2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
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 <linux/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 void 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 BUG_ON(sock_inode_cachep == NULL);
302 static const struct super_operations sockfs_ops = {
303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
309 * sockfs_dname() is called from d_path().
311 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
313 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
314 d_inode(dentry)->i_ino);
317 static const struct dentry_operations sockfs_dentry_operations = {
318 .d_dname = sockfs_dname,
321 static int sockfs_xattr_get(const struct xattr_handler *handler,
322 struct dentry *dentry, struct inode *inode,
323 const char *suffix, void *value, size_t size)
326 if (dentry->d_name.len + 1 > size)
328 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
330 return dentry->d_name.len + 1;
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
337 static const struct xattr_handler sockfs_xattr_handler = {
338 .name = XATTR_NAME_SOCKPROTONAME,
339 .get = sockfs_xattr_get,
342 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343 struct dentry *dentry, struct inode *inode,
344 const char *suffix, const void *value,
345 size_t size, int flags)
347 /* Handled by LSM. */
351 static const struct xattr_handler sockfs_security_xattr_handler = {
352 .prefix = XATTR_SECURITY_PREFIX,
353 .set = sockfs_security_xattr_set,
356 static const struct xattr_handler *sockfs_xattr_handlers[] = {
357 &sockfs_xattr_handler,
358 &sockfs_security_xattr_handler,
362 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363 int flags, const char *dev_name, void *data)
365 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366 sockfs_xattr_handlers,
367 &sockfs_dentry_operations, SOCKFS_MAGIC);
370 static struct vfsmount *sock_mnt __read_mostly;
372 static struct file_system_type sock_fs_type = {
374 .mount = sockfs_mount,
375 .kill_sb = kill_anon_super,
379 * Obtains the first available file descriptor and sets it up for use.
381 * These functions create file structures and maps them to fd space
382 * of the current process. On success it returns file descriptor
383 * and file struct implicitly stored in sock->file.
384 * Note that another thread may close file descriptor before we return
385 * from this function. We use the fact that now we do not refer
386 * to socket after mapping. If one day we will need it, this
387 * function will increment ref. count on file by 1.
389 * In any case returned fd MAY BE not valid!
390 * This race condition is unavoidable
391 * with shared fd spaces, we cannot solve it inside kernel,
392 * but we take care of internal coherence yet.
395 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
397 struct qstr name = { .name = "" };
403 name.len = strlen(name.name);
404 } else if (sock->sk) {
405 name.name = sock->sk->sk_prot_creator->name;
406 name.len = strlen(name.name);
408 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
409 if (unlikely(!path.dentry))
410 return ERR_PTR(-ENOMEM);
411 path.mnt = mntget(sock_mnt);
413 d_instantiate(path.dentry, SOCK_INODE(sock));
415 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
418 /* drop dentry, keep inode */
419 ihold(d_inode(path.dentry));
425 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
426 file->private_data = sock;
429 EXPORT_SYMBOL(sock_alloc_file);
431 static int sock_map_fd(struct socket *sock, int flags)
433 struct file *newfile;
434 int fd = get_unused_fd_flags(flags);
435 if (unlikely(fd < 0))
438 newfile = sock_alloc_file(sock, flags, NULL);
439 if (likely(!IS_ERR(newfile))) {
440 fd_install(fd, newfile);
445 return PTR_ERR(newfile);
448 struct socket *sock_from_file(struct file *file, int *err)
450 if (file->f_op == &socket_file_ops)
451 return file->private_data; /* set in sock_map_fd */
456 EXPORT_SYMBOL(sock_from_file);
459 * sockfd_lookup - Go from a file number to its socket slot
461 * @err: pointer to an error code return
463 * The file handle passed in is locked and the socket it is bound
464 * too is returned. If an error occurs the err pointer is overwritten
465 * with a negative errno code and NULL is returned. The function checks
466 * for both invalid handles and passing a handle which is not a socket.
468 * On a success the socket object pointer is returned.
471 struct socket *sockfd_lookup(int fd, int *err)
482 sock = sock_from_file(file, err);
487 EXPORT_SYMBOL(sockfd_lookup);
489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
491 struct fd f = fdget(fd);
496 sock = sock_from_file(f.file, err);
498 *fput_needed = f.flags;
506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
512 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
522 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
527 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
536 int err = simple_setattr(dentry, iattr);
538 if (!err && (iattr->ia_valid & ATTR_UID)) {
539 struct socket *sock = SOCKET_I(d_inode(dentry));
541 sock->sk->sk_uid = iattr->ia_uid;
547 static const struct inode_operations sockfs_inode_ops = {
548 .listxattr = sockfs_listxattr,
549 .setattr = sockfs_setattr,
553 * sock_alloc - allocate a socket
555 * Allocate a new inode and socket object. The two are bound together
556 * and initialised. The socket is then returned. If we are out of inodes
560 struct socket *sock_alloc(void)
565 inode = new_inode_pseudo(sock_mnt->mnt_sb);
569 sock = SOCKET_I(inode);
571 kmemcheck_annotate_bitfield(sock, type);
572 inode->i_ino = get_next_ino();
573 inode->i_mode = S_IFSOCK | S_IRWXUGO;
574 inode->i_uid = current_fsuid();
575 inode->i_gid = current_fsgid();
576 inode->i_op = &sockfs_inode_ops;
578 this_cpu_add(sockets_in_use, 1);
581 EXPORT_SYMBOL(sock_alloc);
584 * sock_release - close a socket
585 * @sock: socket to close
587 * The socket is released from the protocol stack if it has a release
588 * callback, and the inode is then released if the socket is bound to
589 * an inode not a file.
592 void sock_release(struct socket *sock)
595 struct module *owner = sock->ops->owner;
597 sock->ops->release(sock);
602 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
603 pr_err("%s: fasync list not empty!\n", __func__);
605 this_cpu_sub(sockets_in_use, 1);
607 iput(SOCK_INODE(sock));
612 EXPORT_SYMBOL(sock_release);
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
616 u8 flags = *tx_flags;
618 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619 flags |= SKBTX_HW_TSTAMP;
621 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622 flags |= SKBTX_SW_TSTAMP;
624 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625 flags |= SKBTX_SCHED_TSTAMP;
629 EXPORT_SYMBOL(__sock_tx_timestamp);
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
633 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634 BUG_ON(ret == -EIOCBQUEUED);
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
640 int err = security_socket_sendmsg(sock, msg,
643 return err ?: sock_sendmsg_nosec(sock, msg);
645 EXPORT_SYMBOL(sock_sendmsg);
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 struct kvec *vec, size_t num, size_t size)
650 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651 return sock_sendmsg(sock, msg);
653 EXPORT_SYMBOL(kernel_sendmsg);
655 static bool skb_is_err_queue(const struct sk_buff *skb)
657 /* pkt_type of skbs enqueued on the error queue are set to
658 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
659 * in recvmsg, since skbs received on a local socket will never
660 * have a pkt_type of PACKET_OUTGOING.
