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)
391 struct qstr name = { .name = "" };
397 name.len = strlen(name.name);
398 } else if (sock->sk) {
399 name.name = sock->sk->sk_prot_creator->name;
400 name.len = strlen(name.name);
402 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
403 if (unlikely(!path.dentry)) {
405 return ERR_PTR(-ENOMEM);
407 path.mnt = mntget(sock_mnt);
409 d_instantiate(path.dentry, SOCK_INODE(sock));
411 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
414 /* drop dentry, keep inode for a bit */
415 ihold(d_inode(path.dentry));
417 /* ... and now kill it properly */
423 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
424 file->private_data = sock;
427 EXPORT_SYMBOL(sock_alloc_file);
429 static int sock_map_fd(struct socket *sock, int flags)
431 struct file *newfile;
432 int fd = get_unused_fd_flags(flags);
433 if (unlikely(fd < 0)) {
438 newfile = sock_alloc_file(sock, flags, NULL);
439 if (likely(!IS_ERR(newfile))) {
440 fd_install(fd, newfile);
445 return PTR_ERR(newfile);
448 struct socket *sock_from_file(struct file *file, int *err)
450 if (file->f_op == &socket_file_ops)
451 return file->private_data; /* set in sock_map_fd */
456 EXPORT_SYMBOL(sock_from_file);
459 * sockfd_lookup - Go from a file number to its socket slot
461 * @err: pointer to an error code return
463 * The file handle passed in is locked and the socket it is bound
464 * to is returned. If an error occurs the err pointer is overwritten
465 * with a negative errno code and NULL is returned. The function checks
466 * for both invalid handles and passing a handle which is not a socket.
468 * On a success the socket object pointer is returned.
471 struct socket *sockfd_lookup(int fd, int *err)
482 sock = sock_from_file(file, err);
487 EXPORT_SYMBOL(sockfd_lookup);
489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
491 struct fd f = fdget(fd);
496 sock = sock_from_file(f.file, err);
498 *fput_needed = f.flags;
506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
512 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
522 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
527 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
536 int err = simple_setattr(dentry, iattr);
538 if (!err && (iattr->ia_valid & ATTR_UID)) {
539 struct socket *sock = SOCKET_I(d_inode(dentry));
542 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 static void __sock_release(struct socket *sock, struct inode *inode)
596 struct module *owner = sock->ops->owner;
600 sock->ops->release(sock);
607 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
608 pr_err("%s: fasync list not empty!\n", __func__);
611 iput(SOCK_INODE(sock));
617 void sock_release(struct socket *sock)
619 __sock_release(sock, NULL);
621 EXPORT_SYMBOL(sock_release);
623 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
625 u8 flags = *tx_flags;
627 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
628 flags |= SKBTX_HW_TSTAMP;
630 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
631 flags |= SKBTX_SW_TSTAMP;
633 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
634 flags |= SKBTX_SCHED_TSTAMP;
638 EXPORT_SYMBOL(__sock_tx_timestamp);
640 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
642 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
643 BUG_ON(ret == -EIOCBQUEUED);
647 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
649 int err = security_socket_sendmsg(sock, msg,
652 return err ?: sock_sendmsg_nosec(sock, msg);
654 EXPORT_SYMBOL(sock_sendmsg);
656 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
657 struct kvec *vec, size_t num, size_t size)
659 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
660 return sock_sendmsg(sock, msg);
662 EXPORT_SYMBOL(kernel_sendmsg);
664 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
665 struct kvec *vec, size_t num, size_t size)
667 struct socket *sock = sk->sk_socket;
669 if (!sock->ops->sendmsg_locked)
670 return sock_no_sendmsg_locked(sk, msg, size);
672 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
674 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
676 EXPORT_SYMBOL(kernel_sendmsg_locked);
678 static bool skb_is_err_queue(const struct sk_buff *skb)
680 /* pkt_type of skbs enqueued on the error queue are set to
681 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
682 * in recvmsg, since skbs received on a local socket will never
683 * have a pkt_type of PACKET_OUTGOING.
685 return skb->pkt_type == PACKET_OUTGOING;
688 /* On transmit, software and hardware timestamps are returned independently.
689 * As the two skb clones share the hardware timestamp, which may be updated
690 * before the software timestamp is received, a hardware TX timestamp may be
691 * returned only if there is no software TX timestamp. Ignore false software
692 * timestamps, which may be made in the __sock_recv_timestamp() call when the
693 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
694 * hardware timestamp.
696 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
698 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
701 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
703 struct scm_ts_pktinfo ts_pktinfo;
704 struct net_device *orig_dev;
706 if (!skb_mac_header_was_set(skb))
709 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
712 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
714 ts_pktinfo.if_index = orig_dev->ifindex;
717 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
718 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
719 sizeof(ts_pktinfo), &ts_pktinfo);
723 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
725 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
728 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
729 struct scm_timestamping tss;
730 int empty = 1, false_tstamp = 0;
731 struct skb_shared_hwtstamps *shhwtstamps =
734 /* Race occurred between timestamp enabling and packet
735 receiving. Fill in the current time for now. */
736 if (need_software_tstamp && skb->tstamp == 0) {
737 __net_timestamp(skb);
741 if (need_software_tstamp) {
742 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
744 skb_get_timestamp(skb, &tv);
745 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
749 skb_get_timestampns(skb, &ts);
750 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
755 memset(&tss, 0, sizeof(tss));
756 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
757 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
760 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
761 !skb_is_swtx_tstamp(skb, false_tstamp) &&
762 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
764 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
765 !skb_is_err_queue(skb))
766 put_ts_pktinfo(msg, skb);
769 put_cmsg(msg, SOL_SOCKET,
770 SCM_TIMESTAMPING, sizeof(tss), &tss);
772 if (skb_is_err_queue(skb) && skb->len &&
773 SKB_EXT_ERR(skb)->opt_stats)
774 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
775 skb->len, skb->data);
778 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
780 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
785 if (!sock_flag(sk, SOCK_WIFI_STATUS))
787 if (!skb->wifi_acked_valid)
790 ack = skb->wifi_acked;
792 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
794 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
796 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
799 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
800 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
801 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
804 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
807 sock_recv_timestamp(msg, sk, skb);
808 sock_recv_drops(msg, sk, skb);
810 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
812 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
815 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
818 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
820 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
822 return err ?: sock_recvmsg_nosec(sock, msg, flags);
824 EXPORT_SYMBOL(sock_recvmsg);
827 * kernel_recvmsg - Receive a message from a socket (kernel space)
828 * @sock: The socket to receive the message from
829 * @msg: Received message
830 * @vec: Input s/g array for message data
831 * @num: Size of input s/g array
832 * @size: Number of bytes to read
833 * @flags: Message flags (MSG_DONTWAIT, etc...)
