2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.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 __read_mostly;
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));
541 sock->sk->sk_uid = iattr->ia_uid;
547 static const struct inode_operations sockfs_inode_ops = {
548 .listxattr = sockfs_listxattr,
549 .setattr = sockfs_setattr,
553 * sock_alloc - allocate a socket
555 * Allocate a new inode and socket object. The two are bound together
556 * and initialised. The socket is then returned. If we are out of inodes
560 struct socket *sock_alloc(void)
565 inode = new_inode_pseudo(sock_mnt->mnt_sb);
569 sock = SOCKET_I(inode);
571 inode->i_ino = get_next_ino();
572 inode->i_mode = S_IFSOCK | S_IRWXUGO;
573 inode->i_uid = current_fsuid();
574 inode->i_gid = current_fsgid();
575 inode->i_op = &sockfs_inode_ops;
579 EXPORT_SYMBOL(sock_alloc);
582 * sock_release - close a socket
583 * @sock: socket to close
585 * The socket is released from the protocol stack if it has a release
586 * callback, and the inode is then released if the socket is bound to
587 * an inode not a file.
590 void sock_release(struct socket *sock)
593 struct module *owner = sock->ops->owner;
595 sock->ops->release(sock);
600 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
601 pr_err("%s: fasync list not empty!\n", __func__);
604 iput(SOCK_INODE(sock));
609 EXPORT_SYMBOL(sock_release);
611 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
613 u8 flags = *tx_flags;
615 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
616 flags |= SKBTX_HW_TSTAMP;
618 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
619 flags |= SKBTX_SW_TSTAMP;
621 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
622 flags |= SKBTX_SCHED_TSTAMP;
626 EXPORT_SYMBOL(__sock_tx_timestamp);
628 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
630 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
631 BUG_ON(ret == -EIOCBQUEUED);
635 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
637 int err = security_socket_sendmsg(sock, msg,
640 return err ?: sock_sendmsg_nosec(sock, msg);
642 EXPORT_SYMBOL(sock_sendmsg);
644 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
645 struct kvec *vec, size_t num, size_t size)
647 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
648 return sock_sendmsg(sock, msg);
650 EXPORT_SYMBOL(kernel_sendmsg);
652 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
653 struct kvec *vec, size_t num, size_t size)
655 struct socket *sock = sk->sk_socket;
657 if (!sock->ops->sendmsg_locked)
658 return sock_no_sendmsg_locked(sk, msg, size);
660 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
662 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
664 EXPORT_SYMBOL(kernel_sendmsg_locked);
666 static bool skb_is_err_queue(const struct sk_buff *skb)
668 /* pkt_type of skbs enqueued on the error queue are set to
669 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
670 * in recvmsg, since skbs received on a local socket will never
671 * have a pkt_type of PACKET_OUTGOING.
673 return skb->pkt_type == PACKET_OUTGOING;
676 /* On transmit, software and hardware timestamps are returned independently.
677 * As the two skb clones share the hardware timestamp, which may be updated
678 * before the software timestamp is received, a hardware TX timestamp may be
679 * returned only if there is no software TX timestamp. Ignore false software
680 * timestamps, which may be made in the __sock_recv_timestamp() call when the
681 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
682 * hardware timestamp.
684 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
686 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
689 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
691 struct scm_ts_pktinfo ts_pktinfo;
692 struct net_device *orig_dev;
694 if (!skb_mac_header_was_set(skb))
697 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
700 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
702 ts_pktinfo.if_index = orig_dev->ifindex;
705 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
706 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
707 sizeof(ts_pktinfo), &ts_pktinfo);
711 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
713 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
716 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
717 struct scm_timestamping tss;
718 int empty = 1, false_tstamp = 0;
719 struct skb_shared_hwtstamps *shhwtstamps =
722 /* Race occurred between timestamp enabling and packet
723 receiving. Fill in the current time for now. */
724 if (need_software_tstamp && skb->tstamp == 0) {
725 __net_timestamp(skb);
729 if (need_software_tstamp) {
730 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
732 skb_get_timestamp(skb, &tv);
733 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
737 skb_get_timestampns(skb, &ts);
738 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
743 memset(&tss, 0, sizeof(tss));
744 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
745 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
748 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
749 !skb_is_swtx_tstamp(skb, false_tstamp) &&
750 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
752 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
753 !skb_is_err_queue(skb))
754 put_ts_pktinfo(msg, skb);
757 put_cmsg(msg, SOL_SOCKET,
758 SCM_TIMESTAMPING, sizeof(tss), &tss);
760 if (skb_is_err_queue(skb) && skb->len &&
761 SKB_EXT_ERR(skb)->opt_stats)
762 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
763 skb->len, skb->data);
766 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
768 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
773 if (!sock_flag(sk, SOCK_WIFI_STATUS))
775 if (!skb->wifi_acked_valid)
778 ack = skb->wifi_acked;
780 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
782 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
784 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
787 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
788 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
789 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
792 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
795 sock_recv_timestamp(msg, sk, skb);
796 sock_recv_drops(msg, sk, skb);
798 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
800 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
803 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
806 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
808 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
810 return err ?: sock_recvmsg_nosec(sock, msg, flags);
812 EXPORT_SYMBOL(sock_recvmsg);
815 * kernel_recvmsg - Receive a message from a socket (kernel space)
816 * @sock: The socket to receive the message from
817 * @msg: Received message
818 * @vec: Input s/g array for message data
819 * @num: Size of input s/g array
820 * @size: Number of bytes to read
821 * @flags: Message flags (MSG_DONTWAIT, etc...)
823 * On return the msg structure contains the scatter/gather array passed in the
824 * vec argument. The array is modified so that it consists of the unfilled
825 * portion of the original array.
827 * The returned value is the total number of bytes received, or an error.
