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))
410 return ERR_PTR(-ENOMEM);
411 path.mnt = mntget(sock_mnt);
413 d_instantiate(path.dentry, SOCK_INODE(sock));
415 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
418 /* drop dentry, keep inode */
419 ihold(d_inode(path.dentry));
425 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
426 file->private_data = sock;
429 EXPORT_SYMBOL(sock_alloc_file);
431 static int sock_map_fd(struct socket *sock, int flags)
433 struct file *newfile;
434 int fd = get_unused_fd_flags(flags);
435 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 kmemcheck_annotate_bitfield(sock, type);
572 inode->i_ino = get_next_ino();
573 inode->i_mode = S_IFSOCK | S_IRWXUGO;
574 inode->i_uid = current_fsuid();
575 inode->i_gid = current_fsgid();
576 inode->i_op = &sockfs_inode_ops;
578 this_cpu_add(sockets_in_use, 1);
581 EXPORT_SYMBOL(sock_alloc);
584 * sock_release - close a socket
585 * @sock: socket to close
587 * The socket is released from the protocol stack if it has a release
588 * callback, and the inode is then released if the socket is bound to
589 * an inode not a file.
592 void sock_release(struct socket *sock)
595 struct module *owner = sock->ops->owner;
597 sock->ops->release(sock);
602 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
603 pr_err("%s: fasync list not empty!\n", __func__);
605 this_cpu_sub(sockets_in_use, 1);
607 iput(SOCK_INODE(sock));
612 EXPORT_SYMBOL(sock_release);
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
616 u8 flags = *tx_flags;
618 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619 flags |= SKBTX_HW_TSTAMP;
621 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622 flags |= SKBTX_SW_TSTAMP;
624 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625 flags |= SKBTX_SCHED_TSTAMP;
629 EXPORT_SYMBOL(__sock_tx_timestamp);
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
633 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634 BUG_ON(ret == -EIOCBQUEUED);
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
640 int err = security_socket_sendmsg(sock, msg,
643 return err ?: sock_sendmsg_nosec(sock, msg);
645 EXPORT_SYMBOL(sock_sendmsg);
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 struct kvec *vec, size_t num, size_t size)
650 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651 return sock_sendmsg(sock, msg);
653 EXPORT_SYMBOL(kernel_sendmsg);
655 static bool skb_is_err_queue(const struct sk_buff *skb)
657 /* pkt_type of skbs enqueued on the error queue are set to
658 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
659 * in recvmsg, since skbs received on a local socket will never
660 * have a pkt_type of PACKET_OUTGOING.
662 return skb->pkt_type == PACKET_OUTGOING;
665 /* On transmit, software and hardware timestamps are returned independently.
666 * As the two skb clones share the hardware timestamp, which may be updated
667 * before the software timestamp is received, a hardware TX timestamp may be
668 * returned only if there is no software TX timestamp. Ignore false software
669 * timestamps, which may be made in the __sock_recv_timestamp() call when the
670 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
671 * hardware timestamp.
673 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
675 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
678 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
680 struct scm_ts_pktinfo ts_pktinfo;
681 struct net_device *orig_dev;
683 if (!skb_mac_header_was_set(skb))
686 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
689 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
691 ts_pktinfo.if_index = orig_dev->ifindex;
694 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
695 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
696 sizeof(ts_pktinfo), &ts_pktinfo);
700 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
702 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
705 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
706 struct scm_timestamping tss;
707 int empty = 1, false_tstamp = 0;
708 struct skb_shared_hwtstamps *shhwtstamps =
711 /* Race occurred between timestamp enabling and packet
712 receiving. Fill in the current time for now. */
713 if (need_software_tstamp && skb->tstamp == 0) {
714 __net_timestamp(skb);
718 if (need_software_tstamp) {
719 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
721 skb_get_timestamp(skb, &tv);
722 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
726 skb_get_timestampns(skb, &ts);
727 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
732 memset(&tss, 0, sizeof(tss));
733 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
734 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
737 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
738 !skb_is_swtx_tstamp(skb, false_tstamp) &&
739 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
741 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
742 !skb_is_err_queue(skb))
743 put_ts_pktinfo(msg, skb);
746 put_cmsg(msg, SOL_SOCKET,
747 SCM_TIMESTAMPING, sizeof(tss), &tss);
749 if (skb_is_err_queue(skb) && skb->len &&
750 SKB_EXT_ERR(skb)->opt_stats)
751 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
752 skb->len, skb->data);
755 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
757 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
762 if (!sock_flag(sk, SOCK_WIFI_STATUS))
764 if (!skb->wifi_acked_valid)
767 ack = skb->wifi_acked;
769 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
771 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
773 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
776 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
777 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
778 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
781 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
784 sock_recv_timestamp(msg, sk, skb);
785 sock_recv_drops(msg, sk, skb);
787 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
789 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
792 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
795 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
797 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
799 return err ?: sock_recvmsg_nosec(sock, msg, flags);
801 EXPORT_SYMBOL(sock_recvmsg);
804 * kernel_recvmsg - Receive a message from a socket (kernel space)
805 * @sock: The socket to receive the message from
806 * @msg: Received message
807 * @vec: Input s/g array for message data
808 * @num: Size of input s/g array
809 * @size: Number of bytes to read
810 * @flags: Message flags (MSG_DONTWAIT, etc...)
