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)
392 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
394 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
395 O_RDWR | (flags & O_NONBLOCK),
403 file->private_data = sock;
406 EXPORT_SYMBOL(sock_alloc_file);
408 static int sock_map_fd(struct socket *sock, int flags)
410 struct file *newfile;
411 int fd = get_unused_fd_flags(flags);
412 if (unlikely(fd < 0)) {
417 newfile = sock_alloc_file(sock, flags, NULL);
418 if (likely(!IS_ERR(newfile))) {
419 fd_install(fd, newfile);
424 return PTR_ERR(newfile);
427 struct socket *sock_from_file(struct file *file, int *err)
429 if (file->f_op == &socket_file_ops)
430 return file->private_data; /* set in sock_map_fd */
435 EXPORT_SYMBOL(sock_from_file);
438 * sockfd_lookup - Go from a file number to its socket slot
440 * @err: pointer to an error code return
442 * The file handle passed in is locked and the socket it is bound
443 * to is returned. If an error occurs the err pointer is overwritten
444 * with a negative errno code and NULL is returned. The function checks
445 * for both invalid handles and passing a handle which is not a socket.
447 * On a success the socket object pointer is returned.
450 struct socket *sockfd_lookup(int fd, int *err)
461 sock = sock_from_file(file, err);
466 EXPORT_SYMBOL(sockfd_lookup);
468 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
470 struct fd f = fdget(fd);
475 sock = sock_from_file(f.file, err);
477 *fput_needed = f.flags;
485 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
491 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
501 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
506 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
513 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
515 int err = simple_setattr(dentry, iattr);
517 if (!err && (iattr->ia_valid & ATTR_UID)) {
518 struct socket *sock = SOCKET_I(d_inode(dentry));
521 sock->sk->sk_uid = iattr->ia_uid;
529 static const struct inode_operations sockfs_inode_ops = {
530 .listxattr = sockfs_listxattr,
531 .setattr = sockfs_setattr,
535 * sock_alloc - allocate a socket
537 * Allocate a new inode and socket object. The two are bound together
538 * and initialised. The socket is then returned. If we are out of inodes
542 struct socket *sock_alloc(void)
547 inode = new_inode_pseudo(sock_mnt->mnt_sb);
551 sock = SOCKET_I(inode);
553 inode->i_ino = get_next_ino();
554 inode->i_mode = S_IFSOCK | S_IRWXUGO;
555 inode->i_uid = current_fsuid();
556 inode->i_gid = current_fsgid();
557 inode->i_op = &sockfs_inode_ops;
561 EXPORT_SYMBOL(sock_alloc);
564 * sock_release - close a socket
565 * @sock: socket to close
567 * The socket is released from the protocol stack if it has a release
568 * callback, and the inode is then released if the socket is bound to
569 * an inode not a file.
572 static void __sock_release(struct socket *sock, struct inode *inode)
575 struct module *owner = sock->ops->owner;
579 sock->ops->release(sock);
586 if (sock->wq->fasync_list)
587 pr_err("%s: fasync list not empty!\n", __func__);
590 iput(SOCK_INODE(sock));
596 void sock_release(struct socket *sock)
598 __sock_release(sock, NULL);
600 EXPORT_SYMBOL(sock_release);
602 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
604 u8 flags = *tx_flags;
606 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
607 flags |= SKBTX_HW_TSTAMP;
609 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
610 flags |= SKBTX_SW_TSTAMP;
612 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
613 flags |= SKBTX_SCHED_TSTAMP;
617 EXPORT_SYMBOL(__sock_tx_timestamp);
619 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
621 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
622 BUG_ON(ret == -EIOCBQUEUED);
626 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
628 int err = security_socket_sendmsg(sock, msg,
631 return err ?: sock_sendmsg_nosec(sock, msg);
633 EXPORT_SYMBOL(sock_sendmsg);
635 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
636 struct kvec *vec, size_t num, size_t size)
638 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
639 return sock_sendmsg(sock, msg);
641 EXPORT_SYMBOL(kernel_sendmsg);
643 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
644 struct kvec *vec, size_t num, size_t size)
646 struct socket *sock = sk->sk_socket;
648 if (!sock->ops->sendmsg_locked)
649 return sock_no_sendmsg_locked(sk, msg, size);
651 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
653 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
655 EXPORT_SYMBOL(kernel_sendmsg_locked);
657 static bool skb_is_err_queue(const struct sk_buff *skb)
659 /* pkt_type of skbs enqueued on the error queue are set to
660 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
661 * in recvmsg, since skbs received on a local socket will never
662 * have a pkt_type of PACKET_OUTGOING.
664 return skb->pkt_type == PACKET_OUTGOING;
667 /* On transmit, software and hardware timestamps are returned independently.
668 * As the two skb clones share the hardware timestamp, which may be updated
669 * before the software timestamp is received, a hardware TX timestamp may be
670 * returned only if there is no software TX timestamp. Ignore false software
671 * timestamps, which may be made in the __sock_recv_timestamp() call when the
672 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
673 * hardware timestamp.
675 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
677 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
680 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
682 struct scm_ts_pktinfo ts_pktinfo;
683 struct net_device *orig_dev;
685 if (!skb_mac_header_was_set(skb))
688 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
691 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
693 ts_pktinfo.if_index = orig_dev->ifindex;
696 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
697 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
698 sizeof(ts_pktinfo), &ts_pktinfo);
702 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
704 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
707 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
708 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
709 struct scm_timestamping_internal tss;
711 int empty = 1, false_tstamp = 0;
712 struct skb_shared_hwtstamps *shhwtstamps =
715 /* Race occurred between timestamp enabling and packet
716 receiving. Fill in the current time for now. */
717 if (need_software_tstamp && skb->tstamp == 0) {
718 __net_timestamp(skb);
722 if (need_software_tstamp) {
723 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
725 struct __kernel_sock_timeval tv;
727 skb_get_new_timestamp(skb, &tv);
728 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
731 struct __kernel_old_timeval tv;
733 skb_get_timestamp(skb, &tv);
734 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
739 struct __kernel_timespec ts;
741 skb_get_new_timestampns(skb, &ts);
742 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
747 skb_get_timestampns(skb, &ts);
748 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
754 memset(&tss, 0, sizeof(tss));
755 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
756 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
759 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
760 !skb_is_swtx_tstamp(skb, false_tstamp) &&
761 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
763 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
764 !skb_is_err_queue(skb))
765 put_ts_pktinfo(msg, skb);
768 if (sock_flag(sk, SOCK_TSTAMP_NEW))
769 put_cmsg_scm_timestamping64(msg, &tss);
771 put_cmsg_scm_timestamping(msg, &tss);
773 if (skb_is_err_queue(skb) && skb->len &&
774 SKB_EXT_ERR(skb)->opt_stats)
775 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
776 skb->len, skb->data);
779 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
781 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
786 if (!sock_flag(sk, SOCK_WIFI_STATUS))
788 if (!skb->wifi_acked_valid)
791 ack = skb->wifi_acked;
793 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
795 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
797 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
800 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
801 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
802 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
805 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
808 sock_recv_timestamp(msg, sk, skb);
809 sock_recv_drops(msg, sk, skb);
811 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
813 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
816 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
819 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
821 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
823 return err ?: sock_recvmsg_nosec(sock, msg, flags);
825 EXPORT_SYMBOL(sock_recvmsg);
828 * kernel_recvmsg - Receive a message from a socket (kernel space)
829 * @sock: The socket to receive the message from
830 * @msg: Received message
831 * @vec: Input s/g array for message data
832 * @num: Size of input s/g array
833 * @size: Number of bytes to read
834 * @flags: Message flags (MSG_DONTWAIT, etc...)
