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;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
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;
269 struct socket_wq *wq;
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
272 wq = rcu_dereference_protected(ei->socket.wq, 1);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static void init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
291 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
293 BUG_ON(sock_inode_cachep == NULL);
296 static const struct super_operations sockfs_ops = {
297 .alloc_inode = sock_alloc_inode,
298 .destroy_inode = sock_destroy_inode,
299 .statfs = simple_statfs,
303 * sockfs_dname() is called from d_path().
305 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
307 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
308 d_inode(dentry)->i_ino);
311 static const struct dentry_operations sockfs_dentry_operations = {
312 .d_dname = sockfs_dname,
315 static int sockfs_xattr_get(const struct xattr_handler *handler,
316 struct dentry *dentry, struct inode *inode,
317 const char *suffix, void *value, size_t size)
320 if (dentry->d_name.len + 1 > size)
322 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
324 return dentry->d_name.len + 1;
327 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
331 static const struct xattr_handler sockfs_xattr_handler = {
332 .name = XATTR_NAME_SOCKPROTONAME,
333 .get = sockfs_xattr_get,
336 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337 struct dentry *dentry, struct inode *inode,
338 const char *suffix, const void *value,
339 size_t size, int flags)
341 /* Handled by LSM. */
345 static const struct xattr_handler sockfs_security_xattr_handler = {
346 .prefix = XATTR_SECURITY_PREFIX,
347 .set = sockfs_security_xattr_set,
350 static const struct xattr_handler *sockfs_xattr_handlers[] = {
351 &sockfs_xattr_handler,
352 &sockfs_security_xattr_handler,
356 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
357 int flags, const char *dev_name, void *data)
359 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
360 sockfs_xattr_handlers,
361 &sockfs_dentry_operations, SOCKFS_MAGIC);
364 static struct vfsmount *sock_mnt __read_mostly;
366 static struct file_system_type sock_fs_type = {
368 .mount = sockfs_mount,
369 .kill_sb = kill_anon_super,
373 * Obtains the first available file descriptor and sets it up for use.
375 * These functions create file structures and maps them to fd space
376 * of the current process. On success it returns file descriptor
377 * and file struct implicitly stored in sock->file.
378 * Note that another thread may close file descriptor before we return
379 * from this function. We use the fact that now we do not refer
380 * to socket after mapping. If one day we will need it, this
381 * function will increment ref. count on file by 1.
383 * In any case returned fd MAY BE not valid!
384 * This race condition is unavoidable
385 * with shared fd spaces, we cannot solve it inside kernel,
386 * but we take care of internal coherence yet.
389 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
394 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
396 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
397 O_RDWR | (flags & O_NONBLOCK),
405 file->private_data = sock;
408 EXPORT_SYMBOL(sock_alloc_file);
410 static int sock_map_fd(struct socket *sock, int flags)
412 struct file *newfile;
413 int fd = get_unused_fd_flags(flags);
414 if (unlikely(fd < 0)) {
419 newfile = sock_alloc_file(sock, flags, NULL);
420 if (likely(!IS_ERR(newfile))) {
421 fd_install(fd, newfile);
426 return PTR_ERR(newfile);
429 struct socket *sock_from_file(struct file *file, int *err)
431 if (file->f_op == &socket_file_ops)
432 return file->private_data; /* set in sock_map_fd */
437 EXPORT_SYMBOL(sock_from_file);
440 * sockfd_lookup - Go from a file number to its socket slot
442 * @err: pointer to an error code return
444 * The file handle passed in is locked and the socket it is bound
445 * to is returned. If an error occurs the err pointer is overwritten
446 * with a negative errno code and NULL is returned. The function checks
447 * for both invalid handles and passing a handle which is not a socket.
449 * On a success the socket object pointer is returned.
452 struct socket *sockfd_lookup(int fd, int *err)
463 sock = sock_from_file(file, err);
468 EXPORT_SYMBOL(sockfd_lookup);
470 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
472 struct fd f = fdget(fd);
477 sock = sock_from_file(f.file, err);
479 *fput_needed = f.flags;
487 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
493 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
503 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
508 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
515 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
517 int err = simple_setattr(dentry, iattr);
519 if (!err && (iattr->ia_valid & ATTR_UID)) {
520 struct socket *sock = SOCKET_I(d_inode(dentry));
523 sock->sk->sk_uid = iattr->ia_uid;
531 static const struct inode_operations sockfs_inode_ops = {
532 .listxattr = sockfs_listxattr,
533 .setattr = sockfs_setattr,
537 * sock_alloc - allocate a socket
539 * Allocate a new inode and socket object. The two are bound together
540 * and initialised. The socket is then returned. If we are out of inodes
544 struct socket *sock_alloc(void)
549 inode = new_inode_pseudo(sock_mnt->mnt_sb);
553 sock = SOCKET_I(inode);
555 inode->i_ino = get_next_ino();
556 inode->i_mode = S_IFSOCK | S_IRWXUGO;
557 inode->i_uid = current_fsuid();
558 inode->i_gid = current_fsgid();
559 inode->i_op = &sockfs_inode_ops;
563 EXPORT_SYMBOL(sock_alloc);
566 * sock_release - close a socket
567 * @sock: socket to close
569 * The socket is released from the protocol stack if it has a release
570 * callback, and the inode is then released if the socket is bound to
571 * an inode not a file.
574 static void __sock_release(struct socket *sock, struct inode *inode)
577 struct module *owner = sock->ops->owner;
581 sock->ops->release(sock);
588 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
589 pr_err("%s: fasync list not empty!\n", __func__);
592 iput(SOCK_INODE(sock));
598 void sock_release(struct socket *sock)
600 __sock_release(sock, NULL);
602 EXPORT_SYMBOL(sock_release);
604 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
606 u8 flags = *tx_flags;
608 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
609 flags |= SKBTX_HW_TSTAMP;
611 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
612 flags |= SKBTX_SW_TSTAMP;
614 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
615 flags |= SKBTX_SCHED_TSTAMP;
619 EXPORT_SYMBOL(__sock_tx_timestamp);
621 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
623 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
624 BUG_ON(ret == -EIOCBQUEUED);
628 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
630 int err = security_socket_sendmsg(sock, msg,
633 return err ?: sock_sendmsg_nosec(sock, msg);
635 EXPORT_SYMBOL(sock_sendmsg);
637 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
638 struct kvec *vec, size_t num, size_t size)
640 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
641 return sock_sendmsg(sock, msg);
643 EXPORT_SYMBOL(kernel_sendmsg);
645 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
646 struct kvec *vec, size_t num, size_t size)
648 struct socket *sock = sk->sk_socket;
650 if (!sock->ops->sendmsg_locked)
651 return sock_no_sendmsg_locked(sk, msg, size);
653 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
655 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
657 EXPORT_SYMBOL(kernel_sendmsg_locked);
659 static bool skb_is_err_queue(const struct sk_buff *skb)
661 /* pkt_type of skbs enqueued on the error queue are set to
662 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
663 * in recvmsg, since skbs received on a local socket will never
664 * have a pkt_type of PACKET_OUTGOING.