835 * On return the msg structure contains the scatter/gather array passed in the
836 * vec argument. The array is modified so that it consists of the unfilled
837 * portion of the original array.
839 * The returned value is the total number of bytes received, or an error.
841 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
842 struct kvec *vec, size_t num, size_t size, int flags)
844 mm_segment_t oldfs = get_fs();
847 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
849 result = sock_recvmsg(sock, msg, flags);
853 EXPORT_SYMBOL(kernel_recvmsg);
855 static ssize_t sock_sendpage(struct file *file, struct page *page,
856 int offset, size_t size, loff_t *ppos, int more)
861 sock = file->private_data;
863 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
864 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
867 return kernel_sendpage(sock, page, offset, size, flags);
870 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
871 struct pipe_inode_info *pipe, size_t len,
874 struct socket *sock = file->private_data;
876 if (unlikely(!sock->ops->splice_read))
879 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
882 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
884 struct file *file = iocb->ki_filp;
885 struct socket *sock = file->private_data;
886 struct msghdr msg = {.msg_iter = *to,
890 if (file->f_flags & O_NONBLOCK)
891 msg.msg_flags = MSG_DONTWAIT;
893 if (iocb->ki_pos != 0)
896 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
899 res = sock_recvmsg(sock, &msg, msg.msg_flags);
904 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
906 struct file *file = iocb->ki_filp;
907 struct socket *sock = file->private_data;
908 struct msghdr msg = {.msg_iter = *from,
912 if (iocb->ki_pos != 0)
915 if (file->f_flags & O_NONBLOCK)
916 msg.msg_flags = MSG_DONTWAIT;
918 if (sock->type == SOCK_SEQPACKET)
919 msg.msg_flags |= MSG_EOR;
921 res = sock_sendmsg(sock, &msg);
922 *from = msg.msg_iter;
927 * Atomic setting of ioctl hooks to avoid race
928 * with module unload.
931 static DEFINE_MUTEX(br_ioctl_mutex);
932 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
934 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
936 mutex_lock(&br_ioctl_mutex);
937 br_ioctl_hook = hook;
938 mutex_unlock(&br_ioctl_mutex);
940 EXPORT_SYMBOL(brioctl_set);
942 static DEFINE_MUTEX(vlan_ioctl_mutex);
943 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
945 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
947 mutex_lock(&vlan_ioctl_mutex);
948 vlan_ioctl_hook = hook;
949 mutex_unlock(&vlan_ioctl_mutex);
951 EXPORT_SYMBOL(vlan_ioctl_set);
953 static DEFINE_MUTEX(dlci_ioctl_mutex);
954 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
956 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
958 mutex_lock(&dlci_ioctl_mutex);
959 dlci_ioctl_hook = hook;
960 mutex_unlock(&dlci_ioctl_mutex);
962 EXPORT_SYMBOL(dlci_ioctl_set);
964 static long sock_do_ioctl(struct net *net, struct socket *sock,
965 unsigned int cmd, unsigned long arg)
968 void __user *argp = (void __user *)arg;
970 err = sock->ops->ioctl(sock, cmd, arg);
973 * If this ioctl is unknown try to hand it down
976 if (err != -ENOIOCTLCMD)
979 if (cmd == SIOCGIFCONF) {
981 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
984 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
986 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
991 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
993 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
994 if (!err && need_copyout)
995 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1002 * With an ioctl, arg may well be a user mode pointer, but we don't know
1003 * what to do with it - that's up to the protocol still.
1006 struct ns_common *get_net_ns(struct ns_common *ns)
1008 return &get_net(container_of(ns, struct net, ns))->ns;
1010 EXPORT_SYMBOL_GPL(get_net_ns);
1012 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1014 struct socket *sock;
1016 void __user *argp = (void __user *)arg;
1020 sock = file->private_data;
1023 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1026 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1028 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1029 if (!err && need_copyout)
1030 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1033 #ifdef CONFIG_WEXT_CORE
1034 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1035 err = wext_handle_ioctl(net, cmd, argp);
1042 if (get_user(pid, (int __user *)argp))
1044 err = f_setown(sock->file, pid, 1);
1048 err = put_user(f_getown(sock->file),
1049 (int __user *)argp);
1057 request_module("bridge");
1059 mutex_lock(&br_ioctl_mutex);
1061 err = br_ioctl_hook(net, cmd, argp);
1062 mutex_unlock(&br_ioctl_mutex);
1067 if (!vlan_ioctl_hook)
1068 request_module("8021q");
1070 mutex_lock(&vlan_ioctl_mutex);
1071 if (vlan_ioctl_hook)
1072 err = vlan_ioctl_hook(net, argp);
1073 mutex_unlock(&vlan_ioctl_mutex);
1078 if (!dlci_ioctl_hook)
1079 request_module("dlci");
1081 mutex_lock(&dlci_ioctl_mutex);
1082 if (dlci_ioctl_hook)
1083 err = dlci_ioctl_hook(cmd, argp);
1084 mutex_unlock(&dlci_ioctl_mutex);
1088 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1091 err = open_related_ns(&net->ns, get_net_ns);
1094 err = sock_do_ioctl(net, sock, cmd, arg);
1100 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1103 struct socket *sock = NULL;
1105 err = security_socket_create(family, type, protocol, 1);
1109 sock = sock_alloc();
1116 err = security_socket_post_create(sock, family, type, protocol, 1);
1128 EXPORT_SYMBOL(sock_create_lite);
1130 /* No kernel lock held - perfect */
1131 static __poll_t sock_poll(struct file *file, poll_table *wait)
1133 struct socket *sock = file->private_data;
1134 __poll_t events = poll_requested_events(wait);
1136 sock_poll_busy_loop(sock, events);
1137 if (!sock->ops->poll)
1139 return sock->ops->poll(file, sock, wait) | sock_poll_busy_flag(sock);
1142 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1144 struct socket *sock = file->private_data;
1146 return sock->ops->mmap(file, sock, vma);
1149 static int sock_close(struct inode *inode, struct file *filp)
1151 __sock_release(SOCKET_I(inode), inode);
1156 * Update the socket async list
1158 * Fasync_list locking strategy.
1160 * 1. fasync_list is modified only under process context socket lock
1161 * i.e. under semaphore.
