2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <net/busy_poll.h>
108 #include <linux/errqueue.h>
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly;
112 unsigned int sysctl_net_busy_poll __read_mostly;
115 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
116 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
117 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119 static int sock_close(struct inode *inode, struct file *file);
120 static struct wait_queue_head *sock_get_poll_head(struct file *file,
122 static __poll_t sock_poll_mask(struct file *file, __poll_t);
123 static __poll_t sock_poll(struct file *file, struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 static long compat_sock_ioctl(struct file *file,
127 unsigned int cmd, unsigned long arg);
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131 int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133 struct pipe_inode_info *pipe, size_t len,
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
141 static const struct file_operations socket_file_ops = {
142 .owner = THIS_MODULE,
144 .read_iter = sock_read_iter,
145 .write_iter = sock_write_iter,
146 .get_poll_head = sock_get_poll_head,
147 .poll_mask = sock_poll_mask,
149 .unlocked_ioctl = sock_ioctl,
151 .compat_ioctl = compat_sock_ioctl,
154 .release = sock_close,
155 .fasync = sock_fasync,
156 .sendpage = sock_sendpage,
157 .splice_write = generic_splice_sendpage,
158 .splice_read = sock_splice_read,
162 * The protocol list. Each protocol is registered in here.
165 static DEFINE_SPINLOCK(net_family_lock);
166 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
170 * Move socket addresses back and forth across the kernel/user
171 * divide and look after the messy bits.
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
185 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191 if (copy_from_user(kaddr, uaddr, ulen))
193 return audit_sockaddr(ulen, kaddr);
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
213 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
214 void __user *uaddr, int __user *ulen)
219 BUG_ON(klen > sizeof(struct sockaddr_storage));
220 err = get_user(len, ulen);
228 if (audit_sockaddr(klen, kaddr))
230 if (copy_to_user(uaddr, kaddr, len))
234 * "fromlen shall refer to the value before truncation.."
237 return __put_user(klen, ulen);
240 static struct kmem_cache *sock_inode_cachep __ro_after_init;
242 static struct inode *sock_alloc_inode(struct super_block *sb)
244 struct socket_alloc *ei;
245 struct socket_wq *wq;
247 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
250 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
252 kmem_cache_free(sock_inode_cachep, ei);
255 init_waitqueue_head(&wq->wait);
256 wq->fasync_list = NULL;
258 RCU_INIT_POINTER(ei->socket.wq, wq);
260 ei->socket.state = SS_UNCONNECTED;
261 ei->socket.flags = 0;
262 ei->socket.ops = NULL;
263 ei->socket.sk = NULL;
264 ei->socket.file = NULL;
266 return &ei->vfs_inode;
269 static void sock_destroy_inode(struct inode *inode)
271 struct socket_alloc *ei;
272 struct socket_wq *wq;
274 ei = container_of(inode, struct socket_alloc, vfs_inode);
275 wq = rcu_dereference_protected(ei->socket.wq, 1);
277 kmem_cache_free(sock_inode_cachep, ei);
280 static void init_once(void *foo)
282 struct socket_alloc *ei = (struct socket_alloc *)foo;
284 inode_init_once(&ei->vfs_inode);
287 static void init_inodecache(void)
289 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
290 sizeof(struct socket_alloc),
292 (SLAB_HWCACHE_ALIGN |
293 SLAB_RECLAIM_ACCOUNT |
294 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
296 BUG_ON(sock_inode_cachep == NULL);
299 static const struct super_operations sockfs_ops = {
300 .alloc_inode = sock_alloc_inode,
301 .destroy_inode = sock_destroy_inode,
302 .statfs = simple_statfs,
306 * sockfs_dname() is called from d_path().
308 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
310 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
311 d_inode(dentry)->i_ino);
314 static const struct dentry_operations sockfs_dentry_operations = {
315 .d_dname = sockfs_dname,
318 static int sockfs_xattr_get(const struct xattr_handler *handler,
319 struct dentry *dentry, struct inode *inode,
320 const char *suffix, void *value, size_t size)
323 if (dentry->d_name.len + 1 > size)
325 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
327 return dentry->d_name.len + 1;
330 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
331 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
332 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
334 static const struct xattr_handler sockfs_xattr_handler = {
335 .name = XATTR_NAME_SOCKPROTONAME,
336 .get = sockfs_xattr_get,
339 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
340 struct dentry *dentry, struct inode *inode,
341 const char *suffix, const void *value,
342 size_t size, int flags)
344 /* Handled by LSM. */
348 static const struct xattr_handler sockfs_security_xattr_handler = {
349 .prefix = XATTR_SECURITY_PREFIX,
350 .set = sockfs_security_xattr_set,
353 static const struct xattr_handler *sockfs_xattr_handlers[] = {
354 &sockfs_xattr_handler,
355 &sockfs_security_xattr_handler,
359 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
360 int flags, const char *dev_name, void *data)
362 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
363 sockfs_xattr_handlers,
364 &sockfs_dentry_operations, SOCKFS_MAGIC);
367 static struct vfsmount *sock_mnt __read_mostly;
369 static struct file_system_type sock_fs_type = {
371 .mount = sockfs_mount,
372 .kill_sb = kill_anon_super,
376 * Obtains the first available file descriptor and sets it up for use.
378 * These functions create file structures and maps them to fd space
379 * of the current process. On success it returns file descriptor
380 * and file struct implicitly stored in sock->file.
381 * Note that another thread may close file descriptor before we return
382 * from this function. We use the fact that now we do not refer
383 * to socket after mapping. If one day we will need it, this
384 * function will increment ref. count on file by 1.
386 * In any case returned fd MAY BE not valid!
387 * This race condition is unavoidable
388 * with shared fd spaces, we cannot solve it inside kernel,
389 * but we take care of internal coherence yet.
392 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
394 struct qstr name = { .name = "" };
400 name.len = strlen(name.name);
401 } else if (sock->sk) {
402 name.name = sock->sk->sk_prot_creator->name;
403 name.len = strlen(name.name);
405 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
406 if (unlikely(!path.dentry)) {
408 return ERR_PTR(-ENOMEM);
410 path.mnt = mntget(sock_mnt);
412 d_instantiate(path.dentry, SOCK_INODE(sock));
414 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
417 /* drop dentry, keep inode for a bit */
418 ihold(d_inode(path.dentry));
420 /* ... and now kill it properly */
426 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
427 file->private_data = sock;
430 EXPORT_SYMBOL(sock_alloc_file);
432 static int sock_map_fd(struct socket *sock, int flags)
434 struct file *newfile;
435 int fd = get_unused_fd_flags(flags);
436 if (unlikely(fd < 0)) {
441 newfile = sock_alloc_file(sock, flags, NULL);
442 if (likely(!IS_ERR(newfile))) {
443 fd_install(fd, newfile);
448 return PTR_ERR(newfile);
451 struct socket *sock_from_file(struct file *file, int *err)
453 if (file->f_op == &socket_file_ops)
454 return file->private_data; /* set in sock_map_fd */
459 EXPORT_SYMBOL(sock_from_file);
462 * sockfd_lookup - Go from a file number to its socket slot
464 * @err: pointer to an error code return
466 * The file handle passed in is locked and the socket it is bound
467 * to is returned. If an error occurs the err pointer is overwritten
468 * with a negative errno code and NULL is returned. The function checks
469 * for both invalid handles and passing a handle which is not a socket.
471 * On a success the socket object pointer is returned.