662 return skb->pkt_type == PACKET_OUTGOING;
666 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
668 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
671 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
672 struct scm_timestamping tss;
674 struct skb_shared_hwtstamps *shhwtstamps =
677 /* Race occurred between timestamp enabling and packet
678 receiving. Fill in the current time for now. */
679 if (need_software_tstamp && skb->tstamp == 0)
680 __net_timestamp(skb);
682 if (need_software_tstamp) {
683 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
685 skb_get_timestamp(skb, &tv);
686 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
690 skb_get_timestampns(skb, &ts);
691 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
696 memset(&tss, 0, sizeof(tss));
697 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
698 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
701 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
702 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
705 put_cmsg(msg, SOL_SOCKET,
706 SCM_TIMESTAMPING, sizeof(tss), &tss);
708 if (skb_is_err_queue(skb) && skb->len &&
709 SKB_EXT_ERR(skb)->opt_stats)
710 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
711 skb->len, skb->data);
714 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
716 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
721 if (!sock_flag(sk, SOCK_WIFI_STATUS))
723 if (!skb->wifi_acked_valid)
726 ack = skb->wifi_acked;
728 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
730 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
732 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
735 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
736 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
737 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
740 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
743 sock_recv_timestamp(msg, sk, skb);
744 sock_recv_drops(msg, sk, skb);
746 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
748 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
751 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
754 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
756 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
758 return err ?: sock_recvmsg_nosec(sock, msg, flags);
760 EXPORT_SYMBOL(sock_recvmsg);
763 * kernel_recvmsg - Receive a message from a socket (kernel space)
764 * @sock: The socket to receive the message from
765 * @msg: Received message
766 * @vec: Input s/g array for message data
767 * @num: Size of input s/g array
768 * @size: Number of bytes to read
769 * @flags: Message flags (MSG_DONTWAIT, etc...)
771 * On return the msg structure contains the scatter/gather array passed in the
772 * vec argument. The array is modified so that it consists of the unfilled
773 * portion of the original array.
775 * The returned value is the total number of bytes received, or an error.
777 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
778 struct kvec *vec, size_t num, size_t size, int flags)
780 mm_segment_t oldfs = get_fs();
783 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
785 result = sock_recvmsg(sock, msg, flags);
789 EXPORT_SYMBOL(kernel_recvmsg);
791 static ssize_t sock_sendpage(struct file *file, struct page *page,
792 int offset, size_t size, loff_t *ppos, int more)
797 sock = file->private_data;
799 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
800 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
803 return kernel_sendpage(sock, page, offset, size, flags);
806 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
807 struct pipe_inode_info *pipe, size_t len,
810 struct socket *sock = file->private_data;
812 if (unlikely(!sock->ops->splice_read))
815 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
818 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
820 struct file *file = iocb->ki_filp;
821 struct socket *sock = file->private_data;
822 struct msghdr msg = {.msg_iter = *to,
826 if (file->f_flags & O_NONBLOCK)
827 msg.msg_flags = MSG_DONTWAIT;
829 if (iocb->ki_pos != 0)
832 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
835 res = sock_recvmsg(sock, &msg, msg.msg_flags);
840 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
842 struct file *file = iocb->ki_filp;
843 struct socket *sock = file->private_data;
844 struct msghdr msg = {.msg_iter = *from,
848 if (iocb->ki_pos != 0)
851 if (file->f_flags & O_NONBLOCK)
852 msg.msg_flags = MSG_DONTWAIT;
854 if (sock->type == SOCK_SEQPACKET)
855 msg.msg_flags |= MSG_EOR;
857 res = sock_sendmsg(sock, &msg);
858 *from = msg.msg_iter;
863 * Atomic setting of ioctl hooks to avoid race
864 * with module unload.
867 static DEFINE_MUTEX(br_ioctl_mutex);
868 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
870 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
872 mutex_lock(&br_ioctl_mutex);
873 br_ioctl_hook = hook;
874 mutex_unlock(&br_ioctl_mutex);
876 EXPORT_SYMBOL(brioctl_set);
878 static DEFINE_MUTEX(vlan_ioctl_mutex);
879 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
881 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
883 mutex_lock(&vlan_ioctl_mutex);
884 vlan_ioctl_hook = hook;
885 mutex_unlock(&vlan_ioctl_mutex);
887 EXPORT_SYMBOL(vlan_ioctl_set);
889 static DEFINE_MUTEX(dlci_ioctl_mutex);
890 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
892 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
894 mutex_lock(&dlci_ioctl_mutex);
895 dlci_ioctl_hook = hook;
896 mutex_unlock(&dlci_ioctl_mutex);
898 EXPORT_SYMBOL(dlci_ioctl_set);
900 static long sock_do_ioctl(struct net *net, struct socket *sock,
901 unsigned int cmd, unsigned long arg)
904 void __user *argp = (void __user *)arg;
906 err = sock->ops->ioctl(sock, cmd, arg);
909 * If this ioctl is unknown try to hand it down
912 if (err == -ENOIOCTLCMD)
913 err = dev_ioctl(net, cmd, argp);
919 * With an ioctl, arg may well be a user mode pointer, but we don't know
920 * what to do with it - that's up to the protocol still.
923 static struct ns_common *get_net_ns(struct ns_common *ns)
925 return &get_net(container_of(ns, struct net, ns))->ns;
928 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
932 void __user *argp = (void __user *)arg;
936 sock = file->private_data;
939 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
940 err = dev_ioctl(net, cmd, argp);
942 #ifdef CONFIG_WEXT_CORE
943 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
944 err = dev_ioctl(net, cmd, argp);
951 if (get_user(pid, (int __user *)argp))
953 f_setown(sock->file, pid, 1);
958 err = put_user(f_getown(sock->file),
967 request_module("bridge");
969 mutex_lock(&br_ioctl_mutex);
971 err = br_ioctl_hook(net, cmd, argp);
972 mutex_unlock(&br_ioctl_mutex);
977 if (!vlan_ioctl_hook)
978 request_module("8021q");
980 mutex_lock(&vlan_ioctl_mutex);
982 err = vlan_ioctl_hook(net, argp);
983 mutex_unlock(&vlan_ioctl_mutex);
988 if (!dlci_ioctl_hook)
989 request_module("dlci");
991 mutex_lock(&dlci_ioctl_mutex);
993 err = dlci_ioctl_hook(cmd, argp);
994 mutex_unlock(&dlci_ioctl_mutex);
998 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1001 err = open_related_ns(&net->ns, get_net_ns);
1004 err = sock_do_ioctl(net, sock, cmd, arg);
1010 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1013 struct socket *sock = NULL;
1015 err = security_socket_create(family, type, protocol, 1);
1019 sock = sock_alloc();
1026 err = security_socket_post_create(sock, family, type, protocol, 1);
1038 EXPORT_SYMBOL(sock_create_lite);
1040 /* No kernel lock held - perfect */
1041 static unsigned int sock_poll(struct file *file, poll_table *wait)
1043 unsigned int busy_flag = 0;
1044 struct socket *sock;
1047 * We can't return errors to poll, so it's either yes or no.