1162 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1163 * or under socket lock
1166 static int sock_fasync(int fd, struct file *filp, int on)
1168 struct socket *sock = filp->private_data;
1169 struct sock *sk = sock->sk;
1170 struct socket_wq *wq;
1176 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1177 fasync_helper(fd, filp, on, &wq->fasync_list);
1179 if (!wq->fasync_list)
1180 sock_reset_flag(sk, SOCK_FASYNC);
1182 sock_set_flag(sk, SOCK_FASYNC);
1188 /* This function may be called only under rcu_lock */
1190 int sock_wake_async(struct socket_wq *wq, int how, int band)
1192 if (!wq || !wq->fasync_list)
1196 case SOCK_WAKE_WAITD:
1197 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1200 case SOCK_WAKE_SPACE:
1201 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1206 kill_fasync(&wq->fasync_list, SIGIO, band);
1209 kill_fasync(&wq->fasync_list, SIGURG, band);
1214 EXPORT_SYMBOL(sock_wake_async);
1216 int __sock_create(struct net *net, int family, int type, int protocol,
1217 struct socket **res, int kern)
1220 struct socket *sock;
1221 const struct net_proto_family *pf;
1224 * Check protocol is in range
1226 if (family < 0 || family >= NPROTO)
1227 return -EAFNOSUPPORT;
1228 if (type < 0 || type >= SOCK_MAX)
1233 This uglymoron is moved from INET layer to here to avoid
1234 deadlock in module load.
1236 if (family == PF_INET && type == SOCK_PACKET) {
1237 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1242 err = security_socket_create(family, type, protocol, kern);
1247 * Allocate the socket and allow the family to set things up. if
1248 * the protocol is 0, the family is instructed to select an appropriate
1251 sock = sock_alloc();
1253 net_warn_ratelimited("socket: no more sockets\n");
1254 return -ENFILE; /* Not exactly a match, but its the
1255 closest posix thing */
1260 #ifdef CONFIG_MODULES
1261 /* Attempt to load a protocol module if the find failed.
1263 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1264 * requested real, full-featured networking support upon configuration.
1265 * Otherwise module support will break!
1267 if (rcu_access_pointer(net_families[family]) == NULL)
1268 request_module("net-pf-%d", family);
1272 pf = rcu_dereference(net_families[family]);
1273 err = -EAFNOSUPPORT;
1278 * We will call the ->create function, that possibly is in a loadable
1279 * module, so we have to bump that loadable module refcnt first.
1281 if (!try_module_get(pf->owner))
1284 /* Now protected by module ref count */
1287 err = pf->create(net, sock, protocol, kern);
1289 goto out_module_put;
1292 * Now to bump the refcnt of the [loadable] module that owns this
1293 * socket at sock_release time we decrement its refcnt.
1295 if (!try_module_get(sock->ops->owner))
1296 goto out_module_busy;
1299 * Now that we're done with the ->create function, the [loadable]
1300 * module can have its refcnt decremented
1302 module_put(pf->owner);
1303 err = security_socket_post_create(sock, family, type, protocol, kern);
1305 goto out_sock_release;
1311 err = -EAFNOSUPPORT;
1314 module_put(pf->owner);
1321 goto out_sock_release;
1323 EXPORT_SYMBOL(__sock_create);
1325 int sock_create(int family, int type, int protocol, struct socket **res)
1327 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1329 EXPORT_SYMBOL(sock_create);
1331 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1333 return __sock_create(net, family, type, protocol, res, 1);
1335 EXPORT_SYMBOL(sock_create_kern);
1337 int __sys_socket(int family, int type, int protocol)
1340 struct socket *sock;
1343 /* Check the SOCK_* constants for consistency. */
1344 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1345 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1346 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1347 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1349 flags = type & ~SOCK_TYPE_MASK;
1350 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1352 type &= SOCK_TYPE_MASK;
1354 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1355 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1357 retval = sock_create(family, type, protocol, &sock);
1361 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1364 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1366 return __sys_socket(family, type, protocol);
1370 * Create a pair of connected sockets.
1373 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1375 struct socket *sock1, *sock2;
1377 struct file *newfile1, *newfile2;
1380 flags = type & ~SOCK_TYPE_MASK;
1381 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1383 type &= SOCK_TYPE_MASK;
1385 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1386 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1389 * reserve descriptors and make sure we won't fail
1390 * to return them to userland.
1392 fd1 = get_unused_fd_flags(flags);
1393 if (unlikely(fd1 < 0))
1396 fd2 = get_unused_fd_flags(flags);
1397 if (unlikely(fd2 < 0)) {
1402 err = put_user(fd1, &usockvec[0]);
1406 err = put_user(fd2, &usockvec[1]);
1411 * Obtain the first socket and check if the underlying protocol
1412 * supports the socketpair call.
1415 err = sock_create(family, type, protocol, &sock1);
1416 if (unlikely(err < 0))
1419 err = sock_create(family, type, protocol, &sock2);
1420 if (unlikely(err < 0)) {
1421 sock_release(sock1);
1425 err = security_socket_socketpair(sock1, sock2);
1426 if (unlikely(err)) {
1427 sock_release(sock2);
1428 sock_release(sock1);
1432 err = sock1->ops->socketpair(sock1, sock2);
1433 if (unlikely(err < 0)) {
1434 sock_release(sock2);
1435 sock_release(sock1);
1439 newfile1 = sock_alloc_file(sock1, flags, NULL);
1440 if (IS_ERR(newfile1)) {
1441 err = PTR_ERR(newfile1);
1442 sock_release(sock2);
1446 newfile2 = sock_alloc_file(sock2, flags, NULL);
1447 if (IS_ERR(newfile2)) {
1448 err = PTR_ERR(newfile2);
1453 audit_fd_pair(fd1, fd2);
1455 fd_install(fd1, newfile1);
1456 fd_install(fd2, newfile2);
1465 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1466 int __user *, usockvec)
1468 return __sys_socketpair(family, type, protocol, usockvec);
1472 * Bind a name to a socket. Nothing much to do here since it's
1473 * the protocol's responsibility to handle the local address.
1475 * We move the socket address to kernel space before we call
1476 * the protocol layer (having also checked the address is ok).
1479 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1481 struct socket *sock;
1482 struct sockaddr_storage address;
1483 int err, fput_needed;
1485 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1487 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1489 err = security_socket_bind(sock,
1490 (struct sockaddr *)&address,
1493 err = sock->ops->bind(sock,
1497 fput_light(sock->file, fput_needed);
1502 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1504 return __sys_bind(fd, umyaddr, addrlen);
1508 * Perform a listen. Basically, we allow the protocol to do anything
1509 * necessary for a listen, and if that works, we mark the socket as
1510 * ready for listening.
1513 int __sys_listen(int fd, int backlog)
1515 struct socket *sock;
1516 int err, fput_needed;
1519 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1521 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1522 if ((unsigned int)backlog > somaxconn)
1523 backlog = somaxconn;
1525 err = security_socket_listen(sock, backlog);
1527 err = sock->ops->listen(sock, backlog);
1529 fput_light(sock->file, fput_needed);
1534 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1536 return __sys_listen(fd, backlog);
1540 * For accept, we attempt to create a new socket, set up the link
1541 * with the client, wake up the client, then return the new
1542 * connected fd. We collect the address of the connector in kernel
1543 * space and move it to user at the very end. This is unclean because
1544 * we open the socket then return an error.