666 return skb->pkt_type == PACKET_OUTGOING;
669 /* On transmit, software and hardware timestamps are returned independently.
670 * As the two skb clones share the hardware timestamp, which may be updated
671 * before the software timestamp is received, a hardware TX timestamp may be
672 * returned only if there is no software TX timestamp. Ignore false software
673 * timestamps, which may be made in the __sock_recv_timestamp() call when the
674 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
675 * hardware timestamp.
677 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
679 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
682 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
684 struct scm_ts_pktinfo ts_pktinfo;
685 struct net_device *orig_dev;
687 if (!skb_mac_header_was_set(skb))
690 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
693 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
695 ts_pktinfo.if_index = orig_dev->ifindex;
698 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
699 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
700 sizeof(ts_pktinfo), &ts_pktinfo);
704 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
706 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
709 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
710 struct scm_timestamping 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 skb_get_timestamp(skb, &tv);
726 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
730 skb_get_timestampns(skb, &ts);
731 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
736 memset(&tss, 0, sizeof(tss));
737 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
738 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
741 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
742 !skb_is_swtx_tstamp(skb, false_tstamp) &&
743 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
745 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
746 !skb_is_err_queue(skb))
747 put_ts_pktinfo(msg, skb);
750 put_cmsg(msg, SOL_SOCKET,
751 SCM_TIMESTAMPING, sizeof(tss), &tss);
753 if (skb_is_err_queue(skb) && skb->len &&
754 SKB_EXT_ERR(skb)->opt_stats)
755 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
756 skb->len, skb->data);
759 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
761 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
766 if (!sock_flag(sk, SOCK_WIFI_STATUS))
768 if (!skb->wifi_acked_valid)
771 ack = skb->wifi_acked;
773 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
775 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
777 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
780 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
781 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
782 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
785 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
788 sock_recv_timestamp(msg, sk, skb);
789 sock_recv_drops(msg, sk, skb);
791 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
793 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
796 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
799 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
801 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
803 return err ?: sock_recvmsg_nosec(sock, msg, flags);
805 EXPORT_SYMBOL(sock_recvmsg);
808 * kernel_recvmsg - Receive a message from a socket (kernel space)
809 * @sock: The socket to receive the message from
810 * @msg: Received message
811 * @vec: Input s/g array for message data
812 * @num: Size of input s/g array
813 * @size: Number of bytes to read
814 * @flags: Message flags (MSG_DONTWAIT, etc...)
816 * On return the msg structure contains the scatter/gather array passed in the
817 * vec argument. The array is modified so that it consists of the unfilled
818 * portion of the original array.
820 * The returned value is the total number of bytes received, or an error.
822 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
823 struct kvec *vec, size_t num, size_t size, int flags)
825 mm_segment_t oldfs = get_fs();
828 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
830 result = sock_recvmsg(sock, msg, flags);
834 EXPORT_SYMBOL(kernel_recvmsg);
836 static ssize_t sock_sendpage(struct file *file, struct page *page,
837 int offset, size_t size, loff_t *ppos, int more)
842 sock = file->private_data;
844 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
845 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
848 return kernel_sendpage(sock, page, offset, size, flags);
851 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
852 struct pipe_inode_info *pipe, size_t len,
855 struct socket *sock = file->private_data;
857 if (unlikely(!sock->ops->splice_read))
860 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
863 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
865 struct file *file = iocb->ki_filp;
866 struct socket *sock = file->private_data;
867 struct msghdr msg = {.msg_iter = *to,
871 if (file->f_flags & O_NONBLOCK)
872 msg.msg_flags = MSG_DONTWAIT;
874 if (iocb->ki_pos != 0)
877 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
880 res = sock_recvmsg(sock, &msg, msg.msg_flags);
885 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
887 struct file *file = iocb->ki_filp;
888 struct socket *sock = file->private_data;
889 struct msghdr msg = {.msg_iter = *from,
893 if (iocb->ki_pos != 0)
896 if (file->f_flags & O_NONBLOCK)
897 msg.msg_flags = MSG_DONTWAIT;
899 if (sock->type == SOCK_SEQPACKET)
900 msg.msg_flags |= MSG_EOR;
902 res = sock_sendmsg(sock, &msg);
903 *from = msg.msg_iter;
908 * Atomic setting of ioctl hooks to avoid race
909 * with module unload.
912 static DEFINE_MUTEX(br_ioctl_mutex);
913 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
915 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
917 mutex_lock(&br_ioctl_mutex);
918 br_ioctl_hook = hook;
919 mutex_unlock(&br_ioctl_mutex);
921 EXPORT_SYMBOL(brioctl_set);
923 static DEFINE_MUTEX(vlan_ioctl_mutex);
924 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
926 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
928 mutex_lock(&vlan_ioctl_mutex);
929 vlan_ioctl_hook = hook;
930 mutex_unlock(&vlan_ioctl_mutex);
932 EXPORT_SYMBOL(vlan_ioctl_set);
934 static DEFINE_MUTEX(dlci_ioctl_mutex);
935 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
937 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
939 mutex_lock(&dlci_ioctl_mutex);
940 dlci_ioctl_hook = hook;
941 mutex_unlock(&dlci_ioctl_mutex);
943 EXPORT_SYMBOL(dlci_ioctl_set);
945 static long sock_do_ioctl(struct net *net, struct socket *sock,
946 unsigned int cmd, unsigned long arg)
949 void __user *argp = (void __user *)arg;
951 err = sock->ops->ioctl(sock, cmd, arg);
954 * If this ioctl is unknown try to hand it down
957 if (err != -ENOIOCTLCMD)
960 if (cmd == SIOCGIFCONF) {
962 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
965 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
967 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
972 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
974 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
975 if (!err && need_copyout)
976 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
983 * With an ioctl, arg may well be a user mode pointer, but we don't know
984 * what to do with it - that's up to the protocol still.
