2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92 #include <linux/nospec.h>
94 #include <linux/uaccess.h>
95 #include <asm/unistd.h>
97 #include <net/compat.h>
99 #include <net/cls_cgroup.h>
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <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 __poll_t 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;
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
172 * move_addr_to_kernel - copy a socket address into kernel space
173 * @uaddr: Address in user space
174 * @kaddr: Address in kernel space
175 * @ulen: Length in user space
177 * The address is copied into kernel space. If the provided address is
178 * too long an error code of -EINVAL is returned. If the copy gives
179 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
184 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
188 if (copy_from_user(kaddr, uaddr, ulen))
190 return audit_sockaddr(ulen, kaddr);
194 * move_addr_to_user - copy an address to user space
195 * @kaddr: kernel space address
196 * @klen: length of address in kernel
197 * @uaddr: user space address
198 * @ulen: pointer to user length field
200 * The value pointed to by ulen on entry is the buffer length available.
201 * This is overwritten with the buffer space used. -EINVAL is returned
202 * if an overlong buffer is specified or a negative buffer size. -EFAULT
203 * is returned if either the buffer or the length field are not
205 * After copying the data up to the limit the user specifies, the true
206 * length of the data is written over the length limit the user
207 * specified. Zero is returned for a success.
210 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
211 void __user *uaddr, int __user *ulen)
216 BUG_ON(klen > sizeof(struct sockaddr_storage));
217 err = get_user(len, ulen);
225 if (audit_sockaddr(klen, kaddr))
227 if (copy_to_user(uaddr, kaddr, len))
231 * "fromlen shall refer to the value before truncation.."
234 return __put_user(klen, ulen);
237 static struct kmem_cache *sock_inode_cachep __ro_after_init;
239 static struct inode *sock_alloc_inode(struct super_block *sb)
241 struct socket_alloc *ei;
242 struct socket_wq *wq;
244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
247 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249 kmem_cache_free(sock_inode_cachep, ei);
252 init_waitqueue_head(&wq->wait);
253 wq->fasync_list = NULL;
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 struct socket_alloc *ei;
270 ei = container_of(inode, struct socket_alloc, vfs_inode);
271 kfree_rcu(ei->socket.wq, rcu);
272 kmem_cache_free(sock_inode_cachep, ei);
275 static void init_once(void *foo)
277 struct socket_alloc *ei = (struct socket_alloc *)foo;
279 inode_init_once(&ei->vfs_inode);
282 static void init_inodecache(void)
284 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
285 sizeof(struct socket_alloc),
287 (SLAB_HWCACHE_ALIGN |
288 SLAB_RECLAIM_ACCOUNT |
289 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
291 BUG_ON(sock_inode_cachep == NULL);
294 static const struct super_operations sockfs_ops = {
295 .alloc_inode = sock_alloc_inode,
296 .destroy_inode = sock_destroy_inode,
297 .statfs = simple_statfs,
301 * sockfs_dname() is called from d_path().
303 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
305 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
306 d_inode(dentry)->i_ino);
309 static const struct dentry_operations sockfs_dentry_operations = {
310 .d_dname = sockfs_dname,
313 static int sockfs_xattr_get(const struct xattr_handler *handler,
314 struct dentry *dentry, struct inode *inode,
315 const char *suffix, void *value, size_t size)
318 if (dentry->d_name.len + 1 > size)
320 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
322 return dentry->d_name.len + 1;
325 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
326 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
327 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
329 static const struct xattr_handler sockfs_xattr_handler = {
330 .name = XATTR_NAME_SOCKPROTONAME,
331 .get = sockfs_xattr_get,
334 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
335 struct dentry *dentry, struct inode *inode,
336 const char *suffix, const void *value,
337 size_t size, int flags)
339 /* Handled by LSM. */
343 static const struct xattr_handler sockfs_security_xattr_handler = {
344 .prefix = XATTR_SECURITY_PREFIX,
345 .set = sockfs_security_xattr_set,
348 static const struct xattr_handler *sockfs_xattr_handlers[] = {
349 &sockfs_xattr_handler,
350 &sockfs_security_xattr_handler,
354 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
355 int flags, const char *dev_name, void *data)
357 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
358 sockfs_xattr_handlers,
359 &sockfs_dentry_operations, SOCKFS_MAGIC);
362 static struct vfsmount *sock_mnt __read_mostly;
364 static struct file_system_type sock_fs_type = {
366 .mount = sockfs_mount,
367 .kill_sb = kill_anon_super,
371 * Obtains the first available file descriptor and sets it up for use.
373 * These functions create file structures and maps them to fd space
374 * of the current process. On success it returns file descriptor
375 * and file struct implicitly stored in sock->file.
376 * Note that another thread may close file descriptor before we return
377 * from this function. We use the fact that now we do not refer
378 * to socket after mapping. If one day we will need it, this
379 * function will increment ref. count on file by 1.
381 * In any case returned fd MAY BE not valid!
382 * This race condition is unavoidable
383 * with shared fd spaces, we cannot solve it inside kernel,
384 * but we take care of internal coherence yet.
387 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
389 struct qstr name = { .name = "" };
395 name.len = strlen(name.name);
396 } else if (sock->sk) {
397 name.name = sock->sk->sk_prot_creator->name;
398 name.len = strlen(name.name);
400 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
401 if (unlikely(!path.dentry)) {
403 return ERR_PTR(-ENOMEM);
405 path.mnt = mntget(sock_mnt);
407 d_instantiate(path.dentry, SOCK_INODE(sock));
409 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
412 /* drop dentry, keep inode for a bit */
413 ihold(d_inode(path.dentry));
415 /* ... and now kill it properly */
421 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
422 file->private_data = sock;
425 EXPORT_SYMBOL(sock_alloc_file);
427 static int sock_map_fd(struct socket *sock, int flags)
429 struct file *newfile;
430 int fd = get_unused_fd_flags(flags);
431 if (unlikely(fd < 0)) {
436 newfile = sock_alloc_file(sock, flags, NULL);
437 if (likely(!IS_ERR(newfile))) {
438 fd_install(fd, newfile);
443 return PTR_ERR(newfile);
446 struct socket *sock_from_file(struct file *file, int *err)
448 if (file->f_op == &socket_file_ops)
449 return file->private_data; /* set in sock_map_fd */
454 EXPORT_SYMBOL(sock_from_file);
457 * sockfd_lookup - Go from a file number to its socket slot
459 * @err: pointer to an error code return
461 * The file handle passed in is locked and the socket it is bound
462 * to is returned. If an error occurs the err pointer is overwritten
463 * with a negative errno code and NULL is returned. The function checks
464 * for both invalid handles and passing a handle which is not a socket.
466 * On a success the socket object pointer is returned.
469 struct socket *sockfd_lookup(int fd, int *err)
480 sock = sock_from_file(file, err);
485 EXPORT_SYMBOL(sockfd_lookup);
487 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
489 struct fd f = fdget(fd);
494 sock = sock_from_file(f.file, err);
496 *fput_needed = f.flags;
504 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
510 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
520 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
525 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
532 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
534 int err = simple_setattr(dentry, iattr);
536 if (!err && (iattr->ia_valid & ATTR_UID)) {
537 struct socket *sock = SOCKET_I(d_inode(dentry));
540 sock->sk->sk_uid = iattr->ia_uid;
548 static const struct inode_operations sockfs_inode_ops = {
549 .listxattr = sockfs_listxattr,
550 .setattr = sockfs_setattr,
554 * sock_alloc - allocate a socket
556 * Allocate a new inode and socket object. The two are bound together
557 * and initialised. The socket is then returned. If we are out of inodes
561 struct socket *sock_alloc(void)
566 inode = new_inode_pseudo(sock_mnt->mnt_sb);
570 sock = SOCKET_I(inode);
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;
580 EXPORT_SYMBOL(sock_alloc);
583 * sock_release - close a socket
584 * @sock: socket to close
586 * The socket is released from the protocol stack if it has a release
587 * callback, and the inode is then released if the socket is bound to
588 * an inode not a file.