474 struct socket *sockfd_lookup(int fd, int *err)
485 sock = sock_from_file(file, err);
490 EXPORT_SYMBOL(sockfd_lookup);
492 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
494 struct fd f = fdget(fd);
499 sock = sock_from_file(f.file, err);
501 *fput_needed = f.flags;
509 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
515 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
525 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
530 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
537 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
539 int err = simple_setattr(dentry, iattr);
541 if (!err && (iattr->ia_valid & ATTR_UID)) {
542 struct socket *sock = SOCKET_I(d_inode(dentry));
544 sock->sk->sk_uid = iattr->ia_uid;
550 static const struct inode_operations sockfs_inode_ops = {
551 .listxattr = sockfs_listxattr,
552 .setattr = sockfs_setattr,
556 * sock_alloc - allocate a socket
558 * Allocate a new inode and socket object. The two are bound together
559 * and initialised. The socket is then returned. If we are out of inodes
563 struct socket *sock_alloc(void)
568 inode = new_inode_pseudo(sock_mnt->mnt_sb);
572 sock = SOCKET_I(inode);
574 inode->i_ino = get_next_ino();
575 inode->i_mode = S_IFSOCK | S_IRWXUGO;
576 inode->i_uid = current_fsuid();
577 inode->i_gid = current_fsgid();
578 inode->i_op = &sockfs_inode_ops;
582 EXPORT_SYMBOL(sock_alloc);
585 * sock_release - close a socket
586 * @sock: socket to close
588 * The socket is released from the protocol stack if it has a release
589 * callback, and the inode is then released if the socket is bound to
590 * an inode not a file.
593 void sock_release(struct socket *sock)
596 struct module *owner = sock->ops->owner;
598 sock->ops->release(sock);
603 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
604 pr_err("%s: fasync list not empty!\n", __func__);
607 iput(SOCK_INODE(sock));
612 EXPORT_SYMBOL(sock_release);
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
616 u8 flags = *tx_flags;
618 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619 flags |= SKBTX_HW_TSTAMP;
621 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622 flags |= SKBTX_SW_TSTAMP;
624 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625 flags |= SKBTX_SCHED_TSTAMP;
629 EXPORT_SYMBOL(__sock_tx_timestamp);
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
633 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634 BUG_ON(ret == -EIOCBQUEUED);
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
640 int err = security_socket_sendmsg(sock, msg,
643 return err ?: sock_sendmsg_nosec(sock, msg);
645 EXPORT_SYMBOL(sock_sendmsg);
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 struct kvec *vec, size_t num, size_t size)
650 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651 return sock_sendmsg(sock, msg);
653 EXPORT_SYMBOL(kernel_sendmsg);
655 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
656 struct kvec *vec, size_t num, size_t size)
658 struct socket *sock = sk->sk_socket;
660 if (!sock->ops->sendmsg_locked)
661 return sock_no_sendmsg_locked(sk, msg, size);
663 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
665 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
667 EXPORT_SYMBOL(kernel_sendmsg_locked);
669 static bool skb_is_err_queue(const struct sk_buff *skb)
671 /* pkt_type of skbs enqueued on the error queue are set to
672 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
673 * in recvmsg, since skbs received on a local socket will never
674 * have a pkt_type of PACKET_OUTGOING.
676 return skb->pkt_type == PACKET_OUTGOING;
679 /* On transmit, software and hardware timestamps are returned independently.
680 * As the two skb clones share the hardware timestamp, which may be updated
681 * before the software timestamp is received, a hardware TX timestamp may be
682 * returned only if there is no software TX timestamp. Ignore false software
683 * timestamps, which may be made in the __sock_recv_timestamp() call when the
684 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
685 * hardware timestamp.
687 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
689 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
692 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
694 struct scm_ts_pktinfo ts_pktinfo;
695 struct net_device *orig_dev;
697 if (!skb_mac_header_was_set(skb))
700 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
703 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
705 ts_pktinfo.if_index = orig_dev->ifindex;
708 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
709 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
710 sizeof(ts_pktinfo), &ts_pktinfo);
714 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
716 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
719 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
720 struct scm_timestamping tss;
721 int empty = 1, false_tstamp = 0;
722 struct skb_shared_hwtstamps *shhwtstamps =
725 /* Race occurred between timestamp enabling and packet
726 receiving. Fill in the current time for now. */
727 if (need_software_tstamp && skb->tstamp == 0) {
728 __net_timestamp(skb);
732 if (need_software_tstamp) {
733 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
735 skb_get_timestamp(skb, &tv);
736 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
740 skb_get_timestampns(skb, &ts);
741 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
746 memset(&tss, 0, sizeof(tss));
747 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
748 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
751 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
752 !skb_is_swtx_tstamp(skb, false_tstamp) &&
753 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
755 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
756 !skb_is_err_queue(skb))
757 put_ts_pktinfo(msg, skb);
760 put_cmsg(msg, SOL_SOCKET,
761 SCM_TIMESTAMPING, sizeof(tss), &tss);
763 if (skb_is_err_queue(skb) && skb->len &&
764 SKB_EXT_ERR(skb)->opt_stats)
765 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
766 skb->len, skb->data);
769 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
771 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
776 if (!sock_flag(sk, SOCK_WIFI_STATUS))
778 if (!skb->wifi_acked_valid)
781 ack = skb->wifi_acked;
783 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
785 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
787 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
790 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
791 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
792 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
795 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
798 sock_recv_timestamp(msg, sk, skb);
799 sock_recv_drops(msg, sk, skb);
801 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
803 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
806 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
809 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
811 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
813 return err ?: sock_recvmsg_nosec(sock, msg, flags);
815 EXPORT_SYMBOL(sock_recvmsg);
818 * kernel_recvmsg - Receive a message from a socket (kernel space)
819 * @sock: The socket to receive the message from
820 * @msg: Received message
821 * @vec: Input s/g array for message data
822 * @num: Size of input s/g array
823 * @size: Number of bytes to read
824 * @flags: Message flags (MSG_DONTWAIT, etc...)
826 * On return the msg structure contains the scatter/gather array passed in the
827 * vec argument. The array is modified so that it consists of the unfilled
828 * portion of the original array.
830 * The returned value is the total number of bytes received, or an error.
832 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
833 struct kvec *vec, size_t num, size_t size, int flags)
835 mm_segment_t oldfs = get_fs();
838 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
840 result = sock_recvmsg(sock, msg, flags);
844 EXPORT_SYMBOL(kernel_recvmsg);
846 static ssize_t sock_sendpage(struct file *file, struct page *page,
847 int offset, size_t size, loff_t *ppos, int more)
852 sock = file->private_data;
854 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
855 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
858 return kernel_sendpage(sock, page, offset, size, flags);
861 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
862 struct pipe_inode_info *pipe, size_t len,
865 struct socket *sock = file->private_data;
867 if (unlikely(!sock->ops->splice_read))
870 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
873 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
875 struct file *file = iocb->ki_filp;
876 struct socket *sock = file->private_data;
877 struct msghdr msg = {.msg_iter = *to,
881 if (file->f_flags & O_NONBLOCK)
882 msg.msg_flags = MSG_DONTWAIT;
884 if (iocb->ki_pos != 0)
887 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
890 res = sock_recvmsg(sock, &msg, msg.msg_flags);
895 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
897 struct file *file = iocb->ki_filp;
898 struct socket *sock = file->private_data;
899 struct msghdr msg = {.msg_iter = *from,
903 if (iocb->ki_pos != 0)
906 if (file->f_flags & O_NONBLOCK)
907 msg.msg_flags = MSG_DONTWAIT;
909 if (sock->type == SOCK_SEQPACKET)
910 msg.msg_flags |= MSG_EOR;
912 res = sock_sendmsg(sock, &msg);
913 *from = msg.msg_iter;
918 * Atomic setting of ioctl hooks to avoid race
919 * with module unload.
