if (sock->file->f_flags & O_NONBLOCK)
1654 flags |= MSG_DONTWAIT;
1655 msg.msg_flags = flags;
1656 err = sock_sendmsg(sock, &msg);
1659 fput_light(sock->file, fput_needed);
1665 * Send a datagram down a socket.
1668 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1669 unsigned int, flags)
1671 return sys_sendto(fd, buff, len, flags, NULL, 0);
1675 * Receive a frame from the socket and optionally record the address of the
1676 * sender. We verify the buffers are writable and if needed move the
1677 * sender address from kernel to user space.
1680 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1681 unsigned int, flags, struct sockaddr __user *, addr,
1682 int __user *, addr_len)
1684 struct socket *sock;
1687 struct sockaddr_storage address;
1691 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1694 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1698 msg.msg_control = NULL;
1699 msg.msg_controllen = 0;
1700 /* Save some cycles and don't copy the address if not needed */
1701 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1702 /* We assume all kernel code knows the size of sockaddr_storage */
1703 msg.msg_namelen = 0;
1704 msg.msg_iocb = NULL;
1706 if (sock->file->f_flags & O_NONBLOCK)
1707 flags |= MSG_DONTWAIT;
1708 err = sock_recvmsg(sock, &msg, flags);
1710 if (err >= 0 && addr != NULL) {
1711 err2 = move_addr_to_user(&address,
1712 msg.msg_namelen, addr, addr_len);
1717 fput_light(sock->file, fput_needed);
1723 * Receive a datagram from a socket.
1726 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1727 unsigned int, flags)
1729 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1733 * Set a socket option. Because we don't know the option lengths we have
1734 * to pass the user mode parameter for the protocols to sort out.
1737 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1738 char __user *, optval, int, optlen)
1740 int err, fput_needed;
1741 struct socket *sock;
1746 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1748 err = security_socket_setsockopt(sock, level, optname);
1752 if (level == SOL_SOCKET)
1754 sock_setsockopt(sock, level, optname, optval,
1758 sock->ops->setsockopt(sock, level, optname, optval,
1761 fput_light(sock->file, fput_needed);
1767 * Get a socket option. Because we don't know the option lengths we have
1768 * to pass a user mode parameter for the protocols to sort out.
1771 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1772 char __user *, optval, int __user *, optlen)
1774 int err, fput_needed;
1775 struct socket *sock;
1777 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1779 err = security_socket_getsockopt(sock, level, optname);
1783 if (level == SOL_SOCKET)
1785 sock_getsockopt(sock, level, optname, optval,
1789 sock->ops->getsockopt(sock, level, optname, optval,
1792 fput_light(sock->file, fput_needed);
1798 * Shutdown a socket.
1801 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1803 int err, fput_needed;
1804 struct socket *sock;
1806 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1808 err = security_socket_shutdown(sock, how);
1810 err = sock->ops->shutdown(sock, how);
1811 fput_light(sock->file, fput_needed);
1816 /* A couple of helpful macros for getting the address of the 32/64 bit
1817 * fields which are the same type (int / unsigned) on our platforms.
1819 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1820 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1821 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1823 struct used_address {
1824 struct sockaddr_storage name;
1825 unsigned int name_len;
1828 static int copy_msghdr_from_user(struct msghdr *kmsg,
1829 struct user_msghdr __user *umsg,
1830 struct sockaddr __user **save_addr,
1833 struct sockaddr __user *uaddr;
1834 struct iovec __user *uiov;
1838 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1839 __get_user(uaddr, &umsg->msg_name) ||
1840 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1841 __get_user(uiov, &umsg->msg_iov) ||
1842 __get_user(nr_segs, &umsg->msg_iovlen) ||
1843 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1844 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1845 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1849 kmsg->msg_namelen = 0;
1851 if (kmsg->msg_namelen < 0)
1854 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1855 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1860 if (uaddr && kmsg->msg_namelen) {
1862 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1868 kmsg->msg_name = NULL;
1869 kmsg->msg_namelen = 0;
1872 if (nr_segs > UIO_MAXIOV)
1875 kmsg->msg_iocb = NULL;
1877 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1878 UIO_FASTIOV, iov, &kmsg->msg_iter);
1881 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1882 struct msghdr *msg_sys, unsigned int flags,
1883 struct used_address *used_address,
1884 unsigned int allowed_msghdr_flags)
1886 struct compat_msghdr __user *msg_compat =
1887 (struct compat_msghdr __user *)msg;
1888 struct sockaddr_storage address;
1889 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1890 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1891 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1892 /* 20 is size of ipv6_pktinfo */
1893 unsigned char *ctl_buf = ctl;
1897 msg_sys->msg_name = &address;
1899 if (MSG_CMSG_COMPAT & flags)
1900 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1902 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1908 if (msg_sys->msg_controllen > INT_MAX)
1910 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1911 ctl_len = msg_sys->msg_controllen;
1912 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1914 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1918 ctl_buf = msg_sys->msg_control;
1919 ctl_len = msg_sys->msg_controllen;
1920 } else if (ctl_len) {
1921 if (ctl_len > sizeof(ctl)) {
1922 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1923 if (ctl_buf == NULL)
1928 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1929 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1930 * checking falls down on this.