829 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
830 struct kvec *vec, size_t num, size_t size, int flags)
832 mm_segment_t oldfs = get_fs();
835 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
837 result = sock_recvmsg(sock, msg, flags);
841 EXPORT_SYMBOL(kernel_recvmsg);
843 static ssize_t sock_sendpage(struct file *file, struct page *page,
844 int offset, size_t size, loff_t *ppos, int more)
849 sock = file->private_data;
851 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
852 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
855 return kernel_sendpage(sock, page, offset, size, flags);
858 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
859 struct pipe_inode_info *pipe, size_t len,
862 struct socket *sock = file->private_data;
864 if (unlikely(!sock->ops->splice_read))
867 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
870 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
872 struct file *file = iocb->ki_filp;
873 struct socket *sock = file->private_data;
874 struct msghdr msg = {.msg_iter = *to,
878 if (file->f_flags & O_NONBLOCK)
879 msg.msg_flags = MSG_DONTWAIT;
881 if (iocb->ki_pos != 0)
884 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
887 res = sock_recvmsg(sock, &msg, msg.msg_flags);
892 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
894 struct file *file = iocb->ki_filp;
895 struct socket *sock = file->private_data;
896 struct msghdr msg = {.msg_iter = *from,
900 if (iocb->ki_pos != 0)
903 if (file->f_flags & O_NONBLOCK)
904 msg.msg_flags = MSG_DONTWAIT;
906 if (sock->type == SOCK_SEQPACKET)
907 msg.msg_flags |= MSG_EOR;
909 res = sock_sendmsg(sock, &msg);
910 *from = msg.msg_iter;
915 * Atomic setting of ioctl hooks to avoid race
916 * with module unload.
919 static DEFINE_MUTEX(br_ioctl_mutex);
920 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
922 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
924 mutex_lock(&br_ioctl_mutex);
925 br_ioctl_hook = hook;
926 mutex_unlock(&br_ioctl_mutex);
928 EXPORT_SYMBOL(brioctl_set);
930 static DEFINE_MUTEX(vlan_ioctl_mutex);
931 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
933 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
935 mutex_lock(&vlan_ioctl_mutex);
936 vlan_ioctl_hook = hook;
937 mutex_unlock(&vlan_ioctl_mutex);
939 EXPORT_SYMBOL(vlan_ioctl_set);
941 static DEFINE_MUTEX(dlci_ioctl_mutex);
942 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
944 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
946 mutex_lock(&dlci_ioctl_mutex);
947 dlci_ioctl_hook = hook;
948 mutex_unlock(&dlci_ioctl_mutex);
950 EXPORT_SYMBOL(dlci_ioctl_set);
952 static long sock_do_ioctl(struct net *net, struct socket *sock,
953 unsigned int cmd, unsigned long arg)
956 void __user *argp = (void __user *)arg;
958 err = sock->ops->ioctl(sock, cmd, arg);
961 * If this ioctl is unknown try to hand it down
964 if (err != -ENOIOCTLCMD)
967 if (cmd == SIOCGIFCONF) {
969 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
972 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
974 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
979 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
981 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
982 if (!err && need_copyout)
983 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
990 * With an ioctl, arg may well be a user mode pointer, but we don't know
991 * what to do with it - that's up to the protocol still.
994 static struct ns_common *get_net_ns(struct ns_common *ns)
996 return &get_net(container_of(ns, struct net, ns))->ns;
999 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1001 struct socket *sock;
1003 void __user *argp = (void __user *)arg;
1007 sock = file->private_data;
1010 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1013 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1015 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1016 if (!err && need_copyout)
1017 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1020 #ifdef CONFIG_WEXT_CORE
1021 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1022 err = wext_handle_ioctl(net, cmd, argp);
1029 if (get_user(pid, (int __user *)argp))
1031 err = f_setown(sock->file, pid, 1);
1035 err = put_user(f_getown(sock->file),
1036 (int __user *)argp);
1044 request_module("bridge");
1046 mutex_lock(&br_ioctl_mutex);
1048 err = br_ioctl_hook(net, cmd, argp);
1049 mutex_unlock(&br_ioctl_mutex);
1054 if (!vlan_ioctl_hook)
1055 request_module("8021q");
1057 mutex_lock(&vlan_ioctl_mutex);
1058 if (vlan_ioctl_hook)
1059 err = vlan_ioctl_hook(net, argp);
1060 mutex_unlock(&vlan_ioctl_mutex);
1065 if (!dlci_ioctl_hook)
1066 request_module("dlci");
1068 mutex_lock(&dlci_ioctl_mutex);
1069 if (dlci_ioctl_hook)
1070 err = dlci_ioctl_hook(cmd, argp);
1071 mutex_unlock(&dlci_ioctl_mutex);
1075 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1078 err = open_related_ns(&net->ns, get_net_ns);
1081 err = sock_do_ioctl(net, sock, cmd, arg);
1087 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1090 struct socket *sock = NULL;
1092 err = security_socket_create(family, type, protocol, 1);
1096 sock = sock_alloc();
1103 err = security_socket_post_create(sock, family, type, protocol, 1);
1115 EXPORT_SYMBOL(sock_create_lite);
1117 /* No kernel lock held - perfect */
1118 static __poll_t sock_poll(struct file *file, poll_table *wait)
1120 __poll_t busy_flag = 0;
1121 struct socket *sock;
1124 * We can't return errors to poll, so it's either yes or no.
1126 sock = file->private_data;
1128 if (sk_can_busy_loop(sock->sk)) {
1129 /* this socket can poll_ll so tell the system call */
1130 busy_flag = POLL_BUSY_LOOP;
1132 /* once, only if requested by syscall */
1133 if (wait && (wait->_key & POLL_BUSY_LOOP))
1134 sk_busy_loop(sock->sk, 1);
1137 return busy_flag | sock->ops->poll(file, sock, wait);
1140 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1142 struct socket *sock = file->private_data;
1144 return sock->ops->mmap(file, sock, vma);
1147 static int sock_close(struct inode *inode, struct file *filp)
1149 sock_release(SOCKET_I(inode));
1154 * Update the socket async list
1156 * Fasync_list locking strategy.
1158 * 1. fasync_list is modified only under process context socket lock
1159 * i.e. under semaphore.
1160 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1161 * or under socket lock
1164 static int sock_fasync(int fd, struct file *filp, int on)
1166 struct socket *sock = filp->private_data;
1167 struct sock *sk = sock->sk;
1168 struct socket_wq *wq;
1174 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1175 fasync_helper(fd, filp, on, &wq->fasync_list);
1177 if (!wq->fasync_list)
1178 sock_reset_flag(sk, SOCK_FASYNC);
1180 sock_set_flag(sk, SOCK_FASYNC);
1186 /* This function may be called only under rcu_lock */
1188 int sock_wake_async(struct socket_wq *wq, int how, int band)
1190 if (!wq || !wq->fasync_list)
1194 case SOCK_WAKE_WAITD:
1195 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1198 case SOCK_WAKE_SPACE:
1199 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1204 kill_fasync(&wq->fasync_list, SIGIO, band);
1207 kill_fasync(&wq->fasync_list, SIGURG, band);
1212 EXPORT_SYMBOL(sock_wake_async);
1214 int __sock_create(struct net *net, int family, int type, int protocol,
1215 struct socket **res, int kern)
1218 struct socket *sock;
1219 const struct net_proto_family *pf;
1222 * Check protocol is in range
1224 if (family < 0 || family >= NPROTO)
1225 return -EAFNOSUPPORT;
1226 if (type < 0 || type >= SOCK_MAX)
1231 This uglymoron is moved from INET layer to here to avoid
1232 deadlock in module load.