591 static void __sock_release(struct socket *sock, struct inode *inode)
594 struct module *owner = sock->ops->owner;
598 sock->ops->release(sock);
605 if (sock->wq->fasync_list)
606 pr_err("%s: fasync list not empty!\n", __func__);
609 iput(SOCK_INODE(sock));
615 void sock_release(struct socket *sock)
617 __sock_release(sock, NULL);
619 EXPORT_SYMBOL(sock_release);
621 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
623 u8 flags = *tx_flags;
625 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
626 flags |= SKBTX_HW_TSTAMP;
628 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
629 flags |= SKBTX_SW_TSTAMP;
631 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
632 flags |= SKBTX_SCHED_TSTAMP;
636 EXPORT_SYMBOL(__sock_tx_timestamp);
638 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
640 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
641 BUG_ON(ret == -EIOCBQUEUED);
645 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
647 int err = security_socket_sendmsg(sock, msg,
650 return err ?: sock_sendmsg_nosec(sock, msg);
652 EXPORT_SYMBOL(sock_sendmsg);
654 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
655 struct kvec *vec, size_t num, size_t size)
657 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
658 return sock_sendmsg(sock, msg);
660 EXPORT_SYMBOL(kernel_sendmsg);
662 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
663 struct kvec *vec, size_t num, size_t size)
665 struct socket *sock = sk->sk_socket;
667 if (!sock->ops->sendmsg_locked)
668 return sock_no_sendmsg_locked(sk, msg, size);
670 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
672 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
674 EXPORT_SYMBOL(kernel_sendmsg_locked);
676 static bool skb_is_err_queue(const struct sk_buff *skb)
678 /* pkt_type of skbs enqueued on the error queue are set to
679 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
680 * in recvmsg, since skbs received on a local socket will never
681 * have a pkt_type of PACKET_OUTGOING.
683 return skb->pkt_type == PACKET_OUTGOING;
686 /* On transmit, software and hardware timestamps are returned independently.
687 * As the two skb clones share the hardware timestamp, which may be updated
688 * before the software timestamp is received, a hardware TX timestamp may be
689 * returned only if there is no software TX timestamp. Ignore false software
690 * timestamps, which may be made in the __sock_recv_timestamp() call when the
691 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
692 * hardware timestamp.
694 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
696 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
699 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
701 struct scm_ts_pktinfo ts_pktinfo;
702 struct net_device *orig_dev;
704 if (!skb_mac_header_was_set(skb))
707 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
710 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
712 ts_pktinfo.if_index = orig_dev->ifindex;
715 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
716 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
717 sizeof(ts_pktinfo), &ts_pktinfo);
721 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
723 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
726 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
727 struct scm_timestamping tss;
728 int empty = 1, false_tstamp = 0;
729 struct skb_shared_hwtstamps *shhwtstamps =
732 /* Race occurred between timestamp enabling and packet
733 receiving. Fill in the current time for now. */
734 if (need_software_tstamp && skb->tstamp == 0) {
735 __net_timestamp(skb);
739 if (need_software_tstamp) {
740 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
742 skb_get_timestamp(skb, &tv);
743 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
747 skb_get_timestampns(skb, &ts);
748 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
753 memset(&tss, 0, sizeof(tss));
754 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
755 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
758 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
759 !skb_is_swtx_tstamp(skb, false_tstamp) &&
760 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
762 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
763 !skb_is_err_queue(skb))
764 put_ts_pktinfo(msg, skb);
767 put_cmsg(msg, SOL_SOCKET,
768 SCM_TIMESTAMPING, sizeof(tss), &tss);
770 if (skb_is_err_queue(skb) && skb->len &&
771 SKB_EXT_ERR(skb)->opt_stats)
772 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
773 skb->len, skb->data);
776 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
778 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
783 if (!sock_flag(sk, SOCK_WIFI_STATUS))
785 if (!skb->wifi_acked_valid)
788 ack = skb->wifi_acked;
790 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
792 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
794 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
797 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
798 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
799 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
802 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
805 sock_recv_timestamp(msg, sk, skb);
806 sock_recv_drops(msg, sk, skb);
808 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
810 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
813 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
816 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
818 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
820 return err ?: sock_recvmsg_nosec(sock, msg, flags);
822 EXPORT_SYMBOL(sock_recvmsg);
825 * kernel_recvmsg - Receive a message from a socket (kernel space)
826 * @sock: The socket to receive the message from
827 * @msg: Received message
828 * @vec: Input s/g array for message data
829 * @num: Size of input s/g array
830 * @size: Number of bytes to read
831 * @flags: Message flags (MSG_DONTWAIT, etc...)
833 * On return the msg structure contains the scatter/gather array passed in the
834 * vec argument. The array is modified so that it consists of the unfilled
835 * portion of the original array.
837 * The returned value is the total number of bytes received, or an error.
839 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
840 struct kvec *vec, size_t num, size_t size, int flags)
842 mm_segment_t oldfs = get_fs();
845 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
847 result = sock_recvmsg(sock, msg, flags);
851 EXPORT_SYMBOL(kernel_recvmsg);
853 static ssize_t sock_sendpage(struct file *file, struct page *page,
854 int offset, size_t size, loff_t *ppos, int more)
859 sock = file->private_data;
861 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
862 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
865 return kernel_sendpage(sock, page, offset, size, flags);
868 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
869 struct pipe_inode_info *pipe, size_t len,
872 struct socket *sock = file->private_data;
874 if (unlikely(!sock->ops->splice_read))
877 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
880 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
882 struct file *file = iocb->ki_filp;
883 struct socket *sock = file->private_data;
884 struct msghdr msg = {.msg_iter = *to,
888 if (file->f_flags & O_NONBLOCK)
889 msg.msg_flags = MSG_DONTWAIT;
891 if (iocb->ki_pos != 0)
894 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
897 res = sock_recvmsg(sock, &msg, msg.msg_flags);
902 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
904 struct file *file = iocb->ki_filp;
905 struct socket *sock = file->private_data;
906 struct msghdr msg = {.msg_iter = *from,
910 if (iocb->ki_pos != 0)
913 if (file->f_flags & O_NONBLOCK)
914 msg.msg_flags = MSG_DONTWAIT;
916 if (sock->type == SOCK_SEQPACKET)
917 msg.msg_flags |= MSG_EOR;
919 res = sock_sendmsg(sock, &msg);
920 *from = msg.msg_iter;
925 * Atomic setting of ioctl hooks to avoid race
926 * with module unload.