1049 sock = file->private_data;
1051 if (sk_can_busy_loop(sock->sk)) {
1052 /* this socket can poll_ll so tell the system call */
1053 busy_flag = POLL_BUSY_LOOP;
1055 /* once, only if requested by syscall */
1056 if (wait && (wait->_key & POLL_BUSY_LOOP))
1057 sk_busy_loop(sock->sk, 1);
1060 return busy_flag | sock->ops->poll(file, sock, wait);
1063 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1065 struct socket *sock = file->private_data;
1067 return sock->ops->mmap(file, sock, vma);
1070 static int sock_close(struct inode *inode, struct file *filp)
1072 sock_release(SOCKET_I(inode));
1077 * Update the socket async list
1079 * Fasync_list locking strategy.
1081 * 1. fasync_list is modified only under process context socket lock
1082 * i.e. under semaphore.
1083 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1084 * or under socket lock
1087 static int sock_fasync(int fd, struct file *filp, int on)
1089 struct socket *sock = filp->private_data;
1090 struct sock *sk = sock->sk;
1091 struct socket_wq *wq;
1097 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1098 fasync_helper(fd, filp, on, &wq->fasync_list);
1100 if (!wq->fasync_list)
1101 sock_reset_flag(sk, SOCK_FASYNC);
1103 sock_set_flag(sk, SOCK_FASYNC);
1109 /* This function may be called only under rcu_lock */
1111 int sock_wake_async(struct socket_wq *wq, int how, int band)
1113 if (!wq || !wq->fasync_list)
1117 case SOCK_WAKE_WAITD:
1118 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1121 case SOCK_WAKE_SPACE:
1122 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1127 kill_fasync(&wq->fasync_list, SIGIO, band);
1130 kill_fasync(&wq->fasync_list, SIGURG, band);
1135 EXPORT_SYMBOL(sock_wake_async);
1137 int __sock_create(struct net *net, int family, int type, int protocol,
1138 struct socket **res, int kern)
1141 struct socket *sock;
1142 const struct net_proto_family *pf;
1145 * Check protocol is in range
1147 if (family < 0 || family >= NPROTO)
1148 return -EAFNOSUPPORT;
1149 if (type < 0 || type >= SOCK_MAX)
1154 This uglymoron is moved from INET layer to here to avoid
1155 deadlock in module load.
1157 if (family == PF_INET && type == SOCK_PACKET) {
1158 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1163 err = security_socket_create(family, type, protocol, kern);
1168 * Allocate the socket and allow the family to set things up. if
1169 * the protocol is 0, the family is instructed to select an appropriate
1172 sock = sock_alloc();
1174 net_warn_ratelimited("socket: no more sockets\n");
1175 return -ENFILE; /* Not exactly a match, but its the
1176 closest posix thing */
1181 #ifdef CONFIG_MODULES
1182 /* Attempt to load a protocol module if the find failed.
1184 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1185 * requested real, full-featured networking support upon configuration.
1186 * Otherwise module support will break!
1188 if (rcu_access_pointer(net_families[family]) == NULL)
1189 request_module("net-pf-%d", family);
1193 pf = rcu_dereference(net_families[family]);
1194 err = -EAFNOSUPPORT;
1199 * We will call the ->create function, that possibly is in a loadable
1200 * module, so we have to bump that loadable module refcnt first.
1202 if (!try_module_get(pf->owner))
1205 /* Now protected by module ref count */
1208 err = pf->create(net, sock, protocol, kern);
1210 goto out_module_put;
1213 * Now to bump the refcnt of the [loadable] module that owns this
1214 * socket at sock_release time we decrement its refcnt.
1216 if (!try_module_get(sock->ops->owner))
1217 goto out_module_busy;
1220 * Now that we're done with the ->create function, the [loadable]
1221 * module can have its refcnt decremented
1223 module_put(pf->owner);
1224 err = security_socket_post_create(sock, family, type, protocol, kern);
1226 goto out_sock_release;
1232 err = -EAFNOSUPPORT;
1235 module_put(pf->owner);
1242 goto out_sock_release;
1244 EXPORT_SYMBOL(__sock_create);
1246 int sock_create(int family, int type, int protocol, struct socket **res)
1248 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1250 EXPORT_SYMBOL(sock_create);
1252 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1254 return __sock_create(net, family, type, protocol, res, 1);
1256 EXPORT_SYMBOL(sock_create_kern);
1258 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1261 struct socket *sock;
1264 /* Check the SOCK_* constants for consistency. */
1265 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1266 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1267 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1268 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1270 flags = type & ~SOCK_TYPE_MASK;
1271 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1273 type &= SOCK_TYPE_MASK;
1275 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1276 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1278 retval = sock_create(family, type, protocol, &sock);
1282 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1287 /* It may be already another descriptor 8) Not kernel problem. */
1296 * Create a pair of connected sockets.
1299 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1300 int __user *, usockvec)
1302 struct socket *sock1, *sock2;
1304 struct file *newfile1, *newfile2;
1307 flags = type & ~SOCK_TYPE_MASK;
1308 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1310 type &= SOCK_TYPE_MASK;
1312 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1313 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1316 * Obtain the first socket and check if the underlying protocol
1317 * supports the socketpair call.
1320 err = sock_create(family, type, protocol, &sock1);
1324 err = sock_create(family, type, protocol, &sock2);
1328 err = sock1->ops->socketpair(sock1, sock2);
1330 goto out_release_both;
1332 fd1 = get_unused_fd_flags(flags);
1333 if (unlikely(fd1 < 0)) {
1335 goto out_release_both;
1338 fd2 = get_unused_fd_flags(flags);
1339 if (unlikely(fd2 < 0)) {
1341 goto out_put_unused_1;
1344 newfile1 = sock_alloc_file(sock1, flags, NULL);
1345 if (IS_ERR(newfile1)) {
1346 err = PTR_ERR(newfile1);
1347 goto out_put_unused_both;
1350 newfile2 = sock_alloc_file(sock2, flags, NULL);
1351 if (IS_ERR(newfile2)) {
1352 err = PTR_ERR(newfile2);
1356 err = put_user(fd1, &usockvec[0]);
1360 err = put_user(fd2, &usockvec[1]);
1364 audit_fd_pair(fd1, fd2);
1366 fd_install(fd1, newfile1);
1367 fd_install(fd2, newfile2);
1368 /* fd1 and fd2 may be already another descriptors.
1369 * Not kernel problem.
1385 sock_release(sock2);
1388 out_put_unused_both:
1393 sock_release(sock2);
1395 sock_release(sock1);
1401 * Bind a name to a socket. Nothing much to do here since it's
1402 * the protocol's responsibility to handle the local address.
1404 * We move the socket address to kernel space before we call
1405 * the protocol layer (having also checked the address is ok).
1408 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1410 struct socket *sock;
1411 struct sockaddr_storage address;
1412 int err, fput_needed;
1414 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1416 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1418 err = security_socket_bind(sock,
1419 (struct sockaddr *)&address,
1422 err = sock->ops->bind(sock,
1426 fput_light(sock->file, fput_needed);
1432 * Perform a listen. Basically, we allow the protocol to do anything
1433 * necessary for a listen, and if that works, we mark the socket as
1434 * ready for listening.