1932 if (copy_from_user(ctl_buf,
1933 (void __user __force *)msg_sys->msg_control,
1936 msg_sys->msg_control = ctl_buf;
1938 msg_sys->msg_flags = flags;
1940 if (sock->file->f_flags & O_NONBLOCK)
1941 msg_sys->msg_flags |= MSG_DONTWAIT;
1943 * If this is sendmmsg() and current destination address is same as
1944 * previously succeeded address, omit asking LSM's decision.
1945 * used_address->name_len is initialized to UINT_MAX so that the first
1946 * destination address never matches.
1948 if (used_address && msg_sys->msg_name &&
1949 used_address->name_len == msg_sys->msg_namelen &&
1950 !memcmp(&used_address->name, msg_sys->msg_name,
1951 used_address->name_len)) {
1952 err = sock_sendmsg_nosec(sock, msg_sys);
1955 err = sock_sendmsg(sock, msg_sys);
1957 * If this is sendmmsg() and sending to current destination address was
1958 * successful, remember it.
1960 if (used_address && err >= 0) {
1961 used_address->name_len = msg_sys->msg_namelen;
1962 if (msg_sys->msg_name)
1963 memcpy(&used_address->name, msg_sys->msg_name,
1964 used_address->name_len);
1969 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1976 * BSD sendmsg interface
1979 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1981 int fput_needed, err;
1982 struct msghdr msg_sys;
1983 struct socket *sock;
1985 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1989 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
1991 fput_light(sock->file, fput_needed);
1996 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1998 if (flags & MSG_CMSG_COMPAT)
2000 return __sys_sendmsg(fd, msg, flags);
2004 * Linux sendmmsg interface
2007 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2010 int fput_needed, err, datagrams;
2011 struct socket *sock;
2012 struct mmsghdr __user *entry;
2013 struct compat_mmsghdr __user *compat_entry;
2014 struct msghdr msg_sys;
2015 struct used_address used_address;
2016 unsigned int oflags = flags;
2018 if (vlen > UIO_MAXIOV)
2023 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2027 used_address.name_len = UINT_MAX;
2029 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2033 while (datagrams < vlen) {
2034 if (datagrams == vlen - 1)
2037 if (MSG_CMSG_COMPAT & flags) {
2038 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2039 &msg_sys, flags, &used_address, MSG_EOR);
2042 err = __put_user(err, &compat_entry->msg_len);
2045 err = ___sys_sendmsg(sock,
2046 (struct user_msghdr __user *)entry,
2047 &msg_sys, flags, &used_address, MSG_EOR);
2050 err = put_user(err, &entry->msg_len);
2057 if (msg_data_left(&msg_sys))
2062 fput_light(sock->file, fput_needed);
2064 /* We only return an error if no datagrams were able to be sent */
2071 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2072 unsigned int, vlen, unsigned int, flags)
2074 if (flags & MSG_CMSG_COMPAT)
2076 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2079 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2080 struct msghdr *msg_sys, unsigned int flags, int nosec)
2082 struct compat_msghdr __user *msg_compat =
2083 (struct compat_msghdr __user *)msg;
2084 struct iovec iovstack[UIO_FASTIOV];
2085 struct iovec *iov = iovstack;
2086 unsigned long cmsg_ptr;
2090 /* kernel mode address */
2091 struct sockaddr_storage addr;
2093 /* user mode address pointers */
2094 struct sockaddr __user *uaddr;
2095 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2097 msg_sys->msg_name = &addr;
2099 if (MSG_CMSG_COMPAT & flags)
2100 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2102 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2106 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2107 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2109 /* We assume all kernel code knows the size of sockaddr_storage */
2110 msg_sys->msg_namelen = 0;
2112 if (sock->file->f_flags & O_NONBLOCK)
2113 flags |= MSG_DONTWAIT;
