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 unsigned int 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;
166 * Statistics counters of the socket lists
169 static DEFINE_PER_CPU(int, sockets_in_use);
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
190 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
194 if (copy_from_user(kaddr, uaddr, ulen))
196 return audit_sockaddr(ulen, kaddr);
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217 void __user *uaddr, int __user *ulen)
222 BUG_ON(klen > sizeof(struct sockaddr_storage));
223 err = get_user(len, ulen);
231 if (audit_sockaddr(klen, kaddr))
233 if (copy_to_user(uaddr, kaddr, len))
237 * "fromlen shall refer to the value before truncation.."
240 return __put_user(klen, ulen);
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
245 static struct inode *sock_alloc_inode(struct super_block *sb)
247 struct socket_alloc *ei;
248 struct socket_wq *wq;
250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 kmem_cache_free(sock_inode_cachep, ei);
258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
269 return &ei->vfs_inode;
272 static void sock_destroy_inode(struct inode *inode)
274 struct socket_alloc *ei;
275 struct socket_wq *wq;
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
278 wq = rcu_dereference_protected(ei->socket.wq, 1);
280 kmem_cache_free(sock_inode_cachep, ei);
283 static void init_once(void *foo)
285 struct socket_alloc *ei = (struct socket_alloc *)foo;
287 inode_init_once(&ei->vfs_inode);
290 static void init_inodecache(void)
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293 sizeof(struct socket_alloc),
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
297 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
299 BUG_ON(sock_inode_cachep == NULL);
302 static const struct super_operations sockfs_ops = {
303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
309 * sockfs_dname() is called from d_path().
311 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
313 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
314 d_inode(dentry)->i_ino);
317 static const struct dentry_operations sockfs_dentry_operations = {
318 .d_dname = sockfs_dname,
321 static int sockfs_xattr_get(const struct xattr_handler *handler,
322 struct dentry *dentry, struct inode *inode,
323 const char *suffix, void *value, size_t size)
326 if (dentry->d_name.len + 1 > size)
328 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
330 return dentry->d_name.len + 1;
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
337 static const struct xattr_handler sockfs_xattr_handler = {
338 .name = XATTR_NAME_SOCKPROTONAME,
339 .get = sockfs_xattr_get,
342 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343 struct dentry *dentry, struct inode *inode,
344 const char *suffix, const void *value,
345 size_t size, int flags)
347 /* Handled by LSM. */
351 static const struct xattr_handler sockfs_security_xattr_handler = {
352 .prefix = XATTR_SECURITY_PREFIX,
353 .set = sockfs_security_xattr_set,
356 static const struct xattr_handler *sockfs_xattr_handlers[] = {
357 &sockfs_xattr_handler,
358 &sockfs_security_xattr_handler,
362 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363 int flags, const char *dev_name, void *data)
365 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366 sockfs_xattr_handlers,
367 &sockfs_dentry_operations, SOCKFS_MAGIC);
370 static struct vfsmount *sock_mnt __read_mostly;
372 static struct file_system_type sock_fs_type = {
374 .mount = sockfs_mount,
375 .kill_sb = kill_anon_super,
379 * Obtains the first available file descriptor and sets it up for use.
381 * These functions create file structures and maps them to fd space
382 * of the current process. On success it returns file descriptor
383 * and file struct implicitly stored in sock->file.
384 * Note that another thread may close file descriptor before we return
385 * from this function. We use the fact that now we do not refer
386 * to socket after mapping. If one day we will need it, this
387 * function will increment ref. count on file by 1.
389 * In any case returned fd MAY BE not valid!
390 * This race condition is unavoidable
391 * with shared fd spaces, we cannot solve it inside kernel,
392 * but we take care of internal coherence yet.
395 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
397 struct qstr name = { .name = "" };
403 name.len = strlen(name.name);
404 } else if (sock->sk) {
405 name.name = sock->sk->sk_prot_creator->name;
406 name.len = strlen(name.name);
408 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
409 if (unlikely(!path.dentry)) {
411 return ERR_PTR(-ENOMEM);
413 path.mnt = mntget(sock_mnt);
415 d_instantiate(path.dentry, SOCK_INODE(sock));
417 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
420 /* drop dentry, keep inode for a bit */
421 ihold(d_inode(path.dentry));
423 /* ... and now kill it properly */
429 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
430 file->private_data = sock;
433 EXPORT_SYMBOL(sock_alloc_file);
435 static int sock_map_fd(struct socket *sock, int flags)
437 struct file *newfile;
438 int fd = get_unused_fd_flags(flags);
439 if (unlikely(fd < 0))
442 newfile = sock_alloc_file(sock, flags, NULL);
443 if (likely(!IS_ERR(newfile))) {
444 fd_install(fd, newfile);
449 return PTR_ERR(newfile);
452 struct socket *sock_from_file(struct file *file, int *err)
454 if (file->f_op == &socket_file_ops)
455 return file->private_data; /* set in sock_map_fd */
460 EXPORT_SYMBOL(sock_from_file);
463 * sockfd_lookup - Go from a file number to its socket slot
465 * @err: pointer to an error code return
467 * The file handle passed in is locked and the socket it is bound
468 * to is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
472 * On a success the socket object pointer is returned.
475 struct socket *sockfd_lookup(int fd, int *err)
486 sock = sock_from_file(file, err);
491 EXPORT_SYMBOL(sockfd_lookup);
493 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
495 struct fd f = fdget(fd);
500 sock = sock_from_file(f.file, err);
502 *fput_needed = f.flags;
510 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
516 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
526 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
531 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
538 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
540 int err = simple_setattr(dentry, iattr);
542 if (!err && (iattr->ia_valid & ATTR_UID)) {
543 struct socket *sock = SOCKET_I(d_inode(dentry));
545 sock->sk->sk_uid = iattr->ia_uid;
551 static const struct inode_operations sockfs_inode_ops = {
552 .listxattr = sockfs_listxattr,
553 .setattr = sockfs_setattr,
557 * sock_alloc - allocate a socket
559 * Allocate a new inode and socket object. The two are bound together
560 * and initialised. The socket is then returned. If we are out of inodes
564 struct socket *sock_alloc(void)
569 inode = new_inode_pseudo(sock_mnt->mnt_sb);
573 sock = SOCKET_I(inode);
575 inode->i_ino = get_next_ino();
576 inode->i_mode = S_IFSOCK | S_IRWXUGO;
577 inode->i_uid = current_fsuid();
578 inode->i_gid = current_fsgid();
579 inode->i_op = &sockfs_inode_ops;
581 this_cpu_add(sockets_in_use, 1);
584 EXPORT_SYMBOL(sock_alloc);
587 * sock_release - close a socket
588 * @sock: socket to close
590 * The socket is released from the protocol stack if it has a release
591 * callback, and the inode is then released if the socket is bound to
592 * an inode not a file.