1437 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1439 struct socket *sock;
1440 int err, fput_needed;
1443 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1445 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1446 if ((unsigned int)backlog > somaxconn)
1447 backlog = somaxconn;
1449 err = security_socket_listen(sock, backlog);
1451 err = sock->ops->listen(sock, backlog);
1453 fput_light(sock->file, fput_needed);
1459 * For accept, we attempt to create a new socket, set up the link
1460 * with the client, wake up the client, then return the new
1461 * connected fd. We collect the address of the connector in kernel
1462 * space and move it to user at the very end. This is unclean because
1463 * we open the socket then return an error.
1465 * 1003.1g adds the ability to recvmsg() to query connection pending
1466 * status to recvmsg. We need to add that support in a way thats
1467 * clean when we restucture accept also.
1470 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1471 int __user *, upeer_addrlen, int, flags)
1473 struct socket *sock, *newsock;
1474 struct file *newfile;
1475 int err, len, newfd, fput_needed;
1476 struct sockaddr_storage address;
1478 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1481 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1482 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1489 newsock = sock_alloc();
1493 newsock->type = sock->type;
1494 newsock->ops = sock->ops;
1497 * We don't need try_module_get here, as the listening socket (sock)
1498 * has the protocol module (sock->ops->owner) held.
1500 __module_get(newsock->ops->owner);
1502 newfd = get_unused_fd_flags(flags);
1503 if (unlikely(newfd < 0)) {
1505 sock_release(newsock);
1508 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1509 if (IS_ERR(newfile)) {
1510 err = PTR_ERR(newfile);
1511 put_unused_fd(newfd);
1512 sock_release(newsock);
1516 err = security_socket_accept(sock, newsock);
1520 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1524 if (upeer_sockaddr) {
1525 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1527 err = -ECONNABORTED;
1530 err = move_addr_to_user(&address,
1531 len, upeer_sockaddr, upeer_addrlen);
1536 /* File flags are not inherited via accept() unlike another OSes. */
1538 fd_install(newfd, newfile);
1542 fput_light(sock->file, fput_needed);
1547 put_unused_fd(newfd);
1551 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1552 int __user *, upeer_addrlen)
1554 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1558 * Attempt to connect to a socket with the server address. The address
1559 * is in user space so we verify it is OK and move it to kernel space.
1561 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1564 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1565 * other SEQPACKET protocols that take time to connect() as it doesn't
1566 * include the -EINPROGRESS status for such sockets.
1569 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1572 struct socket *sock;
1573 struct sockaddr_storage address;
1574 int err, fput_needed;
1576 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1579 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1584 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1588 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1589 sock->file->f_flags);
1591 fput_light(sock->file, fput_needed);
1597 * Get the local address ('name') of a socket object. Move the obtained
1598 * name to user space.
1601 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1602 int __user *, usockaddr_len)
1604 struct socket *sock;
1605 struct sockaddr_storage address;
1606 int len, err, fput_needed;
1608 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1612 err = security_socket_getsockname(sock);
1616 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1619 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1622 fput_light(sock->file, fput_needed);
1628 * Get the remote address ('name') of a socket object. Move the obtained
1629 * name to user space.
1632 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1633 int __user *, usockaddr_len)
1635 struct socket *sock;
1636 struct sockaddr_storage address;
1637 int len, err, fput_needed;
1639 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641 err = security_socket_getpeername(sock);
1643 fput_light(sock->file, fput_needed);
1648 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1651 err = move_addr_to_user(&address, len, usockaddr,
1653 fput_light(sock->file, fput_needed);
1659 * Send a datagram to a given address. We move the address into kernel
1660 * space and check the user space data area is readable before invoking
1664 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1665 unsigned int, flags, struct sockaddr __user *, addr,
1668 struct socket *sock;
1669 struct sockaddr_storage address;
1675 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1678 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1682 msg.msg_name = NULL;
1683 msg.msg_control = NULL;
1684 msg.msg_controllen = 0;
1685 msg.msg_namelen = 0;
1687 err = move_addr_to_kernel(addr, addr_len, &address);
1690 msg.msg_name = (struct sockaddr *)&address;
1691 msg.msg_namelen = addr_len;
1693 if (sock->file->f_flags & O_NONBLOCK)
1694 flags |= MSG_DONTWAIT;
1695 msg.msg_flags = flags;
1696 err = sock_sendmsg(sock, &msg);
1699 fput_light(sock->file, fput_needed);
1705 * Send a datagram down a socket.
1708 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1709 unsigned int, flags)
1711 return sys_sendto(fd, buff, len, flags, NULL, 0);
1715 * Receive a frame from the socket and optionally record the address of the
1716 * sender. We verify the buffers are writable and if needed move the
1717 * sender address from kernel to user space.
1720 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1721 unsigned int, flags, struct sockaddr __user *, addr,
1722 int __user *, addr_len)
1724 struct socket *sock;
1727 struct sockaddr_storage address;
1731 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1734 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1738 msg.msg_control = NULL;
1739 msg.msg_controllen = 0;
1740 /* Save some cycles and don't copy the address if not needed */
1741 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1742 /* We assume all kernel code knows the size of sockaddr_storage */
1743 msg.msg_namelen = 0;
1744 msg.msg_iocb = NULL;
1746 if (sock->file->f_flags & O_NONBLOCK)
1747 flags |= MSG_DONTWAIT;
1748 err = sock_recvmsg(sock, &msg, flags);
1750 if (err >= 0 && addr != NULL) {
1751 err2 = move_addr_to_user(&address,
1752 msg.msg_namelen, addr, addr_len);
1757 fput_light(sock->file, fput_needed);
1763 * Receive a datagram from a socket.
1766 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1767 unsigned int, flags)
1769 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1773 * Set a socket option. Because we don't know the option lengths we have
1774 * to pass the user mode parameter for the protocols to sort out.
1777 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1778 char __user *, optval, int, optlen)
1780 int err, fput_needed;
1781 struct socket *sock;
1786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1788 err = security_socket_setsockopt(sock, level, optname);
1792 if (level == SOL_SOCKET)
1794 sock_setsockopt(sock, level, optname, optval,
1798 sock->ops->setsockopt(sock, level, optname, optval,
1801 fput_light(sock->file, fput_needed);
1807 * Get a socket option. Because we don't know the option lengths we have
1808 * to pass a user mode parameter for the protocols to sort out.
1811 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1812 char __user *, optval, int __user *, optlen)
1814 int err, fput_needed;
1815 struct socket *sock;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1819 err = security_socket_getsockopt(sock, level, optname);
1823 if (level == SOL_SOCKET)
1825 sock_getsockopt(sock, level, optname, optval,
1829 sock->ops->getsockopt(sock, level, optname, optval,
1832 fput_light(sock->file, fput_needed);
1838 * Shutdown a socket.