836 * On return the msg structure contains the scatter/gather array passed in the
837 * vec argument. The array is modified so that it consists of the unfilled
838 * portion of the original array.
840 * The returned value is the total number of bytes received, or an error.
842 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
843 struct kvec *vec, size_t num, size_t size, int flags)
845 mm_segment_t oldfs = get_fs();
848 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
850 result = sock_recvmsg(sock, msg, flags);
854 EXPORT_SYMBOL(kernel_recvmsg);
856 static ssize_t sock_sendpage(struct file *file, struct page *page,
857 int offset, size_t size, loff_t *ppos, int more)
862 sock = file->private_data;
864 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
865 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
868 return kernel_sendpage(sock, page, offset, size, flags);
871 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
872 struct pipe_inode_info *pipe, size_t len,
875 struct socket *sock = file->private_data;
877 if (unlikely(!sock->ops->splice_read))
878 return generic_file_splice_read(file, ppos, pipe, len, flags);
880 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
883 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
885 struct file *file = iocb->ki_filp;
886 struct socket *sock = file->private_data;
887 struct msghdr msg = {.msg_iter = *to,
891 if (file->f_flags & O_NONBLOCK)
892 msg.msg_flags = MSG_DONTWAIT;
894 if (iocb->ki_pos != 0)
897 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
900 res = sock_recvmsg(sock, &msg, msg.msg_flags);
905 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
907 struct file *file = iocb->ki_filp;
908 struct socket *sock = file->private_data;
909 struct msghdr msg = {.msg_iter = *from,
913 if (iocb->ki_pos != 0)
916 if (file->f_flags & O_NONBLOCK)
917 msg.msg_flags = MSG_DONTWAIT;
919 if (sock->type == SOCK_SEQPACKET)
920 msg.msg_flags |= MSG_EOR;
922 res = sock_sendmsg(sock, &msg);
923 *from = msg.msg_iter;
928 * Atomic setting of ioctl hooks to avoid race
929 * with module unload.
932 static DEFINE_MUTEX(br_ioctl_mutex);
933 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
935 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
937 mutex_lock(&br_ioctl_mutex);
938 br_ioctl_hook = hook;
939 mutex_unlock(&br_ioctl_mutex);
941 EXPORT_SYMBOL(brioctl_set);
943 static DEFINE_MUTEX(vlan_ioctl_mutex);
944 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
946 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
948 mutex_lock(&vlan_ioctl_mutex);
949 vlan_ioctl_hook = hook;
950 mutex_unlock(&vlan_ioctl_mutex);
952 EXPORT_SYMBOL(vlan_ioctl_set);
954 static DEFINE_MUTEX(dlci_ioctl_mutex);
955 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
957 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
959 mutex_lock(&dlci_ioctl_mutex);
960 dlci_ioctl_hook = hook;
961 mutex_unlock(&dlci_ioctl_mutex);
963 EXPORT_SYMBOL(dlci_ioctl_set);
965 static long sock_do_ioctl(struct net *net, struct socket *sock,
966 unsigned int cmd, unsigned long arg)
969 void __user *argp = (void __user *)arg;
971 err = sock->ops->ioctl(sock, cmd, arg);
974 * If this ioctl is unknown try to hand it down
977 if (err != -ENOIOCTLCMD)
980 if (cmd == SIOCGIFCONF) {
982 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
985 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
987 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
992 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
994 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
995 if (!err && need_copyout)
996 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1003 * With an ioctl, arg may well be a user mode pointer, but we don't know
1004 * what to do with it - that's up to the protocol still.
1007 struct ns_common *get_net_ns(struct ns_common *ns)
1009 return &get_net(container_of(ns, struct net, ns))->ns;
1011 EXPORT_SYMBOL_GPL(get_net_ns);
1013 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1015 struct socket *sock;
1017 void __user *argp = (void __user *)arg;
1021 sock = file->private_data;
1024 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1027 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1029 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1030 if (!err && need_copyout)
1031 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1034 #ifdef CONFIG_WEXT_CORE
1035 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1036 err = wext_handle_ioctl(net, cmd, argp);
1043 if (get_user(pid, (int __user *)argp))
1045 err = f_setown(sock->file, pid, 1);
1049 err = put_user(f_getown(sock->file),
1050 (int __user *)argp);
1058 request_module("bridge");
1060 mutex_lock(&br_ioctl_mutex);
1062 err = br_ioctl_hook(net, cmd, argp);
1063 mutex_unlock(&br_ioctl_mutex);
1068 if (!vlan_ioctl_hook)
1069 request_module("8021q");
1071 mutex_lock(&vlan_ioctl_mutex);
1072 if (vlan_ioctl_hook)
1073 err = vlan_ioctl_hook(net, argp);
1074 mutex_unlock(&vlan_ioctl_mutex);
1079 if (!dlci_ioctl_hook)
1080 request_module("dlci");
1082 mutex_lock(&dlci_ioctl_mutex);
1083 if (dlci_ioctl_hook)
1084 err = dlci_ioctl_hook(cmd, argp);
1085 mutex_unlock(&dlci_ioctl_mutex);
1089 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1092 err = open_related_ns(&net->ns, get_net_ns);
1095 err = sock_do_ioctl(net, sock, cmd, arg);
1101 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1104 struct socket *sock = NULL;
1106 err = security_socket_create(family, type, protocol, 1);
1110 sock = sock_alloc();
1117 err = security_socket_post_create(sock, family, type, protocol, 1);
1129 EXPORT_SYMBOL(sock_create_lite);
1131 /* No kernel lock held - perfect */
1132 static __poll_t sock_poll(struct file *file, poll_table *wait)
1134 struct socket *sock = file->private_data;
1135 __poll_t events = poll_requested_events(wait), flag = 0;
1137 if (!sock->ops->poll)
1140 if (sk_can_busy_loop(sock->sk)) {
1141 /* poll once if requested by the syscall */
1142 if (events & POLL_BUSY_LOOP)
1143 sk_busy_loop(sock->sk, 1);
1145 /* if this socket can poll_ll, tell the system call */
1146 flag = POLL_BUSY_LOOP;
1149 return sock->ops->poll(file, sock, wait) | flag;
1152 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1154 struct socket *sock = file->private_data;
1156 return sock->ops->mmap(file, sock, vma);
1159 static int sock_close(struct inode *inode, struct file *filp)
1161 __sock_release(SOCKET_I(inode), inode);
1166 * Update the socket async list
1168 * Fasync_list locking strategy.
1170 * 1. fasync_list is modified only under process context socket lock
1171 * i.e. under semaphore.