1234 if (family == PF_INET && type == SOCK_PACKET) {
1235 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1240 err = security_socket_create(family, type, protocol, kern);
1245 * Allocate the socket and allow the family to set things up. if
1246 * the protocol is 0, the family is instructed to select an appropriate
1249 sock = sock_alloc();
1251 net_warn_ratelimited("socket: no more sockets\n");
1252 return -ENFILE; /* Not exactly a match, but its the
1253 closest posix thing */
1258 #ifdef CONFIG_MODULES
1259 /* Attempt to load a protocol module if the find failed.
1261 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1262 * requested real, full-featured networking support upon configuration.
1263 * Otherwise module support will break!
1265 if (rcu_access_pointer(net_families[family]) == NULL)
1266 request_module("net-pf-%d", family);
1270 pf = rcu_dereference(net_families[family]);
1271 err = -EAFNOSUPPORT;
1276 * We will call the ->create function, that possibly is in a loadable
1277 * module, so we have to bump that loadable module refcnt first.
1279 if (!try_module_get(pf->owner))
1282 /* Now protected by module ref count */
1285 err = pf->create(net, sock, protocol, kern);
1287 goto out_module_put;
1290 * Now to bump the refcnt of the [loadable] module that owns this
1291 * socket at sock_release time we decrement its refcnt.
1293 if (!try_module_get(sock->ops->owner))
1294 goto out_module_busy;
1297 * Now that we're done with the ->create function, the [loadable]
1298 * module can have its refcnt decremented
1300 module_put(pf->owner);
1301 err = security_socket_post_create(sock, family, type, protocol, kern);
1303 goto out_sock_release;
1309 err = -EAFNOSUPPORT;
1312 module_put(pf->owner);
1319 goto out_sock_release;
1321 EXPORT_SYMBOL(__sock_create);
1323 int sock_create(int family, int type, int protocol, struct socket **res)
1325 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1327 EXPORT_SYMBOL(sock_create);
1329 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1331 return __sock_create(net, family, type, protocol, res, 1);
1333 EXPORT_SYMBOL(sock_create_kern);
1335 int __sys_socket(int family, int type, int protocol)
1338 struct socket *sock;
1341 /* Check the SOCK_* constants for consistency. */
1342 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1343 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1344 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1345 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1347 flags = type & ~SOCK_TYPE_MASK;
1348 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1350 type &= SOCK_TYPE_MASK;
1352 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1353 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1355 retval = sock_create(family, type, protocol, &sock);
1359 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1362 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1364 return __sys_socket(family, type, protocol);
1368 * Create a pair of connected sockets.
1371 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1373 struct socket *sock1, *sock2;
1375 struct file *newfile1, *newfile2;
1378 flags = type & ~SOCK_TYPE_MASK;
1379 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1381 type &= SOCK_TYPE_MASK;
1383 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1384 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1387 * reserve descriptors and make sure we won't fail
1388 * to return them to userland.
1390 fd1 = get_unused_fd_flags(flags);
1391 if (unlikely(fd1 < 0))
1394 fd2 = get_unused_fd_flags(flags);
1395 if (unlikely(fd2 < 0)) {
1400 err = put_user(fd1, &usockvec[0]);
1404 err = put_user(fd2, &usockvec[1]);
1409 * Obtain the first socket and check if the underlying protocol
1410 * supports the socketpair call.
1413 err = sock_create(family, type, protocol, &sock1);
1414 if (unlikely(err < 0))
1417 err = sock_create(family, type, protocol, &sock2);
1418 if (unlikely(err < 0)) {
1419 sock_release(sock1);
1423 err = sock1->ops->socketpair(sock1, sock2);
1424 if (unlikely(err < 0)) {
1425 sock_release(sock2);
1426 sock_release(sock1);
1430 newfile1 = sock_alloc_file(sock1, flags, NULL);
1431 if (IS_ERR(newfile1)) {
1432 err = PTR_ERR(newfile1);
1433 sock_release(sock2);
1437 newfile2 = sock_alloc_file(sock2, flags, NULL);
1438 if (IS_ERR(newfile2)) {
1439 err = PTR_ERR(newfile2);
1444 audit_fd_pair(fd1, fd2);
1446 fd_install(fd1, newfile1);
1447 fd_install(fd2, newfile2);
1456 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1457 int __user *, usockvec)
1459 return __sys_socketpair(family, type, protocol, usockvec);
1463 * Bind a name to a socket. Nothing much to do here since it's
1464 * the protocol's responsibility to handle the local address.
1466 * We move the socket address to kernel space before we call
1467 * the protocol layer (having also checked the address is ok).
1470 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1472 struct socket *sock;
1473 struct sockaddr_storage address;
1474 int err, fput_needed;
1476 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1478 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1480 err = security_socket_bind(sock,
1481 (struct sockaddr *)&address,
1484 err = sock->ops->bind(sock,
1488 fput_light(sock->file, fput_needed);
1493 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1495 return __sys_bind(fd, umyaddr, addrlen);
1499 * Perform a listen. Basically, we allow the protocol to do anything
1500 * necessary for a listen, and if that works, we mark the socket as
1501 * ready for listening.
1504 int __sys_listen(int fd, int backlog)
1506 struct socket *sock;
1507 int err, fput_needed;
1510 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1512 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1513 if ((unsigned int)backlog > somaxconn)
1514 backlog = somaxconn;
1516 err = security_socket_listen(sock, backlog);
1518 err = sock->ops->listen(sock, backlog);
1520 fput_light(sock->file, fput_needed);
1525 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1527 return __sys_listen(fd, backlog);
1531 * For accept, we attempt to create a new socket, set up the link
1532 * with the client, wake up the client, then return the new
1533 * connected fd. We collect the address of the connector in kernel
1534 * space and move it to user at the very end. This is unclean because
1535 * we open the socket then return an error.
1537 * 1003.1g adds the ability to recvmsg() to query connection pending
1538 * status to recvmsg. We need to add that support in a way thats
1539 * clean when we restucture accept also.
1542 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1543 int __user *upeer_addrlen, int flags)
1545 struct socket *sock, *newsock;
1546 struct file *newfile;
1547 int err, len, newfd, fput_needed;
1548 struct sockaddr_storage address;
1550 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1553 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1554 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1556 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1561 newsock = sock_alloc();
1565 newsock->type = sock->type;
1566 newsock->ops = sock->ops;
1569 * We don't need try_module_get here, as the listening socket (sock)
1570 * has the protocol module (sock->ops->owner) held.