922 static DEFINE_MUTEX(br_ioctl_mutex);
923 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
925 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
927 mutex_lock(&br_ioctl_mutex);
928 br_ioctl_hook = hook;
929 mutex_unlock(&br_ioctl_mutex);
931 EXPORT_SYMBOL(brioctl_set);
933 static DEFINE_MUTEX(vlan_ioctl_mutex);
934 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
936 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
938 mutex_lock(&vlan_ioctl_mutex);
939 vlan_ioctl_hook = hook;
940 mutex_unlock(&vlan_ioctl_mutex);
942 EXPORT_SYMBOL(vlan_ioctl_set);
944 static DEFINE_MUTEX(dlci_ioctl_mutex);
945 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
947 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
949 mutex_lock(&dlci_ioctl_mutex);
950 dlci_ioctl_hook = hook;
951 mutex_unlock(&dlci_ioctl_mutex);
953 EXPORT_SYMBOL(dlci_ioctl_set);
955 static long sock_do_ioctl(struct net *net, struct socket *sock,
956 unsigned int cmd, unsigned long arg)
959 void __user *argp = (void __user *)arg;
961 err = sock->ops->ioctl(sock, cmd, arg);
964 * If this ioctl is unknown try to hand it down
967 if (err != -ENOIOCTLCMD)
970 if (cmd == SIOCGIFCONF) {
972 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
975 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
977 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
982 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
984 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
985 if (!err && need_copyout)
986 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
993 * With an ioctl, arg may well be a user mode pointer, but we don't know
994 * what to do with it - that's up to the protocol still.
997 struct ns_common *get_net_ns(struct ns_common *ns)
999 return &get_net(container_of(ns, struct net, ns))->ns;
1001 EXPORT_SYMBOL_GPL(get_net_ns);
1003 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1005 struct socket *sock;
1007 void __user *argp = (void __user *)arg;
1011 sock = file->private_data;
1014 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1017 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1019 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1020 if (!err && need_copyout)
1021 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1024 #ifdef CONFIG_WEXT_CORE
1025 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1026 err = wext_handle_ioctl(net, cmd, argp);
1033 if (get_user(pid, (int __user *)argp))
1035 err = f_setown(sock->file, pid, 1);
1039 err = put_user(f_getown(sock->file),
1040 (int __user *)argp);
1048 request_module("bridge");
1050 mutex_lock(&br_ioctl_mutex);
1052 err = br_ioctl_hook(net, cmd, argp);
1053 mutex_unlock(&br_ioctl_mutex);
1058 if (!vlan_ioctl_hook)
1059 request_module("8021q");
1061 mutex_lock(&vlan_ioctl_mutex);
1062 if (vlan_ioctl_hook)
1063 err = vlan_ioctl_hook(net, argp);
1064 mutex_unlock(&vlan_ioctl_mutex);
1069 if (!dlci_ioctl_hook)
1070 request_module("dlci");
1072 mutex_lock(&dlci_ioctl_mutex);
1073 if (dlci_ioctl_hook)
1074 err = dlci_ioctl_hook(cmd, argp);
1075 mutex_unlock(&dlci_ioctl_mutex);
1079 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1082 err = open_related_ns(&net->ns, get_net_ns);
1085 err = sock_do_ioctl(net, sock, cmd, arg);
1091 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1094 struct socket *sock = NULL;
1096 err = security_socket_create(family, type, protocol, 1);
1100 sock = sock_alloc();
1107 err = security_socket_post_create(sock, family, type, protocol, 1);
1119 EXPORT_SYMBOL(sock_create_lite);
1121 static struct wait_queue_head *sock_get_poll_head(struct file *file,
1124 struct socket *sock = file->private_data;
1126 if (!sock->ops->poll_mask)
1128 sock_poll_busy_loop(sock, events);
1129 return sk_sleep(sock->sk);
1132 static __poll_t sock_poll_mask(struct file *file, __poll_t events)
1134 struct socket *sock = file->private_data;
1137 * We need to be sure we are in sync with the socket flags modification.
1139 * This memory barrier is paired in the wq_has_sleeper.
1143 /* this socket can poll_ll so tell the system call */
1144 return sock->ops->poll_mask(sock, events) |
1145 (sk_can_busy_loop(sock->sk) ? POLL_BUSY_LOOP : 0);
1148 /* No kernel lock held - perfect */
1149 static __poll_t sock_poll(struct file *file, poll_table *wait)
1151 struct socket *sock = file->private_data;
1152 __poll_t events = poll_requested_events(wait), mask = 0;
1154 if (sock->ops->poll) {
1155 sock_poll_busy_loop(sock, events);
1156 mask = sock->ops->poll(file, sock, wait);
1157 } else if (sock->ops->poll_mask) {
1158 sock_poll_wait(file, sock_get_poll_head(file, events), wait);
1159 mask = sock->ops->poll_mask(sock, events);
1162 return mask | sock_poll_busy_flag(sock);
1165 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1167 struct socket *sock = file->private_data;
1169 return sock->ops->mmap(file, sock, vma);
1172 static int sock_close(struct inode *inode, struct file *filp)
1174 sock_release(SOCKET_I(inode));
1179 * Update the socket async list
1181 * Fasync_list locking strategy.
1183 * 1. fasync_list is modified only under process context socket lock
1184 * i.e. under semaphore.
1185 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1186 * or under socket lock
1189 static int sock_fasync(int fd, struct file *filp, int on)
1191 struct socket *sock = filp->private_data;
1192 struct sock *sk = sock->sk;
1193 struct socket_wq *wq;
1199 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1200 fasync_helper(fd, filp, on, &wq->fasync_list);
1202 if (!wq->fasync_list)
1203 sock_reset_flag(sk, SOCK_FASYNC);
1205 sock_set_flag(sk, SOCK_FASYNC);
1211 /* This function may be called only under rcu_lock */
1213 int sock_wake_async(struct socket_wq *wq, int how, int band)
1215 if (!wq || !wq->fasync_list)
1219 case SOCK_WAKE_WAITD:
1220 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1223 case SOCK_WAKE_SPACE:
1224 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1229 kill_fasync(&wq->fasync_list, SIGIO, band);
1232 kill_fasync(&wq->fasync_list, SIGURG, band);
1237 EXPORT_SYMBOL(sock_wake_async);
1239 int __sock_create(struct net *net, int family, int type, int protocol,
1240 struct socket **res, int kern)
1243 struct socket *sock;
1244 const struct net_proto_family *pf;
1247 * Check protocol is in range
1249 if (family < 0 || family >= NPROTO)
1250 return -EAFNOSUPPORT;
1251 if (type < 0 || type >= SOCK_MAX)
1256 This uglymoron is moved from INET layer to here to avoid
1257 deadlock in module load.
1259 if (family == PF_INET && type == SOCK_PACKET) {
1260 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1265 err = security_socket_create(family, type, protocol, kern);
1270 * Allocate the socket and allow the family to set things up. if
1271 * the protocol is 0, the family is instructed to select an appropriate
1274 sock = sock_alloc();
1276 net_warn_ratelimited("socket: no more sockets\n");
1277 return -ENFILE; /* Not exactly a match, but its the
1278 closest posix thing */
1283 #ifdef CONFIG_MODULES
1284 /* Attempt to load a protocol module if the find failed.
1286 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1287 * requested real, full-featured networking support upon configuration.
1288 * Otherwise module support will break!
1290 if (rcu_access_pointer(net_families[family]) == NULL)
1291 request_module("net-pf-%d", family);
1295 pf = rcu_dereference(net_families[family]);
1296 err = -EAFNOSUPPORT;
1301 * We will call the ->create function, that possibly is in a loadable
1302 * module, so we have to bump that loadable module refcnt first.