1546 * 1003.1g adds the ability to recvmsg() to query connection pending
1547 * status to recvmsg. We need to add that support in a way thats
1548 * clean when we restructure accept also.
1551 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1552 int __user *upeer_addrlen, int flags)
1554 struct socket *sock, *newsock;
1555 struct file *newfile;
1556 int err, len, newfd, fput_needed;
1557 struct sockaddr_storage address;
1559 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1562 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1563 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1565 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1570 newsock = sock_alloc();
1574 newsock->type = sock->type;
1575 newsock->ops = sock->ops;
1578 * We don't need try_module_get here, as the listening socket (sock)
1579 * has the protocol module (sock->ops->owner) held.
1581 __module_get(newsock->ops->owner);
1583 newfd = get_unused_fd_flags(flags);
1584 if (unlikely(newfd < 0)) {
1586 sock_release(newsock);
1589 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1590 if (IS_ERR(newfile)) {
1591 err = PTR_ERR(newfile);
1592 put_unused_fd(newfd);
1596 err = security_socket_accept(sock, newsock);
1600 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1604 if (upeer_sockaddr) {
1605 len = newsock->ops->getname(newsock,
1606 (struct sockaddr *)&address, 2);
1608 err = -ECONNABORTED;
1611 err = move_addr_to_user(&address,
1612 len, upeer_sockaddr, upeer_addrlen);
1617 /* File flags are not inherited via accept() unlike another OSes. */
1619 fd_install(newfd, newfile);
1623 fput_light(sock->file, fput_needed);
1628 put_unused_fd(newfd);
1632 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1633 int __user *, upeer_addrlen, int, flags)
1635 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1638 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1639 int __user *, upeer_addrlen)
1641 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1645 * Attempt to connect to a socket with the server address. The address
1646 * is in user space so we verify it is OK and move it to kernel space.
1648 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1651 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1652 * other SEQPACKET protocols that take time to connect() as it doesn't
1653 * include the -EINPROGRESS status for such sockets.
1656 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1658 struct socket *sock;
1659 struct sockaddr_storage address;
1660 int err, fput_needed;
1662 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1665 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1670 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1674 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1675 sock->file->f_flags);
1677 fput_light(sock->file, fput_needed);
1682 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1685 return __sys_connect(fd, uservaddr, addrlen);
1689 * Get the local address ('name') of a socket object. Move the obtained
1690 * name to user space.
1693 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1694 int __user *usockaddr_len)
1696 struct socket *sock;
1697 struct sockaddr_storage address;
1698 int err, fput_needed;
1700 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1704 err = security_socket_getsockname(sock);
1708 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1711 /* "err" is actually length in this case */
1712 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1715 fput_light(sock->file, fput_needed);
1720 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1721 int __user *, usockaddr_len)
1723 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1727 * Get the remote address ('name') of a socket object. Move the obtained
1728 * name to user space.
1731 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1732 int __user *usockaddr_len)
1734 struct socket *sock;
1735 struct sockaddr_storage address;
1736 int err, fput_needed;
1738 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1740 err = security_socket_getpeername(sock);
1742 fput_light(sock->file, fput_needed);
1746 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1748 /* "err" is actually length in this case */
1749 err = move_addr_to_user(&address, err, usockaddr,
1751 fput_light(sock->file, fput_needed);
1756 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1757 int __user *, usockaddr_len)
1759 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1763 * Send a datagram to a given address. We move the address into kernel
1764 * space and check the user space data area is readable before invoking
1767 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1768 struct sockaddr __user *addr, int addr_len)
1770 struct socket *sock;
1771 struct sockaddr_storage address;
1777 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1780 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1784 msg.msg_name = NULL;
1785 msg.msg_control = NULL;
1786 msg.msg_controllen = 0;
1787 msg.msg_namelen = 0;
1789 err = move_addr_to_kernel(addr, addr_len, &address);
1792 msg.msg_name = (struct sockaddr *)&address;
1793 msg.msg_namelen = addr_len;
1795 if (sock->file->f_flags & O_NONBLOCK)
1796 flags |= MSG_DONTWAIT;
1797 msg.msg_flags = flags;
1798 err = sock_sendmsg(sock, &msg);
1801 fput_light(sock->file, fput_needed);
1806 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1807 unsigned int, flags, struct sockaddr __user *, addr,
1810 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1814 * Send a datagram down a socket.
1817 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1818 unsigned int, flags)
1820 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1824 * Receive a frame from the socket and optionally record the address of the
1825 * sender. We verify the buffers are writable and if needed move the
1826 * sender address from kernel to user space.
1828 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1829 struct sockaddr __user *addr, int __user *addr_len)
1831 struct socket *sock;
1834 struct sockaddr_storage address;
1838 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1841 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1845 msg.msg_control = NULL;
1846 msg.msg_controllen = 0;
1847 /* Save some cycles and don't copy the address if not needed */
1848 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1849 /* We assume all kernel code knows the size of sockaddr_storage */
1850 msg.msg_namelen = 0;
1851 msg.msg_iocb = NULL;
1853 if (sock->file->f_flags & O_NONBLOCK)
1854 flags |= MSG_DONTWAIT;
1855 err = sock_recvmsg(sock, &msg, flags);
1857 if (err >= 0 && addr != NULL) {
1858 err2 = move_addr_to_user(&address,
1859 msg.msg_namelen, addr, addr_len);
1864 fput_light(sock->file, fput_needed);
1869 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1870 unsigned int, flags, struct sockaddr __user *, addr,
1871 int __user *, addr_len)
1873 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1877 * Receive a datagram from a socket.
1880 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1881 unsigned int, flags)
1883 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1887 * Set a socket option. Because we don't know the option lengths we have
1888 * to pass the user mode parameter for the protocols to sort out.
1891 static int __sys_setsockopt(int fd, int level, int optname,
1892 char __user *optval, int optlen)
1894 int err, fput_needed;
1895 struct socket *sock;
1900 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1902 err = security_socket_setsockopt(sock, level, optname);
1906 if (level == SOL_SOCKET)
1908 sock_setsockopt(sock, level, optname, optval,
1912 sock->ops->setsockopt(sock, level, optname, optval,
1915 fput_light(sock->file, fput_needed);
1920 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1921 char __user *, optval, int, optlen)
1923 return __sys_setsockopt(fd, level, optname, optval, optlen);
1927 * Get a socket option. Because we don't know the option lengths we have
1928 * to pass a user mode parameter for the protocols to sort out.