1172 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1173 * or under socket lock
1176 static int sock_fasync(int fd, struct file *filp, int on)
1178 struct socket *sock = filp->private_data;
1179 struct sock *sk = sock->sk;
1180 struct socket_wq *wq;
1187 fasync_helper(fd, filp, on, &wq->fasync_list);
1189 if (!wq->fasync_list)
1190 sock_reset_flag(sk, SOCK_FASYNC);
1192 sock_set_flag(sk, SOCK_FASYNC);
1198 /* This function may be called only under rcu_lock */
1200 int sock_wake_async(struct socket_wq *wq, int how, int band)
1202 if (!wq || !wq->fasync_list)
1206 case SOCK_WAKE_WAITD:
1207 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1210 case SOCK_WAKE_SPACE:
1211 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1216 kill_fasync(&wq->fasync_list, SIGIO, band);
1219 kill_fasync(&wq->fasync_list, SIGURG, band);
1224 EXPORT_SYMBOL(sock_wake_async);
1226 int __sock_create(struct net *net, int family, int type, int protocol,
1227 struct socket **res, int kern)
1230 struct socket *sock;
1231 const struct net_proto_family *pf;
1234 * Check protocol is in range
1236 if (family < 0 || family >= NPROTO)
1237 return -EAFNOSUPPORT;
1238 if (type < 0 || type >= SOCK_MAX)
1243 This uglymoron is moved from INET layer to here to avoid
1244 deadlock in module load.
1246 if (family == PF_INET && type == SOCK_PACKET) {
1247 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1252 err = security_socket_create(family, type, protocol, kern);
1257 * Allocate the socket and allow the family to set things up. if
1258 * the protocol is 0, the family is instructed to select an appropriate
1261 sock = sock_alloc();
1263 net_warn_ratelimited("socket: no more sockets\n");
1264 return -ENFILE; /* Not exactly a match, but its the
1265 closest posix thing */
1270 #ifdef CONFIG_MODULES
1271 /* Attempt to load a protocol module if the find failed.
1273 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1274 * requested real, full-featured networking support upon configuration.
1275 * Otherwise module support will break!
1277 if (rcu_access_pointer(net_families[family]) == NULL)
1278 request_module("net-pf-%d", family);
1282 pf = rcu_dereference(net_families[family]);
1283 err = -EAFNOSUPPORT;
1288 * We will call the ->create function, that possibly is in a loadable
1289 * module, so we have to bump that loadable module refcnt first.
1291 if (!try_module_get(pf->owner))
1294 /* Now protected by module ref count */
1297 err = pf->create(net, sock, protocol, kern);
1299 goto out_module_put;
1302 * Now to bump the refcnt of the [loadable] module that owns this
1303 * socket at sock_release time we decrement its refcnt.
1305 if (!try_module_get(sock->ops->owner))
1306 goto out_module_busy;
1309 * Now that we're done with the ->create function, the [loadable]
1310 * module can have its refcnt decremented
1312 module_put(pf->owner);
1313 err = security_socket_post_create(sock, family, type, protocol, kern);
1315 goto out_sock_release;
1321 err = -EAFNOSUPPORT;
1324 module_put(pf->owner);
1331 goto out_sock_release;
1333 EXPORT_SYMBOL(__sock_create);
1335 int sock_create(int family, int type, int protocol, struct socket **res)
1337 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1339 EXPORT_SYMBOL(sock_create);
1341 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1343 return __sock_create(net, family, type, protocol, res, 1);
1345 EXPORT_SYMBOL(sock_create_kern);
1347 int __sys_socket(int family, int type, int protocol)
1350 struct socket *sock;
1353 /* Check the SOCK_* constants for consistency. */
1354 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1355 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1356 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1357 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1359 flags = type & ~SOCK_TYPE_MASK;
1360 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1362 type &= SOCK_TYPE_MASK;
1364 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1365 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1367 retval = sock_create(family, type, protocol, &sock);
1371 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1374 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1376 return __sys_socket(family, type, protocol);
1380 * Create a pair of connected sockets.
1383 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1385 struct socket *sock1, *sock2;
1387 struct file *newfile1, *newfile2;
1390 flags = type & ~SOCK_TYPE_MASK;
1391 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1393 type &= SOCK_TYPE_MASK;
1395 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1396 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1399 * reserve descriptors and make sure we won't fail
1400 * to return them to userland.
1402 fd1 = get_unused_fd_flags(flags);
1403 if (unlikely(fd1 < 0))
1406 fd2 = get_unused_fd_flags(flags);
1407 if (unlikely(fd2 < 0)) {
1412 err = put_user(fd1, &usockvec[0]);
1416 err = put_user(fd2, &usockvec[1]);
1421 * Obtain the first socket and check if the underlying protocol
1422 * supports the socketpair call.
1425 err = sock_create(family, type, protocol, &sock1);
1426 if (unlikely(err < 0))
1429 err = sock_create(family, type, protocol, &sock2);
1430 if (unlikely(err < 0)) {
1431 sock_release(sock1);
1435 err = security_socket_socketpair(sock1, sock2);
1436 if (unlikely(err)) {
1437 sock_release(sock2);
1438 sock_release(sock1);
1442 err = sock1->ops->socketpair(sock1, sock2);
1443 if (unlikely(err < 0)) {
1444 sock_release(sock2);
1445 sock_release(sock1);
1449 newfile1 = sock_alloc_file(sock1, flags, NULL);
1450 if (IS_ERR(newfile1)) {
1451 err = PTR_ERR(newfile1);
1452 sock_release(sock2);
1456 newfile2 = sock_alloc_file(sock2, flags, NULL);
1457 if (IS_ERR(newfile2)) {
1458 err = PTR_ERR(newfile2);
1463 audit_fd_pair(fd1, fd2);
1465 fd_install(fd1, newfile1);
1466 fd_install(fd2, newfile2);
1475 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1476 int __user *, usockvec)
1478 return __sys_socketpair(family, type, protocol, usockvec);
1482 * Bind a name to a socket. Nothing much to do here since it's
1483 * the protocol's responsibility to handle the local address.
1485 * We move the socket address to kernel space before we call
1486 * the protocol layer (having also checked the address is ok).
1489 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1491 struct socket *sock;
1492 struct sockaddr_storage address;
1493 int err, fput_needed;
1495 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1497 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1499 err = security_socket_bind(sock,
1500 (struct sockaddr *)&address,
1503 err = sock->ops->bind(sock,
1507 fput_light(sock->file, fput_needed);
1512 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1514 return __sys_bind(fd, umyaddr, addrlen);
1518 * Perform a listen. Basically, we allow the protocol to do anything
1519 * necessary for a listen, and if that works, we mark the socket as
1520 * ready for listening.
1523 int __sys_listen(int fd, int backlog)
1525 struct socket *sock;
1526 int err, fput_needed;
1529 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1531 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1532 if ((unsigned int)backlog > somaxconn)
1533 backlog = somaxconn;
1535 err = security_socket_listen(sock, backlog);
1537 err = sock->ops->listen(sock, backlog);
1539 fput_light(sock->file, fput_needed);
1544 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1546 return __sys_listen(fd, backlog);
1550 * For accept, we attempt to create a new socket, set up the link
1551 * with the client, wake up the client, then return the new
1552 * connected fd. We collect the address of the connector in kernel
1553 * space and move it to user at the very end. This is unclean because
1554 * we open the socket then return an error.