595 void sock_release(struct socket *sock)
598 struct module *owner = sock->ops->owner;
600 sock->ops->release(sock);
605 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
606 pr_err("%s: fasync list not empty!\n", __func__);
608 this_cpu_sub(sockets_in_use, 1);
610 iput(SOCK_INODE(sock));
615 EXPORT_SYMBOL(sock_release);
617 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
619 u8 flags = *tx_flags;
621 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
622 flags |= SKBTX_HW_TSTAMP;
624 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
625 flags |= SKBTX_SW_TSTAMP;
627 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
628 flags |= SKBTX_SCHED_TSTAMP;
632 EXPORT_SYMBOL(__sock_tx_timestamp);
634 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
636 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
637 BUG_ON(ret == -EIOCBQUEUED);
641 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
643 int err = security_socket_sendmsg(sock, msg,
646 return err ?: sock_sendmsg_nosec(sock, msg);
648 EXPORT_SYMBOL(sock_sendmsg);
650 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
651 struct kvec *vec, size_t num, size_t size)
653 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
654 return sock_sendmsg(sock, msg);
656 EXPORT_SYMBOL(kernel_sendmsg);
658 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
659 struct kvec *vec, size_t num, size_t size)
661 struct socket *sock = sk->sk_socket;
663 if (!sock->ops->sendmsg_locked)
664 return sock_no_sendmsg_locked(sk, msg, size);
666 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
668 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
670 EXPORT_SYMBOL(kernel_sendmsg_locked);
672 static bool skb_is_err_queue(const struct sk_buff *skb)
674 /* pkt_type of skbs enqueued on the error queue are set to
675 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
676 * in recvmsg, since skbs received on a local socket will never
677 * have a pkt_type of PACKET_OUTGOING.
679 return skb->pkt_type == PACKET_OUTGOING;
682 /* On transmit, software and hardware timestamps are returned independently.
683 * As the two skb clones share the hardware timestamp, which may be updated
684 * before the software timestamp is received, a hardware TX timestamp may be
685 * returned only if there is no software TX timestamp. Ignore false software
686 * timestamps, which may be made in the __sock_recv_timestamp() call when the
687 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
688 * hardware timestamp.
690 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
692 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
695 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
697 struct scm_ts_pktinfo ts_pktinfo;
698 struct net_device *orig_dev;
700 if (!skb_mac_header_was_set(skb))
703 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
706 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
708 ts_pktinfo.if_index = orig_dev->ifindex;
711 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
712 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
713 sizeof(ts_pktinfo), &ts_pktinfo);
717 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
719 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
722 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
723 struct scm_timestamping tss;
724 int empty = 1, false_tstamp = 0;
725 struct skb_shared_hwtstamps *shhwtstamps =
728 /* Race occurred between timestamp enabling and packet
729 receiving. Fill in the current time for now. */
730 if (need_software_tstamp && skb->tstamp == 0) {
731 __net_timestamp(skb);
735 if (need_software_tstamp) {
736 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
738 skb_get_timestamp(skb, &tv);
739 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
743 skb_get_timestampns(skb, &ts);
744 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
749 memset(&tss, 0, sizeof(tss));
750 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
751 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
754 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
755 !skb_is_swtx_tstamp(skb, false_tstamp) &&
756 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
758 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
759 !skb_is_err_queue(skb))
760 put_ts_pktinfo(msg, skb);
763 put_cmsg(msg, SOL_SOCKET,
764 SCM_TIMESTAMPING, sizeof(tss), &tss);
766 if (skb_is_err_queue(skb) && skb->len &&
767 SKB_EXT_ERR(skb)->opt_stats)
768 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
769 skb->len, skb->data);
772 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
774 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
779 if (!sock_flag(sk, SOCK_WIFI_STATUS))
781 if (!skb->wifi_acked_valid)
784 ack = skb->wifi_acked;
786 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
788 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
790 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
793 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
794 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
795 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
798 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
801 sock_recv_timestamp(msg, sk, skb);
802 sock_recv_drops(msg, sk, skb);
804 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
806 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
809 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
812 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
814 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
816 return err ?: sock_recvmsg_nosec(sock, msg, flags);
818 EXPORT_SYMBOL(sock_recvmsg);
821 * kernel_recvmsg - Receive a message from a socket (kernel space)
822 * @sock: The socket to receive the message from
823 * @msg: Received message
824 * @vec: Input s/g array for message data
825 * @num: Size of input s/g array
826 * @size: Number of bytes to read
827 * @flags: Message flags (MSG_DONTWAIT, etc...)
829 * On return the msg structure contains the scatter/gather array passed in the
830 * vec argument. The array is modified so that it consists of the unfilled
831 * portion of the original array.
833 * The returned value is the total number of bytes received, or an error.
835 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
836 struct kvec *vec, size_t num, size_t size, int flags)
838 mm_segment_t oldfs = get_fs();
841 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
843 result = sock_recvmsg(sock, msg, flags);
847 EXPORT_SYMBOL(kernel_recvmsg);
849 static ssize_t sock_sendpage(struct file *file, struct page *page,
850 int offset, size_t size, loff_t *ppos, int more)
855 sock = file->private_data;
857 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
858 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
861 return kernel_sendpage(sock, page, offset, size, flags);
864 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
865 struct pipe_inode_info *pipe, size_t len,
868 struct socket *sock = file->private_data;
870 if (unlikely(!sock->ops->splice_read))
873 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
876 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
878 struct file *file = iocb->ki_filp;
879 struct socket *sock = file->private_data;
880 struct msghdr msg = {.msg_iter = *to,
884 if (file->f_flags & O_NONBLOCK)
885 msg.msg_flags = MSG_DONTWAIT;
887 if (iocb->ki_pos != 0)
890 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
893 res = sock_recvmsg(sock, &msg, msg.msg_flags);
898 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
900 struct file *file = iocb->ki_filp;
901 struct socket *sock = file->private_data;
902 struct msghdr msg = {.msg_iter = *from,
906 if (iocb->ki_pos != 0)
909 if (file->f_flags & O_NONBLOCK)
910 msg.msg_flags = MSG_DONTWAIT;
912 if (sock->type == SOCK_SEQPACKET)
913 msg.msg_flags |= MSG_EOR;
915 res = sock_sendmsg(sock, &msg);
916 *from = msg.msg_iter;
921 * Atomic setting of ioctl hooks to avoid race
922 * with module unload.
925 static DEFINE_MUTEX(br_ioctl_mutex);
926 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
928 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
930 mutex_lock(&br_ioctl_mutex);
931 br_ioctl_hook = hook;
932 mutex_unlock(&br_ioctl_mutex);
934 EXPORT_SYMBOL(brioctl_set);
936 static DEFINE_MUTEX(vlan_ioctl_mutex);
937 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
939 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
941 mutex_lock(&vlan_ioctl_mutex);
942 vlan_ioctl_hook = hook;
943 mutex_unlock(&vlan_ioctl_mutex);
945 EXPORT_SYMBOL(vlan_ioctl_set);
947 static DEFINE_MUTEX(dlci_ioctl_mutex);
948 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
950 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
952 mutex_lock(&dlci_ioctl_mutex);
953 dlci_ioctl_hook = hook;
954 mutex_unlock(&dlci_ioctl_mutex);
956 EXPORT_SYMBOL(dlci_ioctl_set);
958 static long sock_do_ioctl(struct net *net, struct socket *sock,
959 unsigned int cmd, unsigned long arg)
962 void __user *argp = (void __user *)arg;
964 err = sock->ops->ioctl(sock, cmd, arg);
967 * If this ioctl is unknown try to hand it down
970 if (err == -ENOIOCTLCMD)
971 err = dev_ioctl(net, cmd, argp);
977 * With an ioctl, arg may well be a user mode pointer, but we don't know
978 * what to do with it - that's up to the protocol still.