1931 static int __sys_getsockopt(int fd, int level, int optname,
1932 char __user *optval, int __user *optlen)
1934 int err, fput_needed;
1935 struct socket *sock;
1937 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1939 err = security_socket_getsockopt(sock, level, optname);
1943 if (level == SOL_SOCKET)
1945 sock_getsockopt(sock, level, optname, optval,
1949 sock->ops->getsockopt(sock, level, optname, optval,
1952 fput_light(sock->file, fput_needed);
1957 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1958 char __user *, optval, int __user *, optlen)
1960 return __sys_getsockopt(fd, level, optname, optval, optlen);
1964 * Shutdown a socket.
1967 int __sys_shutdown(int fd, int how)
1969 int err, fput_needed;
1970 struct socket *sock;
1972 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1974 err = security_socket_shutdown(sock, how);
1976 err = sock->ops->shutdown(sock, how);
1977 fput_light(sock->file, fput_needed);
1982 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1984 return __sys_shutdown(fd, how);
1987 /* A couple of helpful macros for getting the address of the 32/64 bit
1988 * fields which are the same type (int / unsigned) on our platforms.
1990 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1991 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1992 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1994 struct used_address {
1995 struct sockaddr_storage name;
1996 unsigned int name_len;
1999 static int copy_msghdr_from_user(struct msghdr *kmsg,
2000 struct user_msghdr __user *umsg,
2001 struct sockaddr __user **save_addr,
2004 struct user_msghdr msg;
2007 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2010 kmsg->msg_control = (void __force *)msg.msg_control;
2011 kmsg->msg_controllen = msg.msg_controllen;
2012 kmsg->msg_flags = msg.msg_flags;
2014 kmsg->msg_namelen = msg.msg_namelen;
2016 kmsg->msg_namelen = 0;
2018 if (kmsg->msg_namelen < 0)
2021 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2022 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2025 *save_addr = msg.msg_name;
2027 if (msg.msg_name && kmsg->msg_namelen) {
2029 err = move_addr_to_kernel(msg.msg_name,
2036 kmsg->msg_name = NULL;
2037 kmsg->msg_namelen = 0;
2040 if (msg.msg_iovlen > UIO_MAXIOV)
2043 kmsg->msg_iocb = NULL;
2045 return import_iovec(save_addr ? READ : WRITE,
2046 msg.msg_iov, msg.msg_iovlen,
2047 UIO_FASTIOV, iov, &kmsg->msg_iter);
2050 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2051 struct msghdr *msg_sys, unsigned int flags,
2052 struct used_address *used_address,
2053 unsigned int allowed_msghdr_flags)
2055 struct compat_msghdr __user *msg_compat =
2056 (struct compat_msghdr __user *)msg;
2057 struct sockaddr_storage address;
2058 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2059 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2060 __aligned(sizeof(__kernel_size_t));
2061 /* 20 is size of ipv6_pktinfo */
2062 unsigned char *ctl_buf = ctl;
2066 msg_sys->msg_name = &address;
2068 if (MSG_CMSG_COMPAT & flags)
2069 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2071 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2077 if (msg_sys->msg_controllen > INT_MAX)
2079 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2080 ctl_len = msg_sys->msg_controllen;
2081 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2083 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2087 ctl_buf = msg_sys->msg_control;
2088 ctl_len = msg_sys->msg_controllen;
2089 } else if (ctl_len) {
2090 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2091 CMSG_ALIGN(sizeof(struct cmsghdr)));
2092 if (ctl_len > sizeof(ctl)) {
2093 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2094 if (ctl_buf == NULL)
2099 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2100 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2101 * checking falls down on this.
2103 if (copy_from_user(ctl_buf,
2104 (void __user __force *)msg_sys->msg_control,
2107 msg_sys->msg_control = ctl_buf;
2109 msg_sys->msg_flags = flags;
2111 if (sock->file->f_flags & O_NONBLOCK)
2112 msg_sys->msg_flags |= MSG_DONTWAIT;
2114 * If this is sendmmsg() and current destination address is same as
2115 * previously succeeded address, omit asking LSM's decision.
2116 * used_address->name_len is initialized to UINT_MAX so that the first
2117 * destination address never matches.
2119 if (used_address && msg_sys->msg_name &&
2120 used_address->name_len == msg_sys->msg_namelen &&
2121 !memcmp(&used_address->name, msg_sys->msg_name,
2122 used_address->name_len)) {
2123 err = sock_sendmsg_nosec(sock, msg_sys);
2126 err = sock_sendmsg(sock, msg_sys);
2128 * If this is sendmmsg() and sending to current destination address was
2129 * successful, remember it.
2131 if (used_address && err >= 0) {
2132 used_address->name_len = msg_sys->msg_namelen;
2133 if (msg_sys->msg_name)
2134 memcpy(&used_address->name, msg_sys->msg_name,
2135 used_address->name_len);
2140 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2147 * BSD sendmsg interface
2150 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2151 bool forbid_cmsg_compat)
2153 int fput_needed, err;
2154 struct msghdr msg_sys;
2155 struct socket *sock;
2157 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2160 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2164 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2166 fput_light(sock->file, fput_needed);
2171 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2173 return __sys_sendmsg(fd, msg, flags, true);
2177 * Linux sendmmsg interface
2180 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2181 unsigned int flags, bool forbid_cmsg_compat)
2183 int fput_needed, err, datagrams;
2184 struct socket *sock;
2185 struct mmsghdr __user *entry;
2186 struct compat_mmsghdr __user *compat_entry;
2187 struct msghdr msg_sys;
2188 struct used_address used_address;
2189 unsigned int oflags = flags;
2191 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2194 if (vlen > UIO_MAXIOV)
2199 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2203 used_address.name_len = UINT_MAX;
2205 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2209 while (datagrams < vlen) {
2210 if (datagrams == vlen - 1)
2213 if (MSG_CMSG_COMPAT & flags) {
2214 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2215 &msg_sys, flags, &used_address, MSG_EOR);
2218 err = __put_user(err, &compat_entry->msg_len);
2221 err = ___sys_sendmsg(sock,
2222 (struct user_msghdr __user *)entry,
2223 &msg_sys, flags, &used_address, MSG_EOR);
2226 err = put_user(err, &entry->msg_len);
2233 if (msg_data_left(&msg_sys))
2238 fput_light(sock->file, fput_needed);
2240 /* We only return an error if no datagrams were able to be sent */
2247 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2248 unsigned int, vlen, unsigned int, flags)
2250 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2253 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2254 struct msghdr *msg_sys, unsigned int flags, int nosec)
2256 struct compat_msghdr __user *msg_compat =
2257 (struct compat_msghdr __user *)msg;
2258 struct iovec iovstack[UIO_FASTIOV];
2259 struct iovec *iov = iovstack;
2260 unsigned long cmsg_ptr;
2264 /* kernel mode address */
2265 struct sockaddr_storage addr;
2267 /* user mode address pointers */
2268 struct sockaddr __user *uaddr;
2269 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2271 msg_sys->msg_name = &addr;
2273 if (MSG_CMSG_COMPAT & flags)
2274 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2276 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2280 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2281 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2283 /* We assume all kernel code knows the size of sockaddr_storage */
2284 msg_sys->msg_namelen = 0;
2286 if (sock->file->f_flags & O_NONBLOCK)
2287 flags |= MSG_DONTWAIT;
2288 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2293 if (uaddr != NULL) {
2294 err = move_addr_to_user(&addr,
2295 msg_sys->msg_namelen, uaddr,
2300 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2304 if (MSG_CMSG_COMPAT & flags)
2305 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2306 &msg_compat->msg_controllen);
2308 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2309 &msg->msg_controllen);
2320 * BSD recvmsg interface
2323 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2324 bool forbid_cmsg_compat)
2326 int fput_needed, err;
2327 struct msghdr msg_sys;
2328 struct socket *sock;
2330 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2333 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2337 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2339 fput_light(sock->file, fput_needed);
2344 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2345 unsigned int, flags)
2347 return __sys_recvmsg(fd, msg, flags, true);
2351 * Linux recvmmsg interface
2354 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2355 unsigned int flags, struct timespec *timeout)
2357 int fput_needed, err, datagrams;
2358 struct socket *sock;
2359 struct mmsghdr __user *entry;
2360 struct compat_mmsghdr __user *compat_entry;
2361 struct msghdr msg_sys;
2362 struct timespec64 end_time;
2363 struct timespec64 timeout64;
2366 poll_select_set_timeout(&end_time, timeout->tv_sec,
2372 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2376 if (likely(!(flags & MSG_ERRQUEUE))) {
2377 err = sock_error(sock->sk);
2385 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2387 while (datagrams < vlen) {
2389 * No need to ask LSM for more than the first datagram.