2114 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2119 if (uaddr != NULL) {
2120 err = move_addr_to_user(&addr,
2121 msg_sys->msg_namelen, uaddr,
2126 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2130 if (MSG_CMSG_COMPAT & flags)
2131 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2132 &msg_compat->msg_controllen);
2134 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2135 &msg->msg_controllen);
2146 * BSD recvmsg interface
2149 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2151 int fput_needed, err;
2152 struct msghdr msg_sys;
2153 struct socket *sock;
2155 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2161 fput_light(sock->file, fput_needed);
2166 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2167 unsigned int, flags)
2169 if (flags & MSG_CMSG_COMPAT)
2171 return __sys_recvmsg(fd, msg, flags);
2175 * Linux recvmmsg interface
2178 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2179 unsigned int flags, struct timespec *timeout)
2181 int fput_needed, err, datagrams;
2182 struct socket *sock;
2183 struct mmsghdr __user *entry;
2184 struct compat_mmsghdr __user *compat_entry;
2185 struct msghdr msg_sys;
2186 struct timespec64 end_time;
2187 struct timespec64 timeout64;
2190 poll_select_set_timeout(&end_time, timeout->tv_sec,
2196 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2200 err = sock_error(sock->sk);
2207 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2209 while (datagrams < vlen) {
2211 * No need to ask LSM for more than the first datagram.
2213 if (MSG_CMSG_COMPAT & flags) {
2214 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2215 &msg_sys, flags & ~MSG_WAITFORONE,
2219 err = __put_user(err, &compat_entry->msg_len);
2222 err = ___sys_recvmsg(sock,
2223 (struct user_msghdr __user *)entry,
2224 &msg_sys, flags & ~MSG_WAITFORONE,
2228 err = put_user(err, &entry->msg_len);
2236 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2237 if (flags & MSG_WAITFORONE)
2238 flags |= MSG_DONTWAIT;
2241 ktime_get_ts64(&timeout64);
2242 *timeout = timespec64_to_timespec(
2243 timespec64_sub(end_time, timeout64));
2244 if (timeout->tv_sec < 0) {
2245 timeout->tv_sec = timeout->tv_nsec = 0;
2249 /* Timeout, return less than vlen datagrams */
2250 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2254 /* Out of band data, return right away */
2255 if (msg_sys.msg_flags & MSG_OOB)
2263 if (datagrams == 0) {
2269 * We may return less entries than requested (vlen) if the
2270 * sock is non block and there aren't enough datagrams...
2272 if (err != -EAGAIN) {
2274 * ... or if recvmsg returns an error after we
2275 * received some datagrams, where we record the
2276 * error to return on the next call or if the
2277 * app asks about it using getsockopt(SO_ERROR).
2279 sock->sk->sk_err = -err;
2282 fput_light(sock->file, fput_needed);
2287 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2288 unsigned int, vlen, unsigned int, flags,
2289 struct timespec __user *, timeout)
2292 struct timespec timeout_sys;
2294 if (flags & MSG_CMSG_COMPAT)
2298 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2300 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2303 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2305 if (datagrams > 0 &&
2306 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2307 datagrams = -EFAULT;
2312 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2313 /* Argument list sizes for sys_socketcall */
2314 #define AL(x) ((x) * sizeof(unsigned long))
2315 static const unsigned char nargs[21] = {
2316 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2317 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2318 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2325 * System call vectors.
2327 * Argument checking cleaned up. Saved 20% in size.
2328 * This function doesn't need to set the kernel lock because
2329 * it is set by the callees.