987 struct ns_common *get_net_ns(struct ns_common *ns)
989 return &get_net(container_of(ns, struct net, ns))->ns;
991 EXPORT_SYMBOL_GPL(get_net_ns);
993 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
997 void __user *argp = (void __user *)arg;
1001 sock = file->private_data;
1004 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1007 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1009 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1010 if (!err && need_copyout)
1011 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1014 #ifdef CONFIG_WEXT_CORE
1015 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1016 err = wext_handle_ioctl(net, cmd, argp);
1023 if (get_user(pid, (int __user *)argp))
1025 err = f_setown(sock->file, pid, 1);
1029 err = put_user(f_getown(sock->file),
1030 (int __user *)argp);
1038 request_module("bridge");
1040 mutex_lock(&br_ioctl_mutex);
1042 err = br_ioctl_hook(net, cmd, argp);
1043 mutex_unlock(&br_ioctl_mutex);
1048 if (!vlan_ioctl_hook)
1049 request_module("8021q");
1051 mutex_lock(&vlan_ioctl_mutex);
1052 if (vlan_ioctl_hook)
1053 err = vlan_ioctl_hook(net, argp);
1054 mutex_unlock(&vlan_ioctl_mutex);
1059 if (!dlci_ioctl_hook)
1060 request_module("dlci");
1062 mutex_lock(&dlci_ioctl_mutex);
1063 if (dlci_ioctl_hook)
1064 err = dlci_ioctl_hook(cmd, argp);
1065 mutex_unlock(&dlci_ioctl_mutex);
1069 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1072 err = open_related_ns(&net->ns, get_net_ns);
1075 err = sock_do_ioctl(net, sock, cmd, arg);
1081 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1084 struct socket *sock = NULL;
1086 err = security_socket_create(family, type, protocol, 1);
1090 sock = sock_alloc();
1097 err = security_socket_post_create(sock, family, type, protocol, 1);
1109 EXPORT_SYMBOL(sock_create_lite);
1111 /* No kernel lock held - perfect */
1112 static __poll_t sock_poll(struct file *file, poll_table *wait)
1114 struct socket *sock = file->private_data;
1115 __poll_t events = poll_requested_events(wait);
1117 sock_poll_busy_loop(sock, events);
1118 if (!sock->ops->poll)
1120 return sock->ops->poll(file, sock, wait) | sock_poll_busy_flag(sock);
1123 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1125 struct socket *sock = file->private_data;
1127 return sock->ops->mmap(file, sock, vma);
1130 static int sock_close(struct inode *inode, struct file *filp)
1132 __sock_release(SOCKET_I(inode), inode);
1137 * Update the socket async list
1139 * Fasync_list locking strategy.
1141 * 1. fasync_list is modified only under process context socket lock
1142 * i.e. under semaphore.
1143 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1144 * or under socket lock
1147 static int sock_fasync(int fd, struct file *filp, int on)
1149 struct socket *sock = filp->private_data;
1150 struct sock *sk = sock->sk;
1151 struct socket_wq *wq;
1157 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1158 fasync_helper(fd, filp, on, &wq->fasync_list);
1160 if (!wq->fasync_list)
1161 sock_reset_flag(sk, SOCK_FASYNC);
1163 sock_set_flag(sk, SOCK_FASYNC);
1169 /* This function may be called only under rcu_lock */
1171 int sock_wake_async(struct socket_wq *wq, int how, int band)
1173 if (!wq || !wq->fasync_list)
1177 case SOCK_WAKE_WAITD:
1178 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1181 case SOCK_WAKE_SPACE:
1182 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1187 kill_fasync(&wq->fasync_list, SIGIO, band);
1190 kill_fasync(&wq->fasync_list, SIGURG, band);
1195 EXPORT_SYMBOL(sock_wake_async);
1197 int __sock_create(struct net *net, int family, int type, int protocol,
1198 struct socket **res, int kern)
1201 struct socket *sock;
1202 const struct net_proto_family *pf;
1205 * Check protocol is in range
1207 if (family < 0 || family >= NPROTO)
1208 return -EAFNOSUPPORT;
1209 if (type < 0 || type >= SOCK_MAX)
1214 This uglymoron is moved from INET layer to here to avoid
1215 deadlock in module load.
1217 if (family == PF_INET && type == SOCK_PACKET) {
1218 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1223 err = security_socket_create(family, type, protocol, kern);
1228 * Allocate the socket and allow the family to set things up. if
1229 * the protocol is 0, the family is instructed to select an appropriate
1232 sock = sock_alloc();
1234 net_warn_ratelimited("socket: no more sockets\n");
1235 return -ENFILE; /* Not exactly a match, but its the
1236 closest posix thing */
1241 #ifdef CONFIG_MODULES
1242 /* Attempt to load a protocol module if the find failed.
1244 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1245 * requested real, full-featured networking support upon configuration.
1246 * Otherwise module support will break!
1248 if (rcu_access_pointer(net_families[family]) == NULL)
1249 request_module("net-pf-%d", family);
1253 pf = rcu_dereference(net_families[family]);
1254 err = -EAFNOSUPPORT;
1259 * We will call the ->create function, that possibly is in a loadable
1260 * module, so we have to bump that loadable module refcnt first.
1262 if (!try_module_get(pf->owner))
1265 /* Now protected by module ref count */
1268 err = pf->create(net, sock, protocol, kern);
1270 goto out_module_put;
1273 * Now to bump the refcnt of the [loadable] module that owns this
1274 * socket at sock_release time we decrement its refcnt.
1276 if (!try_module_get(sock->ops->owner))
1277 goto out_module_busy;
1280 * Now that we're done with the ->create function, the [loadable]
1281 * module can have its refcnt decremented
1283 module_put(pf->owner);
1284 err = security_socket_post_create(sock, family, type, protocol, kern);
1286 goto out_sock_release;
1292 err = -EAFNOSUPPORT;
1295 module_put(pf->owner);
1302 goto out_sock_release;
1304 EXPORT_SYMBOL(__sock_create);
1306 int sock_create(int family, int type, int protocol, struct socket **res)
1308 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1310 EXPORT_SYMBOL(sock_create);
1312 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1314 return __sock_create(net, family, type, protocol, res, 1);
1316 EXPORT_SYMBOL(sock_create_kern);
1318 int __sys_socket(int family, int type, int protocol)
1321 struct socket *sock;
1324 /* Check the SOCK_* constants for consistency. */
1325 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1326 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1327 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1328 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1330 flags = type & ~SOCK_TYPE_MASK;
1331 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1333 type &= SOCK_TYPE_MASK;
1335 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1336 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1338 retval = sock_create(family, type, protocol, &sock);
1342 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1345 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1347 return __sys_socket(family, type, protocol);
1351 * Create a pair of connected sockets.
1354 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1356 struct socket *sock1, *sock2;
1358 struct file *newfile1, *newfile2;
1361 flags = type & ~SOCK_TYPE_MASK;
1362 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1364 type &= SOCK_TYPE_MASK;
1366 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1367 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1370 * reserve descriptors and make sure we won't fail
1371 * to return them to userland.
1373 fd1 = get_unused_fd_flags(flags);
1374 if (unlikely(fd1 < 0))
1377 fd2 = get_unused_fd_flags(flags);
1378 if (unlikely(fd2 < 0)) {
1383 err = put_user(fd1, &usockvec[0]);
1387 err = put_user(fd2, &usockvec[1]);
1392 * Obtain the first socket and check if the underlying protocol
1393 * supports the socketpair call.
1396 err = sock_create(family, type, protocol, &sock1);
1397 if (unlikely(err < 0))
1400 err = sock_create(family, type, protocol, &sock2);
1401 if (unlikely(err < 0)) {
1402 sock_release(sock1);
1406 err = security_socket_socketpair(sock1, sock2);
1407 if (unlikely(err)) {
1408 sock_release(sock2);
1409 sock_release(sock1);
1413 err = sock1->ops->socketpair(sock1, sock2);
1414 if (unlikely(err < 0)) {
1415 sock_release(sock2);
1416 sock_release(sock1);
1420 newfile1 = sock_alloc_file(sock1, flags, NULL);
1421 if (IS_ERR(newfile1)) {
1422 err = PTR_ERR(newfile1);
1423 sock_release(sock2);
1427 newfile2 = sock_alloc_file(sock2, flags, NULL);
1428 if (IS_ERR(newfile2)) {
1429 err = PTR_ERR(newfile2);
1434 audit_fd_pair(fd1, fd2);
1436 fd_install(fd1, newfile1);
1437 fd_install(fd2, newfile2);
1446 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1447 int __user *, usockvec)
1449 return __sys_socketpair(family, type, protocol, usockvec);
1453 * Bind a name to a socket. Nothing much to do here since it's
1454 * the protocol's responsibility to handle the local address.