981 static struct ns_common *get_net_ns(struct ns_common *ns)
983 return &get_net(container_of(ns, struct net, ns))->ns;
986 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
990 void __user *argp = (void __user *)arg;
994 sock = file->private_data;
997 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
998 err = dev_ioctl(net, cmd, argp);
1000 #ifdef CONFIG_WEXT_CORE
1001 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1002 err = dev_ioctl(net, cmd, argp);
1009 if (get_user(pid, (int __user *)argp))
1011 err = f_setown(sock->file, pid, 1);
1015 err = put_user(f_getown(sock->file),
1016 (int __user *)argp);
1024 request_module("bridge");
1026 mutex_lock(&br_ioctl_mutex);
1028 err = br_ioctl_hook(net, cmd, argp);
1029 mutex_unlock(&br_ioctl_mutex);
1034 if (!vlan_ioctl_hook)
1035 request_module("8021q");
1037 mutex_lock(&vlan_ioctl_mutex);
1038 if (vlan_ioctl_hook)
1039 err = vlan_ioctl_hook(net, argp);
1040 mutex_unlock(&vlan_ioctl_mutex);
1045 if (!dlci_ioctl_hook)
1046 request_module("dlci");
1048 mutex_lock(&dlci_ioctl_mutex);
1049 if (dlci_ioctl_hook)
1050 err = dlci_ioctl_hook(cmd, argp);
1051 mutex_unlock(&dlci_ioctl_mutex);
1055 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1058 err = open_related_ns(&net->ns, get_net_ns);
1061 err = sock_do_ioctl(net, sock, cmd, arg);
1067 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1070 struct socket *sock = NULL;
1072 err = security_socket_create(family, type, protocol, 1);
1076 sock = sock_alloc();
1083 err = security_socket_post_create(sock, family, type, protocol, 1);
1095 EXPORT_SYMBOL(sock_create_lite);
1097 /* No kernel lock held - perfect */
1098 static unsigned int sock_poll(struct file *file, poll_table *wait)
1100 unsigned int busy_flag = 0;
1101 struct socket *sock;
1104 * We can't return errors to poll, so it's either yes or no.
1106 sock = file->private_data;
1108 if (sk_can_busy_loop(sock->sk)) {
1109 /* this socket can poll_ll so tell the system call */
1110 busy_flag = POLL_BUSY_LOOP;
1112 /* once, only if requested by syscall */
1113 if (wait && (wait->_key & POLL_BUSY_LOOP))
1114 sk_busy_loop(sock->sk, 1);
1117 return busy_flag | sock->ops->poll(file, sock, wait);
1120 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1122 struct socket *sock = file->private_data;
1124 return sock->ops->mmap(file, sock, vma);
1127 static int sock_close(struct inode *inode, struct file *filp)
1129 sock_release(SOCKET_I(inode));
1134 * Update the socket async list
1136 * Fasync_list locking strategy.
1138 * 1. fasync_list is modified only under process context socket lock
1139 * i.e. under semaphore.
1140 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1141 * or under socket lock
1144 static int sock_fasync(int fd, struct file *filp, int on)
1146 struct socket *sock = filp->private_data;
1147 struct sock *sk = sock->sk;
1148 struct socket_wq *wq;
1154 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1155 fasync_helper(fd, filp, on, &wq->fasync_list);
1157 if (!wq->fasync_list)
1158 sock_reset_flag(sk, SOCK_FASYNC);
1160 sock_set_flag(sk, SOCK_FASYNC);
1166 /* This function may be called only under rcu_lock */
1168 int sock_wake_async(struct socket_wq *wq, int how, int band)
1170 if (!wq || !wq->fasync_list)
1174 case SOCK_WAKE_WAITD:
1175 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1178 case SOCK_WAKE_SPACE:
1179 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1184 kill_fasync(&wq->fasync_list, SIGIO, band);
1187 kill_fasync(&wq->fasync_list, SIGURG, band);
1192 EXPORT_SYMBOL(sock_wake_async);
1194 int __sock_create(struct net *net, int family, int type, int protocol,
1195 struct socket **res, int kern)
1198 struct socket *sock;
1199 const struct net_proto_family *pf;
1202 * Check protocol is in range
1204 if (family < 0 || family >= NPROTO)
1205 return -EAFNOSUPPORT;
1206 if (type < 0 || type >= SOCK_MAX)
1211 This uglymoron is moved from INET layer to here to avoid
1212 deadlock in module load.
1214 if (family == PF_INET && type == SOCK_PACKET) {
1215 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1220 err = security_socket_create(family, type, protocol, kern);
1225 * Allocate the socket and allow the family to set things up. if
1226 * the protocol is 0, the family is instructed to select an appropriate
1229 sock = sock_alloc();
1231 net_warn_ratelimited("socket: no more sockets\n");
1232 return -ENFILE; /* Not exactly a match, but its the
1233 closest posix thing */
1238 #ifdef CONFIG_MODULES
1239 /* Attempt to load a protocol module if the find failed.
1241 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1242 * requested real, full-featured networking support upon configuration.
1243 * Otherwise module support will break!
1245 if (rcu_access_pointer(net_families[family]) == NULL)
1246 request_module("net-pf-%d", family);
1250 pf = rcu_dereference(net_families[family]);
1251 err = -EAFNOSUPPORT;
1256 * We will call the ->create function, that possibly is in a loadable
1257 * module, so we have to bump that loadable module refcnt first.
1259 if (!try_module_get(pf->owner))
1262 /* Now protected by module ref count */
1265 err = pf->create(net, sock, protocol, kern);
1267 goto out_module_put;
1270 * Now to bump the refcnt of the [loadable] module that owns this
1271 * socket at sock_release time we decrement its refcnt.
1273 if (!try_module_get(sock->ops->owner))
1274 goto out_module_busy;
1277 * Now that we're done with the ->create function, the [loadable]
1278 * module can have its refcnt decremented
1280 module_put(pf->owner);
1281 err = security_socket_post_create(sock, family, type, protocol, kern);
1283 goto out_sock_release;
1289 err = -EAFNOSUPPORT;
1292 module_put(pf->owner);
1299 goto out_sock_release;
1301 EXPORT_SYMBOL(__sock_create);
1303 int sock_create(int family, int type, int protocol, struct socket **res)
1305 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1307 EXPORT_SYMBOL(sock_create);
1309 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1311 return __sock_create(net, family, type, protocol, res, 1);
1313 EXPORT_SYMBOL(sock_create_kern);
1315 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1318 struct socket *sock;
1321 /* Check the SOCK_* constants for consistency. */
1322 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1323 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1324 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1325 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1327 flags = type & ~SOCK_TYPE_MASK;
1328 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1330 type &= SOCK_TYPE_MASK;
1332 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1333 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1335 retval = sock_create(family, type, protocol, &sock);
1339 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1343 * Create a pair of connected sockets.
1346 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1347 int __user *, usockvec)
1349 struct socket *sock1, *sock2;
1351 struct file *newfile1, *newfile2;
1354 flags = type & ~SOCK_TYPE_MASK;
1355 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1357 type &= SOCK_TYPE_MASK;
1359 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1360 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1363 * reserve descriptors and make sure we won't fail
1364 * to return them to userland.
1366 fd1 = get_unused_fd_flags(flags);
1367 if (unlikely(fd1 < 0))
1370 fd2 = get_unused_fd_flags(flags);
1371 if (unlikely(fd2 < 0)) {
1376 err = put_user(fd1, &usockvec[0]);
1380 err = put_user(fd2, &usockvec[1]);
1385 * Obtain the first socket and check if the underlying protocol
1386 * supports the socketpair call.