812 * On return the msg structure contains the scatter/gather array passed in the
813 * vec argument. The array is modified so that it consists of the unfilled
814 * portion of the original array.
816 * The returned value is the total number of bytes received, or an error.
818 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
819 struct kvec *vec, size_t num, size_t size, int flags)
821 mm_segment_t oldfs = get_fs();
824 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
826 result = sock_recvmsg(sock, msg, flags);
830 EXPORT_SYMBOL(kernel_recvmsg);
832 static ssize_t sock_sendpage(struct file *file, struct page *page,
833 int offset, size_t size, loff_t *ppos, int more)
838 sock = file->private_data;
840 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
841 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
844 return kernel_sendpage(sock, page, offset, size, flags);
847 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
848 struct pipe_inode_info *pipe, size_t len,
851 struct socket *sock = file->private_data;
853 if (unlikely(!sock->ops->splice_read))
856 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
859 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
861 struct file *file = iocb->ki_filp;
862 struct socket *sock = file->private_data;
863 struct msghdr msg = {.msg_iter = *to,
867 if (file->f_flags & O_NONBLOCK)
868 msg.msg_flags = MSG_DONTWAIT;
870 if (iocb->ki_pos != 0)
873 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
876 res = sock_recvmsg(sock, &msg, msg.msg_flags);
881 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
883 struct file *file = iocb->ki_filp;
884 struct socket *sock = file->private_data;
885 struct msghdr msg = {.msg_iter = *from,
889 if (iocb->ki_pos != 0)
892 if (file->f_flags & O_NONBLOCK)
893 msg.msg_flags = MSG_DONTWAIT;
895 if (sock->type == SOCK_SEQPACKET)
896 msg.msg_flags |= MSG_EOR;
898 res = sock_sendmsg(sock, &msg);
899 *from = msg.msg_iter;
904 * Atomic setting of ioctl hooks to avoid race
905 * with module unload.
908 static DEFINE_MUTEX(br_ioctl_mutex);
909 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
911 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
913 mutex_lock(&br_ioctl_mutex);
914 br_ioctl_hook = hook;
915 mutex_unlock(&br_ioctl_mutex);
917 EXPORT_SYMBOL(brioctl_set);
919 static DEFINE_MUTEX(vlan_ioctl_mutex);
920 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
922 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
924 mutex_lock(&vlan_ioctl_mutex);
925 vlan_ioctl_hook = hook;
926 mutex_unlock(&vlan_ioctl_mutex);
928 EXPORT_SYMBOL(vlan_ioctl_set);
930 static DEFINE_MUTEX(dlci_ioctl_mutex);
931 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
933 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
935 mutex_lock(&dlci_ioctl_mutex);
936 dlci_ioctl_hook = hook;
937 mutex_unlock(&dlci_ioctl_mutex);
939 EXPORT_SYMBOL(dlci_ioctl_set);
941 static long sock_do_ioctl(struct net *net, struct socket *sock,
942 unsigned int cmd, unsigned long arg)
945 void __user *argp = (void __user *)arg;
947 err = sock->ops->ioctl(sock, cmd, arg);
950 * If this ioctl is unknown try to hand it down
953 if (err == -ENOIOCTLCMD)
954 err = dev_ioctl(net, cmd, argp);
960 * With an ioctl, arg may well be a user mode pointer, but we don't know
961 * what to do with it - that's up to the protocol still.
964 static struct ns_common *get_net_ns(struct ns_common *ns)
966 return &get_net(container_of(ns, struct net, ns))->ns;
969 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
973 void __user *argp = (void __user *)arg;
977 sock = file->private_data;
980 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
981 err = dev_ioctl(net, cmd, argp);
983 #ifdef CONFIG_WEXT_CORE
984 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
985 err = dev_ioctl(net, cmd, argp);
992 if (get_user(pid, (int __user *)argp))
994 f_setown(sock->file, pid, 1);
999 err = put_user(f_getown(sock->file),
1000 (int __user *)argp);
1008 request_module("bridge");
1010 mutex_lock(&br_ioctl_mutex);
1012 err = br_ioctl_hook(net, cmd, argp);
1013 mutex_unlock(&br_ioctl_mutex);
1018 if (!vlan_ioctl_hook)
1019 request_module("8021q");
1021 mutex_lock(&vlan_ioctl_mutex);
1022 if (vlan_ioctl_hook)
1023 err = vlan_ioctl_hook(net, argp);
1024 mutex_unlock(&vlan_ioctl_mutex);
1029 if (!dlci_ioctl_hook)
1030 request_module("dlci");
1032 mutex_lock(&dlci_ioctl_mutex);
1033 if (dlci_ioctl_hook)
1034 err = dlci_ioctl_hook(cmd, argp);
1035 mutex_unlock(&dlci_ioctl_mutex);
1039 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1042 err = open_related_ns(&net->ns, get_net_ns);
1045 err = sock_do_ioctl(net, sock, cmd, arg);
1051 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1054 struct socket *sock = NULL;
1056 err = security_socket_create(family, type, protocol, 1);
1060 sock = sock_alloc();
1067 err = security_socket_post_create(sock, family, type, protocol, 1);
1079 EXPORT_SYMBOL(sock_create_lite);
1081 /* No kernel lock held - perfect */
1082 static unsigned int sock_poll(struct file *file, poll_table *wait)
1084 unsigned int busy_flag = 0;
1085 struct socket *sock;
1088 * We can't return errors to poll, so it's either yes or no.