929 static DEFINE_MUTEX(br_ioctl_mutex);
930 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
932 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
934 mutex_lock(&br_ioctl_mutex);
935 br_ioctl_hook = hook;
936 mutex_unlock(&br_ioctl_mutex);
938 EXPORT_SYMBOL(brioctl_set);
940 static DEFINE_MUTEX(vlan_ioctl_mutex);
941 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
943 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
945 mutex_lock(&vlan_ioctl_mutex);
946 vlan_ioctl_hook = hook;
947 mutex_unlock(&vlan_ioctl_mutex);
949 EXPORT_SYMBOL(vlan_ioctl_set);
951 static DEFINE_MUTEX(dlci_ioctl_mutex);
952 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
954 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
956 mutex_lock(&dlci_ioctl_mutex);
957 dlci_ioctl_hook = hook;
958 mutex_unlock(&dlci_ioctl_mutex);
960 EXPORT_SYMBOL(dlci_ioctl_set);
962 static long sock_do_ioctl(struct net *net, struct socket *sock,
963 unsigned int cmd, unsigned long arg)
966 void __user *argp = (void __user *)arg;
968 err = sock->ops->ioctl(sock, cmd, arg);
971 * If this ioctl is unknown try to hand it down
974 if (err != -ENOIOCTLCMD)
977 if (cmd == SIOCGIFCONF) {
979 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
982 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
984 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
989 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
991 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
992 if (!err && need_copyout)
993 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1000 * With an ioctl, arg may well be a user mode pointer, but we don't know
1001 * what to do with it - that's up to the protocol still.
1004 struct ns_common *get_net_ns(struct ns_common *ns)
1006 return &get_net(container_of(ns, struct net, ns))->ns;
1008 EXPORT_SYMBOL_GPL(get_net_ns);
1010 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1012 struct socket *sock;
1014 void __user *argp = (void __user *)arg;
1018 sock = file->private_data;
1021 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1024 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1026 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1027 if (!err && need_copyout)
1028 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1031 #ifdef CONFIG_WEXT_CORE
1032 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1033 err = wext_handle_ioctl(net, cmd, argp);
1040 if (get_user(pid, (int __user *)argp))
1042 err = f_setown(sock->file, pid, 1);
1046 err = put_user(f_getown(sock->file),
1047 (int __user *)argp);
1055 request_module("bridge");
1057 mutex_lock(&br_ioctl_mutex);
1059 err = br_ioctl_hook(net, cmd, argp);
1060 mutex_unlock(&br_ioctl_mutex);
1065 if (!vlan_ioctl_hook)
1066 request_module("8021q");
1068 mutex_lock(&vlan_ioctl_mutex);
1069 if (vlan_ioctl_hook)
1070 err = vlan_ioctl_hook(net, argp);
1071 mutex_unlock(&vlan_ioctl_mutex);
1076 if (!dlci_ioctl_hook)
1077 request_module("dlci");
1079 mutex_lock(&dlci_ioctl_mutex);
1080 if (dlci_ioctl_hook)
1081 err = dlci_ioctl_hook(cmd, argp);
1082 mutex_unlock(&dlci_ioctl_mutex);
1086 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1089 err = open_related_ns(&net->ns, get_net_ns);
1092 err = sock_do_ioctl(net, sock, cmd, arg);
1098 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1101 struct socket *sock = NULL;
1103 err = security_socket_create(family, type, protocol, 1);
1107 sock = sock_alloc();
1114 err = security_socket_post_create(sock, family, type, protocol, 1);
1126 EXPORT_SYMBOL(sock_create_lite);
1128 /* No kernel lock held - perfect */
1129 static __poll_t sock_poll(struct file *file, poll_table *wait)
1131 struct socket *sock = file->private_data;
1132 __poll_t events = poll_requested_events(wait), flag = 0;
1134 if (!sock->ops->poll)
1137 if (sk_can_busy_loop(sock->sk)) {
1138 /* poll once if requested by the syscall */
1139 if (events & POLL_BUSY_LOOP)
1140 sk_busy_loop(sock->sk, 1);
1142 /* if this socket can poll_ll, tell the system call */
1143 flag = POLL_BUSY_LOOP;
1146 return sock->ops->poll(file, sock, wait) | flag;
1149 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1151 struct socket *sock = file->private_data;
1153 return sock->ops->mmap(file, sock, vma);
1156 static int sock_close(struct inode *inode, struct file *filp)
1158 __sock_release(SOCKET_I(inode), inode);
1163 * Update the socket async list
1165 * Fasync_list locking strategy.
1167 * 1. fasync_list is modified only under process context socket lock
1168 * i.e. under semaphore.
1169 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1170 * or under socket lock
1173 static int sock_fasync(int fd, struct file *filp, int on)
1175 struct socket *sock = filp->private_data;
1176 struct sock *sk = sock->sk;
1177 struct socket_wq *wq;
1184 fasync_helper(fd, filp, on, &wq->fasync_list);
1186 if (!wq->fasync_list)
1187 sock_reset_flag(sk, SOCK_FASYNC);
1189 sock_set_flag(sk, SOCK_FASYNC);
1195 /* This function may be called only under rcu_lock */
1197 int sock_wake_async(struct socket_wq *wq, int how, int band)
1199 if (!wq || !wq->fasync_list)
1203 case SOCK_WAKE_WAITD:
1204 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1207 case SOCK_WAKE_SPACE:
1208 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1213 kill_fasync(&wq->fasync_list, SIGIO, band);
1216 kill_fasync(&wq->fasync_list, SIGURG, band);
1221 EXPORT_SYMBOL(sock_wake_async);
1223 int __sock_create(struct net *net, int family, int type, int protocol,
1224 struct socket **res, int kern)
1227 struct socket *sock;
1228 const struct net_proto_family *pf;
1231 * Check protocol is in range
1233 if (family < 0 || family >= NPROTO)
1234 return -EAFNOSUPPORT;
1235 if (type < 0 || type >= SOCK_MAX)
1240 This uglymoron is moved from INET layer to here to avoid
1241 deadlock in module load.
1243 if (family == PF_INET && type == SOCK_PACKET) {
1244 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1249 err = security_socket_create(family, type, protocol, kern);
1254 * Allocate the socket and allow the family to set things up. if
1255 * the protocol is 0, the family is instructed to select an appropriate
1258 sock = sock_alloc();
1260 net_warn_ratelimited("socket: no more sockets\n");
1261 return -ENFILE; /* Not exactly a match, but its the
1262 closest posix thing */
1267 #ifdef CONFIG_MODULES
1268 /* Attempt to load a protocol module if the find failed.
1270 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1271 * requested real, full-featured networking support upon configuration.
1272 * Otherwise module support will break!
1274 if (rcu_access_pointer(net_families[family]) == NULL)
1275 request_module("net-pf-%d", family);
1279 pf = rcu_dereference(net_families[family]);
1280 err = -EAFNOSUPPORT;
1285 * We will call the ->create function, that possibly is in a loadable
1286 * module, so we have to bump that loadable module refcnt first.
1288 if (!try_module_get(pf->owner))
1291 /* Now protected by module ref count */
1294 err = pf->create(net, sock, protocol, kern);
1296 goto out_module_put;
1299 * Now to bump the refcnt of the [loadable] module that owns this
1300 * socket at sock_release time we decrement its refcnt.
1302 if (!try_module_get(sock->ops->owner))
1303 goto out_module_busy;
1306 * Now that we're done with the ->create function, the [loadable]
1307 * module can have its refcnt decremented
1309 module_put(pf->owner);
1310 err = security_socket_post_create(sock, family, type, protocol, kern);
1312 goto out_sock_release;
1318 err = -EAFNOSUPPORT;
1321 module_put(pf->owner);
1328 goto out_sock_release;
1330 EXPORT_SYMBOL(__sock_create);
1332 int sock_create(int family, int type, int protocol, struct socket **res)
1334 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1336 EXPORT_SYMBOL(sock_create);
1338 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1340 return __sock_create(net, family, type, protocol, res, 1);
1342 EXPORT_SYMBOL(sock_create_kern);
1344 int __sys_socket(int family, int type, int protocol)
1347 struct socket *sock;
1350 /* Check the SOCK_* constants for consistency. */
1351 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1352 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1353 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1354 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1356 flags = type & ~SOCK_TYPE_MASK;
1357 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1359 type &= SOCK_TYPE_MASK;
1361 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1362 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1364 retval = sock_create(family, type, protocol, &sock);
1368 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1371 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1373 return __sys_socket(family, type, protocol);
1377 * Create a pair of connected sockets.