1456 * We move the socket address to kernel space before we call
1457 * the protocol layer (having also checked the address is ok).
1460 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1462 struct socket *sock;
1463 struct sockaddr_storage address;
1464 int err, fput_needed;
1466 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1468 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1470 err = security_socket_bind(sock,
1471 (struct sockaddr *)&address,
1474 err = sock->ops->bind(sock,
1478 fput_light(sock->file, fput_needed);
1483 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1485 return __sys_bind(fd, umyaddr, addrlen);
1489 * Perform a listen. Basically, we allow the protocol to do anything
1490 * necessary for a listen, and if that works, we mark the socket as
1491 * ready for listening.
1494 int __sys_listen(int fd, int backlog)
1496 struct socket *sock;
1497 int err, fput_needed;
1500 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1502 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1503 if ((unsigned int)backlog > somaxconn)
1504 backlog = somaxconn;
1506 err = security_socket_listen(sock, backlog);
1508 err = sock->ops->listen(sock, backlog);
1510 fput_light(sock->file, fput_needed);
1515 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1517 return __sys_listen(fd, backlog);
1521 * For accept, we attempt to create a new socket, set up the link
1522 * with the client, wake up the client, then return the new
1523 * connected fd. We collect the address of the connector in kernel
1524 * space and move it to user at the very end. This is unclean because
1525 * we open the socket then return an error.
1527 * 1003.1g adds the ability to recvmsg() to query connection pending
1528 * status to recvmsg. We need to add that support in a way thats
1529 * clean when we restructure accept also.
1532 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1533 int __user *upeer_addrlen, int flags)
1535 struct socket *sock, *newsock;
1536 struct file *newfile;
1537 int err, len, newfd, fput_needed;
1538 struct sockaddr_storage address;
1540 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1543 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1544 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1546 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1551 newsock = sock_alloc();
1555 newsock->type = sock->type;
1556 newsock->ops = sock->ops;
1559 * We don't need try_module_get here, as the listening socket (sock)
1560 * has the protocol module (sock->ops->owner) held.
1562 __module_get(newsock->ops->owner);
1564 newfd = get_unused_fd_flags(flags);
1565 if (unlikely(newfd < 0)) {
1567 sock_release(newsock);
1570 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1571 if (IS_ERR(newfile)) {
1572 err = PTR_ERR(newfile);
1573 put_unused_fd(newfd);
1577 err = security_socket_accept(sock, newsock);
1581 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1585 if (upeer_sockaddr) {
1586 len = newsock->ops->getname(newsock,
1587 (struct sockaddr *)&address, 2);
1589 err = -ECONNABORTED;
1592 err = move_addr_to_user(&address,
1593 len, upeer_sockaddr, upeer_addrlen);
1598 /* File flags are not inherited via accept() unlike another OSes. */
1600 fd_install(newfd, newfile);
1604 fput_light(sock->file, fput_needed);
1609 put_unused_fd(newfd);
1613 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1614 int __user *, upeer_addrlen, int, flags)
1616 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1619 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1620 int __user *, upeer_addrlen)
1622 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1626 * Attempt to connect to a socket with the server address. The address
1627 * is in user space so we verify it is OK and move it to kernel space.
1629 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1632 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1633 * other SEQPACKET protocols that take time to connect() as it doesn't
1634 * include the -EINPROGRESS status for such sockets.
1637 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1639 struct socket *sock;
1640 struct sockaddr_storage address;
1641 int err, fput_needed;
1643 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1646 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1651 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1655 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1656 sock->file->f_flags);
1658 fput_light(sock->file, fput_needed);
1663 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1666 return __sys_connect(fd, uservaddr, addrlen);
1670 * Get the local address ('name') of a socket object. Move the obtained
1671 * name to user space.
1674 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1675 int __user *usockaddr_len)
1677 struct socket *sock;
1678 struct sockaddr_storage address;
1679 int err, fput_needed;
1681 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1685 err = security_socket_getsockname(sock);
1689 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1692 /* "err" is actually length in this case */
1693 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1696 fput_light(sock->file, fput_needed);
1701 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1702 int __user *, usockaddr_len)
1704 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1708 * Get the remote address ('name') of a socket object. Move the obtained
1709 * name to user space.
1712 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1713 int __user *usockaddr_len)
1715 struct socket *sock;
1716 struct sockaddr_storage address;
1717 int err, fput_needed;
1719 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1721 err = security_socket_getpeername(sock);
1723 fput_light(sock->file, fput_needed);
1727 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1729 /* "err" is actually length in this case */
1730 err = move_addr_to_user(&address, err, usockaddr,
1732 fput_light(sock->file, fput_needed);
1737 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1738 int __user *, usockaddr_len)
1740 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1744 * Send a datagram to a given address. We move the address into kernel
1745 * space and check the user space data area is readable before invoking
1748 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1749 struct sockaddr __user *addr, int addr_len)
1751 struct socket *sock;
1752 struct sockaddr_storage address;
1758 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1761 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765 msg.msg_name = NULL;
1766 msg.msg_control = NULL;
1767 msg.msg_controllen = 0;
1768 msg.msg_namelen = 0;
1770 err = move_addr_to_kernel(addr, addr_len, &address);
1773 msg.msg_name = (struct sockaddr *)&address;
1774 msg.msg_namelen = addr_len;
1776 if (sock->file->f_flags & O_NONBLOCK)
1777 flags |= MSG_DONTWAIT;
1778 msg.msg_flags = flags;
1779 err = sock_sendmsg(sock, &msg);
1782 fput_light(sock->file, fput_needed);
1787 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1788 unsigned int, flags, struct sockaddr __user *, addr,
1791 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1795 * Send a datagram down a socket.
1798 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1799 unsigned int, flags)
1801 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1805 * Receive a frame from the socket and optionally record the address of the
1806 * sender. We verify the buffers are writable and if needed move the
1807 * sender address from kernel to user space.
1809 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1810 struct sockaddr __user *addr, int __user *addr_len)
1812 struct socket *sock;
1815 struct sockaddr_storage address;
1819 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1822 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1826 msg.msg_control = NULL;
1827 msg.msg_controllen = 0;
1828 /* Save some cycles and don't copy the address if not needed */
1829 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1830 /* We assume all kernel code knows the size of sockaddr_storage */
1831 msg.msg_namelen = 0;
1832 msg.msg_iocb = NULL;
1834 if (sock->file->f_flags & O_NONBLOCK)
1835 flags |= MSG_DONTWAIT;
1836 err = sock_recvmsg(sock, &msg, flags);
1838 if (err >= 0 && addr != NULL) {
1839 err2 = move_addr_to_user(&address,
1840 msg.msg_namelen, addr, addr_len);
1845 fput_light(sock->file, fput_needed);
1850 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1851 unsigned int, flags, struct sockaddr __user *, addr,
1852 int __user *, addr_len)
1854 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1858 * Receive a datagram from a socket.