1090 sock = file->private_data;
1092 if (sk_can_busy_loop(sock->sk)) {
1093 /* this socket can poll_ll so tell the system call */
1094 busy_flag = POLL_BUSY_LOOP;
1096 /* once, only if requested by syscall */
1097 if (wait && (wait->_key & POLL_BUSY_LOOP))
1098 sk_busy_loop(sock->sk, 1);
1101 return busy_flag | sock->ops->poll(file, sock, wait);
1104 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1106 struct socket *sock = file->private_data;
1108 return sock->ops->mmap(file, sock, vma);
1111 static int sock_close(struct inode *inode, struct file *filp)
1113 sock_release(SOCKET_I(inode));
1118 * Update the socket async list
1120 * Fasync_list locking strategy.
1122 * 1. fasync_list is modified only under process context socket lock
1123 * i.e. under semaphore.
1124 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1125 * or under socket lock
1128 static int sock_fasync(int fd, struct file *filp, int on)
1130 struct socket *sock = filp->private_data;
1131 struct sock *sk = sock->sk;
1132 struct socket_wq *wq;
1138 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1139 fasync_helper(fd, filp, on, &wq->fasync_list);
1141 if (!wq->fasync_list)
1142 sock_reset_flag(sk, SOCK_FASYNC);
1144 sock_set_flag(sk, SOCK_FASYNC);
1150 /* This function may be called only under rcu_lock */
1152 int sock_wake_async(struct socket_wq *wq, int how, int band)
1154 if (!wq || !wq->fasync_list)
1158 case SOCK_WAKE_WAITD:
1159 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1162 case SOCK_WAKE_SPACE:
1163 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1168 kill_fasync(&wq->fasync_list, SIGIO, band);
1171 kill_fasync(&wq->fasync_list, SIGURG, band);
1176 EXPORT_SYMBOL(sock_wake_async);
1178 int __sock_create(struct net *net, int family, int type, int protocol,
1179 struct socket **res, int kern)
1182 struct socket *sock;
1183 const struct net_proto_family *pf;
1186 * Check protocol is in range
1188 if (family < 0 || family >= NPROTO)
1189 return -EAFNOSUPPORT;
1190 if (type < 0 || type >= SOCK_MAX)
1195 This uglymoron is moved from INET layer to here to avoid
1196 deadlock in module load.
1198 if (family == PF_INET && type == SOCK_PACKET) {
1199 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1204 err = security_socket_create(family, type, protocol, kern);
1209 * Allocate the socket and allow the family to set things up. if
1210 * the protocol is 0, the family is instructed to select an appropriate
1213 sock = sock_alloc();
1215 net_warn_ratelimited("socket: no more sockets\n");
1216 return -ENFILE; /* Not exactly a match, but its the
1217 closest posix thing */
1222 #ifdef CONFIG_MODULES
1223 /* Attempt to load a protocol module if the find failed.
1225 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1226 * requested real, full-featured networking support upon configuration.
1227 * Otherwise module support will break!
1229 if (rcu_access_pointer(net_families[family]) == NULL)
1230 request_module("net-pf-%d", family);
1234 pf = rcu_dereference(net_families[family]);
1235 err = -EAFNOSUPPORT;
1240 * We will call the ->create function, that possibly is in a loadable
1241 * module, so we have to bump that loadable module refcnt first.
1243 if (!try_module_get(pf->owner))
1246 /* Now protected by module ref count */
1249 err = pf->create(net, sock, protocol, kern);
1251 goto out_module_put;
1254 * Now to bump the refcnt of the [loadable] module that owns this
1255 * socket at sock_release time we decrement its refcnt.
1257 if (!try_module_get(sock->ops->owner))
1258 goto out_module_busy;
1261 * Now that we're done with the ->create function, the [loadable]
1262 * module can have its refcnt decremented
1264 module_put(pf->owner);
1265 err = security_socket_post_create(sock, family, type, protocol, kern);
1267 goto out_sock_release;
1273 err = -EAFNOSUPPORT;
1276 module_put(pf->owner);
1283 goto out_sock_release;
1285 EXPORT_SYMBOL(__sock_create);
1287 int sock_create(int family, int type, int protocol, struct socket **res)
1289 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1291 EXPORT_SYMBOL(sock_create);
1293 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1295 return __sock_create(net, family, type, protocol, res, 1);
1297 EXPORT_SYMBOL(sock_create_kern);
1299 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1302 struct socket *sock;
1305 /* Check the SOCK_* constants for consistency. */
1306 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1307 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1308 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1309 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1311 flags = type & ~SOCK_TYPE_MASK;
1312 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1314 type &= SOCK_TYPE_MASK;
1316 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1317 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1319 retval = sock_create(family, type, protocol, &sock);
1323 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1328 /* It may be already another descriptor 8) Not kernel problem. */
1337 * Create a pair of connected sockets.