1304 if (!try_module_get(pf->owner))
1307 /* Now protected by module ref count */
1310 err = pf->create(net, sock, protocol, kern);
1312 goto out_module_put;
1315 * Now to bump the refcnt of the [loadable] module that owns this
1316 * socket at sock_release time we decrement its refcnt.
1318 if (!try_module_get(sock->ops->owner))
1319 goto out_module_busy;
1322 * Now that we're done with the ->create function, the [loadable]
1323 * module can have its refcnt decremented
1325 module_put(pf->owner);
1326 err = security_socket_post_create(sock, family, type, protocol, kern);
1328 goto out_sock_release;
1334 err = -EAFNOSUPPORT;
1337 module_put(pf->owner);
1344 goto out_sock_release;
1346 EXPORT_SYMBOL(__sock_create);
1348 int sock_create(int family, int type, int protocol, struct socket **res)
1350 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1352 EXPORT_SYMBOL(sock_create);
1354 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1356 return __sock_create(net, family, type, protocol, res, 1);
1358 EXPORT_SYMBOL(sock_create_kern);
1360 int __sys_socket(int family, int type, int protocol)
1363 struct socket *sock;
1366 /* Check the SOCK_* constants for consistency. */
1367 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1368 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1369 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1370 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1372 flags = type & ~SOCK_TYPE_MASK;
1373 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1375 type &= SOCK_TYPE_MASK;
1377 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1378 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1380 retval = sock_create(family, type, protocol, &sock);
1384 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1387 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1389 return __sys_socket(family, type, protocol);
1393 * Create a pair of connected sockets.
1396 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1398 struct socket *sock1, *sock2;
1400 struct file *newfile1, *newfile2;
1403 flags = type & ~SOCK_TYPE_MASK;
1404 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1406 type &= SOCK_TYPE_MASK;
1408 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1409 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1412 * reserve descriptors and make sure we won't fail
1413 * to return them to userland.
1415 fd1 = get_unused_fd_flags(flags);
1416 if (unlikely(fd1 < 0))
1419 fd2 = get_unused_fd_flags(flags);
1420 if (unlikely(fd2 < 0)) {
1425 err = put_user(fd1, &usockvec[0]);
1429 err = put_user(fd2, &usockvec[1]);
1434 * Obtain the first socket and check if the underlying protocol
1435 * supports the socketpair call.
1438 err = sock_create(family, type, protocol, &sock1);
1439 if (unlikely(err < 0))
1442 err = sock_create(family, type, protocol, &sock2);
1443 if (unlikely(err < 0)) {
1444 sock_release(sock1);
1448 err = security_socket_socketpair(sock1, sock2);
1449 if (unlikely(err)) {
1450 sock_release(sock2);
1451 sock_release(sock1);
1455 err = sock1->ops->socketpair(sock1, sock2);
1456 if (unlikely(err < 0)) {
1457 sock_release(sock2);
1458 sock_release(sock1);
1462 newfile1 = sock_alloc_file(sock1, flags, NULL);
1463 if (IS_ERR(newfile1)) {
1464 err = PTR_ERR(newfile1);
1465 sock_release(sock2);
1469 newfile2 = sock_alloc_file(sock2, flags, NULL);
1470 if (IS_ERR(newfile2)) {
1471 err = PTR_ERR(newfile2);
1476 audit_fd_pair(fd1, fd2);
1478 fd_install(fd1, newfile1);
1479 fd_install(fd2, newfile2);
1488 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1489 int __user *, usockvec)
1491 return __sys_socketpair(family, type, protocol, usockvec);
1495 * Bind a name to a socket. Nothing much to do here since it's
1496 * the protocol's responsibility to handle the local address.
1498 * We move the socket address to kernel space before we call
1499 * the protocol layer (having also checked the address is ok).
1502 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1504 struct socket *sock;
1505 struct sockaddr_storage address;
1506 int err, fput_needed;
1508 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1510 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1512 err = security_socket_bind(sock,
1513 (struct sockaddr *)&address,
1516 err = sock->ops->bind(sock,
1520 fput_light(sock->file, fput_needed);
1525 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1527 return __sys_bind(fd, umyaddr, addrlen);
1531 * Perform a listen. Basically, we allow the protocol to do anything
1532 * necessary for a listen, and if that works, we mark the socket as
1533 * ready for listening.
1536 int __sys_listen(int fd, int backlog)
1538 struct socket *sock;
1539 int err, fput_needed;
1542 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1544 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1545 if ((unsigned int)backlog > somaxconn)
1546 backlog = somaxconn;
1548 err = security_socket_listen(sock, backlog);
1550 err = sock->ops->listen(sock, backlog);
1552 fput_light(sock->file, fput_needed);
1557 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1559 return __sys_listen(fd, backlog);
1563 * For accept, we attempt to create a new socket, set up the link
1564 * with the client, wake up the client, then return the new
1565 * connected fd. We collect the address of the connector in kernel
1566 * space and move it to user at the very end. This is unclean because
1567 * we open the socket then return an error.
1569 * 1003.1g adds the ability to recvmsg() to query connection pending
1570 * status to recvmsg. We need to add that support in a way thats
1571 * clean when we restructure accept also.
1574 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1575 int __user *upeer_addrlen, int flags)
1577 struct socket *sock, *newsock;
1578 struct file *newfile;
1579 int err, len, newfd, fput_needed;
1580 struct sockaddr_storage address;
1582 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1585 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1586 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1588 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1593 newsock = sock_alloc();
1597 newsock->type = sock->type;
1598 newsock->ops = sock->ops;
1601 * We don't need try_module_get here, as the listening socket (sock)
1602 * has the protocol module (sock->ops->owner) held.
1604 __module_get(newsock->ops->owner);
1606 newfd = get_unused_fd_flags(flags);
1607 if (unlikely(newfd < 0)) {
1609 sock_release(newsock);
1612 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1613 if (IS_ERR(newfile)) {
1614 err = PTR_ERR(newfile);
1615 put_unused_fd(newfd);
1619 err = security_socket_accept(sock, newsock);
1623 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1627 if (upeer_sockaddr) {
1628 len = newsock->ops->getname(newsock,
1629 (struct sockaddr *)&address, 2);
1631 err = -ECONNABORTED;
1634 err = move_addr_to_user(&address,
1635 len, upeer_sockaddr, upeer_addrlen);
1640 /* File flags are not inherited via accept() unlike another OSes. */
1642 fd_install(newfd, newfile);
1646 fput_light(sock->file, fput_needed);
1651 put_unused_fd(newfd);
1655 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1656 int __user *, upeer_addrlen, int, flags)
1658 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1661 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1662 int __user *, upeer_addrlen)
1664 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1668 * Attempt to connect to a socket with the server address. The address
1669 * is in user space so we verify it is OK and move it to kernel space.
1671 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1674 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1675 * other SEQPACKET protocols that take time to connect() as it doesn't
1676 * include the -EINPROGRESS status for such sockets.
1679 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1681 struct socket *sock;
1682 struct sockaddr_storage address;
1683 int err, fput_needed;
1685 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1688 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1693 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1697 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1698 sock->file->f_flags);
1700 fput_light(sock->file, fput_needed);
1705 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1708 return __sys_connect(fd, uservaddr, addrlen);
1712 * Get the local address ('name') of a socket object. Move the obtained
1713 * name to user space.
1716 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1717 int __user *usockaddr_len)
1719 struct socket *sock;
1720 struct sockaddr_storage address;
1721 int err, fput_needed;
1723 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1727 err = security_socket_getsockname(sock);
1731 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1734 /* "err" is actually length in this case */
1735 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1738 fput_light(sock->file, fput_needed);
1743 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1744 int __user *, usockaddr_len)
1746 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1750 * Get the remote address ('name') of a socket object. Move the obtained
1751 * name to user space.