1841 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1843 int err, fput_needed;
1844 struct socket *sock;
1846 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1848 err = security_socket_shutdown(sock, how);
1850 err = sock->ops->shutdown(sock, how);
1851 fput_light(sock->file, fput_needed);
1856 /* A couple of helpful macros for getting the address of the 32/64 bit
1857 * fields which are the same type (int / unsigned) on our platforms.
1859 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1860 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1861 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1863 struct used_address {
1864 struct sockaddr_storage name;
1865 unsigned int name_len;
1868 static int copy_msghdr_from_user(struct msghdr *kmsg,
1869 struct user_msghdr __user *umsg,
1870 struct sockaddr __user **save_addr,
1873 struct sockaddr __user *uaddr;
1874 struct iovec __user *uiov;
1878 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1879 __get_user(uaddr, &umsg->msg_name) ||
1880 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1881 __get_user(uiov, &umsg->msg_iov) ||
1882 __get_user(nr_segs, &umsg->msg_iovlen) ||
1883 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1884 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1885 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1889 kmsg->msg_namelen = 0;
1891 if (kmsg->msg_namelen < 0)
1894 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1895 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1900 if (uaddr && kmsg->msg_namelen) {
1902 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1908 kmsg->msg_name = NULL;
1909 kmsg->msg_namelen = 0;
1912 if (nr_segs > UIO_MAXIOV)
1915 kmsg->msg_iocb = NULL;
1917 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1918 UIO_FASTIOV, iov, &kmsg->msg_iter);
1921 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1922 struct msghdr *msg_sys, unsigned int flags,
1923 struct used_address *used_address,
1924 unsigned int allowed_msghdr_flags)
1926 struct compat_msghdr __user *msg_compat =
1927 (struct compat_msghdr __user *)msg;
1928 struct sockaddr_storage address;
1929 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1930 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1931 __aligned(sizeof(__kernel_size_t));
1932 /* 20 is size of ipv6_pktinfo */
1933 unsigned char *ctl_buf = ctl;
1937 msg_sys->msg_name = &address;
1939 if (MSG_CMSG_COMPAT & flags)
1940 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1942 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1948 if (msg_sys->msg_controllen > INT_MAX)
1950 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1951 ctl_len = msg_sys->msg_controllen;
1952 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1954 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1958 ctl_buf = msg_sys->msg_control;
1959 ctl_len = msg_sys->msg_controllen;
1960 } else if (ctl_len) {
1961 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
1962 CMSG_ALIGN(sizeof(struct cmsghdr)));
1963 if (ctl_len > sizeof(ctl)) {
1964 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1965 if (ctl_buf == NULL)
1970 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1971 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1972 * checking falls down on this.
1974 if (copy_from_user(ctl_buf,
1975 (void __user __force *)msg_sys->msg_control,
1978 msg_sys->msg_control = ctl_buf;
1980 msg_sys->msg_flags = flags;
1982 if (sock->file->f_flags & O_NONBLOCK)
1983 msg_sys->msg_flags |= MSG_DONTWAIT;
1985 * If this is sendmmsg() and current destination address is same as
1986 * previously succeeded address, omit asking LSM's decision.
1987 * used_address->name_len is initialized to UINT_MAX so that the first
1988 * destination address never matches.
1990 if (used_address && msg_sys->msg_name &&
1991 used_address->name_len == msg_sys->msg_namelen &&
1992 !memcmp(&used_address->name, msg_sys->msg_name,
1993 used_address->name_len)) {
1994 err = sock_sendmsg_nosec(sock, msg_sys);
1997 err = sock_sendmsg(sock, msg_sys);
1999 * If this is sendmmsg() and sending to current destination address was
2000 * successful, remember it.
2002 if (used_address && err >= 0) {
2003 used_address->name_len = msg_sys->msg_namelen;
2004 if (msg_sys->msg_name)
2005 memcpy(&used_address->name, msg_sys->msg_name,
2006 used_address->name_len);
2011 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2018 * BSD sendmsg interface
2021 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2023 int fput_needed, err;
2024 struct msghdr msg_sys;
2025 struct socket *sock;
2027 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2031 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2033 fput_light(sock->file, fput_needed);
2038 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2040 if (flags & MSG_CMSG_COMPAT)
2042 return __sys_sendmsg(fd, msg, flags);
2046 * Linux sendmmsg interface
2049 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2052 int fput_needed, err, datagrams;
2053 struct socket *sock;
2054 struct mmsghdr __user *entry;
2055 struct compat_mmsghdr __user *compat_entry;
2056 struct msghdr msg_sys;
2057 struct used_address used_address;
2058 unsigned int oflags = flags;
2060 if (vlen > UIO_MAXIOV)
2065 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2069 used_address.name_len = UINT_MAX;
2071 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2075 while (datagrams < vlen) {
2076 if (datagrams == vlen - 1)
2079 if (MSG_CMSG_COMPAT & flags) {
2080 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2081 &msg_sys, flags, &used_address, MSG_EOR);
2084 err = __put_user(err, &compat_entry->msg_len);
2087 err = ___sys_sendmsg(sock,
2088 (struct user_msghdr __user *)entry,
2089 &msg_sys, flags, &used_address, MSG_EOR);
2092 err = put_user(err, &entry->msg_len);
2099 if (msg_data_left(&msg_sys))
2104 fput_light(sock->file, fput_needed);
2106 /* We only return an error if no datagrams were able to be sent */
2113 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2114 unsigned int, vlen, unsigned int, flags)
2116 if (flags & MSG_CMSG_COMPAT)
2118 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2121 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2122 struct msghdr *msg_sys, unsigned int flags, int nosec)
2124 struct compat_msghdr __user *msg_compat =
2125 (struct compat_msghdr __user *)msg;
2126 struct iovec iovstack[UIO_FASTIOV];
2127 struct iovec *iov = iovstack;
2128 unsigned long cmsg_ptr;
2132 /* kernel mode address */
2133 struct sockaddr_storage addr;
2135 /* user mode address pointers */
2136 struct sockaddr __user *uaddr;
2137 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2139 msg_sys->msg_name = &addr;
2141 if (MSG_CMSG_COMPAT & flags)
2142 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2144 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2148 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2149 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2151 /* We assume all kernel code knows the size of sockaddr_storage */
2152 msg_sys->msg_namelen = 0;
2154 if (sock->file->f_flags & O_NONBLOCK)
2155 flags |= MSG_DONTWAIT;
2156 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2161 if (uaddr != NULL) {
2162 err = move_addr_to_user(&addr,
2163 msg_sys->msg_namelen, uaddr,
2168 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2172 if (MSG_CMSG_COMPAT & flags)
2173 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2174 &msg_compat->msg_controllen);
2176 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2177 &msg->msg_controllen);
2188 * BSD recvmsg interface
2191 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2193 int fput_needed, err;
2194 struct msghdr msg_sys;
2195 struct socket *sock;
2197 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2201 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2203 fput_light(sock->file, fput_needed);
2208 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2209 unsigned int, flags)
2211 if (flags & MSG_CMSG_COMPAT)
2213 return __sys_recvmsg(fd, msg, flags);
2217 * Linux recvmmsg interface
2220 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2221 unsigned int flags, struct timespec *timeout)
2223 int fput_needed, err, datagrams;
2224 struct socket *sock;
2225 struct mmsghdr __user *entry;
2226 struct compat_mmsghdr __user *compat_entry;
2227 struct msghdr msg_sys;
2228 struct timespec64 end_time;
2229 struct timespec64 timeout64;
2232 poll_select_set_timeout(&end_time, timeout->tv_sec,
2238 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2242 err = sock_error(sock->sk);
2249 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2251 while (datagrams < vlen) {
2253 * No need to ask LSM for more than the first datagram.