1340 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1341 int __user *, usockvec)
1343 struct socket *sock1, *sock2;
1345 struct file *newfile1, *newfile2;
1348 flags = type & ~SOCK_TYPE_MASK;
1349 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1351 type &= SOCK_TYPE_MASK;
1353 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1354 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1357 * Obtain the first socket and check if the underlying protocol
1358 * supports the socketpair call.
1361 err = sock_create(family, type, protocol, &sock1);
1365 err = sock_create(family, type, protocol, &sock2);
1369 err = sock1->ops->socketpair(sock1, sock2);
1371 goto out_release_both;
1373 fd1 = get_unused_fd_flags(flags);
1374 if (unlikely(fd1 < 0)) {
1376 goto out_release_both;
1379 fd2 = get_unused_fd_flags(flags);
1380 if (unlikely(fd2 < 0)) {
1382 goto out_put_unused_1;
1385 newfile1 = sock_alloc_file(sock1, flags, NULL);
1386 if (IS_ERR(newfile1)) {
1387 err = PTR_ERR(newfile1);
1388 goto out_put_unused_both;
1391 newfile2 = sock_alloc_file(sock2, flags, NULL);
1392 if (IS_ERR(newfile2)) {
1393 err = PTR_ERR(newfile2);
1397 err = put_user(fd1, &usockvec[0]);
1401 err = put_user(fd2, &usockvec[1]);
1405 audit_fd_pair(fd1, fd2);
1407 fd_install(fd1, newfile1);
1408 fd_install(fd2, newfile2);
1409 /* fd1 and fd2 may be already another descriptors.
1410 * Not kernel problem.
1426 sock_release(sock2);
1429 out_put_unused_both:
1434 sock_release(sock2);
1436 sock_release(sock1);
1442 * Bind a name to a socket. Nothing much to do here since it's
1443 * the protocol's responsibility to handle the local address.
1445 * We move the socket address to kernel space before we call
1446 * the protocol layer (having also checked the address is ok).
1449 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1451 struct socket *sock;
1452 struct sockaddr_storage address;
1453 int err, fput_needed;
1455 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1457 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1459 err = security_socket_bind(sock,
1460 (struct sockaddr *)&address,
1463 err = sock->ops->bind(sock,
1467 fput_light(sock->file, fput_needed);
1473 * Perform a listen. Basically, we allow the protocol to do anything
1474 * necessary for a listen, and if that works, we mark the socket as
1475 * ready for listening.
1478 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1480 struct socket *sock;
1481 int err, fput_needed;
1484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1486 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1487 if ((unsigned int)backlog > somaxconn)
1488 backlog = somaxconn;
1490 err = security_socket_listen(sock, backlog);
1492 err = sock->ops->listen(sock, backlog);
1494 fput_light(sock->file, fput_needed);
1500 * For accept, we attempt to create a new socket, set up the link
1501 * with the client, wake up the client, then return the new
1502 * connected fd. We collect the address of the connector in kernel
1503 * space and move it to user at the very end. This is unclean because
1504 * we open the socket then return an error.
1506 * 1003.1g adds the ability to recvmsg() to query connection pending
1507 * status to recvmsg. We need to add that support in a way thats
1508 * clean when we restucture accept also.
1511 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1512 int __user *, upeer_addrlen, int, flags)
1514 struct socket *sock, *newsock;
1515 struct file *newfile;
1516 int err, len, newfd, fput_needed;
1517 struct sockaddr_storage address;
1519 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1522 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1523 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1525 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1530 newsock = sock_alloc();
1534 newsock->type = sock->type;
1535 newsock->ops = sock->ops;
1538 * We don't need try_module_get here, as the listening socket (sock)
1539 * has the protocol module (sock->ops->owner) held.
1541 __module_get(newsock->ops->owner);
1543 newfd = get_unused_fd_flags(flags);
1544 if (unlikely(newfd < 0)) {
1546 sock_release(newsock);
1549 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1550 if (IS_ERR(newfile)) {
1551 err = PTR_ERR(newfile);
1552 put_unused_fd(newfd);
1553 sock_release(newsock);
1557 err = security_socket_accept(sock, newsock);
1561 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1565 if (upeer_sockaddr) {
1566 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1568 err = -ECONNABORTED;
1571 err = move_addr_to_user(&address,
1572 len, upeer_sockaddr, upeer_addrlen);
1577 /* File flags are not inherited via accept() unlike another OSes. */
1579 fd_install(newfd, newfile);
1583 fput_light(sock->file, fput_needed);
1588 put_unused_fd(newfd);
1592 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1593 int __user *, upeer_addrlen)
1595 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1599 * Attempt to connect to a socket with the server address. The address
1600 * is in user space so we verify it is OK and move it to kernel space.