1754 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1755 int __user *usockaddr_len)
1757 struct socket *sock;
1758 struct sockaddr_storage address;
1759 int err, fput_needed;
1761 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1763 err = security_socket_getpeername(sock);
1765 fput_light(sock->file, fput_needed);
1769 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1771 /* "err" is actually length in this case */
1772 err = move_addr_to_user(&address, err, usockaddr,
1774 fput_light(sock->file, fput_needed);
1779 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1780 int __user *, usockaddr_len)
1782 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1786 * Send a datagram to a given address. We move the address into kernel
1787 * space and check the user space data area is readable before invoking
1790 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1791 struct sockaddr __user *addr, int addr_len)
1793 struct socket *sock;
1794 struct sockaddr_storage address;
1800 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1803 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1807 msg.msg_name = NULL;
1808 msg.msg_control = NULL;
1809 msg.msg_controllen = 0;
1810 msg.msg_namelen = 0;
1812 err = move_addr_to_kernel(addr, addr_len, &address);
1815 msg.msg_name = (struct sockaddr *)&address;
1816 msg.msg_namelen = addr_len;
1818 if (sock->file->f_flags & O_NONBLOCK)
1819 flags |= MSG_DONTWAIT;
1820 msg.msg_flags = flags;
1821 err = sock_sendmsg(sock, &msg);
1824 fput_light(sock->file, fput_needed);
1829 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1830 unsigned int, flags, struct sockaddr __user *, addr,
1833 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1837 * Send a datagram down a socket.
1840 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1841 unsigned int, flags)
1843 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1847 * Receive a frame from the socket and optionally record the address of the
1848 * sender. We verify the buffers are writable and if needed move the
1849 * sender address from kernel to user space.
1851 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1852 struct sockaddr __user *addr, int __user *addr_len)
1854 struct socket *sock;
1857 struct sockaddr_storage address;
1861 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1864 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1868 msg.msg_control = NULL;
1869 msg.msg_controllen = 0;
1870 /* Save some cycles and don't copy the address if not needed */
1871 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1872 /* We assume all kernel code knows the size of sockaddr_storage */
1873 msg.msg_namelen = 0;
1874 msg.msg_iocb = NULL;
1876 if (sock->file->f_flags & O_NONBLOCK)
1877 flags |= MSG_DONTWAIT;
1878 err = sock_recvmsg(sock, &msg, flags);
1880 if (err >= 0 && addr != NULL) {
1881 err2 = move_addr_to_user(&address,
1882 msg.msg_namelen, addr, addr_len);
1887 fput_light(sock->file, fput_needed);
1892 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1893 unsigned int, flags, struct sockaddr __user *, addr,
1894 int __user *, addr_len)
1896 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1900 * Receive a datagram from a socket.
1903 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1904 unsigned int, flags)
1906 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1910 * Set a socket option. Because we don't know the option lengths we have
1911 * to pass the user mode parameter for the protocols to sort out.
1914 static int __sys_setsockopt(int fd, int level, int optname,
1915 char __user *optval, int optlen)
1917 int err, fput_needed;
1918 struct socket *sock;
1923 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1925 err = security_socket_setsockopt(sock, level, optname);
1929 if (level == SOL_SOCKET)
1931 sock_setsockopt(sock, level, optname, optval,
1935 sock->ops->setsockopt(sock, level, optname, optval,
1938 fput_light(sock->file, fput_needed);
1943 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1944 char __user *, optval, int, optlen)
1946 return __sys_setsockopt(fd, level, optname, optval, optlen);
1950 * Get a socket option. Because we don't know the option lengths we have
1951 * to pass a user mode parameter for the protocols to sort out.
1954 static int __sys_getsockopt(int fd, int level, int optname,
1955 char __user *optval, int __user *optlen)
1957 int err, fput_needed;
1958 struct socket *sock;
1960 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1962 err = security_socket_getsockopt(sock, level, optname);
1966 if (level == SOL_SOCKET)
1968 sock_getsockopt(sock, level, optname, optval,
1972 sock->ops->getsockopt(sock, level, optname, optval,
1975 fput_light(sock->file, fput_needed);
1980 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1981 char __user *, optval, int __user *, optlen)
1983 return __sys_getsockopt(fd, level, optname, optval, optlen);
1987 * Shutdown a socket.
1990 int __sys_shutdown(int fd, int how)
1992 int err, fput_needed;
1993 struct socket *sock;
1995 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1997 err = security_socket_shutdown(sock, how);
1999 err = sock->ops->shutdown(sock, how);
2000 fput_light(sock->file, fput_needed);
2005 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2007 return __sys_shutdown(fd, how);
2010 /* A couple of helpful macros for getting the address of the 32/64 bit
2011 * fields which are the same type (int / unsigned) on our platforms.
2013 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2014 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2015 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2017 struct used_address {
2018 struct sockaddr_storage name;
2019 unsigned int name_len;
2022 static int copy_msghdr_from_user(struct msghdr *kmsg,
2023 struct user_msghdr __user *umsg,
2024 struct sockaddr __user **save_addr,
2027 struct user_msghdr msg;
2030 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2033 kmsg->msg_control = (void __force *)msg.msg_control;
2034 kmsg->msg_controllen = msg.msg_controllen;
2035 kmsg->msg_flags = msg.msg_flags;
2037 kmsg->msg_namelen = msg.msg_namelen;
2039 kmsg->msg_namelen = 0;
2041 if (kmsg->msg_namelen < 0)
2044 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2045 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2048 *save_addr = msg.msg_name;
2050 if (msg.msg_name && kmsg->msg_namelen) {
2052 err = move_addr_to_kernel(msg.msg_name,
2059 kmsg->msg_name = NULL;
2060 kmsg->msg_namelen = 0;
2063 if (msg.msg_iovlen > UIO_MAXIOV)
2066 kmsg->msg_iocb = NULL;
2068 return import_iovec(save_addr ? READ : WRITE,
2069 msg.msg_iov, msg.msg_iovlen,
2070 UIO_FASTIOV, iov, &kmsg->msg_iter);
2073 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2074 struct msghdr *msg_sys, unsigned int flags,
2075 struct used_address *used_address,
2076 unsigned int allowed_msghdr_flags)
2078 struct compat_msghdr __user *msg_compat =
2079 (struct compat_msghdr __user *)msg;
2080 struct sockaddr_storage address;
2081 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2082 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2083 __aligned(sizeof(__kernel_size_t));
2084 /* 20 is size of ipv6_pktinfo */
2085 unsigned char *ctl_buf = ctl;
2089 msg_sys->msg_name = &address;
2091 if (MSG_CMSG_COMPAT & flags)
2092 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2094 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2100 if (msg_sys->msg_controllen > INT_MAX)
2102 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2103 ctl_len = msg_sys->msg_controllen;
2104 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2106 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2110 ctl_buf = msg_sys->msg_control;
2111 ctl_len = msg_sys->msg_controllen;
2112 } else if (ctl_len) {
2113 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2114 CMSG_ALIGN(sizeof(struct cmsghdr)));
2115 if (ctl_len > sizeof(ctl)) {
2116 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2117 if (ctl_buf == NULL)
2122 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2123 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2124 * checking falls down on this.
2126 if (copy_from_user(ctl_buf,
2127 (void __user __force *)msg_sys->msg_control,
2130 msg_sys->msg_control = ctl_buf;
2132 msg_sys->msg_flags = flags;
2134 if (sock->file->f_flags & O_NONBLOCK)
2135 msg_sys->msg_flags |= MSG_DONTWAIT;
2137 * If this is sendmmsg() and current destination address is same as
2138 * previously succeeded address, omit asking LSM's decision.
2139 * used_address->name_len is initialized to UINT_MAX so that the first
2140 * destination address never matches.