2255 if (MSG_CMSG_COMPAT & flags) {
2256 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2257 &msg_sys, flags & ~MSG_WAITFORONE,
2261 err = __put_user(err, &compat_entry->msg_len);
2264 err = ___sys_recvmsg(sock,
2265 (struct user_msghdr __user *)entry,
2266 &msg_sys, flags & ~MSG_WAITFORONE,
2270 err = put_user(err, &entry->msg_len);
2278 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2279 if (flags & MSG_WAITFORONE)
2280 flags |= MSG_DONTWAIT;
2283 ktime_get_ts64(&timeout64);
2284 *timeout = timespec64_to_timespec(
2285 timespec64_sub(end_time, timeout64));
2286 if (timeout->tv_sec < 0) {
2287 timeout->tv_sec = timeout->tv_nsec = 0;
2291 /* Timeout, return less than vlen datagrams */
2292 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2296 /* Out of band data, return right away */
2297 if (msg_sys.msg_flags & MSG_OOB)
2305 if (datagrams == 0) {
2311 * We may return less entries than requested (vlen) if the
2312 * sock is non block and there aren't enough datagrams...
2314 if (err != -EAGAIN) {
2316 * ... or if recvmsg returns an error after we
2317 * received some datagrams, where we record the
2318 * error to return on the next call or if the
2319 * app asks about it using getsockopt(SO_ERROR).
2321 sock->sk->sk_err = -err;
2324 fput_light(sock->file, fput_needed);
2329 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2330 unsigned int, vlen, unsigned int, flags,
2331 struct timespec __user *, timeout)
2334 struct timespec timeout_sys;
2336 if (flags & MSG_CMSG_COMPAT)
2340 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2342 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2345 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2347 if (datagrams > 0 &&
2348 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2349 datagrams = -EFAULT;
2354 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2355 /* Argument list sizes for sys_socketcall */
2356 #define AL(x) ((x) * sizeof(unsigned long))
2357 static const unsigned char nargs[21] = {
2358 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2359 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2360 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2367 * System call vectors.
2369 * Argument checking cleaned up. Saved 20% in size.
2370 * This function doesn't need to set the kernel lock because
2371 * it is set by the callees.
2374 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2376 unsigned long a[AUDITSC_ARGS];
2377 unsigned long a0, a1;
2381 if (call < 1 || call > SYS_SENDMMSG)
2385 if (len > sizeof(a))
2388 /* copy_from_user should be SMP safe. */
2389 if (copy_from_user(a, args, len))
2392 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2401 err = sys_socket(a0, a1, a[2]);
2404 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2407 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2410 err = sys_listen(a0, a1);
2413 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2414 (int __user *)a[2], 0);
2416 case SYS_GETSOCKNAME:
2418 sys_getsockname(a0, (struct sockaddr __user *)a1,
2419 (int __user *)a[2]);
2421 case SYS_GETPEERNAME:
2423 sys_getpeername(a0, (struct sockaddr __user *)a1,
2424 (int __user *)a[2]);
2426 case SYS_SOCKETPAIR:
2427 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2430 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2433 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2434 (struct sockaddr __user *)a[4], a[5]);
2437 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2440 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2441 (struct sockaddr __user *)a[4],
2442 (int __user *)a[5]);
2445 err = sys_shutdown(a0, a1);
2447 case SYS_SETSOCKOPT:
2448 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2450 case SYS_GETSOCKOPT:
2452 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2453 (int __user *)a[4]);
2456 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2459 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2462 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2465 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2466 (struct timespec __user *)a[4]);
2469 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2470 (int __user *)a[2], a[3]);
2479 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2482 * sock_register - add a socket protocol handler
2483 * @ops: description of protocol
2485 * This function is called by a protocol handler that wants to
2486 * advertise its address family, and have it linked into the
2487 * socket interface. The value ops->family corresponds to the
2488 * socket system call protocol family.
2490 int sock_register(const struct net_proto_family *ops)
2494 if (ops->family >= NPROTO) {
2495 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2499 spin_lock(&net_family_lock);
2500 if (rcu_dereference_protected(net_families[ops->family],
2501 lockdep_is_held(&net_family_lock)))
2504 rcu_assign_pointer(net_families[ops->family], ops);
2507 spin_unlock(&net_family_lock);
2509 pr_info("NET: Registered protocol family %d\n", ops->family);
2512 EXPORT_SYMBOL(sock_register);
2515 * sock_unregister - remove a protocol handler
2516 * @family: protocol family to remove
2518 * This function is called by a protocol handler that wants to
2519 * remove its address family, and have it unlinked from the
2520 * new socket creation.
2522 * If protocol handler is a module, then it can use module reference
2523 * counts to protect against new references. If protocol handler is not
2524 * a module then it needs to provide its own protection in
2525 * the ops->create routine.
2527 void sock_unregister(int family)
2529 BUG_ON(family < 0 || family >= NPROTO);
2531 spin_lock(&net_family_lock);
2532 RCU_INIT_POINTER(net_families[family], NULL);
2533 spin_unlock(&net_family_lock);
2537 pr_info("NET: Unregistered protocol family %d\n", family);
2539 EXPORT_SYMBOL(sock_unregister);
2541 static int __init sock_init(void)
2545 * Initialize the network sysctl infrastructure.
2547 err = net_sysctl_init();
2552 * Initialize skbuff SLAB cache
2557 * Initialize the protocols module.
2562 err = register_filesystem(&sock_fs_type);
2565 sock_mnt = kern_mount(&sock_fs_type);
2566 if (IS_ERR(sock_mnt)) {
2567 err = PTR_ERR(sock_mnt);
2571 /* The real protocol initialization is performed in later initcalls.