1556 * 1003.1g adds the ability to recvmsg() to query connection pending
1557 * status to recvmsg. We need to add that support in a way thats
1558 * clean when we restructure accept also.
1561 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1562 int __user *upeer_addrlen, int flags)
1564 struct socket *sock, *newsock;
1565 struct file *newfile;
1566 int err, len, newfd, fput_needed;
1567 struct sockaddr_storage address;
1569 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1572 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1573 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1575 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1580 newsock = sock_alloc();
1584 newsock->type = sock->type;
1585 newsock->ops = sock->ops;
1588 * We don't need try_module_get here, as the listening socket (sock)
1589 * has the protocol module (sock->ops->owner) held.
1591 __module_get(newsock->ops->owner);
1593 newfd = get_unused_fd_flags(flags);
1594 if (unlikely(newfd < 0)) {
1596 sock_release(newsock);
1599 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1600 if (IS_ERR(newfile)) {
1601 err = PTR_ERR(newfile);
1602 put_unused_fd(newfd);
1606 err = security_socket_accept(sock, newsock);
1610 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1614 if (upeer_sockaddr) {
1615 len = newsock->ops->getname(newsock,
1616 (struct sockaddr *)&address, 2);
1618 err = -ECONNABORTED;
1621 err = move_addr_to_user(&address,
1622 len, upeer_sockaddr, upeer_addrlen);
1627 /* File flags are not inherited via accept() unlike another OSes. */
1629 fd_install(newfd, newfile);
1633 fput_light(sock->file, fput_needed);
1638 put_unused_fd(newfd);
1642 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1643 int __user *, upeer_addrlen, int, flags)
1645 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1648 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1649 int __user *, upeer_addrlen)
1651 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1655 * Attempt to connect to a socket with the server address. The address
1656 * is in user space so we verify it is OK and move it to kernel space.
1658 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1661 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1662 * other SEQPACKET protocols that take time to connect() as it doesn't
1663 * include the -EINPROGRESS status for such sockets.
1666 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1668 struct socket *sock;
1669 struct sockaddr_storage address;
1670 int err, fput_needed;
1672 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1675 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1680 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1684 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1685 sock->file->f_flags);
1687 fput_light(sock->file, fput_needed);
1692 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1695 return __sys_connect(fd, uservaddr, addrlen);
1699 * Get the local address ('name') of a socket object. Move the obtained
1700 * name to user space.
1703 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1704 int __user *usockaddr_len)
1706 struct socket *sock;
1707 struct sockaddr_storage address;
1708 int err, fput_needed;
1710 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1714 err = security_socket_getsockname(sock);
1718 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1721 /* "err" is actually length in this case */
1722 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1725 fput_light(sock->file, fput_needed);
1730 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1731 int __user *, usockaddr_len)
1733 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1737 * Get the remote address ('name') of a socket object. Move the obtained
1738 * name to user space.
1741 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1742 int __user *usockaddr_len)
1744 struct socket *sock;
1745 struct sockaddr_storage address;
1746 int err, fput_needed;
1748 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1750 err = security_socket_getpeername(sock);
1752 fput_light(sock->file, fput_needed);
1756 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1758 /* "err" is actually length in this case */
1759 err = move_addr_to_user(&address, err, usockaddr,
1761 fput_light(sock->file, fput_needed);
1766 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1767 int __user *, usockaddr_len)
1769 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1773 * Send a datagram to a given address. We move the address into kernel
1774 * space and check the user space data area is readable before invoking
1777 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1778 struct sockaddr __user *addr, int addr_len)
1780 struct socket *sock;
1781 struct sockaddr_storage address;
1787 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1790 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1794 msg.msg_name = NULL;
1795 msg.msg_control = NULL;
1796 msg.msg_controllen = 0;
1797 msg.msg_namelen = 0;
1799 err = move_addr_to_kernel(addr, addr_len, &address);
1802 msg.msg_name = (struct sockaddr *)&address;
1803 msg.msg_namelen = addr_len;
1805 if (sock->file->f_flags & O_NONBLOCK)
1806 flags |= MSG_DONTWAIT;
1807 msg.msg_flags = flags;
1808 err = sock_sendmsg(sock, &msg);
1811 fput_light(sock->file, fput_needed);
1816 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1817 unsigned int, flags, struct sockaddr __user *, addr,
1820 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1824 * Send a datagram down a socket.
1827 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1828 unsigned int, flags)
1830 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1834 * Receive a frame from the socket and optionally record the address of the
1835 * sender. We verify the buffers are writable and if needed move the
1836 * sender address from kernel to user space.
1838 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1839 struct sockaddr __user *addr, int __user *addr_len)
1841 struct socket *sock;
1844 struct sockaddr_storage address;
1848 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1851 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1855 msg.msg_control = NULL;
1856 msg.msg_controllen = 0;
1857 /* Save some cycles and don't copy the address if not needed */
1858 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1859 /* We assume all kernel code knows the size of sockaddr_storage */
1860 msg.msg_namelen = 0;
1861 msg.msg_iocb = NULL;
1863 if (sock->file->f_flags & O_NONBLOCK)
1864 flags |= MSG_DONTWAIT;
1865 err = sock_recvmsg(sock, &msg, flags);
1867 if (err >= 0 && addr != NULL) {
1868 err2 = move_addr_to_user(&address,
1869 msg.msg_namelen, addr, addr_len);
1874 fput_light(sock->file, fput_needed);
1879 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1880 unsigned int, flags, struct sockaddr __user *, addr,
1881 int __user *, addr_len)
1883 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1887 * Receive a datagram from a socket.
1890 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1891 unsigned int, flags)
1893 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1897 * Set a socket option. Because we don't know the option lengths we have
1898 * to pass the user mode parameter for the protocols to sort out.
1901 static int __sys_setsockopt(int fd, int level, int optname,
1902 char __user *optval, int optlen)
1904 int err, fput_needed;
1905 struct socket *sock;
1910 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1912 err = security_socket_setsockopt(sock, level, optname);
1916 if (level == SOL_SOCKET)
1918 sock_setsockopt(sock, level, optname, optval,
1922 sock->ops->setsockopt(sock, level, optname, optval,
1925 fput_light(sock->file, fput_needed);
1930 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1931 char __user *, optval, int, optlen)
1933 return __sys_setsockopt(fd, level, optname, optval, optlen);
1937 * Get a socket option. Because we don't know the option lengths we have
1938 * to pass a user mode parameter for the protocols to sort out.
1941 static int __sys_getsockopt(int fd, int level, int optname,
1942 char __user *optval, int __user *optlen)
1944 int err, fput_needed;
1945 struct socket *sock;
1947 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1949 err = security_socket_getsockopt(sock, level, optname);
1953 if (level == SOL_SOCKET)
1955 sock_getsockopt(sock, level, optname, optval,
1959 sock->ops->getsockopt(sock, level, optname, optval,
1962 fput_light(sock->file, fput_needed);
1967 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1968 char __user *, optval, int __user *, optlen)
1970 return __sys_getsockopt(fd, level, optname, optval, optlen);
1974 * Shutdown a socket.