2142 if (used_address && msg_sys->msg_name &&
2143 used_address->name_len == msg_sys->msg_namelen &&
2144 !memcmp(&used_address->name, msg_sys->msg_name,
2145 used_address->name_len)) {
2146 err = sock_sendmsg_nosec(sock, msg_sys);
2149 err = sock_sendmsg(sock, msg_sys);
2151 * If this is sendmmsg() and sending to current destination address was
2152 * successful, remember it.
2154 if (used_address && err >= 0) {
2155 used_address->name_len = msg_sys->msg_namelen;
2156 if (msg_sys->msg_name)
2157 memcpy(&used_address->name, msg_sys->msg_name,
2158 used_address->name_len);
2163 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2170 * BSD sendmsg interface
2173 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2174 bool forbid_cmsg_compat)
2176 int fput_needed, err;
2177 struct msghdr msg_sys;
2178 struct socket *sock;
2180 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2183 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2187 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2189 fput_light(sock->file, fput_needed);
2194 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2196 return __sys_sendmsg(fd, msg, flags, true);
2200 * Linux sendmmsg interface
2203 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2204 unsigned int flags, bool forbid_cmsg_compat)
2206 int fput_needed, err, datagrams;
2207 struct socket *sock;
2208 struct mmsghdr __user *entry;
2209 struct compat_mmsghdr __user *compat_entry;
2210 struct msghdr msg_sys;
2211 struct used_address used_address;
2212 unsigned int oflags = flags;
2214 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2217 if (vlen > UIO_MAXIOV)
2222 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2226 used_address.name_len = UINT_MAX;
2228 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2232 while (datagrams < vlen) {
2233 if (datagrams == vlen - 1)
2236 if (MSG_CMSG_COMPAT & flags) {
2237 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2238 &msg_sys, flags, &used_address, MSG_EOR);
2241 err = __put_user(err, &compat_entry->msg_len);
2244 err = ___sys_sendmsg(sock,
2245 (struct user_msghdr __user *)entry,
2246 &msg_sys, flags, &used_address, MSG_EOR);
2249 err = put_user(err, &entry->msg_len);
2256 if (msg_data_left(&msg_sys))
2261 fput_light(sock->file, fput_needed);
2263 /* We only return an error if no datagrams were able to be sent */
2270 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2271 unsigned int, vlen, unsigned int, flags)
2273 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2276 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2277 struct msghdr *msg_sys, unsigned int flags, int nosec)
2279 struct compat_msghdr __user *msg_compat =
2280 (struct compat_msghdr __user *)msg;
2281 struct iovec iovstack[UIO_FASTIOV];
2282 struct iovec *iov = iovstack;
2283 unsigned long cmsg_ptr;
2287 /* kernel mode address */
2288 struct sockaddr_storage addr;
2290 /* user mode address pointers */
2291 struct sockaddr __user *uaddr;
2292 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2294 msg_sys->msg_name = &addr;
2296 if (MSG_CMSG_COMPAT & flags)
2297 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2299 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2303 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2304 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2306 /* We assume all kernel code knows the size of sockaddr_storage */
2307 msg_sys->msg_namelen = 0;
2309 if (sock->file->f_flags & O_NONBLOCK)
2310 flags |= MSG_DONTWAIT;
2311 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2316 if (uaddr != NULL) {
2317 err = move_addr_to_user(&addr,
2318 msg_sys->msg_namelen, uaddr,
2323 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2327 if (MSG_CMSG_COMPAT & flags)
2328 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2329 &msg_compat->msg_controllen);
2331 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2332 &msg->msg_controllen);
2343 * BSD recvmsg interface
2346 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2347 bool forbid_cmsg_compat)
2349 int fput_needed, err;
2350 struct msghdr msg_sys;
2351 struct socket *sock;
2353 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2356 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2360 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2362 fput_light(sock->file, fput_needed);
2367 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2368 unsigned int, flags)
2370 return __sys_recvmsg(fd, msg, flags, true);
2374 * Linux recvmmsg interface
2377 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2378 unsigned int flags, struct timespec *timeout)
2380 int fput_needed, err, datagrams;
2381 struct socket *sock;
2382 struct mmsghdr __user *entry;
2383 struct compat_mmsghdr __user *compat_entry;
2384 struct msghdr msg_sys;
2385 struct timespec64 end_time;
2386 struct timespec64 timeout64;
2389 poll_select_set_timeout(&end_time, timeout->tv_sec,
2395 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2399 if (likely(!(flags & MSG_ERRQUEUE))) {
2400 err = sock_error(sock->sk);
2408 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2410 while (datagrams < vlen) {
2412 * No need to ask LSM for more than the first datagram.
2414 if (MSG_CMSG_COMPAT & flags) {
2415 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2416 &msg_sys, flags & ~MSG_WAITFORONE,
2420 err = __put_user(err, &compat_entry->msg_len);
2423 err = ___sys_recvmsg(sock,
2424 (struct user_msghdr __user *)entry,
2425 &msg_sys, flags & ~MSG_WAITFORONE,
2429 err = put_user(err, &entry->msg_len);
2437 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2438 if (flags & MSG_WAITFORONE)
2439 flags |= MSG_DONTWAIT;
2442 ktime_get_ts64(&timeout64);
2443 *timeout = timespec64_to_timespec(
2444 timespec64_sub(end_time, timeout64));
2445 if (timeout->tv_sec < 0) {
2446 timeout->tv_sec = timeout->tv_nsec = 0;
2450 /* Timeout, return less than vlen datagrams */
2451 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2455 /* Out of band data, return right away */
2456 if (msg_sys.msg_flags & MSG_OOB)
2464 if (datagrams == 0) {
2470 * We may return less entries than requested (vlen) if the
2471 * sock is non block and there aren't enough datagrams...
2473 if (err != -EAGAIN) {
2475 * ... or if recvmsg returns an error after we
2476 * received some datagrams, where we record the
2477 * error to return on the next call or if the
2478 * app asks about it using getsockopt(SO_ERROR).
2480 sock->sk->sk_err = -err;
2483 fput_light(sock->file, fput_needed);
2488 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2489 unsigned int vlen, unsigned int flags,
2490 struct timespec __user *timeout)
2493 struct timespec timeout_sys;
2495 if (flags & MSG_CMSG_COMPAT)
2499 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2501 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2504 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2506 if (datagrams > 0 &&
2507 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2508 datagrams = -EFAULT;
2513 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2514 unsigned int, vlen, unsigned int, flags,
2515 struct timespec __user *, timeout)
2517 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2520 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2521 /* Argument list sizes for sys_socketcall */
2522 #define AL(x) ((x) * sizeof(unsigned long))
2523 static const unsigned char nargs[21] = {
2524 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2525 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2526 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2533 * System call vectors.
2535 * Argument checking cleaned up. Saved 20% in size.
2536 * This function doesn't need to set the kernel lock because
2537 * it is set by the callees.
2540 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2542 unsigned long a[AUDITSC_ARGS];
2543 unsigned long a0, a1;
2547 if (call < 1 || call > SYS_SENDMMSG)
2551 if (len > sizeof(a))
2554 /* copy_from_user should be SMP safe. */
2555 if (copy_from_user(a, args, len))
2558 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2567 err = __sys_socket(a0, a1, a[2]);
2570 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2573 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2576 err = __sys_listen(a0, a1);
2579 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2580 (int __user *)a[2], 0);
2582 case SYS_GETSOCKNAME:
2584 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2585 (int __user *)a[2]);
2587 case SYS_GETPEERNAME:
2589 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2590 (int __user *)a[2]);
2592 case SYS_SOCKETPAIR:
2593 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2596 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2600 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2601 (struct sockaddr __user *)a[4], a[5]);
2604 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2608 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2609 (struct sockaddr __user *)a[4],
2610 (int __user *)a[5]);
2613 err = __sys_shutdown(a0, a1);
2615 case SYS_SETSOCKOPT:
2616 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2619 case SYS_GETSOCKOPT:
2621 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2622 (int __user *)a[4]);
2625 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2629 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2633 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2637 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2638 a[3], (struct timespec __user *)a[4]);
2641 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2642 (int __user *)a[2], a[3]);
2651 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2654 * sock_register - add a socket protocol handler
2655 * @ops: description of protocol
2657 * This function is called by a protocol handler that wants to
2658 * advertise its address family, and have it linked into the
2659 * socket interface. The value ops->family corresponds to the
2660 * socket system call protocol family.