1572 __module_get(newsock->ops->owner);
1574 newfd = get_unused_fd_flags(flags);
1575 if (unlikely(newfd < 0)) {
1577 sock_release(newsock);
1580 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1581 if (IS_ERR(newfile)) {
1582 err = PTR_ERR(newfile);
1583 put_unused_fd(newfd);
1587 err = security_socket_accept(sock, newsock);
1591 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1595 if (upeer_sockaddr) {
1596 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1598 err = -ECONNABORTED;
1601 err = move_addr_to_user(&address,
1602 len, upeer_sockaddr, upeer_addrlen);
1607 /* File flags are not inherited via accept() unlike another OSes. */
1609 fd_install(newfd, newfile);
1613 fput_light(sock->file, fput_needed);
1618 put_unused_fd(newfd);
1622 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1623 int __user *, upeer_addrlen, int, flags)
1625 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1628 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1629 int __user *, upeer_addrlen)
1631 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1635 * Attempt to connect to a socket with the server address. The address
1636 * is in user space so we verify it is OK and move it to kernel space.
1638 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1641 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1642 * other SEQPACKET protocols that take time to connect() as it doesn't
1643 * include the -EINPROGRESS status for such sockets.
1646 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1648 struct socket *sock;
1649 struct sockaddr_storage address;
1650 int err, fput_needed;
1652 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1655 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1660 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1664 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1665 sock->file->f_flags);
1667 fput_light(sock->file, fput_needed);
1672 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1675 return __sys_connect(fd, uservaddr, addrlen);
1679 * Get the local address ('name') of a socket object. Move the obtained
1680 * name to user space.
1683 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1684 int __user *usockaddr_len)
1686 struct socket *sock;
1687 struct sockaddr_storage address;
1688 int len, err, fput_needed;
1690 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1694 err = security_socket_getsockname(sock);
1698 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1701 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1704 fput_light(sock->file, fput_needed);
1709 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1710 int __user *, usockaddr_len)
1712 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1716 * Get the remote address ('name') of a socket object. Move the obtained
1717 * name to user space.
1720 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1721 int __user *usockaddr_len)
1723 struct socket *sock;
1724 struct sockaddr_storage address;
1725 int len, err, fput_needed;
1727 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1729 err = security_socket_getpeername(sock);
1731 fput_light(sock->file, fput_needed);
1736 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1739 err = move_addr_to_user(&address, len, usockaddr,
1741 fput_light(sock->file, fput_needed);
1746 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1747 int __user *, usockaddr_len)
1749 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1753 * Send a datagram to a given address. We move the address into kernel
1754 * space and check the user space data area is readable before invoking
1757 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1758 struct sockaddr __user *addr, int addr_len)
1760 struct socket *sock;
1761 struct sockaddr_storage address;
1767 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1770 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1774 msg.msg_name = NULL;
1775 msg.msg_control = NULL;
1776 msg.msg_controllen = 0;
1777 msg.msg_namelen = 0;
1779 err = move_addr_to_kernel(addr, addr_len, &address);
1782 msg.msg_name = (struct sockaddr *)&address;
1783 msg.msg_namelen = addr_len;
1785 if (sock->file->f_flags & O_NONBLOCK)
1786 flags |= MSG_DONTWAIT;
1787 msg.msg_flags = flags;
1788 err = sock_sendmsg(sock, &msg);
1791 fput_light(sock->file, fput_needed);
1796 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1797 unsigned int, flags, struct sockaddr __user *, addr,
1800 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1804 * Send a datagram down a socket.
1807 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1808 unsigned int, flags)
1810 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1814 * Receive a frame from the socket and optionally record the address of the
1815 * sender. We verify the buffers are writable and if needed move the
1816 * sender address from kernel to user space.
1818 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1819 struct sockaddr __user *addr, int __user *addr_len)
1821 struct socket *sock;
1824 struct sockaddr_storage address;
1828 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1831 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1835 msg.msg_control = NULL;
1836 msg.msg_controllen = 0;
1837 /* Save some cycles and don't copy the address if not needed */
1838 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1839 /* We assume all kernel code knows the size of sockaddr_storage */
1840 msg.msg_namelen = 0;
1841 msg.msg_iocb = NULL;
1843 if (sock->file->f_flags & O_NONBLOCK)
1844 flags |= MSG_DONTWAIT;
1845 err = sock_recvmsg(sock, &msg, flags);
1847 if (err >= 0 && addr != NULL) {
1848 err2 = move_addr_to_user(&address,
1849 msg.msg_namelen, addr, addr_len);
1854 fput_light(sock->file, fput_needed);
1859 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1860 unsigned int, flags, struct sockaddr __user *, addr,
1861 int __user *, addr_len)
1863 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1867 * Receive a datagram from a socket.
1870 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1871 unsigned int, flags)
1873 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1877 * Set a socket option. Because we don't know the option lengths we have
1878 * to pass the user mode parameter for the protocols to sort out.
1881 static int __sys_setsockopt(int fd, int level, int optname,
1882 char __user *optval, int optlen)
1884 int err, fput_needed;
1885 struct socket *sock;
1890 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1892 err = security_socket_setsockopt(sock, level, optname);
1896 if (level == SOL_SOCKET)
1898 sock_setsockopt(sock, level, optname, optval,
1902 sock->ops->setsockopt(sock, level, optname, optval,
1905 fput_light(sock->file, fput_needed);
1910 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1911 char __user *, optval, int, optlen)
1913 return __sys_setsockopt(fd, level, optname, optval, optlen);
1917 * Get a socket option. Because we don't know the option lengths we have
1918 * to pass a user mode parameter for the protocols to sort out.
1921 static int __sys_getsockopt(int fd, int level, int optname,
1922 char __user *optval, int __user *optlen)
1924 int err, fput_needed;
1925 struct socket *sock;
1927 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1929 err = security_socket_getsockopt(sock, level, optname);
1933 if (level == SOL_SOCKET)
1935 sock_getsockopt(sock, level, optname, optval,
1939 sock->ops->getsockopt(sock, level, optname, optval,
1942 fput_light(sock->file, fput_needed);
1947 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1948 char __user *, optval, int __user *, optlen)
1950 return __sys_getsockopt(fd, level, optname, optval, optlen);
1954 * Shutdown a socket.