1380 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1382 struct socket *sock1, *sock2;
1384 struct file *newfile1, *newfile2;
1387 flags = type & ~SOCK_TYPE_MASK;
1388 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1390 type &= SOCK_TYPE_MASK;
1392 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1393 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1396 * reserve descriptors and make sure we won't fail
1397 * to return them to userland.
1399 fd1 = get_unused_fd_flags(flags);
1400 if (unlikely(fd1 < 0))
1403 fd2 = get_unused_fd_flags(flags);
1404 if (unlikely(fd2 < 0)) {
1409 err = put_user(fd1, &usockvec[0]);
1413 err = put_user(fd2, &usockvec[1]);
1418 * Obtain the first socket and check if the underlying protocol
1419 * supports the socketpair call.
1422 err = sock_create(family, type, protocol, &sock1);
1423 if (unlikely(err < 0))
1426 err = sock_create(family, type, protocol, &sock2);
1427 if (unlikely(err < 0)) {
1428 sock_release(sock1);
1432 err = security_socket_socketpair(sock1, sock2);
1433 if (unlikely(err)) {
1434 sock_release(sock2);
1435 sock_release(sock1);
1439 err = sock1->ops->socketpair(sock1, sock2);
1440 if (unlikely(err < 0)) {
1441 sock_release(sock2);
1442 sock_release(sock1);
1446 newfile1 = sock_alloc_file(sock1, flags, NULL);
1447 if (IS_ERR(newfile1)) {
1448 err = PTR_ERR(newfile1);
1449 sock_release(sock2);
1453 newfile2 = sock_alloc_file(sock2, flags, NULL);
1454 if (IS_ERR(newfile2)) {
1455 err = PTR_ERR(newfile2);
1460 audit_fd_pair(fd1, fd2);
1462 fd_install(fd1, newfile1);
1463 fd_install(fd2, newfile2);
1472 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1473 int __user *, usockvec)
1475 return __sys_socketpair(family, type, protocol, usockvec);
1479 * Bind a name to a socket. Nothing much to do here since it's
1480 * the protocol's responsibility to handle the local address.
1482 * We move the socket address to kernel space before we call
1483 * the protocol layer (having also checked the address is ok).
1486 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1488 struct socket *sock;
1489 struct sockaddr_storage address;
1490 int err, fput_needed;
1492 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1494 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1496 err = security_socket_bind(sock,
1497 (struct sockaddr *)&address,
1500 err = sock->ops->bind(sock,
1504 fput_light(sock->file, fput_needed);
1509 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1511 return __sys_bind(fd, umyaddr, addrlen);
1515 * Perform a listen. Basically, we allow the protocol to do anything
1516 * necessary for a listen, and if that works, we mark the socket as
1517 * ready for listening.
1520 int __sys_listen(int fd, int backlog)
1522 struct socket *sock;
1523 int err, fput_needed;
1526 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1528 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1529 if ((unsigned int)backlog > somaxconn)
1530 backlog = somaxconn;
1532 err = security_socket_listen(sock, backlog);
1534 err = sock->ops->listen(sock, backlog);
1536 fput_light(sock->file, fput_needed);
1541 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1543 return __sys_listen(fd, backlog);
1547 * For accept, we attempt to create a new socket, set up the link
1548 * with the client, wake up the client, then return the new
1549 * connected fd. We collect the address of the connector in kernel
1550 * space and move it to user at the very end. This is unclean because
1551 * we open the socket then return an error.
1553 * 1003.1g adds the ability to recvmsg() to query connection pending
1554 * status to recvmsg. We need to add that support in a way thats
1555 * clean when we restructure accept also.
1558 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1559 int __user *upeer_addrlen, int flags)
1561 struct socket *sock, *newsock;
1562 struct file *newfile;
1563 int err, len, newfd, fput_needed;
1564 struct sockaddr_storage address;
1566 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1569 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1570 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1572 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1577 newsock = sock_alloc();
1581 newsock->type = sock->type;
1582 newsock->ops = sock->ops;
1585 * We don't need try_module_get here, as the listening socket (sock)
1586 * has the protocol module (sock->ops->owner) held.
1588 __module_get(newsock->ops->owner);
1590 newfd = get_unused_fd_flags(flags);
1591 if (unlikely(newfd < 0)) {
1593 sock_release(newsock);
1596 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1597 if (IS_ERR(newfile)) {
1598 err = PTR_ERR(newfile);
1599 put_unused_fd(newfd);
1603 err = security_socket_accept(sock, newsock);
1607 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1611 if (upeer_sockaddr) {
1612 len = newsock->ops->getname(newsock,
1613 (struct sockaddr *)&address, 2);
1615 err = -ECONNABORTED;
1618 err = move_addr_to_user(&address,
1619 len, upeer_sockaddr, upeer_addrlen);
1624 /* File flags are not inherited via accept() unlike another OSes. */
1626 fd_install(newfd, newfile);
1630 fput_light(sock->file, fput_needed);
1635 put_unused_fd(newfd);
1639 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1640 int __user *, upeer_addrlen, int, flags)
1642 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1645 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1646 int __user *, upeer_addrlen)
1648 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1652 * Attempt to connect to a socket with the server address. The address
1653 * is in user space so we verify it is OK and move it to kernel space.
1655 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1658 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1659 * other SEQPACKET protocols that take time to connect() as it doesn't
1660 * include the -EINPROGRESS status for such sockets.
1663 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1665 struct socket *sock;
1666 struct sockaddr_storage address;
1667 int err, fput_needed;
1669 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1677 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1681 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1682 sock->file->f_flags);
1684 fput_light(sock->file, fput_needed);
1689 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1692 return __sys_connect(fd, uservaddr, addrlen);
1696 * Get the local address ('name') of a socket object. Move the obtained
1697 * name to user space.
1700 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1701 int __user *usockaddr_len)
1703 struct socket *sock;
1704 struct sockaddr_storage address;
1705 int err, fput_needed;
1707 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1711 err = security_socket_getsockname(sock);
1715 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1718 /* "err" is actually length in this case */
1719 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1722 fput_light(sock->file, fput_needed);
1727 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1728 int __user *, usockaddr_len)
1730 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1734 * Get the remote address ('name') of a socket object. Move the obtained
1735 * name to user space.
1738 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1739 int __user *usockaddr_len)
1741 struct socket *sock;
1742 struct sockaddr_storage address;
1743 int err, fput_needed;
1745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1747 err = security_socket_getpeername(sock);
1749 fput_light(sock->file, fput_needed);
1753 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1755 /* "err" is actually length in this case */
1756 err = move_addr_to_user(&address, err, usockaddr,
1758 fput_light(sock->file, fput_needed);
1763 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1764 int __user *, usockaddr_len)
1766 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1770 * Send a datagram to a given address. We move the address into kernel
1771 * space and check the user space data area is readable before invoking
1774 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1775 struct sockaddr __user *addr, int addr_len)
1777 struct socket *sock;
1778 struct sockaddr_storage address;
1784 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1787 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1791 msg.msg_name = NULL;
1792 msg.msg_control = NULL;
1793 msg.msg_controllen = 0;
1794 msg.msg_namelen = 0;
1796 err = move_addr_to_kernel(addr, addr_len, &address);
1799 msg.msg_name = (struct sockaddr *)&address;
1800 msg.msg_namelen = addr_len;
1802 if (sock->file->f_flags & O_NONBLOCK)
1803 flags |= MSG_DONTWAIT;
1804 msg.msg_flags = flags;
1805 err = sock_sendmsg(sock, &msg);
1808 fput_light(sock->file, fput_needed);
1813 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1814 unsigned int, flags, struct sockaddr __user *, addr,
1817 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1821 * Send a datagram down a socket.