1977 int __sys_shutdown(int fd, int how)
1979 int err, fput_needed;
1980 struct socket *sock;
1982 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1984 err = security_socket_shutdown(sock, how);
1986 err = sock->ops->shutdown(sock, how);
1987 fput_light(sock->file, fput_needed);
1992 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1994 return __sys_shutdown(fd, how);
1997 /* A couple of helpful macros for getting the address of the 32/64 bit
1998 * fields which are the same type (int / unsigned) on our platforms.
2000 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2001 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2002 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2004 struct used_address {
2005 struct sockaddr_storage name;
2006 unsigned int name_len;
2009 static int copy_msghdr_from_user(struct msghdr *kmsg,
2010 struct user_msghdr __user *umsg,
2011 struct sockaddr __user **save_addr,
2014 struct user_msghdr msg;
2017 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2020 kmsg->msg_control = (void __force *)msg.msg_control;
2021 kmsg->msg_controllen = msg.msg_controllen;
2022 kmsg->msg_flags = msg.msg_flags;
2024 kmsg->msg_namelen = msg.msg_namelen;
2026 kmsg->msg_namelen = 0;
2028 if (kmsg->msg_namelen < 0)
2031 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2032 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2035 *save_addr = msg.msg_name;
2037 if (msg.msg_name && kmsg->msg_namelen) {
2039 err = move_addr_to_kernel(msg.msg_name,
2046 kmsg->msg_name = NULL;
2047 kmsg->msg_namelen = 0;
2050 if (msg.msg_iovlen > UIO_MAXIOV)
2053 kmsg->msg_iocb = NULL;
2055 return import_iovec(save_addr ? READ : WRITE,
2056 msg.msg_iov, msg.msg_iovlen,
2057 UIO_FASTIOV, iov, &kmsg->msg_iter);
2060 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2061 struct msghdr *msg_sys, unsigned int flags,
2062 struct used_address *used_address,
2063 unsigned int allowed_msghdr_flags)
2065 struct compat_msghdr __user *msg_compat =
2066 (struct compat_msghdr __user *)msg;
2067 struct sockaddr_storage address;
2068 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2069 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2070 __aligned(sizeof(__kernel_size_t));
2071 /* 20 is size of ipv6_pktinfo */
2072 unsigned char *ctl_buf = ctl;
2076 msg_sys->msg_name = &address;
2078 if (MSG_CMSG_COMPAT & flags)
2079 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2081 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2087 if (msg_sys->msg_controllen > INT_MAX)
2089 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2090 ctl_len = msg_sys->msg_controllen;
2091 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2093 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2097 ctl_buf = msg_sys->msg_control;
2098 ctl_len = msg_sys->msg_controllen;
2099 } else if (ctl_len) {
2100 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2101 CMSG_ALIGN(sizeof(struct cmsghdr)));
2102 if (ctl_len > sizeof(ctl)) {
2103 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2104 if (ctl_buf == NULL)
2109 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2110 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2111 * checking falls down on this.
2113 if (copy_from_user(ctl_buf,
2114 (void __user __force *)msg_sys->msg_control,
2117 msg_sys->msg_control = ctl_buf;
2119 msg_sys->msg_flags = flags;
2121 if (sock->file->f_flags & O_NONBLOCK)
2122 msg_sys->msg_flags |= MSG_DONTWAIT;
2124 * If this is sendmmsg() and current destination address is same as
2125 * previously succeeded address, omit asking LSM's decision.
2126 * used_address->name_len is initialized to UINT_MAX so that the first
2127 * destination address never matches.
2129 if (used_address && msg_sys->msg_name &&
2130 used_address->name_len == msg_sys->msg_namelen &&
2131 !memcmp(&used_address->name, msg_sys->msg_name,
2132 used_address->name_len)) {
2133 err = sock_sendmsg_nosec(sock, msg_sys);
2136 err = sock_sendmsg(sock, msg_sys);
2138 * If this is sendmmsg() and sending to current destination address was
2139 * successful, remember it.
2141 if (used_address && err >= 0) {
2142 used_address->name_len = msg_sys->msg_namelen;
2143 if (msg_sys->msg_name)
2144 memcpy(&used_address->name, msg_sys->msg_name,
2145 used_address->name_len);
2150 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2157 * BSD sendmsg interface
2160 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2161 bool forbid_cmsg_compat)
2163 int fput_needed, err;
2164 struct msghdr msg_sys;
2165 struct socket *sock;
2167 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2170 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2174 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2176 fput_light(sock->file, fput_needed);
2181 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2183 return __sys_sendmsg(fd, msg, flags, true);
2187 * Linux sendmmsg interface
2190 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2191 unsigned int flags, bool forbid_cmsg_compat)
2193 int fput_needed, err, datagrams;
2194 struct socket *sock;
2195 struct mmsghdr __user *entry;
2196 struct compat_mmsghdr __user *compat_entry;
2197 struct msghdr msg_sys;
2198 struct used_address used_address;
2199 unsigned int oflags = flags;
2201 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2204 if (vlen > UIO_MAXIOV)
2209 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2213 used_address.name_len = UINT_MAX;
2215 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2219 while (datagrams < vlen) {
2220 if (datagrams == vlen - 1)
2223 if (MSG_CMSG_COMPAT & flags) {
2224 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2225 &msg_sys, flags, &used_address, MSG_EOR);
2228 err = __put_user(err, &compat_entry->msg_len);
2231 err = ___sys_sendmsg(sock,
2232 (struct user_msghdr __user *)entry,
2233 &msg_sys, flags, &used_address, MSG_EOR);
2236 err = put_user(err, &entry->msg_len);
2243 if (msg_data_left(&msg_sys))
2248 fput_light(sock->file, fput_needed);
2250 /* We only return an error if no datagrams were able to be sent */
2257 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2258 unsigned int, vlen, unsigned int, flags)
2260 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2263 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2264 struct msghdr *msg_sys, unsigned int flags, int nosec)
2266 struct compat_msghdr __user *msg_compat =
2267 (struct compat_msghdr __user *)msg;
2268 struct iovec iovstack[UIO_FASTIOV];
2269 struct iovec *iov = iovstack;
2270 unsigned long cmsg_ptr;
2274 /* kernel mode address */
2275 struct sockaddr_storage addr;
2277 /* user mode address pointers */
2278 struct sockaddr __user *uaddr;
2279 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2281 msg_sys->msg_name = &addr;
2283 if (MSG_CMSG_COMPAT & flags)
2284 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2286 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2290 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2291 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2293 /* We assume all kernel code knows the size of sockaddr_storage */
2294 msg_sys->msg_namelen = 0;
2296 if (sock->file->f_flags & O_NONBLOCK)
2297 flags |= MSG_DONTWAIT;
2298 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2303 if (uaddr != NULL) {
2304 err = move_addr_to_user(&addr,
2305 msg_sys->msg_namelen, uaddr,
2310 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2314 if (MSG_CMSG_COMPAT & flags)
2315 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2316 &msg_compat->msg_controllen);
2318 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2319 &msg->msg_controllen);
2330 * BSD recvmsg interface
2333 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2334 bool forbid_cmsg_compat)
2336 int fput_needed, err;
2337 struct msghdr msg_sys;
2338 struct socket *sock;
2340 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2343 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2347 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2349 fput_light(sock->file, fput_needed);
2354 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2355 unsigned int, flags)
2357 return __sys_recvmsg(fd, msg, flags, true);
2361 * Linux recvmmsg interface
2364 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2365 unsigned int vlen, unsigned int flags,
2366 struct timespec64 *timeout)
2368 int fput_needed, err, datagrams;
2369 struct socket *sock;
2370 struct mmsghdr __user *entry;
2371 struct compat_mmsghdr __user *compat_entry;
2372 struct msghdr msg_sys;
2373 struct timespec64 end_time;
2374 struct timespec64 timeout64;
2377 poll_select_set_timeout(&end_time, timeout->tv_sec,
2383 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2387 if (likely(!(flags & MSG_ERRQUEUE))) {
2388 err = sock_error(sock->sk);
2396 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2398 while (datagrams < vlen) {
2400 * No need to ask LSM for more than the first datagram.