1602 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1605 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1606 * other SEQPACKET protocols that take time to connect() as it doesn't
1607 * include the -EINPROGRESS status for such sockets.
1610 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1613 struct socket *sock;
1614 struct sockaddr_storage address;
1615 int err, fput_needed;
1617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1620 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1625 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1629 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1630 sock->file->f_flags);
1632 fput_light(sock->file, fput_needed);
1638 * Get the local address ('name') of a socket object. Move the obtained
1639 * name to user space.
1642 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1643 int __user *, usockaddr_len)
1645 struct socket *sock;
1646 struct sockaddr_storage address;
1647 int len, err, fput_needed;
1649 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1653 err = security_socket_getsockname(sock);
1657 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1660 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1663 fput_light(sock->file, fput_needed);
1669 * Get the remote address ('name') of a socket object. Move the obtained
1670 * name to user space.
1673 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1674 int __user *, usockaddr_len)
1676 struct socket *sock;
1677 struct sockaddr_storage address;
1678 int len, err, fput_needed;
1680 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1682 err = security_socket_getpeername(sock);
1684 fput_light(sock->file, fput_needed);
1689 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1692 err = move_addr_to_user(&address, len, usockaddr,
1694 fput_light(sock->file, fput_needed);
1700 * Send a datagram to a given address. We move the address into kernel
1701 * space and check the user space data area is readable before invoking
1705 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1706 unsigned int, flags, struct sockaddr __user *, addr,
1709 struct socket *sock;
1710 struct sockaddr_storage address;
1716 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1719 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1723 msg.msg_name = NULL;
1724 msg.msg_control = NULL;
1725 msg.msg_controllen = 0;
1726 msg.msg_namelen = 0;
1728 err = move_addr_to_kernel(addr, addr_len, &address);
1731 msg.msg_name = (struct sockaddr *)&address;
1732 msg.msg_namelen = addr_len;
1734 if (sock->file->f_flags & O_NONBLOCK)
1735 flags |= MSG_DONTWAIT;
1736 msg.msg_flags = flags;
1737 err = sock_sendmsg(sock, &msg);
1740 fput_light(sock->file, fput_needed);
1746 * Send a datagram down a socket.
1749 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1750 unsigned int, flags)
1752 return sys_sendto(fd, buff, len, flags, NULL, 0);
1756 * Receive a frame from the socket and optionally record the address of the
1757 * sender. We verify the buffers are writable and if needed move the
1758 * sender address from kernel to user space.
1761 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1762 unsigned int, flags, struct sockaddr __user *, addr,
1763 int __user *, addr_len)
1765 struct socket *sock;
1768 struct sockaddr_storage address;
1772 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1775 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1779 msg.msg_control = NULL;
1780 msg.msg_controllen = 0;
1781 /* Save some cycles and don't copy the address if not needed */
1782 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1783 /* We assume all kernel code knows the size of sockaddr_storage */
1784 msg.msg_namelen = 0;
1785 msg.msg_iocb = NULL;
1787 if (sock->file->f_flags & O_NONBLOCK)
1788 flags |= MSG_DONTWAIT;
1789 err = sock_recvmsg(sock, &msg, flags);
1791 if (err >= 0 && addr != NULL) {
1792 err2 = move_addr_to_user(&address,
1793 msg.msg_namelen, addr, addr_len);
1798 fput_light(sock->file, fput_needed);
1804 * Receive a datagram from a socket.
1807 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1808 unsigned int, flags)
1810 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1814 * Set a socket option. Because we don't know the option lengths we have
1815 * to pass the user mode parameter for the protocols to sort out.
1818 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1819 char __user *, optval, int, optlen)
1821 int err, fput_needed;
1822 struct socket *sock;
1827 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1829 err = security_socket_setsockopt(sock, level, optname);
1833 if (level == SOL_SOCKET)
1835 sock_setsockopt(sock, level, optname, optval,
1839 sock->ops->setsockopt(sock, level, optname, optval,
1842 fput_light(sock->file, fput_needed);
1848 * Get a socket option. Because we don't know the option lengths we have
1849 * to pass a user mode parameter for the protocols to sort out.
1852 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1853 char __user *, optval, int __user *, optlen)
1855 int err, fput_needed;
1856 struct socket *sock;
1858 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1860 err = security_socket_getsockopt(sock, level, optname);
1864 if (level == SOL_SOCKET)
1866 sock_getsockopt(sock, level, optname, optval,
1870 sock->ops->getsockopt(sock, level, optname, optval,
1873 fput_light(sock->file, fput_needed);
1879 * Shutdown a socket.