2332 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2334 unsigned long a[AUDITSC_ARGS];
2335 unsigned long a0, a1;
2339 if (call < 1 || call > SYS_SENDMMSG)
2343 if (len > sizeof(a))
2346 /* copy_from_user should be SMP safe. */
2347 if (copy_from_user(a, args, len))
2350 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2359 err = sys_socket(a0, a1, a[2]);
2362 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2365 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2368 err = sys_listen(a0, a1);
2371 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2372 (int __user *)a[2], 0);
2374 case SYS_GETSOCKNAME:
2376 sys_getsockname(a0, (struct sockaddr __user *)a1,
2377 (int __user *)a[2]);
2379 case SYS_GETPEERNAME:
2381 sys_getpeername(a0, (struct sockaddr __user *)a1,
2382 (int __user *)a[2]);
2384 case SYS_SOCKETPAIR:
2385 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2388 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2391 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2392 (struct sockaddr __user *)a[4], a[5]);
2395 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2398 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2399 (struct sockaddr __user *)a[4],
2400 (int __user *)a[5]);
2403 err = sys_shutdown(a0, a1);
2405 case SYS_SETSOCKOPT:
2406 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2408 case SYS_GETSOCKOPT:
2410 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2411 (int __user *)a[4]);
2414 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2417 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2420 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2423 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2424 (struct timespec __user *)a[4]);
2427 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2428 (int __user *)a[2], a[3]);
2437 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2440 * sock_register - add a socket protocol handler
2441 * @ops: description of protocol
2443 * This function is called by a protocol handler that wants to
2444 * advertise its address family, and have it linked into the
2445 * socket interface. The value ops->family corresponds to the
2446 * socket system call protocol family.
2448 int sock_register(const struct net_proto_family *ops)
2452 if (ops->family >= NPROTO) {
2453 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2457 spin_lock(&net_family_lock);
2458 if (rcu_dereference_protected(net_families[ops->family],
2459 lockdep_is_held(&net_family_lock)))
2462 rcu_assign_pointer(net_families[ops->family], ops);
2465 spin_unlock(&net_family_lock);
2467 pr_info("NET: Registered protocol family %d\n", ops->family);
2470 EXPORT_SYMBOL(sock_register);
2473 * sock_unregister - remove a protocol handler
2474 * @family: protocol family to remove
2476 * This function is called by a protocol handler that wants to
2477 * remove its address family, and have it unlinked from the
2478 * new socket creation.
2480 * If protocol handler is a module, then it can use module reference
2481 * counts to protect against new references. If protocol handler is not
2482 * a module then it needs to provide its own protection in
2483 * the ops->create routine.
2485 void sock_unregister(int family)
2487 BUG_ON(family < 0 || family >= NPROTO);
2489 spin_lock(&net_family_lock);
2490 RCU_INIT_POINTER(net_families[family], NULL);
2491 spin_unlock(&net_family_lock);
2495 pr_info("NET: Unregistered protocol family %d\n", family);
2497 EXPORT_SYMBOL(sock_unregister);
2499 static int __init sock_init(void)
2503 * Initialize the network sysctl infrastructure.
2505 err = net_sysctl_init();
2510 * Initialize skbuff SLAB cache
2515 * Initialize the protocols module.
2520 err = register_filesystem(&sock_fs_type);
2523 sock_mnt = kern_mount(&sock_fs_type);
2524 if (IS_ERR(sock_mnt)) {
2525 err = PTR_ERR(sock_mnt);
2529 /* The real protocol initialization is performed in later initcalls.