2402 if (MSG_CMSG_COMPAT & flags) {
2403 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2404 &msg_sys, flags & ~MSG_WAITFORONE,
2408 err = __put_user(err, &compat_entry->msg_len);
2411 err = ___sys_recvmsg(sock,
2412 (struct user_msghdr __user *)entry,
2413 &msg_sys, flags & ~MSG_WAITFORONE,
2417 err = put_user(err, &entry->msg_len);
2425 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2426 if (flags & MSG_WAITFORONE)
2427 flags |= MSG_DONTWAIT;
2430 ktime_get_ts64(&timeout64);
2431 *timeout = timespec64_sub(end_time, timeout64);
2432 if (timeout->tv_sec < 0) {
2433 timeout->tv_sec = timeout->tv_nsec = 0;
2437 /* Timeout, return less than vlen datagrams */
2438 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2442 /* Out of band data, return right away */
2443 if (msg_sys.msg_flags & MSG_OOB)
2451 if (datagrams == 0) {
2457 * We may return less entries than requested (vlen) if the
2458 * sock is non block and there aren't enough datagrams...
2460 if (err != -EAGAIN) {
2462 * ... or if recvmsg returns an error after we
2463 * received some datagrams, where we record the
2464 * error to return on the next call or if the
2465 * app asks about it using getsockopt(SO_ERROR).
2467 sock->sk->sk_err = -err;
2470 fput_light(sock->file, fput_needed);
2475 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2476 unsigned int vlen, unsigned int flags,
2477 struct __kernel_timespec __user *timeout,
2478 struct old_timespec32 __user *timeout32)
2481 struct timespec64 timeout_sys;
2483 if (timeout && get_timespec64(&timeout_sys, timeout))
2486 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2489 if (!timeout && !timeout32)
2490 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2492 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2497 if (timeout && put_timespec64(&timeout_sys, timeout))
2498 datagrams = -EFAULT;
2500 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2501 datagrams = -EFAULT;
2506 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2507 unsigned int, vlen, unsigned int, flags,
2508 struct __kernel_timespec __user *, timeout)
2510 if (flags & MSG_CMSG_COMPAT)
2513 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2516 #ifdef CONFIG_COMPAT_32BIT_TIME
2517 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2518 unsigned int, vlen, unsigned int, flags,
2519 struct old_timespec32 __user *, timeout)
2521 if (flags & MSG_CMSG_COMPAT)
2524 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2528 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2529 /* Argument list sizes for sys_socketcall */
2530 #define AL(x) ((x) * sizeof(unsigned long))
2531 static const unsigned char nargs[21] = {
2532 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2533 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2534 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2541 * System call vectors.
2543 * Argument checking cleaned up. Saved 20% in size.
2544 * This function doesn't need to set the kernel lock because
2545 * it is set by the callees.
2548 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2550 unsigned long a[AUDITSC_ARGS];
2551 unsigned long a0, a1;
2555 if (call < 1 || call > SYS_SENDMMSG)
2557 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2560 if (len > sizeof(a))
2563 /* copy_from_user should be SMP safe. */
2564 if (copy_from_user(a, args, len))
2567 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2576 err = __sys_socket(a0, a1, a[2]);
2579 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2582 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2585 err = __sys_listen(a0, a1);
2588 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2589 (int __user *)a[2], 0);
2591 case SYS_GETSOCKNAME:
2593 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2594 (int __user *)a[2]);
2596 case SYS_GETPEERNAME:
2598 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2599 (int __user *)a[2]);
2601 case SYS_SOCKETPAIR:
2602 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2605 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2609 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2610 (struct sockaddr __user *)a[4], a[5]);
2613 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2617 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2618 (struct sockaddr __user *)a[4],
2619 (int __user *)a[5]);
2622 err = __sys_shutdown(a0, a1);
2624 case SYS_SETSOCKOPT:
2625 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2628 case SYS_GETSOCKOPT:
2630 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2631 (int __user *)a[4]);
2634 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2638 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2642 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2646 if (IS_ENABLED(CONFIG_64BIT) || !IS_ENABLED(CONFIG_64BIT_TIME))
2647 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2649 (struct __kernel_timespec __user *)a[4],
2652 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2654 (struct old_timespec32 __user *)a[4]);
2657 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2658 (int __user *)a[2], a[3]);
2667 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2670 * sock_register - add a socket protocol handler
2671 * @ops: description of protocol
2673 * This function is called by a protocol handler that wants to
2674 * advertise its address family, and have it linked into the
2675 * socket interface. The value ops->family corresponds to the
2676 * socket system call protocol family.
2678 int sock_register(const struct net_proto_family *ops)
2682 if (ops->family >= NPROTO) {
2683 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2687 spin_lock(&net_family_lock);
2688 if (rcu_dereference_protected(net_families[ops->family],
2689 lockdep_is_held(&net_family_lock)))
2692 rcu_assign_pointer(net_families[ops->family], ops);
2695 spin_unlock(&net_family_lock);
2697 pr_info("NET: Registered protocol family %d\n", ops->family);
2700 EXPORT_SYMBOL(sock_register);
2703 * sock_unregister - remove a protocol handler
2704 * @family: protocol family to remove
2706 * This function is called by a protocol handler that wants to
2707 * remove its address family, and have it unlinked from the
2708 * new socket creation.
2710 * If protocol handler is a module, then it can use module reference
2711 * counts to protect against new references. If protocol handler is not
2712 * a module then it needs to provide its own protection in
2713 * the ops->create routine.
2715 void sock_unregister(int family)
2717 BUG_ON(family < 0 || family >= NPROTO);
2719 spin_lock(&net_family_lock);
2720 RCU_INIT_POINTER(net_families[family], NULL);
2721 spin_unlock(&net_family_lock);
2725 pr_info("NET: Unregistered protocol family %d\n", family);
2727 EXPORT_SYMBOL(sock_unregister);
2729 bool sock_is_registered(int family)
2731 return family < NPROTO && rcu_access_pointer(net_families[family]);
2734 static int __init sock_init(void)
2738 * Initialize the network sysctl infrastructure.