1861 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1862 unsigned int, flags)
1864 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1868 * Set a socket option. Because we don't know the option lengths we have
1869 * to pass the user mode parameter for the protocols to sort out.
1872 static int __sys_setsockopt(int fd, int level, int optname,
1873 char __user *optval, int optlen)
1875 int err, fput_needed;
1876 struct socket *sock;
1881 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1883 err = security_socket_setsockopt(sock, level, optname);
1887 if (level == SOL_SOCKET)
1889 sock_setsockopt(sock, level, optname, optval,
1893 sock->ops->setsockopt(sock, level, optname, optval,
1896 fput_light(sock->file, fput_needed);
1901 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1902 char __user *, optval, int, optlen)
1904 return __sys_setsockopt(fd, level, optname, optval, optlen);
1908 * Get a socket option. Because we don't know the option lengths we have
1909 * to pass a user mode parameter for the protocols to sort out.
1912 static int __sys_getsockopt(int fd, int level, int optname,
1913 char __user *optval, int __user *optlen)
1915 int err, fput_needed;
1916 struct socket *sock;
1918 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1920 err = security_socket_getsockopt(sock, level, optname);
1924 if (level == SOL_SOCKET)
1926 sock_getsockopt(sock, level, optname, optval,
1930 sock->ops->getsockopt(sock, level, optname, optval,
1933 fput_light(sock->file, fput_needed);
1938 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1939 char __user *, optval, int __user *, optlen)
1941 return __sys_getsockopt(fd, level, optname, optval, optlen);
1945 * Shutdown a socket.
1948 int __sys_shutdown(int fd, int how)
1950 int err, fput_needed;
1951 struct socket *sock;
1953 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1955 err = security_socket_shutdown(sock, how);
1957 err = sock->ops->shutdown(sock, how);
1958 fput_light(sock->file, fput_needed);
1963 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1965 return __sys_shutdown(fd, how);
1968 /* A couple of helpful macros for getting the address of the 32/64 bit
1969 * fields which are the same type (int / unsigned) on our platforms.
1971 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1972 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1973 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1975 struct used_address {
1976 struct sockaddr_storage name;
1977 unsigned int name_len;
1980 static int copy_msghdr_from_user(struct msghdr *kmsg,
1981 struct user_msghdr __user *umsg,
1982 struct sockaddr __user **save_addr,
1985 struct user_msghdr msg;
1988 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1991 kmsg->msg_control = (void __force *)msg.msg_control;
1992 kmsg->msg_controllen = msg.msg_controllen;
1993 kmsg->msg_flags = msg.msg_flags;
1995 kmsg->msg_namelen = msg.msg_namelen;
1997 kmsg->msg_namelen = 0;
1999 if (kmsg->msg_namelen < 0)
2002 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2003 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2006 *save_addr = msg.msg_name;
2008 if (msg.msg_name && kmsg->msg_namelen) {
2010 err = move_addr_to_kernel(msg.msg_name,
2017 kmsg->msg_name = NULL;
2018 kmsg->msg_namelen = 0;
2021 if (msg.msg_iovlen > UIO_MAXIOV)
2024 kmsg->msg_iocb = NULL;
2026 return import_iovec(save_addr ? READ : WRITE,
2027 msg.msg_iov, msg.msg_iovlen,
2028 UIO_FASTIOV, iov, &kmsg->msg_iter);
2031 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2032 struct msghdr *msg_sys, unsigned int flags,
2033 struct used_address *used_address,
2034 unsigned int allowed_msghdr_flags)
2036 struct compat_msghdr __user *msg_compat =
2037 (struct compat_msghdr __user *)msg;
2038 struct sockaddr_storage address;
2039 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2040 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2041 __aligned(sizeof(__kernel_size_t));
2042 /* 20 is size of ipv6_pktinfo */
2043 unsigned char *ctl_buf = ctl;
2047 msg_sys->msg_name = &address;
2049 if (MSG_CMSG_COMPAT & flags)
2050 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2052 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2058 if (msg_sys->msg_controllen > INT_MAX)
2060 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2061 ctl_len = msg_sys->msg_controllen;
2062 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2064 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2068 ctl_buf = msg_sys->msg_control;
2069 ctl_len = msg_sys->msg_controllen;
2070 } else if (ctl_len) {
2071 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2072 CMSG_ALIGN(sizeof(struct cmsghdr)));
2073 if (ctl_len > sizeof(ctl)) {
2074 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2075 if (ctl_buf == NULL)
2080 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2081 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2082 * checking falls down on this.
2084 if (copy_from_user(ctl_buf,
2085 (void __user __force *)msg_sys->msg_control,
2088 msg_sys->msg_control = ctl_buf;
2090 msg_sys->msg_flags = flags;
2092 if (sock->file->f_flags & O_NONBLOCK)
2093 msg_sys->msg_flags |= MSG_DONTWAIT;
2095 * If this is sendmmsg() and current destination address is same as
2096 * previously succeeded address, omit asking LSM's decision.
2097 * used_address->name_len is initialized to UINT_MAX so that the first
2098 * destination address never matches.
2100 if (used_address && msg_sys->msg_name &&
2101 used_address->name_len == msg_sys->msg_namelen &&
2102 !memcmp(&used_address->name, msg_sys->msg_name,
2103 used_address->name_len)) {
2104 err = sock_sendmsg_nosec(sock, msg_sys);
2107 err = sock_sendmsg(sock, msg_sys);
2109 * If this is sendmmsg() and sending to current destination address was
2110 * successful, remember it.