2574 #ifdef CONFIG_NETFILTER
2575 err = netfilter_init();
2580 ptp_classifier_init();
2586 unregister_filesystem(&sock_fs_type);
2591 core_initcall(sock_init); /* early initcall */
2593 #ifdef CONFIG_PROC_FS
2594 void socket_seq_show(struct seq_file *seq)
2599 for_each_possible_cpu(cpu)
2600 counter += per_cpu(sockets_in_use, cpu);
2602 /* It can be negative, by the way. 8) */
2606 seq_printf(seq, "sockets: used %d\n", counter);
2608 #endif /* CONFIG_PROC_FS */
2610 #ifdef CONFIG_COMPAT
2611 static int do_siocgstamp(struct net *net, struct socket *sock,
2612 unsigned int cmd, void __user *up)
2614 mm_segment_t old_fs = get_fs();
2619 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2622 err = compat_put_timeval(&ktv, up);
2627 static int do_siocgstampns(struct net *net, struct socket *sock,
2628 unsigned int cmd, void __user *up)
2630 mm_segment_t old_fs = get_fs();
2631 struct timespec kts;
2635 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2638 err = compat_put_timespec(&kts, up);
2643 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2645 struct ifreq __user *uifr;
2648 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2649 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2652 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2656 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2662 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2664 struct compat_ifconf ifc32;
2666 struct ifconf __user *uifc;
2667 struct compat_ifreq __user *ifr32;
2668 struct ifreq __user *ifr;
2672 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2675 memset(&ifc, 0, sizeof(ifc));
2676 if (ifc32.ifcbuf == 0) {
2680 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2682 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2683 sizeof(struct ifreq);
2684 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2686 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2687 ifr32 = compat_ptr(ifc32.ifcbuf);
2688 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2689 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2695 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2698 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2702 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2706 ifr32 = compat_ptr(ifc32.ifcbuf);
2708 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2709 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2710 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2716 if (ifc32.ifcbuf == 0) {
2717 /* Translate from 64-bit structure multiple to
2721 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2726 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2732 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2734 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2735 bool convert_in = false, convert_out = false;
2736 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2737 struct ethtool_rxnfc __user *rxnfc;
2738 struct ifreq __user *ifr;
2739 u32 rule_cnt = 0, actual_rule_cnt;
2744 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2747 compat_rxnfc = compat_ptr(data);
2749 if (get_user(ethcmd, &compat_rxnfc->cmd))
2752 /* Most ethtool structures are defined without padding.
2753 * Unfortunately struct ethtool_rxnfc is an exception.
2758 case ETHTOOL_GRXCLSRLALL:
2759 /* Buffer size is variable */
2760 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2762 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2764 buf_size += rule_cnt * sizeof(u32);
2766 case ETHTOOL_GRXRINGS:
2767 case ETHTOOL_GRXCLSRLCNT:
2768 case ETHTOOL_GRXCLSRULE:
2769 case ETHTOOL_SRXCLSRLINS:
2772 case ETHTOOL_SRXCLSRLDEL:
2773 buf_size += sizeof(struct ethtool_rxnfc);
2778 ifr = compat_alloc_user_space(buf_size);
2779 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2781 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2784 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2785 &ifr->ifr_ifru.ifru_data))
2789 /* We expect there to be holes between fs.m_ext and
2790 * fs.ring_cookie and at the end of fs, but nowhere else.
2792 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2793 sizeof(compat_rxnfc->fs.m_ext) !=
2794 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2795 sizeof(rxnfc->fs.m_ext));
2797 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2798 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2799 offsetof(struct ethtool_rxnfc, fs.location) -
2800 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2802 if (copy_in_user(rxnfc, compat_rxnfc,
2803 (void __user *)(&rxnfc->fs.m_ext + 1) -
2804 (void __user *)rxnfc) ||
2805 copy_in_user(&rxnfc->fs.ring_cookie,
2806 &compat_rxnfc->fs.ring_cookie,
2807 (void __user *)(&rxnfc->fs.location + 1) -
2808 (void __user *)&rxnfc->fs.ring_cookie) ||
2809 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2810 sizeof(rxnfc->rule_cnt)))
2814 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2819 if (copy_in_user(compat_rxnfc, rxnfc,
2820 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2821 (const void __user *)rxnfc) ||
2822 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2823 &rxnfc->fs.ring_cookie,
2824 (const void __user *)(&rxnfc->fs.location + 1) -
2825 (const void __user *)&rxnfc->fs.ring_cookie) ||
2826 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2827 sizeof(rxnfc->rule_cnt)))
2830 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2831 /* As an optimisation, we only copy the actual
2832 * number of rules that the underlying
2833 * function returned. Since Mallory might
2834 * change the rule count in user memory, we
2835 * check that it is less than the rule count
2836 * originally given (as the user buffer size),
2837 * which has been range-checked.
2839 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2841 if (actual_rule_cnt < rule_cnt)
2842 rule_cnt = actual_rule_cnt;
2843 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2844 &rxnfc->rule_locs[0],
2845 rule_cnt * sizeof(u32)))
2853 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2856 compat_uptr_t uptr32;
2857 struct ifreq __user *uifr;
2859 uifr = compat_alloc_user_space(sizeof(*uifr));
2860 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2863 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2866 uptr = compat_ptr(uptr32);
2868 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2871 return dev_ioctl(net, SIOCWANDEV, uifr);
2874 static int bond_ioctl(struct net *net, unsigned int cmd,
2875 struct compat_ifreq __user *ifr32)
2878 mm_segment_t old_fs;
2882 case SIOCBONDENSLAVE:
2883 case SIOCBONDRELEASE:
2884 case SIOCBONDSETHWADDR:
2885 case SIOCBONDCHANGEACTIVE:
2886 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2891 err = dev_ioctl(net, cmd,
2892 (struct ifreq __user __force *) &kifr);
2897 return -ENOIOCTLCMD;
2901 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2902 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2903 struct compat_ifreq __user *u_ifreq32)
2905 struct ifreq __user *u_ifreq64;
2906 char tmp_buf[IFNAMSIZ];
2907 void __user *data64;
2910 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2913 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2915 data64 = compat_ptr(data32);
2917 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2919 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2922 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2925 return dev_ioctl(net, cmd, u_ifreq64);
2928 static int dev_ifsioc(struct net *net, struct socket *sock,
2929 unsigned int cmd, struct compat_ifreq __user *uifr32)
2931 struct ifreq __user *uifr;
2934 uifr = compat_alloc_user_space(sizeof(*uifr));
2935 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2938 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2949 case SIOCGIFBRDADDR:
2950 case SIOCGIFDSTADDR:
2951 case SIOCGIFNETMASK:
2956 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2964 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2965 struct compat_ifreq __user *uifr32)
2968 struct compat_ifmap __user *uifmap32;
2969 mm_segment_t old_fs;
2972 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2973 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2974 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2975 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2976 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2977 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2978 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2979 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2985 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2988 if (cmd == SIOCGIFMAP && !err) {