2740 err = net_sysctl_init();
2745 * Initialize skbuff SLAB cache
2750 * Initialize the protocols module.
2755 err = register_filesystem(&sock_fs_type);
2758 sock_mnt = kern_mount(&sock_fs_type);
2759 if (IS_ERR(sock_mnt)) {
2760 err = PTR_ERR(sock_mnt);
2764 /* The real protocol initialization is performed in later initcalls.
2767 #ifdef CONFIG_NETFILTER
2768 err = netfilter_init();
2773 ptp_classifier_init();
2779 unregister_filesystem(&sock_fs_type);
2784 core_initcall(sock_init); /* early initcall */
2786 #ifdef CONFIG_PROC_FS
2787 void socket_seq_show(struct seq_file *seq)
2789 seq_printf(seq, "sockets: used %d\n",
2790 sock_inuse_get(seq->private));
2792 #endif /* CONFIG_PROC_FS */
2794 #ifdef CONFIG_COMPAT
2795 static int do_siocgstamp(struct net *net, struct socket *sock,
2796 unsigned int cmd, void __user *up)
2798 mm_segment_t old_fs = get_fs();
2803 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2806 err = compat_put_timeval(&ktv, up);
2811 static int do_siocgstampns(struct net *net, struct socket *sock,
2812 unsigned int cmd, void __user *up)
2814 mm_segment_t old_fs = get_fs();
2815 struct timespec kts;
2819 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2822 err = compat_put_timespec(&kts, up);
2827 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2829 struct compat_ifconf ifc32;
2833 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2836 ifc.ifc_len = ifc32.ifc_len;
2837 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2840 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2845 ifc32.ifc_len = ifc.ifc_len;
2846 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2852 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2854 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2855 bool convert_in = false, convert_out = false;
2856 size_t buf_size = 0;
2857 struct ethtool_rxnfc __user *rxnfc = NULL;
2859 u32 rule_cnt = 0, actual_rule_cnt;
2864 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2867 compat_rxnfc = compat_ptr(data);
2869 if (get_user(ethcmd, &compat_rxnfc->cmd))
2872 /* Most ethtool structures are defined without padding.
2873 * Unfortunately struct ethtool_rxnfc is an exception.
2878 case ETHTOOL_GRXCLSRLALL:
2879 /* Buffer size is variable */
2880 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2882 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2884 buf_size += rule_cnt * sizeof(u32);
2886 case ETHTOOL_GRXRINGS:
2887 case ETHTOOL_GRXCLSRLCNT:
2888 case ETHTOOL_GRXCLSRULE:
2889 case ETHTOOL_SRXCLSRLINS:
2892 case ETHTOOL_SRXCLSRLDEL:
2893 buf_size += sizeof(struct ethtool_rxnfc);
2895 rxnfc = compat_alloc_user_space(buf_size);
2899 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2902 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2905 /* We expect there to be holes between fs.m_ext and
2906 * fs.ring_cookie and at the end of fs, but nowhere else.
2908 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2909 sizeof(compat_rxnfc->fs.m_ext) !=
2910 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2911 sizeof(rxnfc->fs.m_ext));
2913 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2914 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2915 offsetof(struct ethtool_rxnfc, fs.location) -
2916 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2918 if (copy_in_user(rxnfc, compat_rxnfc,
2919 (void __user *)(&rxnfc->fs.m_ext + 1) -
2920 (void __user *)rxnfc) ||
2921 copy_in_user(&rxnfc->fs.ring_cookie,
2922 &compat_rxnfc->fs.ring_cookie,
2923 (void __user *)(&rxnfc->fs.location + 1) -
2924 (void __user *)&rxnfc->fs.ring_cookie))
2926 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2927 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2929 } else if (copy_in_user(&rxnfc->rule_cnt,
2930 &compat_rxnfc->rule_cnt,
2931 sizeof(rxnfc->rule_cnt)))
2935 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2940 if (copy_in_user(compat_rxnfc, rxnfc,
2941 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2942 (const void __user *)rxnfc) ||
2943 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2944 &rxnfc->fs.ring_cookie,
2945 (const void __user *)(&rxnfc->fs.location + 1) -
2946 (const void __user *)&rxnfc->fs.ring_cookie) ||
2947 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2948 sizeof(rxnfc->rule_cnt)))
2951 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2952 /* As an optimisation, we only copy the actual
2953 * number of rules that the underlying
2954 * function returned. Since Mallory might
2955 * change the rule count in user memory, we
2956 * check that it is less than the rule count
2957 * originally given (as the user buffer size),
2958 * which has been range-checked.
2960 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2962 if (actual_rule_cnt < rule_cnt)
2963 rule_cnt = actual_rule_cnt;
2964 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2965 &rxnfc->rule_locs[0],
2966 rule_cnt * sizeof(u32)))
2974 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2976 compat_uptr_t uptr32;
2981 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2984 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2987 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2988 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2990 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2992 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2993 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2999 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3000 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3001 struct compat_ifreq __user *u_ifreq32)
3006 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3008 if (get_user(data32, &u_ifreq32->ifr_data))
3010 ifreq.ifr_data = compat_ptr(data32);
3012 return dev_ioctl(net, cmd, &ifreq, NULL);
3015 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3017 struct compat_ifreq __user *uifr32)
3019 struct ifreq __user *uifr;
3022 /* Handle the fact that while struct ifreq has the same *layout* on
3023 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3024 * which are handled elsewhere, it still has different *size* due to
3025 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3026 * resulting in struct ifreq being 32 and 40 bytes respectively).
3027 * As a result, if the struct happens to be at the end of a page and
3028 * the next page isn't readable/writable, we get a fault. To prevent
3029 * that, copy back and forth to the full size.