2112 if (used_address && err >= 0) {
2113 used_address->name_len = msg_sys->msg_namelen;
2114 if (msg_sys->msg_name)
2115 memcpy(&used_address->name, msg_sys->msg_name,
2116 used_address->name_len);
2121 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2128 * BSD sendmsg interface
2131 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2132 bool forbid_cmsg_compat)
2134 int fput_needed, err;
2135 struct msghdr msg_sys;
2136 struct socket *sock;
2138 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2141 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2145 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2147 fput_light(sock->file, fput_needed);
2152 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2154 return __sys_sendmsg(fd, msg, flags, true);
2158 * Linux sendmmsg interface
2161 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2162 unsigned int flags, bool forbid_cmsg_compat)
2164 int fput_needed, err, datagrams;
2165 struct socket *sock;
2166 struct mmsghdr __user *entry;
2167 struct compat_mmsghdr __user *compat_entry;
2168 struct msghdr msg_sys;
2169 struct used_address used_address;
2170 unsigned int oflags = flags;
2172 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2175 if (vlen > UIO_MAXIOV)
2180 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2184 used_address.name_len = UINT_MAX;
2186 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2190 while (datagrams < vlen) {
2191 if (datagrams == vlen - 1)
2194 if (MSG_CMSG_COMPAT & flags) {
2195 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2196 &msg_sys, flags, &used_address, MSG_EOR);
2199 err = __put_user(err, &compat_entry->msg_len);
2202 err = ___sys_sendmsg(sock,
2203 (struct user_msghdr __user *)entry,
2204 &msg_sys, flags, &used_address, MSG_EOR);
2207 err = put_user(err, &entry->msg_len);
2214 if (msg_data_left(&msg_sys))
2219 fput_light(sock->file, fput_needed);
2221 /* We only return an error if no datagrams were able to be sent */
2228 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2229 unsigned int, vlen, unsigned int, flags)
2231 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2234 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2235 struct msghdr *msg_sys, unsigned int flags, int nosec)
2237 struct compat_msghdr __user *msg_compat =
2238 (struct compat_msghdr __user *)msg;
2239 struct iovec iovstack[UIO_FASTIOV];
2240 struct iovec *iov = iovstack;
2241 unsigned long cmsg_ptr;
2245 /* kernel mode address */
2246 struct sockaddr_storage addr;
2248 /* user mode address pointers */
2249 struct sockaddr __user *uaddr;
2250 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2252 msg_sys->msg_name = &addr;
2254 if (MSG_CMSG_COMPAT & flags)
2255 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2257 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2261 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2262 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2264 /* We assume all kernel code knows the size of sockaddr_storage */
2265 msg_sys->msg_namelen = 0;
2267 if (sock->file->f_flags & O_NONBLOCK)
2268 flags |= MSG_DONTWAIT;
2269 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2274 if (uaddr != NULL) {
2275 err = move_addr_to_user(&addr,
2276 msg_sys->msg_namelen, uaddr,
2281 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2285 if (MSG_CMSG_COMPAT & flags)
2286 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2287 &msg_compat->msg_controllen);
2289 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2290 &msg->msg_controllen);
2301 * BSD recvmsg interface
2304 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2305 bool forbid_cmsg_compat)
2307 int fput_needed, err;
2308 struct msghdr msg_sys;
2309 struct socket *sock;
2311 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2314 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2318 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2320 fput_light(sock->file, fput_needed);
2325 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2326 unsigned int, flags)
2328 return __sys_recvmsg(fd, msg, flags, true);
2332 * Linux recvmmsg interface
2335 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2336 unsigned int flags, struct timespec *timeout)
2338 int fput_needed, err, datagrams;
2339 struct socket *sock;
2340 struct mmsghdr __user *entry;
2341 struct compat_mmsghdr __user *compat_entry;
2342 struct msghdr msg_sys;
2343 struct timespec64 end_time;
2344 struct timespec64 timeout64;
2347 poll_select_set_timeout(&end_time, timeout->tv_sec,
2353 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2357 if (likely(!(flags & MSG_ERRQUEUE))) {
2358 err = sock_error(sock->sk);
2366 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2368 while (datagrams < vlen) {
2370 * No need to ask LSM for more than the first datagram.
2372 if (MSG_CMSG_COMPAT & flags) {
2373 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2374 &msg_sys, flags & ~MSG_WAITFORONE,
2378 err = __put_user(err, &compat_entry->msg_len);
2381 err = ___sys_recvmsg(sock,
2382 (struct user_msghdr __user *)entry,
2383 &msg_sys, flags & ~MSG_WAITFORONE,
2387 err = put_user(err, &entry->msg_len);
2395 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2396 if (flags & MSG_WAITFORONE)
2397 flags |= MSG_DONTWAIT;
2400 ktime_get_ts64(&timeout64);
2401 *timeout = timespec64_to_timespec(
2402 timespec64_sub(end_time, timeout64));
2403 if (timeout->tv_sec < 0) {
2404 timeout->tv_sec = timeout->tv_nsec = 0;
2408 /* Timeout, return less than vlen datagrams */
2409 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2413 /* Out of band data, return right away */
2414 if (msg_sys.msg_flags & MSG_OOB)
2422 if (datagrams == 0) {
2428 * We may return less entries than requested (vlen) if the
2429 * sock is non block and there aren't enough datagrams...
2431 if (err != -EAGAIN) {
2433 * ... or if recvmsg returns an error after we
2434 * received some datagrams, where we record the
2435 * error to return on the next call or if the
2436 * app asks about it using getsockopt(SO_ERROR).
2438 sock->sk->sk_err = -err;
2441 fput_light(sock->file, fput_needed);
2446 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2447 unsigned int vlen, unsigned int flags,
2448 struct timespec __user *timeout)
2451 struct timespec timeout_sys;
2453 if (flags & MSG_CMSG_COMPAT)
2457 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2459 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2462 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2464 if (datagrams > 0 &&
2465 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2466 datagrams = -EFAULT;
2471 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2472 unsigned int, vlen, unsigned int, flags,
2473 struct timespec __user *, timeout)
2475 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2478 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2479 /* Argument list sizes for sys_socketcall */
2480 #define AL(x) ((x) * sizeof(unsigned long))
2481 static const unsigned char nargs[21] = {
2482 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2483 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2484 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2491 * System call vectors.
2493 * Argument checking cleaned up. Saved 20% in size.
2494 * This function doesn't need to set the kernel lock because
2495 * it is set by the callees.
2498 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2500 unsigned long a[AUDITSC_ARGS];
2501 unsigned long a0, a1;
2505 if (call < 1 || call > SYS_SENDMMSG)
2507 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2510 if (len > sizeof(a))
2513 /* copy_from_user should be SMP safe. */
2514 if (copy_from_user(a, args, len))
2517 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2526 err = __sys_socket(a0, a1, a[2]);
2529 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2532 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2535 err = __sys_listen(a0, a1);
2538 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2539 (int __user *)a[2], 0);
2541 case SYS_GETSOCKNAME:
2543 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2544 (int __user *)a[2]);
2546 case SYS_GETPEERNAME:
2548 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2549 (int __user *)a[2]);
2551 case SYS_SOCKETPAIR:
2552 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2555 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2559 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2560 (struct sockaddr __user *)a[4], a[5]);
2563 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2567 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2568 (struct sockaddr __user *)a[4],
2569 (int __user *)a[5]);
2572 err = __sys_shutdown(a0, a1);
2574 case SYS_SETSOCKOPT:
2575 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2578 case SYS_GETSOCKOPT:
2580 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2581 (int __user *)a[4]);
2584 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2588 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2592 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2596 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2597 a[3], (struct timespec __user *)a[4]);
2600 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2601 (int __user *)a[2], a[3]);
2610 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2613 * sock_register - add a socket protocol handler
2614 * @ops: description of protocol
2616 * This function is called by a protocol handler that wants to
2617 * advertise its address family, and have it linked into the
2618 * socket interface. The value ops->family corresponds to the
2619 * socket system call protocol family.
2621 int sock_register(const struct net_proto_family *ops)
2625 if (ops->family >= NPROTO) {
2626 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2630 spin_lock(&net_family_lock);
2631 if (rcu_dereference_protected(net_families[ops->family],
2632 lockdep_is_held(&net_family_lock)))
2635 rcu_assign_pointer(net_families[ops->family], ops);
2638 spin_unlock(&net_family_lock);
2640 pr_info("NET: Registered protocol family %d\n", ops->family);
2643 EXPORT_SYMBOL(sock_register);
2646 * sock_unregister - remove a protocol handler
2647 * @family: protocol family to remove
2649 * This function is called by a protocol handler that wants to
2650 * remove its address family, and have it unlinked from the
2651 * new socket creation.