1824 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1825 unsigned int, flags)
1827 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1831 * Receive a frame from the socket and optionally record the address of the
1832 * sender. We verify the buffers are writable and if needed move the
1833 * sender address from kernel to user space.
1835 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1836 struct sockaddr __user *addr, int __user *addr_len)
1838 struct socket *sock;
1841 struct sockaddr_storage address;
1845 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1848 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1852 msg.msg_control = NULL;
1853 msg.msg_controllen = 0;
1854 /* Save some cycles and don't copy the address if not needed */
1855 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1856 /* We assume all kernel code knows the size of sockaddr_storage */
1857 msg.msg_namelen = 0;
1858 msg.msg_iocb = NULL;
1860 if (sock->file->f_flags & O_NONBLOCK)
1861 flags |= MSG_DONTWAIT;
1862 err = sock_recvmsg(sock, &msg, flags);
1864 if (err >= 0 && addr != NULL) {
1865 err2 = move_addr_to_user(&address,
1866 msg.msg_namelen, addr, addr_len);
1871 fput_light(sock->file, fput_needed);
1876 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1877 unsigned int, flags, struct sockaddr __user *, addr,
1878 int __user *, addr_len)
1880 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1884 * Receive a datagram from a socket.
1887 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1888 unsigned int, flags)
1890 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1894 * Set a socket option. Because we don't know the option lengths we have
1895 * to pass the user mode parameter for the protocols to sort out.
1898 static int __sys_setsockopt(int fd, int level, int optname,
1899 char __user *optval, int optlen)
1901 int err, fput_needed;
1902 struct socket *sock;
1907 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1909 err = security_socket_setsockopt(sock, level, optname);
1913 if (level == SOL_SOCKET)
1915 sock_setsockopt(sock, level, optname, optval,
1919 sock->ops->setsockopt(sock, level, optname, optval,
1922 fput_light(sock->file, fput_needed);
1927 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1928 char __user *, optval, int, optlen)
1930 return __sys_setsockopt(fd, level, optname, optval, optlen);
1934 * Get a socket option. Because we don't know the option lengths we have
1935 * to pass a user mode parameter for the protocols to sort out.
1938 static int __sys_getsockopt(int fd, int level, int optname,
1939 char __user *optval, int __user *optlen)
1941 int err, fput_needed;
1942 struct socket *sock;
1944 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1946 err = security_socket_getsockopt(sock, level, optname);
1950 if (level == SOL_SOCKET)
1952 sock_getsockopt(sock, level, optname, optval,
1956 sock->ops->getsockopt(sock, level, optname, optval,
1959 fput_light(sock->file, fput_needed);
1964 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1965 char __user *, optval, int __user *, optlen)
1967 return __sys_getsockopt(fd, level, optname, optval, optlen);
1971 * Shutdown a socket.
1974 int __sys_shutdown(int fd, int how)
1976 int err, fput_needed;
1977 struct socket *sock;
1979 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1981 err = security_socket_shutdown(sock, how);
1983 err = sock->ops->shutdown(sock, how);
1984 fput_light(sock->file, fput_needed);
1989 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1991 return __sys_shutdown(fd, how);
1994 /* A couple of helpful macros for getting the address of the 32/64 bit
1995 * fields which are the same type (int / unsigned) on our platforms.
1997 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1998 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1999 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2001 struct used_address {
2002 struct sockaddr_storage name;
2003 unsigned int name_len;
2006 static int copy_msghdr_from_user(struct msghdr *kmsg,
2007 struct user_msghdr __user *umsg,
2008 struct sockaddr __user **save_addr,
2011 struct user_msghdr msg;
2014 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2017 kmsg->msg_control = (void __force *)msg.msg_control;
2018 kmsg->msg_controllen = msg.msg_controllen;
2019 kmsg->msg_flags = msg.msg_flags;
2021 kmsg->msg_namelen = msg.msg_namelen;
2023 kmsg->msg_namelen = 0;
2025 if (kmsg->msg_namelen < 0)
2028 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2029 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2032 *save_addr = msg.msg_name;
2034 if (msg.msg_name && kmsg->msg_namelen) {
2036 err = move_addr_to_kernel(msg.msg_name,
2043 kmsg->msg_name = NULL;
2044 kmsg->msg_namelen = 0;
2047 if (msg.msg_iovlen > UIO_MAXIOV)
2050 kmsg->msg_iocb = NULL;
2052 return import_iovec(save_addr ? READ : WRITE,
2053 msg.msg_iov, msg.msg_iovlen,
2054 UIO_FASTIOV, iov, &kmsg->msg_iter);
2057 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2058 struct msghdr *msg_sys, unsigned int flags,
2059 struct used_address *used_address,
2060 unsigned int allowed_msghdr_flags)
2062 struct compat_msghdr __user *msg_compat =
2063 (struct compat_msghdr __user *)msg;
2064 struct sockaddr_storage address;
2065 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2066 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2067 __aligned(sizeof(__kernel_size_t));
2068 /* 20 is size of ipv6_pktinfo */
2069 unsigned char *ctl_buf = ctl;
2073 msg_sys->msg_name = &address;
2075 if (MSG_CMSG_COMPAT & flags)
2076 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2078 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2084 if (msg_sys->msg_controllen > INT_MAX)
2086 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2087 ctl_len = msg_sys->msg_controllen;
2088 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2090 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2094 ctl_buf = msg_sys->msg_control;
2095 ctl_len = msg_sys->msg_controllen;
2096 } else if (ctl_len) {
2097 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2098 CMSG_ALIGN(sizeof(struct cmsghdr)));
2099 if (ctl_len > sizeof(ctl)) {
2100 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2101 if (ctl_buf == NULL)
2106 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2107 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2108 * checking falls down on this.
2110 if (copy_from_user(ctl_buf,
2111 (void __user __force *)msg_sys->msg_control,
2114 msg_sys->msg_control = ctl_buf;
2116 msg_sys->msg_flags = flags;
2118 if (sock->file->f_flags & O_NONBLOCK)
2119 msg_sys->msg_flags |= MSG_DONTWAIT;
2121 * If this is sendmmsg() and current destination address is same as
2122 * previously succeeded address, omit asking LSM's decision.
2123 * used_address->name_len is initialized to UINT_MAX so that the first
2124 * destination address never matches.
2126 if (used_address && msg_sys->msg_name &&
2127 used_address->name_len == msg_sys->msg_namelen &&
2128 !memcmp(&used_address->name, msg_sys->msg_name,
2129 used_address->name_len)) {
2130 err = sock_sendmsg_nosec(sock, msg_sys);
2133 err = sock_sendmsg(sock, msg_sys);
2135 * If this is sendmmsg() and sending to current destination address was
2136 * successful, remember it.