1389 err = sock_create(family, type, protocol, &sock1);
1390 if (unlikely(err < 0))
1393 err = sock_create(family, type, protocol, &sock2);
1394 if (unlikely(err < 0)) {
1395 sock_release(sock1);
1399 err = sock1->ops->socketpair(sock1, sock2);
1400 if (unlikely(err < 0)) {
1401 sock_release(sock2);
1402 sock_release(sock1);
1406 newfile1 = sock_alloc_file(sock1, flags, NULL);
1407 if (IS_ERR(newfile1)) {
1408 err = PTR_ERR(newfile1);
1409 sock_release(sock2);
1413 newfile2 = sock_alloc_file(sock2, flags, NULL);
1414 if (IS_ERR(newfile2)) {
1415 err = PTR_ERR(newfile2);
1420 audit_fd_pair(fd1, fd2);
1422 fd_install(fd1, newfile1);
1423 fd_install(fd2, newfile2);
1433 * Bind a name to a socket. Nothing much to do here since it's
1434 * the protocol's responsibility to handle the local address.
1436 * We move the socket address to kernel space before we call
1437 * the protocol layer (having also checked the address is ok).
1440 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1442 struct socket *sock;
1443 struct sockaddr_storage address;
1444 int err, fput_needed;
1446 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1448 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1450 err = security_socket_bind(sock,
1451 (struct sockaddr *)&address,
1454 err = sock->ops->bind(sock,
1458 fput_light(sock->file, fput_needed);
1464 * Perform a listen. Basically, we allow the protocol to do anything
1465 * necessary for a listen, and if that works, we mark the socket as
1466 * ready for listening.
1469 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1471 struct socket *sock;
1472 int err, fput_needed;
1475 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1477 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1478 if ((unsigned int)backlog > somaxconn)
1479 backlog = somaxconn;
1481 err = security_socket_listen(sock, backlog);
1483 err = sock->ops->listen(sock, backlog);
1485 fput_light(sock->file, fput_needed);
1491 * For accept, we attempt to create a new socket, set up the link
1492 * with the client, wake up the client, then return the new
1493 * connected fd. We collect the address of the connector in kernel
1494 * space and move it to user at the very end. This is unclean because
1495 * we open the socket then return an error.
1497 * 1003.1g adds the ability to recvmsg() to query connection pending
1498 * status to recvmsg. We need to add that support in a way thats
1499 * clean when we restucture accept also.
1502 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1503 int __user *, upeer_addrlen, int, flags)
1505 struct socket *sock, *newsock;
1506 struct file *newfile;
1507 int err, len, newfd, fput_needed;
1508 struct sockaddr_storage address;
1510 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1513 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1514 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1516 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1521 newsock = sock_alloc();
1525 newsock->type = sock->type;
1526 newsock->ops = sock->ops;
1529 * We don't need try_module_get here, as the listening socket (sock)
1530 * has the protocol module (sock->ops->owner) held.
1532 __module_get(newsock->ops->owner);
1534 newfd = get_unused_fd_flags(flags);
1535 if (unlikely(newfd < 0)) {
1537 sock_release(newsock);
1540 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1541 if (IS_ERR(newfile)) {
1542 err = PTR_ERR(newfile);
1543 put_unused_fd(newfd);
1547 err = security_socket_accept(sock, newsock);
1551 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1555 if (upeer_sockaddr) {
1556 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1558 err = -ECONNABORTED;
1561 err = move_addr_to_user(&address,
1562 len, upeer_sockaddr, upeer_addrlen);
1567 /* File flags are not inherited via accept() unlike another OSes. */
1569 fd_install(newfd, newfile);
1573 fput_light(sock->file, fput_needed);
1578 put_unused_fd(newfd);
1582 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1583 int __user *, upeer_addrlen)
1585 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1589 * Attempt to connect to a socket with the server address. The address
1590 * is in user space so we verify it is OK and move it to kernel space.
1592 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1595 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1596 * other SEQPACKET protocols that take time to connect() as it doesn't
1597 * include the -EINPROGRESS status for such sockets.
1600 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1603 struct socket *sock;
1604 struct sockaddr_storage address;
1605 int err, fput_needed;
1607 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1610 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1615 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1619 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1620 sock->file->f_flags);
1622 fput_light(sock->file, fput_needed);
1628 * Get the local address ('name') of a socket object. Move the obtained
1629 * name to user space.
1632 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1633 int __user *, usockaddr_len)
1635 struct socket *sock;
1636 struct sockaddr_storage address;
1637 int len, err, fput_needed;
1639 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1643 err = security_socket_getsockname(sock);
1647 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1650 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1653 fput_light(sock->file, fput_needed);
1659 * Get the remote address ('name') of a socket object. Move the obtained
1660 * name to user space.
1663 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1664 int __user *, usockaddr_len)
1666 struct socket *sock;
1667 struct sockaddr_storage address;
1668 int len, err, fput_needed;
1670 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672 err = security_socket_getpeername(sock);
1674 fput_light(sock->file, fput_needed);
1679 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1682 err = move_addr_to_user(&address, len, usockaddr,
1684 fput_light(sock->file, fput_needed);
1690 * Send a datagram to a given address. We move the address into kernel
1691 * space and check the user space data area is readable before invoking
1695 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1696 unsigned int, flags, struct sockaddr __user *, addr,
1699 struct socket *sock;
1700 struct sockaddr_storage address;
1706 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1709 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1713 msg.msg_name = NULL;
1714 msg.msg_control = NULL;
1715 msg.msg_controllen = 0;
1716 msg.msg_namelen = 0;
1718 err = move_addr_to_kernel(addr, addr_len, &address);
1721 msg.msg_name = (struct sockaddr *)&address;
1722 msg.msg_namelen = addr_len;
1724 if (sock->file->f_flags & O_NONBLOCK)
1725 flags |= MSG_DONTWAIT;
1726 msg.msg_flags = flags;
1727 err = sock_sendmsg(sock, &msg);
1730 fput_light(sock->file, fput_needed);
1736 * Send a datagram down a socket.
1739 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1740 unsigned int, flags)
1742 return sys_sendto(fd, buff, len, flags, NULL, 0);
1746 * Receive a frame from the socket and optionally record the address of the
1747 * sender. We verify the buffers are writable and if needed move the
1748 * sender address from kernel to user space.
1751 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1752 unsigned int, flags, struct sockaddr __user *, addr,
1753 int __user *, addr_len)
1755 struct socket *sock;
1758 struct sockaddr_storage address;
1762 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1765 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 msg.msg_control = NULL;
1770 msg.msg_controllen = 0;
1771 /* Save some cycles and don't copy the address if not needed */
1772 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1773 /* We assume all kernel code knows the size of sockaddr_storage */
1774 msg.msg_namelen = 0;
1775 msg.msg_iocb = NULL;
1777 if (sock->file->f_flags & O_NONBLOCK)
1778 flags |= MSG_DONTWAIT;
1779 err = sock_recvmsg(sock, &msg, flags);
1781 if (err >= 0 && addr != NULL) {
1782 err2 = move_addr_to_user(&address,
1783 msg.msg_namelen, addr, addr_len);
1788 fput_light(sock->file, fput_needed);
1794 * Receive a datagram from a socket.
1797 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1798 unsigned int, flags)
1800 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1804 * Set a socket option. Because we don't know the option lengths we have
1805 * to pass the user mode parameter for the protocols to sort out.