2391 if (MSG_CMSG_COMPAT & flags) {
2392 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2393 &msg_sys, flags & ~MSG_WAITFORONE,
2397 err = __put_user(err, &compat_entry->msg_len);
2400 err = ___sys_recvmsg(sock,
2401 (struct user_msghdr __user *)entry,
2402 &msg_sys, flags & ~MSG_WAITFORONE,
2406 err = put_user(err, &entry->msg_len);
2414 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2415 if (flags & MSG_WAITFORONE)
2416 flags |= MSG_DONTWAIT;
2419 ktime_get_ts64(&timeout64);
2420 *timeout = timespec64_to_timespec(
2421 timespec64_sub(end_time, timeout64));
2422 if (timeout->tv_sec < 0) {
2423 timeout->tv_sec = timeout->tv_nsec = 0;
2427 /* Timeout, return less than vlen datagrams */
2428 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2432 /* Out of band data, return right away */
2433 if (msg_sys.msg_flags & MSG_OOB)
2441 if (datagrams == 0) {
2447 * We may return less entries than requested (vlen) if the
2448 * sock is non block and there aren't enough datagrams...
2450 if (err != -EAGAIN) {
2452 * ... or if recvmsg returns an error after we
2453 * received some datagrams, where we record the
2454 * error to return on the next call or if the
2455 * app asks about it using getsockopt(SO_ERROR).
2457 sock->sk->sk_err = -err;
2460 fput_light(sock->file, fput_needed);
2465 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2466 unsigned int vlen, unsigned int flags,
2467 struct timespec __user *timeout)
2470 struct timespec timeout_sys;
2472 if (flags & MSG_CMSG_COMPAT)
2476 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2478 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2481 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2483 if (datagrams > 0 &&
2484 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2485 datagrams = -EFAULT;
2490 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2491 unsigned int, vlen, unsigned int, flags,
2492 struct timespec __user *, timeout)
2494 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2497 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2498 /* Argument list sizes for sys_socketcall */
2499 #define AL(x) ((x) * sizeof(unsigned long))
2500 static const unsigned char nargs[21] = {
2501 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2502 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2503 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2510 * System call vectors.
2512 * Argument checking cleaned up. Saved 20% in size.
2513 * This function doesn't need to set the kernel lock because
2514 * it is set by the callees.
2517 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2519 unsigned long a[AUDITSC_ARGS];
2520 unsigned long a0, a1;
2524 if (call < 1 || call > SYS_SENDMMSG)
2526 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2529 if (len > sizeof(a))
2532 /* copy_from_user should be SMP safe. */
2533 if (copy_from_user(a, args, len))
2536 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2545 err = __sys_socket(a0, a1, a[2]);
2548 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2551 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2554 err = __sys_listen(a0, a1);
2557 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2558 (int __user *)a[2], 0);
2560 case SYS_GETSOCKNAME:
2562 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2563 (int __user *)a[2]);
2565 case SYS_GETPEERNAME:
2567 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2568 (int __user *)a[2]);
2570 case SYS_SOCKETPAIR:
2571 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2574 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2578 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2579 (struct sockaddr __user *)a[4], a[5]);
2582 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2586 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2587 (struct sockaddr __user *)a[4],
2588 (int __user *)a[5]);
2591 err = __sys_shutdown(a0, a1);
2593 case SYS_SETSOCKOPT:
2594 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2597 case SYS_GETSOCKOPT:
2599 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2600 (int __user *)a[4]);
2603 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2607 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2611 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2615 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2616 a[3], (struct timespec __user *)a[4]);
2619 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2620 (int __user *)a[2], a[3]);
2629 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2632 * sock_register - add a socket protocol handler
2633 * @ops: description of protocol
2635 * This function is called by a protocol handler that wants to
2636 * advertise its address family, and have it linked into the
2637 * socket interface. The value ops->family corresponds to the
2638 * socket system call protocol family.
2640 int sock_register(const struct net_proto_family *ops)
2644 if (ops->family >= NPROTO) {
2645 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2649 spin_lock(&net_family_lock);
2650 if (rcu_dereference_protected(net_families[ops->family],
2651 lockdep_is_held(&net_family_lock)))
2654 rcu_assign_pointer(net_families[ops->family], ops);
2657 spin_unlock(&net_family_lock);
2659 pr_info("NET: Registered protocol family %d\n", ops->family);
2662 EXPORT_SYMBOL(sock_register);
2665 * sock_unregister - remove a protocol handler
2666 * @family: protocol family to remove
2668 * This function is called by a protocol handler that wants to
2669 * remove its address family, and have it unlinked from the
2670 * new socket creation.
2672 * If protocol handler is a module, then it can use module reference
2673 * counts to protect against new references. If protocol handler is not
2674 * a module then it needs to provide its own protection in
2675 * the ops->create routine.