1957 int __sys_shutdown(int fd, int how)
1959 int err, fput_needed;
1960 struct socket *sock;
1962 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1964 err = security_socket_shutdown(sock, how);
1966 err = sock->ops->shutdown(sock, how);
1967 fput_light(sock->file, fput_needed);
1972 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1974 return __sys_shutdown(fd, how);
1977 /* A couple of helpful macros for getting the address of the 32/64 bit
1978 * fields which are the same type (int / unsigned) on our platforms.
1980 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1981 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1982 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1984 struct used_address {
1985 struct sockaddr_storage name;
1986 unsigned int name_len;
1989 static int copy_msghdr_from_user(struct msghdr *kmsg,
1990 struct user_msghdr __user *umsg,
1991 struct sockaddr __user **save_addr,
1994 struct user_msghdr msg;
1997 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2000 kmsg->msg_control = (void __force *)msg.msg_control;
2001 kmsg->msg_controllen = msg.msg_controllen;
2002 kmsg->msg_flags = msg.msg_flags;
2004 kmsg->msg_namelen = msg.msg_namelen;
2006 kmsg->msg_namelen = 0;
2008 if (kmsg->msg_namelen < 0)
2011 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2012 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2015 *save_addr = msg.msg_name;
2017 if (msg.msg_name && kmsg->msg_namelen) {
2019 err = move_addr_to_kernel(msg.msg_name,
2026 kmsg->msg_name = NULL;
2027 kmsg->msg_namelen = 0;
2030 if (msg.msg_iovlen > UIO_MAXIOV)
2033 kmsg->msg_iocb = NULL;
2035 return import_iovec(save_addr ? READ : WRITE,
2036 msg.msg_iov, msg.msg_iovlen,
2037 UIO_FASTIOV, iov, &kmsg->msg_iter);
2040 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2041 struct msghdr *msg_sys, unsigned int flags,
2042 struct used_address *used_address,
2043 unsigned int allowed_msghdr_flags)
2045 struct compat_msghdr __user *msg_compat =
2046 (struct compat_msghdr __user *)msg;
2047 struct sockaddr_storage address;
2048 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2049 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2050 __aligned(sizeof(__kernel_size_t));
2051 /* 20 is size of ipv6_pktinfo */
2052 unsigned char *ctl_buf = ctl;
2056 msg_sys->msg_name = &address;
2058 if (MSG_CMSG_COMPAT & flags)
2059 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2061 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2067 if (msg_sys->msg_controllen > INT_MAX)
2069 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2070 ctl_len = msg_sys->msg_controllen;
2071 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2073 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2077 ctl_buf = msg_sys->msg_control;
2078 ctl_len = msg_sys->msg_controllen;
2079 } else if (ctl_len) {
2080 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2081 CMSG_ALIGN(sizeof(struct cmsghdr)));
2082 if (ctl_len > sizeof(ctl)) {
2083 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2084 if (ctl_buf == NULL)
2089 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2090 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2091 * checking falls down on this.
2093 if (copy_from_user(ctl_buf,
2094 (void __user __force *)msg_sys->msg_control,
2097 msg_sys->msg_control = ctl_buf;
2099 msg_sys->msg_flags = flags;
2101 if (sock->file->f_flags & O_NONBLOCK)
2102 msg_sys->msg_flags |= MSG_DONTWAIT;
2104 * If this is sendmmsg() and current destination address is same as
2105 * previously succeeded address, omit asking LSM's decision.
2106 * used_address->name_len is initialized to UINT_MAX so that the first
2107 * destination address never matches.
2109 if (used_address && msg_sys->msg_name &&
2110 used_address->name_len == msg_sys->msg_namelen &&
2111 !memcmp(&used_address->name, msg_sys->msg_name,
2112 used_address->name_len)) {
2113 err = sock_sendmsg_nosec(sock, msg_sys);
2116 err = sock_sendmsg(sock, msg_sys);
2118 * If this is sendmmsg() and sending to current destination address was
2119 * successful, remember it.
2121 if (used_address && err >= 0) {
2122 used_address->name_len = msg_sys->msg_namelen;
2123 if (msg_sys->msg_name)
2124 memcpy(&used_address->name, msg_sys->msg_name,
2125 used_address->name_len);
2130 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2137 * BSD sendmsg interface
2140 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2142 int fput_needed, err;
2143 struct msghdr msg_sys;
2144 struct socket *sock;
2146 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2150 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2152 fput_light(sock->file, fput_needed);
2157 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2159 if (flags & MSG_CMSG_COMPAT)
2161 return __sys_sendmsg(fd, msg, flags);
2165 * Linux sendmmsg interface
2168 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2171 int fput_needed, err, datagrams;
2172 struct socket *sock;
2173 struct mmsghdr __user *entry;
2174 struct compat_mmsghdr __user *compat_entry;
2175 struct msghdr msg_sys;
2176 struct used_address used_address;
2177 unsigned int oflags = flags;
2179 if (vlen > UIO_MAXIOV)
2184 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2188 used_address.name_len = UINT_MAX;
2190 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2194 while (datagrams < vlen) {
2195 if (datagrams == vlen - 1)
2198 if (MSG_CMSG_COMPAT & flags) {
2199 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2200 &msg_sys, flags, &used_address, MSG_EOR);
2203 err = __put_user(err, &compat_entry->msg_len);
2206 err = ___sys_sendmsg(sock,
2207 (struct user_msghdr __user *)entry,
2208 &msg_sys, flags, &used_address, MSG_EOR);
2211 err = put_user(err, &entry->msg_len);
2218 if (msg_data_left(&msg_sys))
2223 fput_light(sock->file, fput_needed);
2225 /* We only return an error if no datagrams were able to be sent */
2232 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2233 unsigned int, vlen, unsigned int, flags)
2235 if (flags & MSG_CMSG_COMPAT)
2237 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2240 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2241 struct msghdr *msg_sys, unsigned int flags, int nosec)
2243 struct compat_msghdr __user *msg_compat =
2244 (struct compat_msghdr __user *)msg;
2245 struct iovec iovstack[UIO_FASTIOV];
2246 struct iovec *iov = iovstack;
2247 unsigned long cmsg_ptr;
2251 /* kernel mode address */
2252 struct sockaddr_storage addr;
2254 /* user mode address pointers */
2255 struct sockaddr __user *uaddr;
2256 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2258 msg_sys->msg_name = &addr;
2260 if (MSG_CMSG_COMPAT & flags)
2261 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2263 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2267 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2268 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2270 /* We assume all kernel code knows the size of sockaddr_storage */
2271 msg_sys->msg_namelen = 0;
2273 if (sock->file->f_flags & O_NONBLOCK)
2274 flags |= MSG_DONTWAIT;
2275 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2280 if (uaddr != NULL) {
2281 err = move_addr_to_user(&addr,
2282 msg_sys->msg_namelen, uaddr,
2287 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2291 if (MSG_CMSG_COMPAT & flags)
2292 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2293 &msg_compat->msg_controllen);
2295 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2296 &msg->msg_controllen);
2307 * BSD recvmsg interface
2310 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2312 int fput_needed, err;
2313 struct msghdr msg_sys;
2314 struct socket *sock;
2316 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2320 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2322 fput_light(sock->file, fput_needed);
2327 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2328 unsigned int, flags)
2330 if (flags & MSG_CMSG_COMPAT)
2332 return __sys_recvmsg(fd, msg, flags);
2336 * Linux recvmmsg interface
2339 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2340 unsigned int flags, struct timespec *timeout)
2342 int fput_needed, err, datagrams;
2343 struct socket *sock;
2344 struct mmsghdr __user *entry;
2345 struct compat_mmsghdr __user *compat_entry;
2346 struct msghdr msg_sys;
2347 struct timespec64 end_time;
2348 struct timespec64 timeout64;
2351 poll_select_set_timeout(&end_time, timeout->tv_sec,
2357 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2361 err = sock_error(sock->sk);
2368 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2370 while (datagrams < vlen) {
2372 * No need to ask LSM for more than the first datagram.