1808 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1809 char __user *, optval, int, optlen)
1811 int err, fput_needed;
1812 struct socket *sock;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1819 err = security_socket_setsockopt(sock, level, optname);
1823 if (level == SOL_SOCKET)
1825 sock_setsockopt(sock, level, optname, optval,
1829 sock->ops->setsockopt(sock, level, optname, optval,
1832 fput_light(sock->file, fput_needed);
1838 * Get a socket option. Because we don't know the option lengths we have
1839 * to pass a user mode parameter for the protocols to sort out.
1842 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1843 char __user *, optval, int __user *, optlen)
1845 int err, fput_needed;
1846 struct socket *sock;
1848 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1850 err = security_socket_getsockopt(sock, level, optname);
1854 if (level == SOL_SOCKET)
1856 sock_getsockopt(sock, level, optname, optval,
1860 sock->ops->getsockopt(sock, level, optname, optval,
1863 fput_light(sock->file, fput_needed);
1869 * Shutdown a socket.
1872 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1874 int err, fput_needed;
1875 struct socket *sock;
1877 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1879 err = security_socket_shutdown(sock, how);
1881 err = sock->ops->shutdown(sock, how);
1882 fput_light(sock->file, fput_needed);
1887 /* A couple of helpful macros for getting the address of the 32/64 bit
1888 * fields which are the same type (int / unsigned) on our platforms.
1890 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1891 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1892 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1894 struct used_address {
1895 struct sockaddr_storage name;
1896 unsigned int name_len;
1899 static int copy_msghdr_from_user(struct msghdr *kmsg,
1900 struct user_msghdr __user *umsg,
1901 struct sockaddr __user **save_addr,
1904 struct user_msghdr msg;
1907 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1910 kmsg->msg_control = (void __force *)msg.msg_control;
1911 kmsg->msg_controllen = msg.msg_controllen;
1912 kmsg->msg_flags = msg.msg_flags;
1914 kmsg->msg_namelen = msg.msg_namelen;
1916 kmsg->msg_namelen = 0;
1918 if (kmsg->msg_namelen < 0)
1921 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1922 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1925 *save_addr = msg.msg_name;
1927 if (msg.msg_name && kmsg->msg_namelen) {
1929 err = move_addr_to_kernel(msg.msg_name,
1936 kmsg->msg_name = NULL;
1937 kmsg->msg_namelen = 0;
1940 if (msg.msg_iovlen > UIO_MAXIOV)
1943 kmsg->msg_iocb = NULL;
1945 return import_iovec(save_addr ? READ : WRITE,
1946 msg.msg_iov, msg.msg_iovlen,
1947 UIO_FASTIOV, iov, &kmsg->msg_iter);
1950 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1951 struct msghdr *msg_sys, unsigned int flags,
1952 struct used_address *used_address,
1953 unsigned int allowed_msghdr_flags)
1955 struct compat_msghdr __user *msg_compat =
1956 (struct compat_msghdr __user *)msg;
1957 struct sockaddr_storage address;
1958 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1959 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1960 __aligned(sizeof(__kernel_size_t));
1961 /* 20 is size of ipv6_pktinfo */
1962 unsigned char *ctl_buf = ctl;
1966 msg_sys->msg_name = &address;
1968 if (MSG_CMSG_COMPAT & flags)
1969 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1971 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1977 if (msg_sys->msg_controllen > INT_MAX)
1979 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1980 ctl_len = msg_sys->msg_controllen;
1981 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1983 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1987 ctl_buf = msg_sys->msg_control;
1988 ctl_len = msg_sys->msg_controllen;
1989 } else if (ctl_len) {
1990 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
1991 CMSG_ALIGN(sizeof(struct cmsghdr)));
1992 if (ctl_len > sizeof(ctl)) {
1993 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1994 if (ctl_buf == NULL)
1999 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2000 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2001 * checking falls down on this.
2003 if (copy_from_user(ctl_buf,
2004 (void __user __force *)msg_sys->msg_control,
2007 msg_sys->msg_control = ctl_buf;
2009 msg_sys->msg_flags = flags;
2011 if (sock->file->f_flags & O_NONBLOCK)
2012 msg_sys->msg_flags |= MSG_DONTWAIT;
2014 * If this is sendmmsg() and current destination address is same as
2015 * previously succeeded address, omit asking LSM's decision.
2016 * used_address->name_len is initialized to UINT_MAX so that the first
2017 * destination address never matches.
2019 if (used_address && msg_sys->msg_name &&
2020 used_address->name_len == msg_sys->msg_namelen &&
2021 !memcmp(&used_address->name, msg_sys->msg_name,
2022 used_address->name_len)) {
2023 err = sock_sendmsg_nosec(sock, msg_sys);
2026 err = sock_sendmsg(sock, msg_sys);
2028 * If this is sendmmsg() and sending to current destination address was
2029 * successful, remember it.
2031 if (used_address && err >= 0) {
2032 used_address->name_len = msg_sys->msg_namelen;
2033 if (msg_sys->msg_name)
2034 memcpy(&used_address->name, msg_sys->msg_name,
2035 used_address->name_len);
2040 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2047 * BSD sendmsg interface
2050 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2052 int fput_needed, err;
2053 struct msghdr msg_sys;
2054 struct socket *sock;
2056 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2060 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2062 fput_light(sock->file, fput_needed);
2067 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2069 if (flags & MSG_CMSG_COMPAT)
2071 return __sys_sendmsg(fd, msg, flags);
2075 * Linux sendmmsg interface
2078 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2081 int fput_needed, err, datagrams;
2082 struct socket *sock;
2083 struct mmsghdr __user *entry;
2084 struct compat_mmsghdr __user *compat_entry;
2085 struct msghdr msg_sys;
2086 struct used_address used_address;
2087 unsigned int oflags = flags;
2089 if (vlen > UIO_MAXIOV)
2094 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2098 used_address.name_len = UINT_MAX;
2100 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2104 while (datagrams < vlen) {
2105 if (datagrams == vlen - 1)
2108 if (MSG_CMSG_COMPAT & flags) {
2109 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2110 &msg_sys, flags, &used_address, MSG_EOR);
2113 err = __put_user(err, &compat_entry->msg_len);
2116 err = ___sys_sendmsg(sock,
2117 (struct user_msghdr __user *)entry,
2118 &msg_sys, flags, &used_address, MSG_EOR);
2121 err = put_user(err, &entry->msg_len);
2128 if (msg_data_left(&msg_sys))
2133 fput_light(sock->file, fput_needed);
2135 /* We only return an error if no datagrams were able to be sent */
2142 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2143 unsigned int, vlen, unsigned int, flags)
2145 if (flags & MSG_CMSG_COMPAT)
2147 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2150 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2151 struct msghdr *msg_sys, unsigned int flags, int nosec)
2153 struct compat_msghdr __user *msg_compat =
2154 (struct compat_msghdr __user *)msg;
2155 struct iovec iovstack[UIO_FASTIOV];
2156 struct iovec *iov = iovstack;
2157 unsigned long cmsg_ptr;
2161 /* kernel mode address */
2162 struct sockaddr_storage addr;
2164 /* user mode address pointers */
2165 struct sockaddr __user *uaddr;
2166 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2168 msg_sys->msg_name = &addr;
2170 if (MSG_CMSG_COMPAT & flags)
2171 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2173 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2177 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2178 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2180 /* We assume all kernel code knows the size of sockaddr_storage */
2181 msg_sys->msg_namelen = 0;
2183 if (sock->file->f_flags & O_NONBLOCK)
2184 flags |= MSG_DONTWAIT;
2185 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2190 if (uaddr != NULL) {
2191 err = move_addr_to_user(&addr,
2192 msg_sys->msg_namelen, uaddr,
2197 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2201 if (MSG_CMSG_COMPAT & flags)
2202 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2203 &msg_compat->msg_controllen);
2205 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2206 &msg->msg_controllen);
2217 * BSD recvmsg interface
2220 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2222 int fput_needed, err;
2223 struct msghdr msg_sys;
2224 struct socket *sock;
2226 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2230 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2232 fput_light(sock->file, fput_needed);
2237 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2238 unsigned int, flags)
2240 if (flags & MSG_CMSG_COMPAT)
2242 return __sys_recvmsg(fd, msg, flags);
2246 * Linux recvmmsg interface
2249 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2250 unsigned int flags, struct timespec *timeout)
2252 int fput_needed, err, datagrams;
2253 struct socket *sock;
2254 struct mmsghdr __user *entry;
2255 struct compat_mmsghdr __user *compat_entry;
2256 struct msghdr msg_sys;
2257 struct timespec64 end_time;
2258 struct timespec64 timeout64;
2261 poll_select_set_timeout(&end_time, timeout->tv_sec,
2267 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2271 err = sock_error(sock->sk);
2278 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2280 while (datagrams < vlen) {
2282 * No need to ask LSM for more than the first datagram.