2677 void sock_unregister(int family)
2679 BUG_ON(family < 0 || family >= NPROTO);
2681 spin_lock(&net_family_lock);
2682 RCU_INIT_POINTER(net_families[family], NULL);
2683 spin_unlock(&net_family_lock);
2687 pr_info("NET: Unregistered protocol family %d\n", family);
2689 EXPORT_SYMBOL(sock_unregister);
2691 bool sock_is_registered(int family)
2693 return family < NPROTO &&
2694 rcu_access_pointer(net_families[array_index_nospec(family, NPROTO)]);
2697 static int __init sock_init(void)
2701 * Initialize the network sysctl infrastructure.
2703 err = net_sysctl_init();
2708 * Initialize skbuff SLAB cache
2713 * Initialize the protocols module.
2718 err = register_filesystem(&sock_fs_type);
2721 sock_mnt = kern_mount(&sock_fs_type);
2722 if (IS_ERR(sock_mnt)) {
2723 err = PTR_ERR(sock_mnt);
2727 /* The real protocol initialization is performed in later initcalls.
2730 #ifdef CONFIG_NETFILTER
2731 err = netfilter_init();
2736 ptp_classifier_init();
2742 unregister_filesystem(&sock_fs_type);
2747 core_initcall(sock_init); /* early initcall */
2749 #ifdef CONFIG_PROC_FS
2750 void socket_seq_show(struct seq_file *seq)
2752 seq_printf(seq, "sockets: used %d\n",
2753 sock_inuse_get(seq->private));
2755 #endif /* CONFIG_PROC_FS */
2757 #ifdef CONFIG_COMPAT
2758 static int do_siocgstamp(struct net *net, struct socket *sock,
2759 unsigned int cmd, void __user *up)
2761 mm_segment_t old_fs = get_fs();
2766 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2769 err = compat_put_timeval(&ktv, up);
2774 static int do_siocgstampns(struct net *net, struct socket *sock,
2775 unsigned int cmd, void __user *up)
2777 mm_segment_t old_fs = get_fs();
2778 struct timespec kts;
2782 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2785 err = compat_put_timespec(&kts, up);
2790 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2792 struct compat_ifconf ifc32;
2796 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2799 ifc.ifc_len = ifc32.ifc_len;
2800 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2803 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2808 ifc32.ifc_len = ifc.ifc_len;
2809 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2815 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2817 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2818 bool convert_in = false, convert_out = false;
2819 size_t buf_size = 0;
2820 struct ethtool_rxnfc __user *rxnfc = NULL;
2822 u32 rule_cnt = 0, actual_rule_cnt;
2827 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2830 compat_rxnfc = compat_ptr(data);
2832 if (get_user(ethcmd, &compat_rxnfc->cmd))
2835 /* Most ethtool structures are defined without padding.
2836 * Unfortunately struct ethtool_rxnfc is an exception.
2841 case ETHTOOL_GRXCLSRLALL:
2842 /* Buffer size is variable */
2843 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2845 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2847 buf_size += rule_cnt * sizeof(u32);
2849 case ETHTOOL_GRXRINGS:
2850 case ETHTOOL_GRXCLSRLCNT:
2851 case ETHTOOL_GRXCLSRULE:
2852 case ETHTOOL_SRXCLSRLINS:
2855 case ETHTOOL_SRXCLSRLDEL:
2856 buf_size += sizeof(struct ethtool_rxnfc);
2858 rxnfc = compat_alloc_user_space(buf_size);
2862 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2865 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2868 /* We expect there to be holes between fs.m_ext and
2869 * fs.ring_cookie and at the end of fs, but nowhere else.
2871 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2872 sizeof(compat_rxnfc->fs.m_ext) !=
2873 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2874 sizeof(rxnfc->fs.m_ext));
2876 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2877 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2878 offsetof(struct ethtool_rxnfc, fs.location) -
2879 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2881 if (copy_in_user(rxnfc, compat_rxnfc,
2882 (void __user *)(&rxnfc->fs.m_ext + 1) -
2883 (void __user *)rxnfc) ||
2884 copy_in_user(&rxnfc->fs.ring_cookie,
2885 &compat_rxnfc->fs.ring_cookie,
2886 (void __user *)(&rxnfc->fs.location + 1) -
2887 (void __user *)&rxnfc->fs.ring_cookie) ||
2888 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2889 sizeof(rxnfc->rule_cnt)))
2893 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2898 if (copy_in_user(compat_rxnfc, rxnfc,
2899 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2900 (const void __user *)rxnfc) ||
2901 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2902 &rxnfc->fs.ring_cookie,
2903 (const void __user *)(&rxnfc->fs.location + 1) -
2904 (const void __user *)&rxnfc->fs.ring_cookie) ||
2905 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2906 sizeof(rxnfc->rule_cnt)))
2909 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2910 /* As an optimisation, we only copy the actual
2911 * number of rules that the underlying
2912 * function returned. Since Mallory might
2913 * change the rule count in user memory, we
2914 * check that it is less than the rule count
2915 * originally given (as the user buffer size),
2916 * which has been range-checked.