2138 if (used_address && err >= 0) {
2139 used_address->name_len = msg_sys->msg_namelen;
2140 if (msg_sys->msg_name)
2141 memcpy(&used_address->name, msg_sys->msg_name,
2142 used_address->name_len);
2147 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2154 * BSD sendmsg interface
2157 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2158 bool forbid_cmsg_compat)
2160 int fput_needed, err;
2161 struct msghdr msg_sys;
2162 struct socket *sock;
2164 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2167 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2171 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2173 fput_light(sock->file, fput_needed);
2178 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2180 return __sys_sendmsg(fd, msg, flags, true);
2184 * Linux sendmmsg interface
2187 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2188 unsigned int flags, bool forbid_cmsg_compat)
2190 int fput_needed, err, datagrams;
2191 struct socket *sock;
2192 struct mmsghdr __user *entry;
2193 struct compat_mmsghdr __user *compat_entry;
2194 struct msghdr msg_sys;
2195 struct used_address used_address;
2196 unsigned int oflags = flags;
2198 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2201 if (vlen > UIO_MAXIOV)
2206 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2210 used_address.name_len = UINT_MAX;
2212 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2216 while (datagrams < vlen) {
2217 if (datagrams == vlen - 1)
2220 if (MSG_CMSG_COMPAT & flags) {
2221 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2222 &msg_sys, flags, &used_address, MSG_EOR);
2225 err = __put_user(err, &compat_entry->msg_len);
2228 err = ___sys_sendmsg(sock,
2229 (struct user_msghdr __user *)entry,
2230 &msg_sys, flags, &used_address, MSG_EOR);
2233 err = put_user(err, &entry->msg_len);
2240 if (msg_data_left(&msg_sys))
2245 fput_light(sock->file, fput_needed);
2247 /* We only return an error if no datagrams were able to be sent */
2254 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2255 unsigned int, vlen, unsigned int, flags)
2257 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2260 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2261 struct msghdr *msg_sys, unsigned int flags, int nosec)
2263 struct compat_msghdr __user *msg_compat =
2264 (struct compat_msghdr __user *)msg;
2265 struct iovec iovstack[UIO_FASTIOV];
2266 struct iovec *iov = iovstack;
2267 unsigned long cmsg_ptr;
2271 /* kernel mode address */
2272 struct sockaddr_storage addr;
2274 /* user mode address pointers */
2275 struct sockaddr __user *uaddr;
2276 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2278 msg_sys->msg_name = &addr;
2280 if (MSG_CMSG_COMPAT & flags)
2281 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2283 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2287 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2288 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2290 /* We assume all kernel code knows the size of sockaddr_storage */
2291 msg_sys->msg_namelen = 0;
2293 if (sock->file->f_flags & O_NONBLOCK)
2294 flags |= MSG_DONTWAIT;
2295 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2300 if (uaddr != NULL) {
2301 err = move_addr_to_user(&addr,
2302 msg_sys->msg_namelen, uaddr,
2307 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2311 if (MSG_CMSG_COMPAT & flags)
2312 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2313 &msg_compat->msg_controllen);
2315 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2316 &msg->msg_controllen);
2327 * BSD recvmsg interface
2330 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2331 bool forbid_cmsg_compat)
2333 int fput_needed, err;
2334 struct msghdr msg_sys;
2335 struct socket *sock;
2337 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2340 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2344 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2346 fput_light(sock->file, fput_needed);
2351 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2352 unsigned int, flags)
2354 return __sys_recvmsg(fd, msg, flags, true);
2358 * Linux recvmmsg interface
2361 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2362 unsigned int flags, struct timespec *timeout)
2364 int fput_needed, err, datagrams;
2365 struct socket *sock;
2366 struct mmsghdr __user *entry;
2367 struct compat_mmsghdr __user *compat_entry;
2368 struct msghdr msg_sys;
2369 struct timespec64 end_time;
2370 struct timespec64 timeout64;
2373 poll_select_set_timeout(&end_time, timeout->tv_sec,
2379 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2383 if (likely(!(flags & MSG_ERRQUEUE))) {
2384 err = sock_error(sock->sk);
2392 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2394 while (datagrams < vlen) {
2396 * No need to ask LSM for more than the first datagram.
2398 if (MSG_CMSG_COMPAT & flags) {
2399 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2400 &msg_sys, flags & ~MSG_WAITFORONE,
2404 err = __put_user(err, &compat_entry->msg_len);
2407 err = ___sys_recvmsg(sock,
2408 (struct user_msghdr __user *)entry,
2409 &msg_sys, flags & ~MSG_WAITFORONE,
2413 err = put_user(err, &entry->msg_len);
2421 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2422 if (flags & MSG_WAITFORONE)
2423 flags |= MSG_DONTWAIT;
2426 ktime_get_ts64(&timeout64);
2427 *timeout = timespec64_to_timespec(
2428 timespec64_sub(end_time, timeout64));
2429 if (timeout->tv_sec < 0) {
2430 timeout->tv_sec = timeout->tv_nsec = 0;
2434 /* Timeout, return less than vlen datagrams */
2435 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2439 /* Out of band data, return right away */
2440 if (msg_sys.msg_flags & MSG_OOB)
2448 if (datagrams == 0) {
2454 * We may return less entries than requested (vlen) if the
2455 * sock is non block and there aren't enough datagrams...
2457 if (err != -EAGAIN) {
2459 * ... or if recvmsg returns an error after we
2460 * received some datagrams, where we record the
2461 * error to return on the next call or if the
2462 * app asks about it using getsockopt(SO_ERROR).
2464 sock->sk->sk_err = -err;
2467 fput_light(sock->file, fput_needed);
2472 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2473 unsigned int vlen, unsigned int flags,
2474 struct timespec __user *timeout)
2477 struct timespec timeout_sys;
2479 if (flags & MSG_CMSG_COMPAT)
2483 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2485 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2488 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2490 if (datagrams > 0 &&
2491 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2492 datagrams = -EFAULT;
2497 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2498 unsigned int, vlen, unsigned int, flags,
2499 struct timespec __user *, timeout)
2501 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2504 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2505 /* Argument list sizes for sys_socketcall */
2506 #define AL(x) ((x) * sizeof(unsigned long))
2507 static const unsigned char nargs[21] = {
2508 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2509 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2510 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2517 * System call vectors.
2519 * Argument checking cleaned up. Saved 20% in size.
2520 * This function doesn't need to set the kernel lock because
2521 * it is set by the callees.
2524 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2526 unsigned long a[AUDITSC_ARGS];
2527 unsigned long a0, a1;
2531 if (call < 1 || call > SYS_SENDMMSG)
2533 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2536 if (len > sizeof(a))
2539 /* copy_from_user should be SMP safe. */
2540 if (copy_from_user(a, args, len))
2543 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2552 err = __sys_socket(a0, a1, a[2]);
2555 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2558 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2561 err = __sys_listen(a0, a1);
2564 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2565 (int __user *)a[2], 0);
2567 case SYS_GETSOCKNAME:
2569 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2570 (int __user *)a[2]);
2572 case SYS_GETPEERNAME:
2574 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2575 (int __user *)a[2]);
2577 case SYS_SOCKETPAIR:
2578 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2581 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2585 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2586 (struct sockaddr __user *)a[4], a[5]);
2589 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2593 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2594 (struct sockaddr __user *)a[4],
2595 (int __user *)a[5]);
2598 err = __sys_shutdown(a0, a1);
2600 case SYS_SETSOCKOPT:
2601 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2604 case SYS_GETSOCKOPT:
2606 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2607 (int __user *)a[4]);
2610 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2614 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2618 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2622 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2623 a[3], (struct timespec __user *)a[4]);
2626 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2627 (int __user *)a[2], a[3]);
2636 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2639 * sock_register - add a socket protocol handler
2640 * @ops: description of protocol
2642 * This function is called by a protocol handler that wants to
2643 * advertise its address family, and have it linked into the
2644 * socket interface. The value ops->family corresponds to the
2645 * socket system call protocol family.