2532 #ifdef CONFIG_NETFILTER
2533 err = netfilter_init();
2538 ptp_classifier_init();
2544 unregister_filesystem(&sock_fs_type);
2549 core_initcall(sock_init); /* early initcall */
2551 static int __init jit_init(void)
2553 #ifdef CONFIG_BPF_JIT_ALWAYS_ON
2558 pure_initcall(jit_init);
2560 #ifdef CONFIG_PROC_FS
2561 void socket_seq_show(struct seq_file *seq)
2566 for_each_possible_cpu(cpu)
2567 counter += per_cpu(sockets_in_use, cpu);
2569 /* It can be negative, by the way. 8) */
2573 seq_printf(seq, "sockets: used %d\n", counter);
2575 #endif /* CONFIG_PROC_FS */
2577 #ifdef CONFIG_COMPAT
2578 static int do_siocgstamp(struct net *net, struct socket *sock,
2579 unsigned int cmd, void __user *up)
2581 mm_segment_t old_fs = get_fs();
2586 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2589 err = compat_put_timeval(&ktv, up);
2594 static int do_siocgstampns(struct net *net, struct socket *sock,
2595 unsigned int cmd, void __user *up)
2597 mm_segment_t old_fs = get_fs();
2598 struct timespec kts;
2602 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2605 err = compat_put_timespec(&kts, up);
2610 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2612 struct ifreq __user *uifr;
2615 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2616 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2619 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2623 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2629 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2631 struct compat_ifconf ifc32;
2633 struct ifconf __user *uifc;
2634 struct compat_ifreq __user *ifr32;
2635 struct ifreq __user *ifr;
2639 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2642 memset(&ifc, 0, sizeof(ifc));
2643 if (ifc32.ifcbuf == 0) {
2647 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2649 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2650 sizeof(struct ifreq);
2651 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2653 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2654 ifr32 = compat_ptr(ifc32.ifcbuf);
2655 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2656 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2662 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2665 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2669 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2673 ifr32 = compat_ptr(ifc32.ifcbuf);
2675 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2676 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2677 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2683 if (ifc32.ifcbuf == 0) {
2684 /* Translate from 64-bit structure multiple to
2688 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2693 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2699 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2701 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2702 bool convert_in = false, convert_out = false;
2703 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2704 struct ethtool_rxnfc __user *rxnfc;
2705 struct ifreq __user *ifr;
2706 u32 rule_cnt = 0, actual_rule_cnt;
2711 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2714 compat_rxnfc = compat_ptr(data);
2716 if (get_user(ethcmd, &compat_rxnfc->cmd))
2719 /* Most ethtool structures are defined without padding.
2720 * Unfortunately struct ethtool_rxnfc is an exception.
2725 case ETHTOOL_GRXCLSRLALL:
2726 /* Buffer size is variable */
2727 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2729 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2731 buf_size += rule_cnt * sizeof(u32);
2733 case ETHTOOL_GRXRINGS:
2734 case ETHTOOL_GRXCLSRLCNT:
2735 case ETHTOOL_GRXCLSRULE:
2736 case ETHTOOL_SRXCLSRLINS:
2739 case ETHTOOL_SRXCLSRLDEL:
2740 buf_size += sizeof(struct ethtool_rxnfc);
2745 ifr = compat_alloc_user_space(buf_size);
2746 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2748 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2751 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2752 &ifr->ifr_ifru.ifru_data))
2756 /* We expect there to be holes between fs.m_ext and
2757 * fs.ring_cookie and at the end of fs, but nowhere else.
2759 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2760 sizeof(compat_rxnfc->fs.m_ext) !=
2761 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2762 sizeof(rxnfc->fs.m_ext));
2764 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2765 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2766 offsetof(struct ethtool_rxnfc, fs.location) -
2767 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2769 if (copy_in_user(rxnfc, compat_rxnfc,
2770 (void __user *)(&rxnfc->fs.m_ext + 1) -
2771 (void __user *)rxnfc) ||
2772 copy_in_user(&rxnfc->fs.ring_cookie,
2773 &compat_rxnfc->fs.ring_cookie,
2774 (void __user *)(&rxnfc->fs.location + 1) -
2775 (void __user *)&rxnfc->fs.ring_cookie) ||
2776 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2777 sizeof(rxnfc->rule_cnt)))
2781 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2786 if (copy_in_user(compat_rxnfc, rxnfc,
2787 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2788 (const void __user *)rxnfc) ||
2789 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2790 &rxnfc->fs.ring_cookie,
2791 (const void __user *)(&rxnfc->fs.location + 1) -
2792 (const void __user *)&rxnfc->fs.ring_cookie) ||
2793 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2794 sizeof(rxnfc->rule_cnt)))
2797 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2798 /* As an optimisation, we only copy the actual
2799 * number of rules that the underlying
2800 * function returned. Since Mallory might
2801 * change the rule count in user memory, we
2802 * check that it is less than the rule count
2803 * originally given (as the user buffer size),
2804 * which has been range-checked.