1882 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1884 int err, fput_needed;
1885 struct socket *sock;
1887 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889 err = security_socket_shutdown(sock, how);
1891 err = sock->ops->shutdown(sock, how);
1892 fput_light(sock->file, fput_needed);
1897 /* A couple of helpful macros for getting the address of the 32/64 bit
1898 * fields which are the same type (int / unsigned) on our platforms.
1900 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1901 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1902 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1904 struct used_address {
1905 struct sockaddr_storage name;
1906 unsigned int name_len;
1909 static int copy_msghdr_from_user(struct msghdr *kmsg,
1910 struct user_msghdr __user *umsg,
1911 struct sockaddr __user **save_addr,
1914 struct sockaddr __user *uaddr;
1915 struct iovec __user *uiov;
1919 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1920 __get_user(uaddr, &umsg->msg_name) ||
1921 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1922 __get_user(uiov, &umsg->msg_iov) ||
1923 __get_user(nr_segs, &umsg->msg_iovlen) ||
1924 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1925 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1926 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1930 kmsg->msg_namelen = 0;
1932 if (kmsg->msg_namelen < 0)
1935 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1936 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1941 if (uaddr && kmsg->msg_namelen) {
1943 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1949 kmsg->msg_name = NULL;
1950 kmsg->msg_namelen = 0;
1953 if (nr_segs > UIO_MAXIOV)
1956 kmsg->msg_iocb = NULL;
1958 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1959 UIO_FASTIOV, iov, &kmsg->msg_iter);
1962 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1963 struct msghdr *msg_sys, unsigned int flags,
1964 struct used_address *used_address,
1965 unsigned int allowed_msghdr_flags)
1967 struct compat_msghdr __user *msg_compat =
1968 (struct compat_msghdr __user *)msg;
1969 struct sockaddr_storage address;
1970 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1971 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1972 __aligned(sizeof(__kernel_size_t));
1973 /* 20 is size of ipv6_pktinfo */
1974 unsigned char *ctl_buf = ctl;
1978 msg_sys->msg_name = &address;
1980 if (MSG_CMSG_COMPAT & flags)
1981 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1983 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1989 if (msg_sys->msg_controllen > INT_MAX)
1991 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1992 ctl_len = msg_sys->msg_controllen;
1993 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1995 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1999 ctl_buf = msg_sys->msg_control;
2000 ctl_len = msg_sys->msg_controllen;
2001 } else if (ctl_len) {
2002 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2003 CMSG_ALIGN(sizeof(struct cmsghdr)));
2004 if (ctl_len > sizeof(ctl)) {
2005 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2006 if (ctl_buf == NULL)
2011 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2012 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2013 * checking falls down on this.
2015 if (copy_from_user(ctl_buf,
2016 (void __user __force *)msg_sys->msg_control,
2019 msg_sys->msg_control = ctl_buf;
2021 msg_sys->msg_flags = flags;
2023 if (sock->file->f_flags & O_NONBLOCK)
2024 msg_sys->msg_flags |= MSG_DONTWAIT;
2026 * If this is sendmmsg() and current destination address is same as
2027 * previously succeeded address, omit asking LSM's decision.
2028 * used_address->name_len is initialized to UINT_MAX so that the first
2029 * destination address never matches.
2031 if (used_address && msg_sys->msg_name &&
2032 used_address->name_len == msg_sys->msg_namelen &&
2033 !memcmp(&used_address->name, msg_sys->msg_name,
2034 used_address->name_len)) {
2035 err = sock_sendmsg_nosec(sock, msg_sys);
2038 err = sock_sendmsg(sock, msg_sys);
2040 * If this is sendmmsg() and sending to current destination address was
2041 * successful, remember it.