3032 uifr = compat_alloc_user_space(sizeof(*uifr));
3033 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3036 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3047 case SIOCGIFBRDADDR:
3048 case SIOCGIFDSTADDR:
3049 case SIOCGIFNETMASK:
3055 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3063 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3064 struct compat_ifreq __user *uifr32)
3067 struct compat_ifmap __user *uifmap32;
3070 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3071 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3072 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3073 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3074 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3075 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3076 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3077 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3081 err = dev_ioctl(net, cmd, &ifr, NULL);
3083 if (cmd == SIOCGIFMAP && !err) {
3084 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3085 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3086 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3087 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3088 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3089 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3090 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3099 struct sockaddr rt_dst; /* target address */
3100 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3101 struct sockaddr rt_genmask; /* target network mask (IP) */
3102 unsigned short rt_flags;
3105 unsigned char rt_tos;
3106 unsigned char rt_class;
3108 short rt_metric; /* +1 for binary compatibility! */
3109 /* char * */ u32 rt_dev; /* forcing the device at add */
3110 u32 rt_mtu; /* per route MTU/Window */
3111 u32 rt_window; /* Window clamping */
3112 unsigned short rt_irtt; /* Initial RTT */
3115 struct in6_rtmsg32 {
3116 struct in6_addr rtmsg_dst;
3117 struct in6_addr rtmsg_src;
3118 struct in6_addr rtmsg_gateway;
3128 static int routing_ioctl(struct net *net, struct socket *sock,
3129 unsigned int cmd, void __user *argp)
3133 struct in6_rtmsg r6;
3137 mm_segment_t old_fs = get_fs();
3139 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3140 struct in6_rtmsg32 __user *ur6 = argp;
3141 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3142 3 * sizeof(struct in6_addr));
3143 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3144 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3145 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3146 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3147 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3148 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3149 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3153 struct rtentry32 __user *ur4 = argp;
3154 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3155 3 * sizeof(struct sockaddr));
3156 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3157 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3158 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3159 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3160 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3161 ret |= get_user(rtdev, &(ur4->rt_dev));
3163 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3164 r4.rt_dev = (char __user __force *)devname;
3178 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3185 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3186 * for some operations; this forces use of the newer bridge-utils that
3187 * use compatible ioctls
3189 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3193 if (get_user(tmp, argp))
3195 if (tmp == BRCTL_GET_VERSION)
3196 return BRCTL_VERSION + 1;
3200 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3201 unsigned int cmd, unsigned long arg)
3203 void __user *argp = compat_ptr(arg);
3204 struct sock *sk = sock->sk;
3205 struct net *net = sock_net(sk);
3207 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3208 return compat_ifr_data_ioctl(net, cmd, argp);
3213 return old_bridge_ioctl(argp);
3215 return compat_dev_ifconf(net, argp);
3217 return ethtool_ioctl(net, argp);
3219 return compat_siocwandev(net, argp);
3222 return compat_sioc_ifmap(net, cmd, argp);
3225 return routing_ioctl(net, sock, cmd, argp);
3227 return do_siocgstamp(net, sock, cmd, argp);
3229 return do_siocgstampns(net, sock, cmd, argp);
3230 case SIOCBONDSLAVEINFOQUERY:
3231 case SIOCBONDINFOQUERY:
3234 return compat_ifr_data_ioctl(net, cmd, argp);
3247 return sock_ioctl(file, cmd, arg);
3264 case SIOCSIFHWBROADCAST:
3266 case SIOCGIFBRDADDR:
3267 case SIOCSIFBRDADDR:
3268 case SIOCGIFDSTADDR:
3269 case SIOCSIFDSTADDR:
3270 case SIOCGIFNETMASK:
3271 case SIOCSIFNETMASK:
3283 case SIOCBONDENSLAVE:
3284 case SIOCBONDRELEASE:
3285 case SIOCBONDSETHWADDR:
3286 case SIOCBONDCHANGEACTIVE:
3287 return compat_ifreq_ioctl(net, sock, cmd, argp);
3293 return sock_do_ioctl(net, sock, cmd, arg);
3296 return -ENOIOCTLCMD;
3299 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3302 struct socket *sock = file->private_data;
3303 int ret = -ENOIOCTLCMD;
3310 if (sock->ops->compat_ioctl)
3311 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3313 if (ret == -ENOIOCTLCMD &&
3314 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3315 ret = compat_wext_handle_ioctl(net, cmd, arg);
3317 if (ret == -ENOIOCTLCMD)
3318 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3324 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3326 return sock->ops->bind(sock, addr, addrlen);
3328 EXPORT_SYMBOL(kernel_bind);
3330 int kernel_listen(struct socket *sock, int backlog)
3332 return sock->ops->listen(sock, backlog);
3334 EXPORT_SYMBOL(kernel_listen);
3336 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3338 struct sock *sk = sock->sk;
3341 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3346 err = sock->ops->accept(sock, *newsock, flags, true);
3348 sock_release(*newsock);
3353 (*newsock)->ops = sock->ops;
3354 __module_get((*newsock)->ops->owner);
3359 EXPORT_SYMBOL(kernel_accept);
3361 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3364 return sock->ops->connect(sock, addr, addrlen, flags);
3366 EXPORT_SYMBOL(kernel_connect);
3368 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3370 return sock->ops->getname(sock, addr, 0);
3372 EXPORT_SYMBOL(kernel_getsockname);
3374 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3376 return sock->ops->getname(sock, addr, 1);
3378 EXPORT_SYMBOL(kernel_getpeername);
3380 int kernel_getsockopt(struct socket *sock, int level, int optname,
3381 char *optval, int *optlen)
3383 mm_segment_t oldfs = get_fs();
3384 char __user *uoptval;
3385 int __user *uoptlen;
3388 uoptval = (char __user __force *) optval;
3389 uoptlen = (int __user __force *) optlen;
3392 if (level == SOL_SOCKET)
3393 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3395 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3400 EXPORT_SYMBOL(kernel_getsockopt);
3402 int kernel_setsockopt(struct socket *sock, int level, int optname,
3403 char *optval, unsigned int optlen)
3405 mm_segment_t oldfs = get_fs();
3406 char __user *uoptval;
3409 uoptval = (char __user __force *) optval;
3412 if (level == SOL_SOCKET)
3413 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3415 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3420 EXPORT_SYMBOL(kernel_setsockopt);
3422 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3423 size_t size, int flags)
3425 if (sock->ops->sendpage)
3426 return sock->ops->sendpage(sock, page, offset, size, flags);
3428 return sock_no_sendpage(sock, page, offset, size, flags);
3430 EXPORT_SYMBOL(kernel_sendpage);
3432 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3433 size_t size, int flags)
3435 struct socket *sock = sk->sk_socket;
3437 if (sock->ops->sendpage_locked)
3438 return sock->ops->sendpage_locked(sk, page, offset, size,
3441 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3443 EXPORT_SYMBOL(kernel_sendpage_locked);
3445 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3447 return sock->ops->shutdown(sock, how);
3449 EXPORT_SYMBOL(kernel_sock_shutdown);
3451 /* This routine returns the IP overhead imposed by a socket i.e.
3452 * the length of the underlying IP header, depending on whether
3453 * this is an IPv4 or IPv6 socket and the length from IP options turned
3454 * on at the socket. Assumes that the caller has a lock on the socket.
3456 u32 kernel_sock_ip_overhead(struct sock *sk)
3458 struct inet_sock *inet;
3459 struct ip_options_rcu *opt;
3461 #if IS_ENABLED(CONFIG_IPV6)
3462 struct ipv6_pinfo *np;
3463 struct ipv6_txoptions *optv6 = NULL;
3464 #endif /* IS_ENABLED(CONFIG_IPV6) */
3469 switch (sk->sk_family) {
3472 overhead += sizeof(struct iphdr);
3473 opt = rcu_dereference_protected(inet->inet_opt,
3474 sock_owned_by_user(sk));
3476 overhead += opt->opt.optlen;
3478 #if IS_ENABLED(CONFIG_IPV6)
3481 overhead += sizeof(struct ipv6hdr);
3483 optv6 = rcu_dereference_protected(np->opt,
3484 sock_owned_by_user(sk));
3486 overhead += (optv6->opt_flen + optv6->opt_nflen);
3488 #endif /* IS_ENABLED(CONFIG_IPV6) */
3489 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3493 EXPORT_SYMBOL(kernel_sock_ip_overhead);