2806 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2808 if (actual_rule_cnt < rule_cnt)
2809 rule_cnt = actual_rule_cnt;
2810 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2811 &rxnfc->rule_locs[0],
2812 rule_cnt * sizeof(u32)))
2820 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2823 compat_uptr_t uptr32;
2824 struct ifreq __user *uifr;
2826 uifr = compat_alloc_user_space(sizeof(*uifr));
2827 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2830 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2833 uptr = compat_ptr(uptr32);
2835 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2838 return dev_ioctl(net, SIOCWANDEV, uifr);
2841 static int bond_ioctl(struct net *net, unsigned int cmd,
2842 struct compat_ifreq __user *ifr32)
2845 mm_segment_t old_fs;
2849 case SIOCBONDENSLAVE:
2850 case SIOCBONDRELEASE:
2851 case SIOCBONDSETHWADDR:
2852 case SIOCBONDCHANGEACTIVE:
2853 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2858 err = dev_ioctl(net, cmd,
2859 (struct ifreq __user __force *) &kifr);
2864 return -ENOIOCTLCMD;
2868 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2869 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2870 struct compat_ifreq __user *u_ifreq32)
2872 struct ifreq __user *u_ifreq64;
2873 char tmp_buf[IFNAMSIZ];
2874 void __user *data64;
2877 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2880 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2882 data64 = compat_ptr(data32);
2884 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2886 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2889 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2892 return dev_ioctl(net, cmd, u_ifreq64);
2895 static int dev_ifsioc(struct net *net, struct socket *sock,
2896 unsigned int cmd, struct compat_ifreq __user *uifr32)
2898 struct ifreq __user *uifr;
2901 uifr = compat_alloc_user_space(sizeof(*uifr));
2902 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2905 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2916 case SIOCGIFBRDADDR:
2917 case SIOCGIFDSTADDR:
2918 case SIOCGIFNETMASK:
2923 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2931 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2932 struct compat_ifreq __user *uifr32)
2935 struct compat_ifmap __user *uifmap32;
2936 mm_segment_t old_fs;
2939 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2940 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2941 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2942 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2943 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2944 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2945 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2946 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2952 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2955 if (cmd == SIOCGIFMAP && !err) {
2956 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2957 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2958 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2959 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2960 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2961 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2962 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2971 struct sockaddr rt_dst; /* target address */
2972 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2973 struct sockaddr rt_genmask; /* target network mask (IP) */
2974 unsigned short rt_flags;
2977 unsigned char rt_tos;
2978 unsigned char rt_class;
2980 short rt_metric; /* +1 for binary compatibility! */
2981 /* char * */ u32 rt_dev; /* forcing the device at add */
2982 u32 rt_mtu; /* per route MTU/Window */
2983 u32 rt_window; /* Window clamping */
2984 unsigned short rt_irtt; /* Initial RTT */
2987 struct in6_rtmsg32 {
2988 struct in6_addr rtmsg_dst;
2989 struct in6_addr rtmsg_src;
2990 struct in6_addr rtmsg_gateway;
3000 static int routing_ioctl(struct net *net, struct socket *sock,
3001 unsigned int cmd, void __user *argp)
3005 struct in6_rtmsg r6;
3009 mm_segment_t old_fs = get_fs();
3011 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3012 struct in6_rtmsg32 __user *ur6 = argp;
3013 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3014 3 * sizeof(struct in6_addr));
3015 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3016 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3017 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3018 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3019 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3020 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3021 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3025 struct rtentry32 __user *ur4 = argp;
3026 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3027 3 * sizeof(struct sockaddr));