2284 if (MSG_CMSG_COMPAT & flags) {
2285 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2286 &msg_sys, flags & ~MSG_WAITFORONE,
2290 err = __put_user(err, &compat_entry->msg_len);
2293 err = ___sys_recvmsg(sock,
2294 (struct user_msghdr __user *)entry,
2295 &msg_sys, flags & ~MSG_WAITFORONE,
2299 err = put_user(err, &entry->msg_len);
2307 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2308 if (flags & MSG_WAITFORONE)
2309 flags |= MSG_DONTWAIT;
2312 ktime_get_ts64(&timeout64);
2313 *timeout = timespec64_to_timespec(
2314 timespec64_sub(end_time, timeout64));
2315 if (timeout->tv_sec < 0) {
2316 timeout->tv_sec = timeout->tv_nsec = 0;
2320 /* Timeout, return less than vlen datagrams */
2321 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2325 /* Out of band data, return right away */
2326 if (msg_sys.msg_flags & MSG_OOB)
2334 if (datagrams == 0) {
2340 * We may return less entries than requested (vlen) if the
2341 * sock is non block and there aren't enough datagrams...
2343 if (err != -EAGAIN) {
2345 * ... or if recvmsg returns an error after we
2346 * received some datagrams, where we record the
2347 * error to return on the next call or if the
2348 * app asks about it using getsockopt(SO_ERROR).
2350 sock->sk->sk_err = -err;
2353 fput_light(sock->file, fput_needed);
2358 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2359 unsigned int, vlen, unsigned int, flags,
2360 struct timespec __user *, timeout)
2363 struct timespec timeout_sys;
2365 if (flags & MSG_CMSG_COMPAT)
2369 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2371 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2374 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2376 if (datagrams > 0 &&
2377 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2378 datagrams = -EFAULT;
2383 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2384 /* Argument list sizes for sys_socketcall */
2385 #define AL(x) ((x) * sizeof(unsigned long))
2386 static const unsigned char nargs[21] = {
2387 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2388 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2389 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2396 * System call vectors.
2398 * Argument checking cleaned up. Saved 20% in size.
2399 * This function doesn't need to set the kernel lock because
2400 * it is set by the callees.
2403 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2405 unsigned long a[AUDITSC_ARGS];
2406 unsigned long a0, a1;
2410 if (call < 1 || call > SYS_SENDMMSG)
2414 if (len > sizeof(a))
2417 /* copy_from_user should be SMP safe. */
2418 if (copy_from_user(a, args, len))
2421 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2430 err = sys_socket(a0, a1, a[2]);
2433 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2436 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2439 err = sys_listen(a0, a1);
2442 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2443 (int __user *)a[2], 0);
2445 case SYS_GETSOCKNAME:
2447 sys_getsockname(a0, (struct sockaddr __user *)a1,
2448 (int __user *)a[2]);
2450 case SYS_GETPEERNAME:
2452 sys_getpeername(a0, (struct sockaddr __user *)a1,
2453 (int __user *)a[2]);
2455 case SYS_SOCKETPAIR:
2456 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2459 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2462 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2463 (struct sockaddr __user *)a[4], a[5]);
2466 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2469 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2470 (struct sockaddr __user *)a[4],
2471 (int __user *)a[5]);
2474 err = sys_shutdown(a0, a1);
2476 case SYS_SETSOCKOPT:
2477 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2479 case SYS_GETSOCKOPT:
2481 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2482 (int __user *)a[4]);
2485 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2488 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2491 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2494 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2495 (struct timespec __user *)a[4]);
2498 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2499 (int __user *)a[2], a[3]);
2508 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2511 * sock_register - add a socket protocol handler
2512 * @ops: description of protocol
2514 * This function is called by a protocol handler that wants to
2515 * advertise its address family, and have it linked into the
2516 * socket interface. The value ops->family corresponds to the
2517 * socket system call protocol family.
2519 int sock_register(const struct net_proto_family *ops)
2523 if (ops->family >= NPROTO) {
2524 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2528 spin_lock(&net_family_lock);
2529 if (rcu_dereference_protected(net_families[ops->family],
2530 lockdep_is_held(&net_family_lock)))
2533 rcu_assign_pointer(net_families[ops->family], ops);
2536 spin_unlock(&net_family_lock);
2538 pr_info("NET: Registered protocol family %d\n", ops->family);
2541 EXPORT_SYMBOL(sock_register);
2544 * sock_unregister - remove a protocol handler
2545 * @family: protocol family to remove
2547 * This function is called by a protocol handler that wants to
2548 * remove its address family, and have it unlinked from the
2549 * new socket creation.
2551 * If protocol handler is a module, then it can use module reference
2552 * counts to protect against new references. If protocol handler is not
2553 * a module then it needs to provide its own protection in
2554 * the ops->create routine.
2556 void sock_unregister(int family)
2558 BUG_ON(family < 0 || family >= NPROTO);
2560 spin_lock(&net_family_lock);
2561 RCU_INIT_POINTER(net_families[family], NULL);
2562 spin_unlock(&net_family_lock);
2566 pr_info("NET: Unregistered protocol family %d\n", family);
2568 EXPORT_SYMBOL(sock_unregister);
2570 static int __init sock_init(void)
2574 * Initialize the network sysctl infrastructure.
2576 err = net_sysctl_init();
2581 * Initialize skbuff SLAB cache
2586 * Initialize the protocols module.
2591 err = register_filesystem(&sock_fs_type);
2594 sock_mnt = kern_mount(&sock_fs_type);
2595 if (IS_ERR(sock_mnt)) {
2596 err = PTR_ERR(sock_mnt);
2600 /* The real protocol initialization is performed in later initcalls.