2374 if (MSG_CMSG_COMPAT & flags) {
2375 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2376 &msg_sys, flags & ~MSG_WAITFORONE,
2380 err = __put_user(err, &compat_entry->msg_len);
2383 err = ___sys_recvmsg(sock,
2384 (struct user_msghdr __user *)entry,
2385 &msg_sys, flags & ~MSG_WAITFORONE,
2389 err = put_user(err, &entry->msg_len);
2397 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2398 if (flags & MSG_WAITFORONE)
2399 flags |= MSG_DONTWAIT;
2402 ktime_get_ts64(&timeout64);
2403 *timeout = timespec64_to_timespec(
2404 timespec64_sub(end_time, timeout64));
2405 if (timeout->tv_sec < 0) {
2406 timeout->tv_sec = timeout->tv_nsec = 0;
2410 /* Timeout, return less than vlen datagrams */
2411 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2415 /* Out of band data, return right away */
2416 if (msg_sys.msg_flags & MSG_OOB)
2424 if (datagrams == 0) {
2430 * We may return less entries than requested (vlen) if the
2431 * sock is non block and there aren't enough datagrams...
2433 if (err != -EAGAIN) {
2435 * ... or if recvmsg returns an error after we
2436 * received some datagrams, where we record the
2437 * error to return on the next call or if the
2438 * app asks about it using getsockopt(SO_ERROR).
2440 sock->sk->sk_err = -err;
2443 fput_light(sock->file, fput_needed);
2448 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2449 unsigned int, vlen, unsigned int, flags,
2450 struct timespec __user *, timeout)
2453 struct timespec timeout_sys;
2455 if (flags & MSG_CMSG_COMPAT)
2459 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2461 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2464 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2466 if (datagrams > 0 &&
2467 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2468 datagrams = -EFAULT;
2473 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2474 /* Argument list sizes for sys_socketcall */
2475 #define AL(x) ((x) * sizeof(unsigned long))
2476 static const unsigned char nargs[21] = {
2477 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2478 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2479 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2486 * System call vectors.
2488 * Argument checking cleaned up. Saved 20% in size.
2489 * This function doesn't need to set the kernel lock because
2490 * it is set by the callees.
2493 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2495 unsigned long a[AUDITSC_ARGS];
2496 unsigned long a0, a1;
2500 if (call < 1 || call > SYS_SENDMMSG)
2504 if (len > sizeof(a))
2507 /* copy_from_user should be SMP safe. */
2508 if (copy_from_user(a, args, len))
2511 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2520 err = __sys_socket(a0, a1, a[2]);
2523 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2526 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2529 err = __sys_listen(a0, a1);
2532 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2533 (int __user *)a[2], 0);
2535 case SYS_GETSOCKNAME:
2537 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2538 (int __user *)a[2]);
2540 case SYS_GETPEERNAME:
2542 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2543 (int __user *)a[2]);
2545 case SYS_SOCKETPAIR:
2546 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2549 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2552 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2553 (struct sockaddr __user *)a[4], a[5]);
2556 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2559 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2560 (struct sockaddr __user *)a[4],
2561 (int __user *)a[5]);
2564 err = __sys_shutdown(a0, a1);
2566 case SYS_SETSOCKOPT:
2567 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2570 case SYS_GETSOCKOPT:
2572 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2573 (int __user *)a[4]);
2576 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2579 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2582 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2585 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2586 (struct timespec __user *)a[4]);
2589 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2590 (int __user *)a[2], a[3]);
2599 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2602 * sock_register - add a socket protocol handler
2603 * @ops: description of protocol
2605 * This function is called by a protocol handler that wants to
2606 * advertise its address family, and have it linked into the
2607 * socket interface. The value ops->family corresponds to the
2608 * socket system call protocol family.
2610 int sock_register(const struct net_proto_family *ops)
2614 if (ops->family >= NPROTO) {
2615 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2619 spin_lock(&net_family_lock);
2620 if (rcu_dereference_protected(net_families[ops->family],
2621 lockdep_is_held(&net_family_lock)))
2624 rcu_assign_pointer(net_families[ops->family], ops);
2627 spin_unlock(&net_family_lock);
2629 pr_info("NET: Registered protocol family %d\n", ops->family);
2632 EXPORT_SYMBOL(sock_register);
2635 * sock_unregister - remove a protocol handler
2636 * @family: protocol family to remove
2638 * This function is called by a protocol handler that wants to
2639 * remove its address family, and have it unlinked from the
2640 * new socket creation.
2642 * If protocol handler is a module, then it can use module reference
2643 * counts to protect against new references. If protocol handler is not
2644 * a module then it needs to provide its own protection in
2645 * the ops->create routine.
2647 void sock_unregister(int family)
2649 BUG_ON(family < 0 || family >= NPROTO);
2651 spin_lock(&net_family_lock);
2652 RCU_INIT_POINTER(net_families[family], NULL);
2653 spin_unlock(&net_family_lock);
2657 pr_info("NET: Unregistered protocol family %d\n", family);
2659 EXPORT_SYMBOL(sock_unregister);
2661 static int __init sock_init(void)
2665 * Initialize the network sysctl infrastructure.
2667 err = net_sysctl_init();
2672 * Initialize skbuff SLAB cache
2677 * Initialize the protocols module.