2647 int sock_register(const struct net_proto_family *ops)
2651 if (ops->family >= NPROTO) {
2652 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2656 spin_lock(&net_family_lock);
2657 if (rcu_dereference_protected(net_families[ops->family],
2658 lockdep_is_held(&net_family_lock)))
2661 rcu_assign_pointer(net_families[ops->family], ops);
2664 spin_unlock(&net_family_lock);
2666 pr_info("NET: Registered protocol family %d\n", ops->family);
2669 EXPORT_SYMBOL(sock_register);
2672 * sock_unregister - remove a protocol handler
2673 * @family: protocol family to remove
2675 * This function is called by a protocol handler that wants to
2676 * remove its address family, and have it unlinked from the
2677 * new socket creation.
2679 * If protocol handler is a module, then it can use module reference
2680 * counts to protect against new references. If protocol handler is not
2681 * a module then it needs to provide its own protection in
2682 * the ops->create routine.
2684 void sock_unregister(int family)
2686 BUG_ON(family < 0 || family >= NPROTO);
2688 spin_lock(&net_family_lock);
2689 RCU_INIT_POINTER(net_families[family], NULL);
2690 spin_unlock(&net_family_lock);
2694 pr_info("NET: Unregistered protocol family %d\n", family);
2696 EXPORT_SYMBOL(sock_unregister);
2698 bool sock_is_registered(int family)
2700 return family < NPROTO && rcu_access_pointer(net_families[family]);
2703 static int __init sock_init(void)
2707 * Initialize the network sysctl infrastructure.
2709 err = net_sysctl_init();
2714 * Initialize skbuff SLAB cache
2719 * Initialize the protocols module.
2724 err = register_filesystem(&sock_fs_type);
2727 sock_mnt = kern_mount(&sock_fs_type);
2728 if (IS_ERR(sock_mnt)) {
2729 err = PTR_ERR(sock_mnt);
2733 /* The real protocol initialization is performed in later initcalls.
2736 #ifdef CONFIG_NETFILTER
2737 err = netfilter_init();
2742 ptp_classifier_init();
2748 unregister_filesystem(&sock_fs_type);
2753 core_initcall(sock_init); /* early initcall */
2755 #ifdef CONFIG_PROC_FS
2756 void socket_seq_show(struct seq_file *seq)
2758 seq_printf(seq, "sockets: used %d\n",
2759 sock_inuse_get(seq->private));
2761 #endif /* CONFIG_PROC_FS */
2763 #ifdef CONFIG_COMPAT
2764 static int do_siocgstamp(struct net *net, struct socket *sock,
2765 unsigned int cmd, void __user *up)
2767 mm_segment_t old_fs = get_fs();
2772 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2775 err = compat_put_timeval(&ktv, up);
2780 static int do_siocgstampns(struct net *net, struct socket *sock,
2781 unsigned int cmd, void __user *up)
2783 mm_segment_t old_fs = get_fs();
2784 struct timespec kts;
2788 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2791 err = compat_put_timespec(&kts, up);
2796 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2798 struct compat_ifconf ifc32;
2802 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2805 ifc.ifc_len = ifc32.ifc_len;
2806 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2809 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2814 ifc32.ifc_len = ifc.ifc_len;
2815 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2821 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2823 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2824 bool convert_in = false, convert_out = false;
2825 size_t buf_size = 0;
2826 struct ethtool_rxnfc __user *rxnfc = NULL;
2828 u32 rule_cnt = 0, actual_rule_cnt;
2833 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2836 compat_rxnfc = compat_ptr(data);
2838 if (get_user(ethcmd, &compat_rxnfc->cmd))
2841 /* Most ethtool structures are defined without padding.
2842 * Unfortunately struct ethtool_rxnfc is an exception.
2847 case ETHTOOL_GRXCLSRLALL:
2848 /* Buffer size is variable */
2849 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2851 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2853 buf_size += rule_cnt * sizeof(u32);
2855 case ETHTOOL_GRXRINGS:
2856 case ETHTOOL_GRXCLSRLCNT:
2857 case ETHTOOL_GRXCLSRULE:
2858 case ETHTOOL_SRXCLSRLINS:
2861 case ETHTOOL_SRXCLSRLDEL:
2862 buf_size += sizeof(struct ethtool_rxnfc);
2864 rxnfc = compat_alloc_user_space(buf_size);
2868 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2871 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2874 /* We expect there to be holes between fs.m_ext and
2875 * fs.ring_cookie and at the end of fs, but nowhere else.
2877 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2878 sizeof(compat_rxnfc->fs.m_ext) !=
2879 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2880 sizeof(rxnfc->fs.m_ext));
2882 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2883 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2884 offsetof(struct ethtool_rxnfc, fs.location) -
2885 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2887 if (copy_in_user(rxnfc, compat_rxnfc,
2888 (void __user *)(&rxnfc->fs.m_ext + 1) -
2889 (void __user *)rxnfc) ||
2890 copy_in_user(&rxnfc->fs.ring_cookie,
2891 &compat_rxnfc->fs.ring_cookie,
2892 (void __user *)(&rxnfc->fs.location + 1) -
2893 (void __user *)&rxnfc->fs.ring_cookie) ||
2894 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2895 sizeof(rxnfc->rule_cnt)))
2899 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2904 if (copy_in_user(compat_rxnfc, rxnfc,
2905 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2906 (const void __user *)rxnfc) ||
2907 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2908 &rxnfc->fs.ring_cookie,
2909 (const void __user *)(&rxnfc->fs.location + 1) -
2910 (const void __user *)&rxnfc->fs.ring_cookie) ||
2911 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2912 sizeof(rxnfc->rule_cnt)))
2915 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2916 /* As an optimisation, we only copy the actual
2917 * number of rules that the underlying
2918 * function returned. Since Mallory might
2919 * change the rule count in user memory, we
2920 * check that it is less than the rule count
2921 * originally given (as the user buffer size),
2922 * which has been range-checked.