2603 #ifdef CONFIG_NETFILTER
2604 err = netfilter_init();
2609 ptp_classifier_init();
2615 unregister_filesystem(&sock_fs_type);
2620 core_initcall(sock_init); /* early initcall */
2622 #ifdef CONFIG_PROC_FS
2623 void socket_seq_show(struct seq_file *seq)
2628 for_each_possible_cpu(cpu)
2629 counter += per_cpu(sockets_in_use, cpu);
2631 /* It can be negative, by the way. 8) */
2635 seq_printf(seq, "sockets: used %d\n", counter);
2637 #endif /* CONFIG_PROC_FS */
2639 #ifdef CONFIG_COMPAT
2640 static int do_siocgstamp(struct net *net, struct socket *sock,
2641 unsigned int cmd, void __user *up)
2643 mm_segment_t old_fs = get_fs();
2648 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2651 err = compat_put_timeval(&ktv, up);
2656 static int do_siocgstampns(struct net *net, struct socket *sock,
2657 unsigned int cmd, void __user *up)
2659 mm_segment_t old_fs = get_fs();
2660 struct timespec kts;
2664 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2667 err = compat_put_timespec(&kts, up);
2672 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2674 struct ifreq __user *uifr;
2677 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2678 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2681 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2685 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2691 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2693 struct compat_ifconf ifc32;
2695 struct ifconf __user *uifc;
2696 struct compat_ifreq __user *ifr32;
2697 struct ifreq __user *ifr;
2701 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2704 memset(&ifc, 0, sizeof(ifc));
2705 if (ifc32.ifcbuf == 0) {
2709 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2711 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2712 sizeof(struct ifreq);
2713 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2715 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2716 ifr32 = compat_ptr(ifc32.ifcbuf);
2717 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2718 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2724 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2727 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2731 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2735 ifr32 = compat_ptr(ifc32.ifcbuf);
2737 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2738 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2739 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2745 if (ifc32.ifcbuf == 0) {
2746 /* Translate from 64-bit structure multiple to
2750 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2755 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2761 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2763 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2764 bool convert_in = false, convert_out = false;
2765 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2766 struct ethtool_rxnfc __user *rxnfc;
2767 struct ifreq __user *ifr;
2768 u32 rule_cnt = 0, actual_rule_cnt;
2773 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2776 compat_rxnfc = compat_ptr(data);
2778 if (get_user(ethcmd, &compat_rxnfc->cmd))
2781 /* Most ethtool structures are defined without padding.
2782 * Unfortunately struct ethtool_rxnfc is an exception.
2787 case ETHTOOL_GRXCLSRLALL:
2788 /* Buffer size is variable */
2789 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2791 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2793 buf_size += rule_cnt * sizeof(u32);
2795 case ETHTOOL_GRXRINGS:
2796 case ETHTOOL_GRXCLSRLCNT:
2797 case ETHTOOL_GRXCLSRULE:
2798 case ETHTOOL_SRXCLSRLINS:
2801 case ETHTOOL_SRXCLSRLDEL:
2802 buf_size += sizeof(struct ethtool_rxnfc);
2807 ifr = compat_alloc_user_space(buf_size);
2808 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2810 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2813 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2814 &ifr->ifr_ifru.ifru_data))
2818 /* We expect there to be holes between fs.m_ext and
2819 * fs.ring_cookie and at the end of fs, but nowhere else.
2821 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2822 sizeof(compat_rxnfc->fs.m_ext) !=
2823 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2824 sizeof(rxnfc->fs.m_ext));
2826 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2827 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2828 offsetof(struct ethtool_rxnfc, fs.location) -
2829 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2831 if (copy_in_user(rxnfc, compat_rxnfc,
2832 (void __user *)(&rxnfc->fs.m_ext + 1) -
2833 (void __user *)rxnfc) ||
2834 copy_in_user(&rxnfc->fs.ring_cookie,
2835 &compat_rxnfc->fs.ring_cookie,
2836 (void __user *)(&rxnfc->fs.location + 1) -
2837 (void __user *)&rxnfc->fs.ring_cookie) ||
2838 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2839 sizeof(rxnfc->rule_cnt)))
2843 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2848 if (copy_in_user(compat_rxnfc, rxnfc,
2849 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2850 (const void __user *)rxnfc) ||
2851 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2852 &rxnfc->fs.ring_cookie,
2853 (const void __user *)(&rxnfc->fs.location + 1) -
2854 (const void __user *)&rxnfc->fs.ring_cookie) ||
2855 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2856 sizeof(rxnfc->rule_cnt)))
2859 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2860 /* As an optimisation, we only copy the actual
2861 * number of rules that the underlying
2862 * function returned. Since Mallory might
2863 * change the rule count in user memory, we
2864 * check that it is less than the rule count
2865 * originally given (as the user buffer size),
2866 * which has been range-checked.
2868 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2870 if (actual_rule_cnt < rule_cnt)
2871 rule_cnt = actual_rule_cnt;
2872 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2873 &rxnfc->rule_locs[0],
2874 rule_cnt * sizeof(u32)))
2882 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2885 compat_uptr_t uptr32;
2886 struct ifreq __user *uifr;
2888 uifr = compat_alloc_user_space(sizeof(*uifr));
2889 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2892 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2895 uptr = compat_ptr(uptr32);
2897 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2900 return dev_ioctl(net, SIOCWANDEV, uifr);
2903 static int bond_ioctl(struct net *net, unsigned int cmd,
2904 struct compat_ifreq __user *ifr32)
2907 mm_segment_t old_fs;
2911 case SIOCBONDENSLAVE:
2912 case SIOCBONDRELEASE:
2913 case SIOCBONDSETHWADDR:
2914 case SIOCBONDCHANGEACTIVE:
2915 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2920 err = dev_ioctl(net, cmd,
2921 (struct ifreq __user __force *) &kifr);
2926 return -ENOIOCTLCMD;
2930 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2931 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2932 struct compat_ifreq __user *u_ifreq32)
2934 struct ifreq __user *u_ifreq64;
2935 char tmp_buf[IFNAMSIZ];
2936 void __user *data64;
2939 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2942 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2944 data64 = compat_ptr(data32);
2946 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2948 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2951 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2954 return dev_ioctl(net, cmd, u_ifreq64);
2957 static int dev_ifsioc(struct net *net, struct socket *sock,
2958 unsigned int cmd, struct compat_ifreq __user *uifr32)
2960 struct ifreq __user *uifr;
2963 uifr = compat_alloc_user_space(sizeof(*uifr));
2964 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2967 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2978 case SIOCGIFBRDADDR:
2979 case SIOCGIFDSTADDR:
2980 case SIOCGIFNETMASK:
2985 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2993 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2994 struct compat_ifreq __user *uifr32)
2997 struct compat_ifmap __user *uifmap32;
2998 mm_segment_t old_fs;
3001 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3002 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3003 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3004 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3005 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3006 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3007 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3008 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3014 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3017 if (cmd == SIOCGIFMAP && !err) {