2662 int sock_register(const struct net_proto_family *ops)
2666 if (ops->family >= NPROTO) {
2667 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2671 spin_lock(&net_family_lock);
2672 if (rcu_dereference_protected(net_families[ops->family],
2673 lockdep_is_held(&net_family_lock)))
2676 rcu_assign_pointer(net_families[ops->family], ops);
2679 spin_unlock(&net_family_lock);
2681 pr_info("NET: Registered protocol family %d\n", ops->family);
2684 EXPORT_SYMBOL(sock_register);
2687 * sock_unregister - remove a protocol handler
2688 * @family: protocol family to remove
2690 * This function is called by a protocol handler that wants to
2691 * remove its address family, and have it unlinked from the
2692 * new socket creation.
2694 * If protocol handler is a module, then it can use module reference
2695 * counts to protect against new references. If protocol handler is not
2696 * a module then it needs to provide its own protection in
2697 * the ops->create routine.
2699 void sock_unregister(int family)
2701 BUG_ON(family < 0 || family >= NPROTO);
2703 spin_lock(&net_family_lock);
2704 RCU_INIT_POINTER(net_families[family], NULL);
2705 spin_unlock(&net_family_lock);
2709 pr_info("NET: Unregistered protocol family %d\n", family);
2711 EXPORT_SYMBOL(sock_unregister);
2713 bool sock_is_registered(int family)
2715 return family < NPROTO && rcu_access_pointer(net_families[family]);
2718 static int __init sock_init(void)
2722 * Initialize the network sysctl infrastructure.
2724 err = net_sysctl_init();
2729 * Initialize skbuff SLAB cache
2734 * Initialize the protocols module.
2739 err = register_filesystem(&sock_fs_type);
2742 sock_mnt = kern_mount(&sock_fs_type);
2743 if (IS_ERR(sock_mnt)) {
2744 err = PTR_ERR(sock_mnt);
2748 /* The real protocol initialization is performed in later initcalls.
2751 #ifdef CONFIG_NETFILTER
2752 err = netfilter_init();
2757 ptp_classifier_init();
2763 unregister_filesystem(&sock_fs_type);
2768 core_initcall(sock_init); /* early initcall */
2770 #ifdef CONFIG_PROC_FS
2771 void socket_seq_show(struct seq_file *seq)
2773 seq_printf(seq, "sockets: used %d\n",
2774 sock_inuse_get(seq->private));
2776 #endif /* CONFIG_PROC_FS */
2778 #ifdef CONFIG_COMPAT
2779 static int do_siocgstamp(struct net *net, struct socket *sock,
2780 unsigned int cmd, void __user *up)
2782 mm_segment_t old_fs = get_fs();
2787 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2790 err = compat_put_timeval(&ktv, up);
2795 static int do_siocgstampns(struct net *net, struct socket *sock,
2796 unsigned int cmd, void __user *up)
2798 mm_segment_t old_fs = get_fs();
2799 struct timespec kts;
2803 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2806 err = compat_put_timespec(&kts, up);
2811 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2813 struct compat_ifconf ifc32;
2817 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2820 ifc.ifc_len = ifc32.ifc_len;
2821 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2824 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2829 ifc32.ifc_len = ifc.ifc_len;
2830 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2836 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2838 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2839 bool convert_in = false, convert_out = false;
2840 size_t buf_size = 0;
2841 struct ethtool_rxnfc __user *rxnfc = NULL;
2843 u32 rule_cnt = 0, actual_rule_cnt;
2848 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2851 compat_rxnfc = compat_ptr(data);
2853 if (get_user(ethcmd, &compat_rxnfc->cmd))
2856 /* Most ethtool structures are defined without padding.
2857 * Unfortunately struct ethtool_rxnfc is an exception.
2862 case ETHTOOL_GRXCLSRLALL:
2863 /* Buffer size is variable */
2864 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2866 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2868 buf_size += rule_cnt * sizeof(u32);
2870 case ETHTOOL_GRXRINGS:
2871 case ETHTOOL_GRXCLSRLCNT:
2872 case ETHTOOL_GRXCLSRULE:
2873 case ETHTOOL_SRXCLSRLINS:
2876 case ETHTOOL_SRXCLSRLDEL:
2877 buf_size += sizeof(struct ethtool_rxnfc);
2879 rxnfc = compat_alloc_user_space(buf_size);
2883 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2886 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2889 /* We expect there to be holes between fs.m_ext and
2890 * fs.ring_cookie and at the end of fs, but nowhere else.
2892 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2893 sizeof(compat_rxnfc->fs.m_ext) !=
2894 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2895 sizeof(rxnfc->fs.m_ext));
2897 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2898 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2899 offsetof(struct ethtool_rxnfc, fs.location) -
2900 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2902 if (copy_in_user(rxnfc, compat_rxnfc,
2903 (void __user *)(&rxnfc->fs.m_ext + 1) -
2904 (void __user *)rxnfc) ||
2905 copy_in_user(&rxnfc->fs.ring_cookie,
2906 &compat_rxnfc->fs.ring_cookie,
2907 (void __user *)(&rxnfc->fs.location + 1) -
2908 (void __user *)&rxnfc->fs.ring_cookie) ||
2909 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2910 sizeof(rxnfc->rule_cnt)))
2914 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2919 if (copy_in_user(compat_rxnfc, rxnfc,
2920 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2921 (const void __user *)rxnfc) ||
2922 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2923 &rxnfc->fs.ring_cookie,
2924 (const void __user *)(&rxnfc->fs.location + 1) -
2925 (const void __user *)&rxnfc->fs.ring_cookie) ||
2926 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2927 sizeof(rxnfc->rule_cnt)))
2930 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2931 /* As an optimisation, we only copy the actual
2932 * number of rules that the underlying
2933 * function returned. Since Mallory might
2934 * change the rule count in user memory, we
2935 * check that it is less than the rule count
2936 * originally given (as the user buffer size),
2937 * which has been range-checked.