2924 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2926 if (actual_rule_cnt < rule_cnt)
2927 rule_cnt = actual_rule_cnt;
2928 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2929 &rxnfc->rule_locs[0],
2930 rule_cnt * sizeof(u32)))
2938 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2940 compat_uptr_t uptr32;
2945 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2948 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2951 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2952 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2954 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2956 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2957 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2963 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2964 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2965 struct compat_ifreq __user *u_ifreq32)
2970 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2972 if (get_user(data32, &u_ifreq32->ifr_data))
2974 ifreq.ifr_data = compat_ptr(data32);
2976 return dev_ioctl(net, cmd, &ifreq, NULL);
2979 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2980 struct compat_ifreq __user *uifr32)
2983 struct compat_ifmap __user *uifmap32;
2986 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2987 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2988 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2989 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2990 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2991 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2992 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2993 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2997 err = dev_ioctl(net, cmd, &ifr, NULL);
2999 if (cmd == SIOCGIFMAP && !err) {
3000 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3001 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3002 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3003 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3004 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3005 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3006 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3015 struct sockaddr rt_dst; /* target address */
3016 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3017 struct sockaddr rt_genmask; /* target network mask (IP) */
3018 unsigned short rt_flags;
3021 unsigned char rt_tos;
3022 unsigned char rt_class;
3024 short rt_metric; /* +1 for binary compatibility! */
3025 /* char * */ u32 rt_dev; /* forcing the device at add */
3026 u32 rt_mtu; /* per route MTU/Window */
3027 u32 rt_window; /* Window clamping */
3028 unsigned short rt_irtt; /* Initial RTT */
3031 struct in6_rtmsg32 {
3032 struct in6_addr rtmsg_dst;
3033 struct in6_addr rtmsg_src;
3034 struct in6_addr rtmsg_gateway;
3044 static int routing_ioctl(struct net *net, struct socket *sock,
3045 unsigned int cmd, void __user *argp)
3049 struct in6_rtmsg r6;
3053 mm_segment_t old_fs = get_fs();
3055 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3056 struct in6_rtmsg32 __user *ur6 = argp;
3057 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3058 3 * sizeof(struct in6_addr));
3059 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3060 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3061 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3062 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3063 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3064 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3065 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3069 struct rtentry32 __user *ur4 = argp;
3070 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3071 3 * sizeof(struct sockaddr));
3072 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3073 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3074 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3075 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3076 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3077 ret |= get_user(rtdev, &(ur4->rt_dev));
3079 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3080 r4.rt_dev = (char __user __force *)devname;
3094 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3101 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3102 * for some operations; this forces use of the newer bridge-utils that
3103 * use compatible ioctls
3105 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3109 if (get_user(tmp, argp))
3111 if (tmp == BRCTL_GET_VERSION)
3112 return BRCTL_VERSION + 1;
3116 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3117 unsigned int cmd, unsigned long arg)
3119 void __user *argp = compat_ptr(arg);
3120 struct sock *sk = sock->sk;
3121 struct net *net = sock_net(sk);
3123 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3124 return compat_ifr_data_ioctl(net, cmd, argp);
3129 return old_bridge_ioctl(argp);
3131 return compat_dev_ifconf(net, argp);
3133 return ethtool_ioctl(net, argp);
3135 return compat_siocwandev(net, argp);
3138 return compat_sioc_ifmap(net, cmd, argp);
3141 return routing_ioctl(net, sock, cmd, argp);
3143 return do_siocgstamp(net, sock, cmd, argp);
3145 return do_siocgstampns(net, sock, cmd, argp);
3146 case SIOCBONDSLAVEINFOQUERY:
3147 case SIOCBONDINFOQUERY:
3150 return compat_ifr_data_ioctl(net, cmd, argp);
3163 return sock_ioctl(file, cmd, arg);
3180 case SIOCSIFHWBROADCAST:
3182 case SIOCGIFBRDADDR:
3183 case SIOCSIFBRDADDR:
3184 case SIOCGIFDSTADDR:
3185 case SIOCSIFDSTADDR:
3186 case SIOCGIFNETMASK:
3187 case SIOCSIFNETMASK:
3202 case SIOCBONDENSLAVE:
3203 case SIOCBONDRELEASE:
3204 case SIOCBONDSETHWADDR:
3205 case SIOCBONDCHANGEACTIVE:
3207 return sock_do_ioctl(net, sock, cmd, arg);
3210 return -ENOIOCTLCMD;
3213 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3216 struct socket *sock = file->private_data;
3217 int ret = -ENOIOCTLCMD;
3224 if (sock->ops->compat_ioctl)
3225 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3227 if (ret == -ENOIOCTLCMD &&
3228 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3229 ret = compat_wext_handle_ioctl(net, cmd, arg);
3231 if (ret == -ENOIOCTLCMD)
3232 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3238 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3240 return sock->ops->bind(sock, addr, addrlen);
3242 EXPORT_SYMBOL(kernel_bind);
3244 int kernel_listen(struct socket *sock, int backlog)
3246 return sock->ops->listen(sock, backlog);
3248 EXPORT_SYMBOL(kernel_listen);
3250 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3252 struct sock *sk = sock->sk;
3255 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3260 err = sock->ops->accept(sock, *newsock, flags, true);
3262 sock_release(*newsock);
3267 (*newsock)->ops = sock->ops;
3268 __module_get((*newsock)->ops->owner);
3273 EXPORT_SYMBOL(kernel_accept);
3275 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3278 return sock->ops->connect(sock, addr, addrlen, flags);
3280 EXPORT_SYMBOL(kernel_connect);
3282 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3284 return sock->ops->getname(sock, addr, 0);
3286 EXPORT_SYMBOL(kernel_getsockname);
3288 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3290 return sock->ops->getname(sock, addr, 1);
3292 EXPORT_SYMBOL(kernel_getpeername);
3294 int kernel_getsockopt(struct socket *sock, int level, int optname,
3295 char *optval, int *optlen)
3297 mm_segment_t oldfs = get_fs();
3298 char __user *uoptval;
3299 int __user *uoptlen;
3302 uoptval = (char __user __force *) optval;
3303 uoptlen = (int __user __force *) optlen;
3306 if (level == SOL_SOCKET)
3307 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3309 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3314 EXPORT_SYMBOL(kernel_getsockopt);
3316 int kernel_setsockopt(struct socket *sock, int level, int optname,
3317 char *optval, unsigned int optlen)
3319 mm_segment_t oldfs = get_fs();
3320 char __user *uoptval;
3323 uoptval = (char __user __force *) optval;
3326 if (level == SOL_SOCKET)
3327 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3329 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3334 EXPORT_SYMBOL(kernel_setsockopt);
3336 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3337 size_t size, int flags)
3339 if (sock->ops->sendpage)
3340 return sock->ops->sendpage(sock, page, offset, size, flags);
3342 return sock_no_sendpage(sock, page, offset, size, flags);
3344 EXPORT_SYMBOL(kernel_sendpage);
3346 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3347 size_t size, int flags)
3349 struct socket *sock = sk->sk_socket;
3351 if (sock->ops->sendpage_locked)
3352 return sock->ops->sendpage_locked(sk, page, offset, size,
3355 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3357 EXPORT_SYMBOL(kernel_sendpage_locked);
3359 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3361 return sock->ops->shutdown(sock, how);
3363 EXPORT_SYMBOL(kernel_sock_shutdown);
3365 /* This routine returns the IP overhead imposed by a socket i.e.
3366 * the length of the underlying IP header, depending on whether
3367 * this is an IPv4 or IPv6 socket and the length from IP options turned
3368 * on at the socket. Assumes that the caller has a lock on the socket.
3370 u32 kernel_sock_ip_overhead(struct sock *sk)
3372 struct inet_sock *inet;
3373 struct ip_options_rcu *opt;
3375 #if IS_ENABLED(CONFIG_IPV6)
3376 struct ipv6_pinfo *np;
3377 struct ipv6_txoptions *optv6 = NULL;
3378 #endif /* IS_ENABLED(CONFIG_IPV6) */
3383 switch (sk->sk_family) {
3386 overhead += sizeof(struct iphdr);
3387 opt = rcu_dereference_protected(inet->inet_opt,
3388 sock_owned_by_user(sk));
3390 overhead += opt->opt.optlen;
3392 #if IS_ENABLED(CONFIG_IPV6)
3395 overhead += sizeof(struct ipv6hdr);
3397 optv6 = rcu_dereference_protected(np->opt,
3398 sock_owned_by_user(sk));
3400 overhead += (optv6->opt_flen + optv6->opt_nflen);
3402 #endif /* IS_ENABLED(CONFIG_IPV6) */
3403 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3407 EXPORT_SYMBOL(kernel_sock_ip_overhead);