2043 if (used_address && err >= 0) {
2044 used_address->name_len = msg_sys->msg_namelen;
2045 if (msg_sys->msg_name)
2046 memcpy(&used_address->name, msg_sys->msg_name,
2047 used_address->name_len);
2052 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2059 * BSD sendmsg interface
2062 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2064 int fput_needed, err;
2065 struct msghdr msg_sys;
2066 struct socket *sock;
2068 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2072 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2074 fput_light(sock->file, fput_needed);
2079 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2081 if (flags & MSG_CMSG_COMPAT)
2083 return __sys_sendmsg(fd, msg, flags);
2087 * Linux sendmmsg interface
2090 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2093 int fput_needed, err, datagrams;
2094 struct socket *sock;
2095 struct mmsghdr __user *entry;
2096 struct compat_mmsghdr __user *compat_entry;
2097 struct msghdr msg_sys;
2098 struct used_address used_address;
2099 unsigned int oflags = flags;
2101 if (vlen > UIO_MAXIOV)
2106 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2110 used_address.name_len = UINT_MAX;
2112 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2116 while (datagrams < vlen) {
2117 if (datagrams == vlen - 1)
2120 if (MSG_CMSG_COMPAT & flags) {
2121 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2122 &msg_sys, flags, &used_address, MSG_EOR);
2125 err = __put_user(err, &compat_entry->msg_len);
2128 err = ___sys_sendmsg(sock,
2129 (struct user_msghdr __user *)entry,
2130 &msg_sys, flags, &used_address, MSG_EOR);
2133 err = put_user(err, &entry->msg_len);
2140 if (msg_data_left(&msg_sys))
2145 fput_light(sock->file, fput_needed);
2147 /* We only return an error if no datagrams were able to be sent */
2154 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2155 unsigned int, vlen, unsigned int, flags)
2157 if (flags & MSG_CMSG_COMPAT)
2159 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2162 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2163 struct msghdr *msg_sys, unsigned int flags, int nosec)
2165 struct compat_msghdr __user *msg_compat =
2166 (struct compat_msghdr __user *)msg;
2167 struct iovec iovstack[UIO_FASTIOV];
2168 struct iovec *iov = iovstack;
2169 unsigned long cmsg_ptr;
2173 /* kernel mode address */
2174 struct sockaddr_storage addr;
2176 /* user mode address pointers */
2177 struct sockaddr __user *uaddr;
2178 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2180 msg_sys->msg_name = &addr;
2182 if (MSG_CMSG_COMPAT & flags)
2183 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2185 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2189 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2190 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2192 /* We assume all kernel code knows the size of sockaddr_storage */
2193 msg_sys->msg_namelen = 0;
2195 if (sock->file->f_flags & O_NONBLOCK)
2196 flags |= MSG_DONTWAIT;
2197 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2202 if (uaddr != NULL) {
2203 err = move_addr_to_user(&addr,
2204 msg_sys->msg_namelen, uaddr,
2209 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2213 if (MSG_CMSG_COMPAT & flags)
2214 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2215 &msg_compat->msg_controllen);
2217 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2218 &msg->msg_controllen);
2229 * BSD recvmsg interface
2232 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2234 int fput_needed, err;
2235 struct msghdr msg_sys;
2236 struct socket *sock;
2238 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2242 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2244 fput_light(sock->file, fput_needed);
2249 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2250 unsigned int, flags)
2252 if (flags & MSG_CMSG_COMPAT)
2254 return __sys_recvmsg(fd, msg, flags);
2258 * Linux recvmmsg interface
2261 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2262 unsigned int flags, struct timespec *timeout)
2264 int fput_needed, err, datagrams;
2265 struct socket *sock;
2266 struct mmsghdr __user *entry;
2267 struct compat_mmsghdr __user *compat_entry;
2268 struct msghdr msg_sys;
2269 struct timespec64 end_time;
2270 struct timespec64 timeout64;
2273 poll_select_set_timeout(&end_time, timeout->tv_sec,
2279 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2283 err = sock_error(sock->sk);
2290 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2292 while (datagrams < vlen) {
2294 * No need to ask LSM for more than the first datagram.
2296 if (MSG_CMSG_COMPAT & flags) {
2297 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2298 &msg_sys, flags & ~MSG_WAITFORONE,
2302 err = __put_user(err, &compat_entry->msg_len);
2305 err = ___sys_recvmsg(sock,
2306 (struct user_msghdr __user *)entry,
2307 &msg_sys, flags & ~MSG_WAITFORONE,
2311 err = put_user(err, &entry->msg_len);
2319 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2320 if (flags & MSG_WAITFORONE)
2321 flags |= MSG_DONTWAIT;
2324 ktime_get_ts64(&timeout64);
2325 *timeout = timespec64_to_timespec(
2326 timespec64_sub(end_time, timeout64));
2327 if (timeout->tv_sec < 0) {
2328 timeout->tv_sec = timeout->tv_nsec = 0;
2332 /* Timeout, return less than vlen datagrams */
2333 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2337 /* Out of band data, return right away */
2338 if (msg_sys.msg_flags & MSG_OOB)
2346 if (datagrams == 0) {
2352 * We may return less entries than requested (vlen) if the
2353 * sock is non block and there aren't enough datagrams...
2355 if (err != -EAGAIN) {
2357 * ... or if recvmsg returns an error after we
2358 * received some datagrams, where we record the
2359 * error to return on the next call or if the
2360 * app asks about it using getsockopt(SO_ERROR).
2362 sock->sk->sk_err = -err;
2365 fput_light(sock->file, fput_needed);
2370 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2371 unsigned int, vlen, unsigned int, flags,
2372 struct timespec __user *, timeout)
2375 struct timespec timeout_sys;
2377 if (flags & MSG_CMSG_COMPAT)
2381 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2383 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2386 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2388 if (datagrams > 0 &&
2389 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2390 datagrams = -EFAULT;
2395 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2396 /* Argument list sizes for sys_socketcall */
2397 #define AL(x) ((x) * sizeof(unsigned long))
2398 static const unsigned char nargs[21] = {
2399 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2400 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2401 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2408 * System call vectors.
2410 * Argument checking cleaned up. Saved 20% in size.
2411 * This function doesn't need to set the kernel lock because
2412 * it is set by the callees.