2939 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2941 if (actual_rule_cnt < rule_cnt)
2942 rule_cnt = actual_rule_cnt;
2943 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2944 &rxnfc->rule_locs[0],
2945 rule_cnt * sizeof(u32)))
2953 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2955 compat_uptr_t uptr32;
2960 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2963 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2966 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2967 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2969 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2971 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2972 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2978 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2979 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2980 struct compat_ifreq __user *u_ifreq32)
2985 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2987 if (get_user(data32, &u_ifreq32->ifr_data))
2989 ifreq.ifr_data = compat_ptr(data32);
2991 return dev_ioctl(net, cmd, &ifreq, NULL);
2994 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2995 struct compat_ifreq __user *uifr32)
2998 struct compat_ifmap __user *uifmap32;
3001 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3002 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3003 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3004 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3005 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3006 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3007 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3008 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3012 err = dev_ioctl(net, cmd, &ifr, NULL);
3014 if (cmd == SIOCGIFMAP && !err) {
3015 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3016 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3017 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3018 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3019 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3020 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3021 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3030 struct sockaddr rt_dst; /* target address */
3031 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3032 struct sockaddr rt_genmask; /* target network mask (IP) */
3033 unsigned short rt_flags;
3036 unsigned char rt_tos;
3037 unsigned char rt_class;
3039 short rt_metric; /* +1 for binary compatibility! */
3040 /* char * */ u32 rt_dev; /* forcing the device at add */
3041 u32 rt_mtu; /* per route MTU/Window */
3042 u32 rt_window; /* Window clamping */
3043 unsigned short rt_irtt; /* Initial RTT */
3046 struct in6_rtmsg32 {
3047 struct in6_addr rtmsg_dst;
3048 struct in6_addr rtmsg_src;
3049 struct in6_addr rtmsg_gateway;
3059 static int routing_ioctl(struct net *net, struct socket *sock,
3060 unsigned int cmd, void __user *argp)
3064 struct in6_rtmsg r6;
3068 mm_segment_t old_fs = get_fs();
3070 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3071 struct in6_rtmsg32 __user *ur6 = argp;
3072 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3073 3 * sizeof(struct in6_addr));
3074 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3075 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3076 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3077 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3078 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3079 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3080 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3084 struct rtentry32 __user *ur4 = argp;
3085 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3086 3 * sizeof(struct sockaddr));
3087 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3088 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3089 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3090 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3091 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3092 ret |= get_user(rtdev, &(ur4->rt_dev));
3094 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3095 r4.rt_dev = (char __user __force *)devname;
3109 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3116 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3117 * for some operations; this forces use of the newer bridge-utils that
3118 * use compatible ioctls
3120 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3124 if (get_user(tmp, argp))
3126 if (tmp == BRCTL_GET_VERSION)
3127 return BRCTL_VERSION + 1;
3131 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3132 unsigned int cmd, unsigned long arg)
3134 void __user *argp = compat_ptr(arg);
3135 struct sock *sk = sock->sk;
3136 struct net *net = sock_net(sk);
3138 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3139 return compat_ifr_data_ioctl(net, cmd, argp);
3144 return old_bridge_ioctl(argp);
3146 return compat_dev_ifconf(net, argp);
3148 return ethtool_ioctl(net, argp);
3150 return compat_siocwandev(net, argp);
3153 return compat_sioc_ifmap(net, cmd, argp);
3156 return routing_ioctl(net, sock, cmd, argp);
3158 return do_siocgstamp(net, sock, cmd, argp);
3160 return do_siocgstampns(net, sock, cmd, argp);
3161 case SIOCBONDSLAVEINFOQUERY:
3162 case SIOCBONDINFOQUERY:
3165 return compat_ifr_data_ioctl(net, cmd, argp);
3178 return sock_ioctl(file, cmd, arg);
3195 case SIOCSIFHWBROADCAST:
3197 case SIOCGIFBRDADDR:
3198 case SIOCSIFBRDADDR:
3199 case SIOCGIFDSTADDR:
3200 case SIOCSIFDSTADDR:
3201 case SIOCGIFNETMASK:
3202 case SIOCSIFNETMASK:
3217 case SIOCBONDENSLAVE:
3218 case SIOCBONDRELEASE:
3219 case SIOCBONDSETHWADDR:
3220 case SIOCBONDCHANGEACTIVE:
3222 return sock_do_ioctl(net, sock, cmd, arg);
3225 return -ENOIOCTLCMD;
3228 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3231 struct socket *sock = file->private_data;
3232 int ret = -ENOIOCTLCMD;
3239 if (sock->ops->compat_ioctl)
3240 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3242 if (ret == -ENOIOCTLCMD &&
3243 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3244 ret = compat_wext_handle_ioctl(net, cmd, arg);
3246 if (ret == -ENOIOCTLCMD)
3247 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3253 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3255 return sock->ops->bind(sock, addr, addrlen);
3257 EXPORT_SYMBOL(kernel_bind);
3259 int kernel_listen(struct socket *sock, int backlog)
3261 return sock->ops->listen(sock, backlog);
3263 EXPORT_SYMBOL(kernel_listen);
3265 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3267 struct sock *sk = sock->sk;
3270 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3275 err = sock->ops->accept(sock, *newsock, flags, true);
3277 sock_release(*newsock);
3282 (*newsock)->ops = sock->ops;
3283 __module_get((*newsock)->ops->owner);
3288 EXPORT_SYMBOL(kernel_accept);
3290 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3293 return sock->ops->connect(sock, addr, addrlen, flags);
3295 EXPORT_SYMBOL(kernel_connect);
3297 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3299 return sock->ops->getname(sock, addr, 0);
3301 EXPORT_SYMBOL(kernel_getsockname);
3303 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3305 return sock->ops->getname(sock, addr, 1);
3307 EXPORT_SYMBOL(kernel_getpeername);
3309 int kernel_getsockopt(struct socket *sock, int level, int optname,
3310 char *optval, int *optlen)
3312 mm_segment_t oldfs = get_fs();
3313 char __user *uoptval;
3314 int __user *uoptlen;
3317 uoptval = (char __user __force *) optval;
3318 uoptlen = (int __user __force *) optlen;
3321 if (level == SOL_SOCKET)
3322 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3324 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3329 EXPORT_SYMBOL(kernel_getsockopt);
3331 int kernel_setsockopt(struct socket *sock, int level, int optname,
3332 char *optval, unsigned int optlen)
3334 mm_segment_t oldfs = get_fs();
3335 char __user *uoptval;
3338 uoptval = (char __user __force *) optval;
3341 if (level == SOL_SOCKET)
3342 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3344 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3349 EXPORT_SYMBOL(kernel_setsockopt);
3351 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3352 size_t size, int flags)
3354 if (sock->ops->sendpage)
3355 return sock->ops->sendpage(sock, page, offset, size, flags);
3357 return sock_no_sendpage(sock, page, offset, size, flags);
3359 EXPORT_SYMBOL(kernel_sendpage);
3361 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3362 size_t size, int flags)
3364 struct socket *sock = sk->sk_socket;
3366 if (sock->ops->sendpage_locked)
3367 return sock->ops->sendpage_locked(sk, page, offset, size,
3370 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3372 EXPORT_SYMBOL(kernel_sendpage_locked);
3374 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3376 return sock->ops->shutdown(sock, how);
3378 EXPORT_SYMBOL(kernel_sock_shutdown);
3380 /* This routine returns the IP overhead imposed by a socket i.e.
3381 * the length of the underlying IP header, depending on whether
3382 * this is an IPv4 or IPv6 socket and the length from IP options turned
3383 * on at the socket. Assumes that the caller has a lock on the socket.
3385 u32 kernel_sock_ip_overhead(struct sock *sk)
3387 struct inet_sock *inet;
3388 struct ip_options_rcu *opt;
3390 #if IS_ENABLED(CONFIG_IPV6)
3391 struct ipv6_pinfo *np;
3392 struct ipv6_txoptions *optv6 = NULL;
3393 #endif /* IS_ENABLED(CONFIG_IPV6) */
3398 switch (sk->sk_family) {
3401 overhead += sizeof(struct iphdr);
3402 opt = rcu_dereference_protected(inet->inet_opt,
3403 sock_owned_by_user(sk));
3405 overhead += opt->opt.optlen;
3407 #if IS_ENABLED(CONFIG_IPV6)
3410 overhead += sizeof(struct ipv6hdr);
3412 optv6 = rcu_dereference_protected(np->opt,
3413 sock_owned_by_user(sk));
3415 overhead += (optv6->opt_flen + optv6->opt_nflen);
3417 #endif /* IS_ENABLED(CONFIG_IPV6) */
3418 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3422 EXPORT_SYMBOL(kernel_sock_ip_overhead);