2918 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2920 if (actual_rule_cnt < rule_cnt)
2921 rule_cnt = actual_rule_cnt;
2922 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2923 &rxnfc->rule_locs[0],
2924 rule_cnt * sizeof(u32)))
2932 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2934 compat_uptr_t uptr32;
2939 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2942 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2945 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2946 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2948 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2950 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2951 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2957 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2958 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2959 struct compat_ifreq __user *u_ifreq32)
2964 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2966 if (get_user(data32, &u_ifreq32->ifr_data))
2968 ifreq.ifr_data = compat_ptr(data32);
2970 return dev_ioctl(net, cmd, &ifreq, NULL);
2973 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2974 struct compat_ifreq __user *uifr32)
2977 struct compat_ifmap __user *uifmap32;
2980 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2981 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2982 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2983 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2984 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2985 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2986 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2987 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2991 err = dev_ioctl(net, cmd, &ifr, NULL);
2993 if (cmd == SIOCGIFMAP && !err) {
2994 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2995 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2996 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2997 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2998 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2999 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3000 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3009 struct sockaddr rt_dst; /* target address */
3010 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3011 struct sockaddr rt_genmask; /* target network mask (IP) */
3012 unsigned short rt_flags;
3015 unsigned char rt_tos;
3016 unsigned char rt_class;
3018 short rt_metric; /* +1 for binary compatibility! */
3019 /* char * */ u32 rt_dev; /* forcing the device at add */
3020 u32 rt_mtu; /* per route MTU/Window */
3021 u32 rt_window; /* Window clamping */
3022 unsigned short rt_irtt; /* Initial RTT */
3025 struct in6_rtmsg32 {
3026 struct in6_addr rtmsg_dst;
3027 struct in6_addr rtmsg_src;
3028 struct in6_addr rtmsg_gateway;
3038 static int routing_ioctl(struct net *net, struct socket *sock,
3039 unsigned int cmd, void __user *argp)
3043 struct in6_rtmsg r6;
3047 mm_segment_t old_fs = get_fs();
3049 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3050 struct in6_rtmsg32 __user *ur6 = argp;
3051 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3052 3 * sizeof(struct in6_addr));
3053 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3054 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3055 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3056 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3057 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3058 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3059 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3063 struct rtentry32 __user *ur4 = argp;
3064 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3065 3 * sizeof(struct sockaddr));
3066 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3067 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3068 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3069 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3070 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3071 ret |= get_user(rtdev, &(ur4->rt_dev));
3073 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3074 r4.rt_dev = (char __user __force *)devname;
3088 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3095 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3096 * for some operations; this forces use of the newer bridge-utils that
3097 * use compatible ioctls
3099 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3103 if (get_user(tmp, argp))
3105 if (tmp == BRCTL_GET_VERSION)
3106 return BRCTL_VERSION + 1;
3110 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3111 unsigned int cmd, unsigned long arg)
3113 void __user *argp = compat_ptr(arg);
3114 struct sock *sk = sock->sk;
3115 struct net *net = sock_net(sk);
3117 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3118 return compat_ifr_data_ioctl(net, cmd, argp);
3123 return old_bridge_ioctl(argp);
3125 return compat_dev_ifconf(net, argp);
3127 return ethtool_ioctl(net, argp);
3129 return compat_siocwandev(net, argp);
3132 return compat_sioc_ifmap(net, cmd, argp);
3135 return routing_ioctl(net, sock, cmd, argp);
3137 return do_siocgstamp(net, sock, cmd, argp);
3139 return do_siocgstampns(net, sock, cmd, argp);
3140 case SIOCBONDSLAVEINFOQUERY:
3141 case SIOCBONDINFOQUERY:
3144 return compat_ifr_data_ioctl(net, cmd, argp);
3157 return sock_ioctl(file, cmd, arg);
3174 case SIOCSIFHWBROADCAST:
3176 case SIOCGIFBRDADDR:
3177 case SIOCSIFBRDADDR:
3178 case SIOCGIFDSTADDR:
3179 case SIOCSIFDSTADDR:
3180 case SIOCGIFNETMASK:
3181 case SIOCSIFNETMASK:
3196 case SIOCBONDENSLAVE:
3197 case SIOCBONDRELEASE:
3198 case SIOCBONDSETHWADDR:
3199 case SIOCBONDCHANGEACTIVE:
3201 return sock_do_ioctl(net, sock, cmd, arg);
3204 return -ENOIOCTLCMD;
3207 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3210 struct socket *sock = file->private_data;
3211 int ret = -ENOIOCTLCMD;
3218 if (sock->ops->compat_ioctl)
3219 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3221 if (ret == -ENOIOCTLCMD &&
3222 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3223 ret = compat_wext_handle_ioctl(net, cmd, arg);
3225 if (ret == -ENOIOCTLCMD)
3226 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3232 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3234 return sock->ops->bind(sock, addr, addrlen);
3236 EXPORT_SYMBOL(kernel_bind);
3238 int kernel_listen(struct socket *sock, int backlog)
3240 return sock->ops->listen(sock, backlog);
3242 EXPORT_SYMBOL(kernel_listen);
3244 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3246 struct sock *sk = sock->sk;
3249 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3254 err = sock->ops->accept(sock, *newsock, flags, true);
3256 sock_release(*newsock);
3261 (*newsock)->ops = sock->ops;
3262 __module_get((*newsock)->ops->owner);
3267 EXPORT_SYMBOL(kernel_accept);
3269 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3272 return sock->ops->connect(sock, addr, addrlen, flags);
3274 EXPORT_SYMBOL(kernel_connect);
3276 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3278 return sock->ops->getname(sock, addr, 0);
3280 EXPORT_SYMBOL(kernel_getsockname);
3282 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3284 return sock->ops->getname(sock, addr, 1);
3286 EXPORT_SYMBOL(kernel_getpeername);
3288 int kernel_getsockopt(struct socket *sock, int level, int optname,
3289 char *optval, int *optlen)
3291 mm_segment_t oldfs = get_fs();
3292 char __user *uoptval;
3293 int __user *uoptlen;
3296 uoptval = (char __user __force *) optval;
3297 uoptlen = (int __user __force *) optlen;
3300 if (level == SOL_SOCKET)
3301 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3303 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3308 EXPORT_SYMBOL(kernel_getsockopt);
3310 int kernel_setsockopt(struct socket *sock, int level, int optname,
3311 char *optval, unsigned int optlen)
3313 mm_segment_t oldfs = get_fs();
3314 char __user *uoptval;
3317 uoptval = (char __user __force *) optval;
3320 if (level == SOL_SOCKET)
3321 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3323 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3328 EXPORT_SYMBOL(kernel_setsockopt);
3330 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3331 size_t size, int flags)
3333 if (sock->ops->sendpage)
3334 return sock->ops->sendpage(sock, page, offset, size, flags);
3336 return sock_no_sendpage(sock, page, offset, size, flags);
3338 EXPORT_SYMBOL(kernel_sendpage);
3340 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3341 size_t size, int flags)
3343 struct socket *sock = sk->sk_socket;
3345 if (sock->ops->sendpage_locked)
3346 return sock->ops->sendpage_locked(sk, page, offset, size,
3349 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3351 EXPORT_SYMBOL(kernel_sendpage_locked);
3353 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3355 return sock->ops->shutdown(sock, how);
3357 EXPORT_SYMBOL(kernel_sock_shutdown);
3359 /* This routine returns the IP overhead imposed by a socket i.e.
3360 * the length of the underlying IP header, depending on whether
3361 * this is an IPv4 or IPv6 socket and the length from IP options turned
3362 * on at the socket. Assumes that the caller has a lock on the socket.
3364 u32 kernel_sock_ip_overhead(struct sock *sk)
3366 struct inet_sock *inet;
3367 struct ip_options_rcu *opt;
3369 #if IS_ENABLED(CONFIG_IPV6)
3370 struct ipv6_pinfo *np;
3371 struct ipv6_txoptions *optv6 = NULL;
3372 #endif /* IS_ENABLED(CONFIG_IPV6) */
3377 switch (sk->sk_family) {
3380 overhead += sizeof(struct iphdr);
3381 opt = rcu_dereference_protected(inet->inet_opt,
3382 sock_owned_by_user(sk));
3384 overhead += opt->opt.optlen;
3386 #if IS_ENABLED(CONFIG_IPV6)
3389 overhead += sizeof(struct ipv6hdr);
3391 optv6 = rcu_dereference_protected(np->opt,
3392 sock_owned_by_user(sk));
3394 overhead += (optv6->opt_flen + optv6->opt_nflen);
3396 #endif /* IS_ENABLED(CONFIG_IPV6) */
3397 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3401 EXPORT_SYMBOL(kernel_sock_ip_overhead);