2682 err = register_filesystem(&sock_fs_type);
2685 sock_mnt = kern_mount(&sock_fs_type);
2686 if (IS_ERR(sock_mnt)) {
2687 err = PTR_ERR(sock_mnt);
2691 /* The real protocol initialization is performed in later initcalls.
2694 #ifdef CONFIG_NETFILTER
2695 err = netfilter_init();
2700 ptp_classifier_init();
2706 unregister_filesystem(&sock_fs_type);
2711 core_initcall(sock_init); /* early initcall */
2713 #ifdef CONFIG_PROC_FS
2714 void socket_seq_show(struct seq_file *seq)
2716 seq_printf(seq, "sockets: used %d\n",
2717 sock_inuse_get(seq->private));
2719 #endif /* CONFIG_PROC_FS */
2721 #ifdef CONFIG_COMPAT
2722 static int do_siocgstamp(struct net *net, struct socket *sock,
2723 unsigned int cmd, void __user *up)
2725 mm_segment_t old_fs = get_fs();
2730 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2733 err = compat_put_timeval(&ktv, up);
2738 static int do_siocgstampns(struct net *net, struct socket *sock,
2739 unsigned int cmd, void __user *up)
2741 mm_segment_t old_fs = get_fs();
2742 struct timespec kts;
2746 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2749 err = compat_put_timespec(&kts, up);
2754 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2756 struct compat_ifconf ifc32;
2760 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2763 ifc.ifc_len = ifc32.ifc_len;
2764 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2767 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2772 ifc32.ifc_len = ifc.ifc_len;
2773 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2779 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2781 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2782 bool convert_in = false, convert_out = false;
2783 size_t buf_size = 0;
2784 struct ethtool_rxnfc __user *rxnfc = NULL;
2786 u32 rule_cnt = 0, actual_rule_cnt;
2791 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2794 compat_rxnfc = compat_ptr(data);
2796 if (get_user(ethcmd, &compat_rxnfc->cmd))
2799 /* Most ethtool structures are defined without padding.
2800 * Unfortunately struct ethtool_rxnfc is an exception.
2805 case ETHTOOL_GRXCLSRLALL:
2806 /* Buffer size is variable */
2807 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2809 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2811 buf_size += rule_cnt * sizeof(u32);
2813 case ETHTOOL_GRXRINGS:
2814 case ETHTOOL_GRXCLSRLCNT:
2815 case ETHTOOL_GRXCLSRULE:
2816 case ETHTOOL_SRXCLSRLINS:
2819 case ETHTOOL_SRXCLSRLDEL:
2820 buf_size += sizeof(struct ethtool_rxnfc);
2822 rxnfc = compat_alloc_user_space(buf_size);
2826 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2829 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2832 /* We expect there to be holes between fs.m_ext and
2833 * fs.ring_cookie and at the end of fs, but nowhere else.
2835 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2836 sizeof(compat_rxnfc->fs.m_ext) !=
2837 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2838 sizeof(rxnfc->fs.m_ext));
2840 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2841 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2842 offsetof(struct ethtool_rxnfc, fs.location) -
2843 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2845 if (copy_in_user(rxnfc, compat_rxnfc,
2846 (void __user *)(&rxnfc->fs.m_ext + 1) -
2847 (void __user *)rxnfc) ||
2848 copy_in_user(&rxnfc->fs.ring_cookie,
2849 &compat_rxnfc->fs.ring_cookie,
2850 (void __user *)(&rxnfc->fs.location + 1) -
2851 (void __user *)&rxnfc->fs.ring_cookie) ||
2852 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2853 sizeof(rxnfc->rule_cnt)))
2857 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2862 if (copy_in_user(compat_rxnfc, rxnfc,
2863 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2864 (const void __user *)rxnfc) ||
2865 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2866 &rxnfc->fs.ring_cookie,
2867 (const void __user *)(&rxnfc->fs.location + 1) -
2868 (const void __user *)&rxnfc->fs.ring_cookie) ||
2869 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2870 sizeof(rxnfc->rule_cnt)))
2873 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2874 /* As an optimisation, we only copy the actual
2875 * number of rules that the underlying
2876 * function returned. Since Mallory might
2877 * change the rule count in user memory, we
2878 * check that it is less than the rule count
2879 * originally given (as the user buffer size),
2880 * which has been range-checked.
2882 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2884 if (actual_rule_cnt < rule_cnt)
2885 rule_cnt = actual_rule_cnt;
2886 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2887 &rxnfc->rule_locs[0],
2888 rule_cnt * sizeof(u32)))
2896 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2898 compat_uptr_t uptr32;
2903 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2906 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2909 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2910 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2912 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2914 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2915 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2921 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2922 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2923 struct compat_ifreq __user *u_ifreq32)
2928 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2930 if (get_user(data32, &u_ifreq32->ifr_data))
2932 ifreq.ifr_data = compat_ptr(data32);
2934 return dev_ioctl(net, cmd, &ifreq, NULL);
2937 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2938 struct compat_ifreq __user *uifr32)
2941 struct compat_ifmap __user *uifmap32;
2944 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2945 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2946 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2947 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2948 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2949 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2950 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2951 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2955 err = dev_ioctl(net, cmd, &ifr, NULL);
2957 if (cmd == SIOCGIFMAP && !err) {
2958 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2959 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2960 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2961 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2962 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2963 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2964 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2973 struct sockaddr rt_dst; /* target address */
2974 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2975 struct sockaddr rt_genmask; /* target network mask (IP) */
2976 unsigned short rt_flags;
2979 unsigned char rt_tos;
2980 unsigned char rt_class;
2982 short rt_metric; /* +1 for binary compatibility! */
2983 /* char * */ u32 rt_dev; /* forcing the device at add */
2984 u32 rt_mtu; /* per route MTU/Window */
2985 u32 rt_window; /* Window clamping */
2986 unsigned short rt_irtt; /* Initial RTT */
2989 struct in6_rtmsg32 {
2990 struct in6_addr rtmsg_dst;
2991 struct in6_addr rtmsg_src;