3018 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3019 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3020 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3021 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3022 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3023 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3024 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3033 struct sockaddr rt_dst; /* target address */
3034 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3035 struct sockaddr rt_genmask; /* target network mask (IP) */
3036 unsigned short rt_flags;
3039 unsigned char rt_tos;
3040 unsigned char rt_class;
3042 short rt_metric; /* +1 for binary compatibility! */
3043 /* char * */ u32 rt_dev; /* forcing the device at add */
3044 u32 rt_mtu; /* per route MTU/Window */
3045 u32 rt_window; /* Window clamping */
3046 unsigned short rt_irtt; /* Initial RTT */
3049 struct in6_rtmsg32 {
3050 struct in6_addr rtmsg_dst;
3051 struct in6_addr rtmsg_src;
3052 struct in6_addr rtmsg_gateway;
3062 static int routing_ioctl(struct net *net, struct socket *sock,
3063 unsigned int cmd, void __user *argp)
3067 struct in6_rtmsg r6;
3071 mm_segment_t old_fs = get_fs();
3073 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3074 struct in6_rtmsg32 __user *ur6 = argp;
3075 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3076 3 * sizeof(struct in6_addr));
3077 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3078 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3079 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3080 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3081 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3082 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3083 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3087 struct rtentry32 __user *ur4 = argp;
3088 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3089 3 * sizeof(struct sockaddr));
3090 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3091 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3092 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3093 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3094 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3095 ret |= get_user(rtdev, &(ur4->rt_dev));
3097 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3098 r4.rt_dev = (char __user __force *)devname;
3112 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3119 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3120 * for some operations; this forces use of the newer bridge-utils that
3121 * use compatible ioctls
3123 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3127 if (get_user(tmp, argp))
3129 if (tmp == BRCTL_GET_VERSION)
3130 return BRCTL_VERSION + 1;
3134 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3135 unsigned int cmd, unsigned long arg)
3137 void __user *argp = compat_ptr(arg);
3138 struct sock *sk = sock->sk;
3139 struct net *net = sock_net(sk);
3141 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3142 return compat_ifr_data_ioctl(net, cmd, argp);
3147 return old_bridge_ioctl(argp);
3149 return dev_ifname32(net, argp);
3151 return dev_ifconf(net, argp);
3153 return ethtool_ioctl(net, argp);
3155 return compat_siocwandev(net, argp);
3158 return compat_sioc_ifmap(net, cmd, argp);
3159 case SIOCBONDENSLAVE:
3160 case SIOCBONDRELEASE:
3161 case SIOCBONDSETHWADDR:
3162 case SIOCBONDCHANGEACTIVE:
3163 return bond_ioctl(net, cmd, argp);
3166 return routing_ioctl(net, sock, cmd, argp);
3168 return do_siocgstamp(net, sock, cmd, argp);
3170 return do_siocgstampns(net, sock, cmd, argp);
3171 case SIOCBONDSLAVEINFOQUERY:
3172 case SIOCBONDINFOQUERY:
3175 return compat_ifr_data_ioctl(net, cmd, argp);
3188 return sock_ioctl(file, cmd, arg);
3205 case SIOCSIFHWBROADCAST:
3207 case SIOCGIFBRDADDR:
3208 case SIOCSIFBRDADDR:
3209 case SIOCGIFDSTADDR:
3210 case SIOCSIFDSTADDR:
3211 case SIOCGIFNETMASK:
3212 case SIOCSIFNETMASK:
3223 return dev_ifsioc(net, sock, cmd, argp);
3229 return sock_do_ioctl(net, sock, cmd, arg);
3232 return -ENOIOCTLCMD;
3235 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3238 struct socket *sock = file->private_data;
3239 int ret = -ENOIOCTLCMD;
3246 if (sock->ops->compat_ioctl)
3247 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3249 if (ret == -ENOIOCTLCMD &&
3250 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3251 ret = compat_wext_handle_ioctl(net, cmd, arg);
3253 if (ret == -ENOIOCTLCMD)
3254 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3260 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3262 return sock->ops->bind(sock, addr, addrlen);
3264 EXPORT_SYMBOL(kernel_bind);
3266 int kernel_listen(struct socket *sock, int backlog)
3268 return sock->ops->listen(sock, backlog);
3270 EXPORT_SYMBOL(kernel_listen);
3272 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3274 struct sock *sk = sock->sk;
3277 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3282 err = sock->ops->accept(sock, *newsock, flags, true);
3284 sock_release(*newsock);
3289 (*newsock)->ops = sock->ops;
3290 __module_get((*newsock)->ops->owner);
3295 EXPORT_SYMBOL(kernel_accept);
3297 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3300 return sock->ops->connect(sock, addr, addrlen, flags);
3302 EXPORT_SYMBOL(kernel_connect);
3304 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3307 return sock->ops->getname(sock, addr, addrlen, 0);
3309 EXPORT_SYMBOL(kernel_getsockname);
3311 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3314 return sock->ops->getname(sock, addr, addrlen, 1);
3316 EXPORT_SYMBOL(kernel_getpeername);
3318 int kernel_getsockopt(struct socket *sock, int level, int optname,
3319 char *optval, int *optlen)
3321 mm_segment_t oldfs = get_fs();
3322 char __user *uoptval;
3323 int __user *uoptlen;
3326 uoptval = (char __user __force *) optval;
3327 uoptlen = (int __user __force *) optlen;
3330 if (level == SOL_SOCKET)
3331 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3333 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3338 EXPORT_SYMBOL(kernel_getsockopt);
3340 int kernel_setsockopt(struct socket *sock, int level, int optname,
3341 char *optval, unsigned int optlen)
3343 mm_segment_t oldfs = get_fs();
3344 char __user *uoptval;
3347 uoptval = (char __user __force *) optval;
3350 if (level == SOL_SOCKET)
3351 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3353 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3358 EXPORT_SYMBOL(kernel_setsockopt);
3360 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3361 size_t size, int flags)
3363 if (sock->ops->sendpage)
3364 return sock->ops->sendpage(sock, page, offset, size, flags);
3366 return sock_no_sendpage(sock, page, offset, size, flags);
3368 EXPORT_SYMBOL(kernel_sendpage);
3370 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3371 size_t size, int flags)
3373 struct socket *sock = sk->sk_socket;
3375 if (sock->ops->sendpage_locked)
3376 return sock->ops->sendpage_locked(sk, page, offset, size,
3379 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3381 EXPORT_SYMBOL(kernel_sendpage_locked);
3383 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3385 mm_segment_t oldfs = get_fs();
3389 err = sock->ops->ioctl(sock, cmd, arg);
3394 EXPORT_SYMBOL(kernel_sock_ioctl);
3396 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3398 return sock->ops->shutdown(sock, how);
3400 EXPORT_SYMBOL(kernel_sock_shutdown);
3402 /* This routine returns the IP overhead imposed by a socket i.e.
3403 * the length of the underlying IP header, depending on whether
3404 * this is an IPv4 or IPv6 socket and the length from IP options turned
3405 * on at the socket. Assumes that the caller has a lock on the socket.
3407 u32 kernel_sock_ip_overhead(struct sock *sk)
3409 struct inet_sock *inet;
3410 struct ip_options_rcu *opt;
3412 #if IS_ENABLED(CONFIG_IPV6)
3413 struct ipv6_pinfo *np;
3414 struct ipv6_txoptions *optv6 = NULL;
3415 #endif /* IS_ENABLED(CONFIG_IPV6) */
3420 switch (sk->sk_family) {
3423 overhead += sizeof(struct iphdr);
3424 opt = rcu_dereference_protected(inet->inet_opt,
3425 sock_owned_by_user(sk));
3427 overhead += opt->opt.optlen;
3429 #if IS_ENABLED(CONFIG_IPV6)
3432 overhead += sizeof(struct ipv6hdr);
3434 optv6 = rcu_dereference_protected(np->opt,
3435 sock_owned_by_user(sk));
3437 overhead += (optv6->opt_flen + optv6->opt_nflen);
3439 #endif /* IS_ENABLED(CONFIG_IPV6) */
3440 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3444 EXPORT_SYMBOL(kernel_sock_ip_overhead);