3028 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3029 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3030 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3031 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3032 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3033 ret |= get_user(rtdev, &(ur4->rt_dev));
3035 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3036 r4.rt_dev = (char __user __force *)devname;
3050 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3057 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3058 * for some operations; this forces use of the newer bridge-utils that
3059 * use compatible ioctls
3061 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3065 if (get_user(tmp, argp))
3067 if (tmp == BRCTL_GET_VERSION)
3068 return BRCTL_VERSION + 1;
3072 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3073 unsigned int cmd, unsigned long arg)
3075 void __user *argp = compat_ptr(arg);
3076 struct sock *sk = sock->sk;
3077 struct net *net = sock_net(sk);
3079 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3080 return compat_ifr_data_ioctl(net, cmd, argp);
3085 return old_bridge_ioctl(argp);
3087 return dev_ifname32(net, argp);
3089 return dev_ifconf(net, argp);
3091 return ethtool_ioctl(net, argp);
3093 return compat_siocwandev(net, argp);
3096 return compat_sioc_ifmap(net, cmd, argp);
3097 case SIOCBONDENSLAVE:
3098 case SIOCBONDRELEASE:
3099 case SIOCBONDSETHWADDR:
3100 case SIOCBONDCHANGEACTIVE:
3101 return bond_ioctl(net, cmd, argp);
3104 return routing_ioctl(net, sock, cmd, argp);
3106 return do_siocgstamp(net, sock, cmd, argp);
3108 return do_siocgstampns(net, sock, cmd, argp);
3109 case SIOCBONDSLAVEINFOQUERY:
3110 case SIOCBONDINFOQUERY:
3113 return compat_ifr_data_ioctl(net, cmd, argp);
3125 return sock_ioctl(file, cmd, arg);
3142 case SIOCSIFHWBROADCAST:
3144 case SIOCGIFBRDADDR:
3145 case SIOCSIFBRDADDR:
3146 case SIOCGIFDSTADDR:
3147 case SIOCSIFDSTADDR:
3148 case SIOCGIFNETMASK:
3149 case SIOCSIFNETMASK:
3160 return dev_ifsioc(net, sock, cmd, argp);
3166 return sock_do_ioctl(net, sock, cmd, arg);
3169 return -ENOIOCTLCMD;
3172 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3175 struct socket *sock = file->private_data;
3176 int ret = -ENOIOCTLCMD;
3183 if (sock->ops->compat_ioctl)
3184 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3186 if (ret == -ENOIOCTLCMD &&
3187 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3188 ret = compat_wext_handle_ioctl(net, cmd, arg);
3190 if (ret == -ENOIOCTLCMD)
3191 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3197 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3199 return sock->ops->bind(sock, addr, addrlen);
3201 EXPORT_SYMBOL(kernel_bind);
3203 int kernel_listen(struct socket *sock, int backlog)
3205 return sock->ops->listen(sock, backlog);
3207 EXPORT_SYMBOL(kernel_listen);
3209 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3211 struct sock *sk = sock->sk;
3214 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3219 err = sock->ops->accept(sock, *newsock, flags);
3221 sock_release(*newsock);
3226 (*newsock)->ops = sock->ops;
3227 __module_get((*newsock)->ops->owner);
3232 EXPORT_SYMBOL(kernel_accept);
3234 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3237 return sock->ops->connect(sock, addr, addrlen, flags);
3239 EXPORT_SYMBOL(kernel_connect);
3241 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3244 return sock->ops->getname(sock, addr, addrlen, 0);
3246 EXPORT_SYMBOL(kernel_getsockname);
3248 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3251 return sock->ops->getname(sock, addr, addrlen, 1);
3253 EXPORT_SYMBOL(kernel_getpeername);
3255 int kernel_getsockopt(struct socket *sock, int level, int optname,
3256 char *optval, int *optlen)
3258 mm_segment_t oldfs = get_fs();
3259 char __user *uoptval;
3260 int __user *uoptlen;
3263 uoptval = (char __user __force *) optval;
3264 uoptlen = (int __user __force *) optlen;
3267 if (level == SOL_SOCKET)
3268 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3270 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3275 EXPORT_SYMBOL(kernel_getsockopt);
3277 int kernel_setsockopt(struct socket *sock, int level, int optname,
3278 char *optval, unsigned int optlen)
3280 mm_segment_t oldfs = get_fs();
3281 char __user *uoptval;
3284 uoptval = (char __user __force *) optval;
3287 if (level == SOL_SOCKET)
3288 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3290 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3295 EXPORT_SYMBOL(kernel_setsockopt);
3297 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3298 size_t size, int flags)
3300 if (sock->ops->sendpage)
3301 return sock->ops->sendpage(sock, page, offset, size, flags);
3303 return sock_no_sendpage(sock, page, offset, size, flags);
3305 EXPORT_SYMBOL(kernel_sendpage);
3307 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3309 mm_segment_t oldfs = get_fs();
3313 err = sock->ops->ioctl(sock, cmd, arg);
3318 EXPORT_SYMBOL(kernel_sock_ioctl);
3320 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3322 return sock->ops->shutdown(sock, how);
3324 EXPORT_SYMBOL(kernel_sock_shutdown);