2989 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2990 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2991 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2992 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2993 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2994 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2995 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3004 struct sockaddr rt_dst; /* target address */
3005 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3006 struct sockaddr rt_genmask; /* target network mask (IP) */
3007 unsigned short rt_flags;
3010 unsigned char rt_tos;
3011 unsigned char rt_class;
3013 short rt_metric; /* +1 for binary compatibility! */
3014 /* char * */ u32 rt_dev; /* forcing the device at add */
3015 u32 rt_mtu; /* per route MTU/Window */
3016 u32 rt_window; /* Window clamping */
3017 unsigned short rt_irtt; /* Initial RTT */
3020 struct in6_rtmsg32 {
3021 struct in6_addr rtmsg_dst;
3022 struct in6_addr rtmsg_src;
3023 struct in6_addr rtmsg_gateway;
3033 static int routing_ioctl(struct net *net, struct socket *sock,
3034 unsigned int cmd, void __user *argp)
3038 struct in6_rtmsg r6;
3042 mm_segment_t old_fs = get_fs();
3044 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3045 struct in6_rtmsg32 __user *ur6 = argp;
3046 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3047 3 * sizeof(struct in6_addr));
3048 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3049 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3050 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3051 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3052 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3053 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3054 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3058 struct rtentry32 __user *ur4 = argp;
3059 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3060 3 * sizeof(struct sockaddr));
3061 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3062 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3063 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3064 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3065 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3066 ret |= get_user(rtdev, &(ur4->rt_dev));
3068 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3069 r4.rt_dev = (char __user __force *)devname;
3083 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3090 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3091 * for some operations; this forces use of the newer bridge-utils that
3092 * use compatible ioctls
3094 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3098 if (get_user(tmp, argp))
3100 if (tmp == BRCTL_GET_VERSION)
3101 return BRCTL_VERSION + 1;
3105 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3106 unsigned int cmd, unsigned long arg)
3108 void __user *argp = compat_ptr(arg);
3109 struct sock *sk = sock->sk;
3110 struct net *net = sock_net(sk);
3112 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3113 return compat_ifr_data_ioctl(net, cmd, argp);
3118 return old_bridge_ioctl(argp);
3120 return dev_ifname32(net, argp);
3122 return dev_ifconf(net, argp);
3124 return ethtool_ioctl(net, argp);
3126 return compat_siocwandev(net, argp);
3129 return compat_sioc_ifmap(net, cmd, argp);
3130 case SIOCBONDENSLAVE:
3131 case SIOCBONDRELEASE:
3132 case SIOCBONDSETHWADDR:
3133 case SIOCBONDCHANGEACTIVE:
3134 return bond_ioctl(net, cmd, argp);
3137 return routing_ioctl(net, sock, cmd, argp);
3139 return do_siocgstamp(net, sock, cmd, argp);
3141 return do_siocgstampns(net, sock, cmd, argp);
3142 case SIOCBONDSLAVEINFOQUERY:
3143 case SIOCBONDINFOQUERY:
3146 return compat_ifr_data_ioctl(net, cmd, argp);
3159 return sock_ioctl(file, cmd, arg);
3176 case SIOCSIFHWBROADCAST:
3178 case SIOCGIFBRDADDR:
3179 case SIOCSIFBRDADDR:
3180 case SIOCGIFDSTADDR:
3181 case SIOCSIFDSTADDR:
3182 case SIOCGIFNETMASK:
3183 case SIOCSIFNETMASK:
3194 return dev_ifsioc(net, sock, cmd, argp);
3200 return sock_do_ioctl(net, sock, cmd, arg);
3203 return -ENOIOCTLCMD;
3206 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3209 struct socket *sock = file->private_data;
3210 int ret = -ENOIOCTLCMD;
3217 if (sock->ops->compat_ioctl)
3218 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3220 if (ret == -ENOIOCTLCMD &&
3221 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3222 ret = compat_wext_handle_ioctl(net, cmd, arg);
3224 if (ret == -ENOIOCTLCMD)
3225 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3231 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3233 return sock->ops->bind(sock, addr, addrlen);
3235 EXPORT_SYMBOL(kernel_bind);
3237 int kernel_listen(struct socket *sock, int backlog)
3239 return sock->ops->listen(sock, backlog);
3241 EXPORT_SYMBOL(kernel_listen);
3243 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3245 struct sock *sk = sock->sk;
3248 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3253 err = sock->ops->accept(sock, *newsock, flags, true);
3255 sock_release(*newsock);
3260 (*newsock)->ops = sock->ops;
3261 __module_get((*newsock)->ops->owner);
3266 EXPORT_SYMBOL(kernel_accept);
3268 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3271 return sock->ops->connect(sock, addr, addrlen, flags);
3273 EXPORT_SYMBOL(kernel_connect);
3275 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3278 return sock->ops->getname(sock, addr, addrlen, 0);
3280 EXPORT_SYMBOL(kernel_getsockname);
3282 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3285 return sock->ops->getname(sock, addr, addrlen, 1);
3287 EXPORT_SYMBOL(kernel_getpeername);
3289 int kernel_getsockopt(struct socket *sock, int level, int optname,
3290 char *optval, int *optlen)
3292 mm_segment_t oldfs = get_fs();
3293 char __user *uoptval;
3294 int __user *uoptlen;
3297 uoptval = (char __user __force *) optval;
3298 uoptlen = (int __user __force *) optlen;
3301 if (level == SOL_SOCKET)
3302 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3304 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3309 EXPORT_SYMBOL(kernel_getsockopt);
3311 int kernel_setsockopt(struct socket *sock, int level, int optname,
3312 char *optval, unsigned int optlen)
3314 mm_segment_t oldfs = get_fs();
3315 char __user *uoptval;
3318 uoptval = (char __user __force *) optval;
3321 if (level == SOL_SOCKET)
3322 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3324 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3329 EXPORT_SYMBOL(kernel_setsockopt);
3331 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3332 size_t size, int flags)
3334 if (sock->ops->sendpage)
3335 return sock->ops->sendpage(sock, page, offset, size, flags);
3337 return sock_no_sendpage(sock, page, offset, size, flags);
3339 EXPORT_SYMBOL(kernel_sendpage);
3341 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3343 mm_segment_t oldfs = get_fs();
3347 err = sock->ops->ioctl(sock, cmd, arg);
3352 EXPORT_SYMBOL(kernel_sock_ioctl);
3354 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3356 return sock->ops->shutdown(sock, how);
3358 EXPORT_SYMBOL(kernel_sock_shutdown);
3360 /* This routine returns the IP overhead imposed by a socket i.e.
3361 * the length of the underlying IP header, depending on whether
3362 * this is an IPv4 or IPv6 socket and the length from IP options turned
3363 * on at the socket. Assumes that the caller has a lock on the socket.
3365 u32 kernel_sock_ip_overhead(struct sock *sk)
3367 struct inet_sock *inet;
3368 struct ip_options_rcu *opt;
3371 #if IS_ENABLED(CONFIG_IPV6)
3372 struct ipv6_pinfo *np;
3373 struct ipv6_txoptions *optv6 = NULL;
3374 #endif /* IS_ENABLED(CONFIG_IPV6) */
3379 owned_by_user = sock_owned_by_user(sk);
3380 switch (sk->sk_family) {
3383 overhead += sizeof(struct iphdr);
3384 opt = rcu_dereference_protected(inet->inet_opt,
3387 overhead += opt->opt.optlen;
3389 #if IS_ENABLED(CONFIG_IPV6)
3392 overhead += sizeof(struct ipv6hdr);
3394 optv6 = rcu_dereference_protected(np->opt,
3397 overhead += (optv6->opt_flen + optv6->opt_nflen);
3399 #endif /* IS_ENABLED(CONFIG_IPV6) */
3400 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3404 EXPORT_SYMBOL(kernel_sock_ip_overhead);