2653 * If protocol handler is a module, then it can use module reference
2654 * counts to protect against new references. If protocol handler is not
2655 * a module then it needs to provide its own protection in
2656 * the ops->create routine.
2658 void sock_unregister(int family)
2660 BUG_ON(family < 0 || family >= NPROTO);
2662 spin_lock(&net_family_lock);
2663 RCU_INIT_POINTER(net_families[family], NULL);
2664 spin_unlock(&net_family_lock);
2668 pr_info("NET: Unregistered protocol family %d\n", family);
2670 EXPORT_SYMBOL(sock_unregister);
2672 bool sock_is_registered(int family)
2674 return family < NPROTO &&
2675 rcu_access_pointer(net_families[array_index_nospec(family, NPROTO)]);
2678 static int __init sock_init(void)
2682 * Initialize the network sysctl infrastructure.
2684 err = net_sysctl_init();
2689 * Initialize skbuff SLAB cache
2694 * Initialize the protocols module.
2699 err = register_filesystem(&sock_fs_type);
2702 sock_mnt = kern_mount(&sock_fs_type);
2703 if (IS_ERR(sock_mnt)) {
2704 err = PTR_ERR(sock_mnt);
2708 /* The real protocol initialization is performed in later initcalls.
2711 #ifdef CONFIG_NETFILTER
2712 err = netfilter_init();
2717 ptp_classifier_init();
2723 unregister_filesystem(&sock_fs_type);
2728 core_initcall(sock_init); /* early initcall */
2730 #ifdef CONFIG_PROC_FS
2731 void socket_seq_show(struct seq_file *seq)
2733 seq_printf(seq, "sockets: used %d\n",
2734 sock_inuse_get(seq->private));
2736 #endif /* CONFIG_PROC_FS */
2738 #ifdef CONFIG_COMPAT
2739 static int do_siocgstamp(struct net *net, struct socket *sock,
2740 unsigned int cmd, void __user *up)
2742 mm_segment_t old_fs = get_fs();
2747 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2750 err = compat_put_timeval(&ktv, up);
2755 static int do_siocgstampns(struct net *net, struct socket *sock,
2756 unsigned int cmd, void __user *up)
2758 mm_segment_t old_fs = get_fs();
2759 struct timespec kts;
2763 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2766 err = compat_put_timespec(&kts, up);
2771 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2773 struct compat_ifconf ifc32;
2777 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2780 ifc.ifc_len = ifc32.ifc_len;
2781 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2784 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2789 ifc32.ifc_len = ifc.ifc_len;
2790 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2796 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2798 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2799 bool convert_in = false, convert_out = false;
2800 size_t buf_size = 0;
2801 struct ethtool_rxnfc __user *rxnfc = NULL;
2803 u32 rule_cnt = 0, actual_rule_cnt;
2808 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2811 compat_rxnfc = compat_ptr(data);
2813 if (get_user(ethcmd, &compat_rxnfc->cmd))
2816 /* Most ethtool structures are defined without padding.
2817 * Unfortunately struct ethtool_rxnfc is an exception.
2822 case ETHTOOL_GRXCLSRLALL:
2823 /* Buffer size is variable */
2824 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2826 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2828 buf_size += rule_cnt * sizeof(u32);
2830 case ETHTOOL_GRXRINGS:
2831 case ETHTOOL_GRXCLSRLCNT:
2832 case ETHTOOL_GRXCLSRULE:
2833 case ETHTOOL_SRXCLSRLINS:
2836 case ETHTOOL_SRXCLSRLDEL:
2837 buf_size += sizeof(struct ethtool_rxnfc);
2839 rxnfc = compat_alloc_user_space(buf_size);
2843 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2846 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2849 /* We expect there to be holes between fs.m_ext and
2850 * fs.ring_cookie and at the end of fs, but nowhere else.
2852 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2853 sizeof(compat_rxnfc->fs.m_ext) !=
2854 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2855 sizeof(rxnfc->fs.m_ext));
2857 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2858 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2859 offsetof(struct ethtool_rxnfc, fs.location) -
2860 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2862 if (copy_in_user(rxnfc, compat_rxnfc,
2863 (void __user *)(&rxnfc->fs.m_ext + 1) -
2864 (void __user *)rxnfc) ||
2865 copy_in_user(&rxnfc->fs.ring_cookie,
2866 &compat_rxnfc->fs.ring_cookie,
2867 (void __user *)(&rxnfc->fs.location + 1) -
2868 (void __user *)&rxnfc->fs.ring_cookie) ||
2869 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2870 sizeof(rxnfc->rule_cnt)))
2874 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2879 if (copy_in_user(compat_rxnfc, rxnfc,
2880 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2881 (const void __user *)rxnfc) ||
2882 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2883 &rxnfc->fs.ring_cookie,
2884 (const void __user *)(&rxnfc->fs.location + 1) -
2885 (const void __user *)&rxnfc->fs.ring_cookie) ||
2886 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2887 sizeof(rxnfc->rule_cnt)))
2890 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2891 /* As an optimisation, we only copy the actual
2892 * number of rules that the underlying
2893 * function returned. Since Mallory might
2894 * change the rule count in user memory, we
2895 * check that it is less than the rule count
2896 * originally given (as the user buffer size),
2897 * which has been range-checked.