2992 struct in6_addr rtmsg_gateway;
3002 static int routing_ioctl(struct net *net, struct socket *sock,
3003 unsigned int cmd, void __user *argp)
3007 struct in6_rtmsg r6;
3011 mm_segment_t old_fs = get_fs();
3013 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3014 struct in6_rtmsg32 __user *ur6 = argp;
3015 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3016 3 * sizeof(struct in6_addr));
3017 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3018 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3019 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3020 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3021 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3022 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3023 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3027 struct rtentry32 __user *ur4 = argp;
3028 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3029 3 * sizeof(struct sockaddr));
3030 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3031 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3032 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3033 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3034 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3035 ret |= get_user(rtdev, &(ur4->rt_dev));
3037 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3038 r4.rt_dev = (char __user __force *)devname;
3052 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3059 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3060 * for some operations; this forces use of the newer bridge-utils that
3061 * use compatible ioctls
3063 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3067 if (get_user(tmp, argp))
3069 if (tmp == BRCTL_GET_VERSION)
3070 return BRCTL_VERSION + 1;
3074 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3075 unsigned int cmd, unsigned long arg)
3077 void __user *argp = compat_ptr(arg);
3078 struct sock *sk = sock->sk;
3079 struct net *net = sock_net(sk);
3081 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3082 return compat_ifr_data_ioctl(net, cmd, argp);
3087 return old_bridge_ioctl(argp);
3089 return compat_dev_ifconf(net, argp);
3091 return ethtool_ioctl(net, argp);
3093 return compat_siocwandev(net, argp);
3096 return compat_sioc_ifmap(net, cmd, argp);
3099 return routing_ioctl(net, sock, cmd, argp);
3101 return do_siocgstamp(net, sock, cmd, argp);
3103 return do_siocgstampns(net, sock, cmd, argp);
3104 case SIOCBONDSLAVEINFOQUERY:
3105 case SIOCBONDINFOQUERY:
3108 return compat_ifr_data_ioctl(net, cmd, argp);
3121 return sock_ioctl(file, cmd, arg);
3138 case SIOCSIFHWBROADCAST:
3140 case SIOCGIFBRDADDR:
3141 case SIOCSIFBRDADDR:
3142 case SIOCGIFDSTADDR:
3143 case SIOCSIFDSTADDR:
3144 case SIOCGIFNETMASK:
3145 case SIOCSIFNETMASK:
3160 case SIOCBONDENSLAVE:
3161 case SIOCBONDRELEASE:
3162 case SIOCBONDSETHWADDR:
3163 case SIOCBONDCHANGEACTIVE:
3165 return sock_do_ioctl(net, sock, cmd, arg);
3168 return -ENOIOCTLCMD;
3171 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3174 struct socket *sock = file->private_data;
3175 int ret = -ENOIOCTLCMD;
3182 if (sock->ops->compat_ioctl)
3183 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3185 if (ret == -ENOIOCTLCMD &&
3186 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3187 ret = compat_wext_handle_ioctl(net, cmd, arg);
3189 if (ret == -ENOIOCTLCMD)
3190 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3196 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3198 return sock->ops->bind(sock, addr, addrlen);
3200 EXPORT_SYMBOL(kernel_bind);
3202 int kernel_listen(struct socket *sock, int backlog)
3204 return sock->ops->listen(sock, backlog);
3206 EXPORT_SYMBOL(kernel_listen);
3208 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3210 struct sock *sk = sock->sk;
3213 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3218 err = sock->ops->accept(sock, *newsock, flags, true);
3220 sock_release(*newsock);
3225 (*newsock)->ops = sock->ops;
3226 __module_get((*newsock)->ops->owner);
3231 EXPORT_SYMBOL(kernel_accept);
3233 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3236 return sock->ops->connect(sock, addr, addrlen, flags);
3238 EXPORT_SYMBOL(kernel_connect);
3240 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3243 return sock->ops->getname(sock, addr, addrlen, 0);
3245 EXPORT_SYMBOL(kernel_getsockname);
3247 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3250 return sock->ops->getname(sock, addr, addrlen, 1);
3252 EXPORT_SYMBOL(kernel_getpeername);
3254 int kernel_getsockopt(struct socket *sock, int level, int optname,
3255 char *optval, int *optlen)
3257 mm_segment_t oldfs = get_fs();
3258 char __user *uoptval;
3259 int __user *uoptlen;
3262 uoptval = (char __user __force *) optval;
3263 uoptlen = (int __user __force *) optlen;
3266 if (level == SOL_SOCKET)
3267 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3269 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3274 EXPORT_SYMBOL(kernel_getsockopt);
3276 int kernel_setsockopt(struct socket *sock, int level, int optname,
3277 char *optval, unsigned int optlen)
3279 mm_segment_t oldfs = get_fs();
3280 char __user *uoptval;
3283 uoptval = (char __user __force *) optval;
3286 if (level == SOL_SOCKET)
3287 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3289 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3294 EXPORT_SYMBOL(kernel_setsockopt);
3296 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3297 size_t size, int flags)
3299 if (sock->ops->sendpage)
3300 return sock->ops->sendpage(sock, page, offset, size, flags);
3302 return sock_no_sendpage(sock, page, offset, size, flags);
3304 EXPORT_SYMBOL(kernel_sendpage);
3306 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3307 size_t size, int flags)
3309 struct socket *sock = sk->sk_socket;
3311 if (sock->ops->sendpage_locked)
3312 return sock->ops->sendpage_locked(sk, page, offset, size,
3315 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3317 EXPORT_SYMBOL(kernel_sendpage_locked);
3319 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3321 return sock->ops->shutdown(sock, how);
3323 EXPORT_SYMBOL(kernel_sock_shutdown);
3325 /* This routine returns the IP overhead imposed by a socket i.e.
3326 * the length of the underlying IP header, depending on whether
3327 * this is an IPv4 or IPv6 socket and the length from IP options turned
3328 * on at the socket. Assumes that the caller has a lock on the socket.
3330 u32 kernel_sock_ip_overhead(struct sock *sk)
3332 struct inet_sock *inet;
3333 struct ip_options_rcu *opt;
3335 #if IS_ENABLED(CONFIG_IPV6)
3336 struct ipv6_pinfo *np;
3337 struct ipv6_txoptions *optv6 = NULL;
3338 #endif /* IS_ENABLED(CONFIG_IPV6) */
3343 switch (sk->sk_family) {
3346 overhead += sizeof(struct iphdr);
3347 opt = rcu_dereference_protected(inet->inet_opt,
3348 sock_owned_by_user(sk));
3350 overhead += opt->opt.optlen;
3352 #if IS_ENABLED(CONFIG_IPV6)
3355 overhead += sizeof(struct ipv6hdr);
3357 optv6 = rcu_dereference_protected(np->opt,
3358 sock_owned_by_user(sk));
3360 overhead += (optv6->opt_flen + optv6->opt_nflen);
3362 #endif /* IS_ENABLED(CONFIG_IPV6) */
3363 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3367 EXPORT_SYMBOL(kernel_sock_ip_overhead);