2899 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2901 if (actual_rule_cnt < rule_cnt)
2902 rule_cnt = actual_rule_cnt;
2903 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2904 &rxnfc->rule_locs[0],
2905 rule_cnt * sizeof(u32)))
2913 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2915 compat_uptr_t uptr32;
2920 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2923 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2926 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2927 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2929 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2931 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2932 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2938 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2939 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2940 struct compat_ifreq __user *u_ifreq32)
2945 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2947 if (get_user(data32, &u_ifreq32->ifr_data))
2949 ifreq.ifr_data = compat_ptr(data32);
2951 return dev_ioctl(net, cmd, &ifreq, NULL);
2954 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2955 struct compat_ifreq __user *uifr32)
2958 struct compat_ifmap __user *uifmap32;
2961 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2962 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2963 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2964 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2965 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2966 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2967 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2968 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2972 err = dev_ioctl(net, cmd, &ifr, NULL);
2974 if (cmd == SIOCGIFMAP && !err) {
2975 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2976 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2977 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2978 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2979 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2980 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2981 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2990 struct sockaddr rt_dst; /* target address */
2991 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2992 struct sockaddr rt_genmask; /* target network mask (IP) */
2993 unsigned short rt_flags;
2996 unsigned char rt_tos;
2997 unsigned char rt_class;
2999 short rt_metric; /* +1 for binary compatibility! */
3000 /* char * */ u32 rt_dev; /* forcing the device at add */
3001 u32 rt_mtu; /* per route MTU/Window */
3002 u32 rt_window; /* Window clamping */
3003 unsigned short rt_irtt; /* Initial RTT */
3006 struct in6_rtmsg32 {
3007 struct in6_addr rtmsg_dst;
3008 struct in6_addr rtmsg_src;
3009 struct in6_addr rtmsg_gateway;
3019 static int routing_ioctl(struct net *net, struct socket *sock,
3020 unsigned int cmd, void __user *argp)
3024 struct in6_rtmsg r6;
3028 mm_segment_t old_fs = get_fs();
3030 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3031 struct in6_rtmsg32 __user *ur6 = argp;
3032 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3033 3 * sizeof(struct in6_addr));
3034 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3035 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3036 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3037 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3038 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3039 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3040 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3044 struct rtentry32 __user *ur4 = argp;
3045 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3046 3 * sizeof(struct sockaddr));
3047 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3048 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3049 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3050 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3051 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3052 ret |= get_user(rtdev, &(ur4->rt_dev));
3054 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3055 r4.rt_dev = (char __user __force *)devname;
3069 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3076 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3077 * for some operations; this forces use of the newer bridge-utils that
3078 * use compatible ioctls
3080 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3084 if (get_user(tmp, argp))
3086 if (tmp == BRCTL_GET_VERSION)
3087 return BRCTL_VERSION + 1;
3091 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3092 unsigned int cmd, unsigned long arg)
3094 void __user *argp = compat_ptr(arg);
3095 struct sock *sk = sock->sk;
3096 struct net *net = sock_net(sk);
3098 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3099 return compat_ifr_data_ioctl(net, cmd, argp);
3104 return old_bridge_ioctl(argp);
3106 return compat_dev_ifconf(net, argp);
3108 return ethtool_ioctl(net, argp);
3110 return compat_siocwandev(net, argp);
3113 return compat_sioc_ifmap(net, cmd, argp);
3116 return routing_ioctl(net, sock, cmd, argp);
3118 return do_siocgstamp(net, sock, cmd, argp);
3120 return do_siocgstampns(net, sock, cmd, argp);
3121 case SIOCBONDSLAVEINFOQUERY:
3122 case SIOCBONDINFOQUERY:
3125 return compat_ifr_data_ioctl(net, cmd, argp);
3138 return sock_ioctl(file, cmd, arg);
3155 case SIOCSIFHWBROADCAST:
3157 case SIOCGIFBRDADDR:
3158 case SIOCSIFBRDADDR:
3159 case SIOCGIFDSTADDR:
3160 case SIOCSIFDSTADDR:
3161 case SIOCGIFNETMASK:
3162 case SIOCSIFNETMASK:
3177 case SIOCBONDENSLAVE:
3178 case SIOCBONDRELEASE:
3179 case SIOCBONDSETHWADDR:
3180 case SIOCBONDCHANGEACTIVE:
3182 return sock_do_ioctl(net, sock, cmd, arg);
3185 return -ENOIOCTLCMD;
3188 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3191 struct socket *sock = file->private_data;
3192 int ret = -ENOIOCTLCMD;
3199 if (sock->ops->compat_ioctl)
3200 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3202 if (ret == -ENOIOCTLCMD &&
3203 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3204 ret = compat_wext_handle_ioctl(net, cmd, arg);
3206 if (ret == -ENOIOCTLCMD)
3207 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3213 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3215 return sock->ops->bind(sock, addr, addrlen);
3217 EXPORT_SYMBOL(kernel_bind);
3219 int kernel_listen(struct socket *sock, int backlog)
3221 return sock->ops->listen(sock, backlog);
3223 EXPORT_SYMBOL(kernel_listen);
3225 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3227 struct sock *sk = sock->sk;
3230 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3235 err = sock->ops->accept(sock, *newsock, flags, true);
3237 sock_release(*newsock);
3242 (*newsock)->ops = sock->ops;
3243 __module_get((*newsock)->ops->owner);
3248 EXPORT_SYMBOL(kernel_accept);
3250 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3253 return sock->ops->connect(sock, addr, addrlen, flags);
3255 EXPORT_SYMBOL(kernel_connect);
3257 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3259 return sock->ops->getname(sock, addr, 0);
3261 EXPORT_SYMBOL(kernel_getsockname);
3263 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3265 return sock->ops->getname(sock, addr, 1);
3267 EXPORT_SYMBOL(kernel_getpeername);
3269 int kernel_getsockopt(struct socket *sock, int level, int optname,
3270 char *optval, int *optlen)
3272 mm_segment_t oldfs = get_fs();
3273 char __user *uoptval;
3274 int __user *uoptlen;
3277 uoptval = (char __user __force *) optval;
3278 uoptlen = (int __user __force *) optlen;
3281 if (level == SOL_SOCKET)
3282 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3284 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3289 EXPORT_SYMBOL(kernel_getsockopt);
3291 int kernel_setsockopt(struct socket *sock, int level, int optname,
3292 char *optval, unsigned int optlen)
3294 mm_segment_t oldfs = get_fs();
3295 char __user *uoptval;
3298 uoptval = (char __user __force *) optval;
3301 if (level == SOL_SOCKET)
3302 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3304 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3309 EXPORT_SYMBOL(kernel_setsockopt);
3311 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3312 size_t size, int flags)
3314 if (sock->ops->sendpage)
3315 return sock->ops->sendpage(sock, page, offset, size, flags);
3317 return sock_no_sendpage(sock, page, offset, size, flags);
3319 EXPORT_SYMBOL(kernel_sendpage);
3321 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3322 size_t size, int flags)
3324 struct socket *sock = sk->sk_socket;
3326 if (sock->ops->sendpage_locked)
3327 return sock->ops->sendpage_locked(sk, page, offset, size,
3330 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3332 EXPORT_SYMBOL(kernel_sendpage_locked);
3334 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3336 return sock->ops->shutdown(sock, how);
3338 EXPORT_SYMBOL(kernel_sock_shutdown);
3340 /* This routine returns the IP overhead imposed by a socket i.e.
3341 * the length of the underlying IP header, depending on whether
3342 * this is an IPv4 or IPv6 socket and the length from IP options turned
3343 * on at the socket. Assumes that the caller has a lock on the socket.
3345 u32 kernel_sock_ip_overhead(struct sock *sk)
3347 struct inet_sock *inet;
3348 struct ip_options_rcu *opt;
3350 #if IS_ENABLED(CONFIG_IPV6)
3351 struct ipv6_pinfo *np;
3352 struct ipv6_txoptions *optv6 = NULL;
3353 #endif /* IS_ENABLED(CONFIG_IPV6) */
3358 switch (sk->sk_family) {
3361 overhead += sizeof(struct iphdr);
3362 opt = rcu_dereference_protected(inet->inet_opt,
3363 sock_owned_by_user(sk));
3365 overhead += opt->opt.optlen;
3367 #if IS_ENABLED(CONFIG_IPV6)
3370 overhead += sizeof(struct ipv6hdr);
3372 optv6 = rcu_dereference_protected(np->opt,
3373 sock_owned_by_user(sk));
3375 overhead += (optv6->opt_flen + optv6->opt_nflen);
3377 #endif /* IS_ENABLED(CONFIG_IPV6) */
3378 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3382 EXPORT_SYMBOL(kernel_sock_ip_overhead);