2415 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2417 unsigned long a[AUDITSC_ARGS];
2418 unsigned long a0, a1;
2422 if (call < 1 || call > SYS_SENDMMSG)
2426 if (len > sizeof(a))
2429 /* copy_from_user should be SMP safe. */
2430 if (copy_from_user(a, args, len))
2433 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2442 err = sys_socket(a0, a1, a[2]);
2445 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2448 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2451 err = sys_listen(a0, a1);
2454 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2455 (int __user *)a[2], 0);
2457 case SYS_GETSOCKNAME:
2459 sys_getsockname(a0, (struct sockaddr __user *)a1,
2460 (int __user *)a[2]);
2462 case SYS_GETPEERNAME:
2464 sys_getpeername(a0, (struct sockaddr __user *)a1,
2465 (int __user *)a[2]);
2467 case SYS_SOCKETPAIR:
2468 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2471 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2474 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2475 (struct sockaddr __user *)a[4], a[5]);
2478 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2481 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2482 (struct sockaddr __user *)a[4],
2483 (int __user *)a[5]);
2486 err = sys_shutdown(a0, a1);
2488 case SYS_SETSOCKOPT:
2489 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2491 case SYS_GETSOCKOPT:
2493 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2494 (int __user *)a[4]);
2497 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2500 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2503 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2506 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2507 (struct timespec __user *)a[4]);
2510 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2511 (int __user *)a[2], a[3]);
2520 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2523 * sock_register - add a socket protocol handler
2524 * @ops: description of protocol
2526 * This function is called by a protocol handler that wants to
2527 * advertise its address family, and have it linked into the
2528 * socket interface. The value ops->family corresponds to the
2529 * socket system call protocol family.
2531 int sock_register(const struct net_proto_family *ops)
2535 if (ops->family >= NPROTO) {
2536 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2540 spin_lock(&net_family_lock);
2541 if (rcu_dereference_protected(net_families[ops->family],
2542 lockdep_is_held(&net_family_lock)))
2545 rcu_assign_pointer(net_families[ops->family], ops);
2548 spin_unlock(&net_family_lock);
2550 pr_info("NET: Registered protocol family %d\n", ops->family);
2553 EXPORT_SYMBOL(sock_register);
2556 * sock_unregister - remove a protocol handler
2557 * @family: protocol family to remove
2559 * This function is called by a protocol handler that wants to
2560 * remove its address family, and have it unlinked from the
2561 * new socket creation.
2563 * If protocol handler is a module, then it can use module reference
2564 * counts to protect against new references. If protocol handler is not
2565 * a module then it needs to provide its own protection in
2566 * the ops->create routine.
2568 void sock_unregister(int family)
2570 BUG_ON(family < 0 || family >= NPROTO);
2572 spin_lock(&net_family_lock);
2573 RCU_INIT_POINTER(net_families[family], NULL);
2574 spin_unlock(&net_family_lock);
2578 pr_info("NET: Unregistered protocol family %d\n", family);
2580 EXPORT_SYMBOL(sock_unregister);
2582 static int __init sock_init(void)
2586 * Initialize the network sysctl infrastructure.
2588 err = net_sysctl_init();
2593 * Initialize skbuff SLAB cache
2598 * Initialize the protocols module.
2603 err = register_filesystem(&sock_fs_type);
2606 sock_mnt = kern_mount(&sock_fs_type);
2607 if (IS_ERR(sock_mnt)) {
2608 err = PTR_ERR(sock_mnt);
2612 /* The real protocol initialization is performed in later initcalls.
2615 #ifdef CONFIG_NETFILTER
2616 err = netfilter_init();
2621 ptp_classifier_init();
2627 unregister_filesystem(&sock_fs_type);
2632 core_initcall(sock_init); /* early initcall */
2634 #ifdef CONFIG_PROC_FS
2635 void socket_seq_show(struct seq_file *seq)
2640 for_each_possible_cpu(cpu)
2641 counter += per_cpu(sockets_in_use, cpu);
2643 /* It can be negative, by the way. 8) */
2647 seq_printf(seq, "sockets: used %d\n", counter);
2649 #endif /* CONFIG_PROC_FS */
2651 #ifdef CONFIG_COMPAT
2652 static int do_siocgstamp(struct net *net, struct socket *sock,
2653 unsigned int cmd, void __user *up)
2655 mm_segment_t old_fs = get_fs();
2660 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2663 err = compat_put_timeval(&ktv, up);
2668 static int do_siocgstampns(struct net *net, struct socket *sock,
2669 unsigned int cmd, void __user *up)
2671 mm_segment_t old_fs = get_fs();
2672 struct timespec kts;
2676 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2679 err = compat_put_timespec(&kts, up);
2684 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2686 struct ifreq __user *uifr;
2689 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2690 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2693 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2697 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2703 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2705 struct compat_ifconf ifc32;
2707 struct ifconf __user *uifc;
2708 struct compat_ifreq __user *ifr32;
2709 struct ifreq __user *ifr;
2713 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2716 memset(&ifc, 0, sizeof(ifc));
2717 if (ifc32.ifcbuf == 0) {
2721 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2723 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2724 sizeof(struct ifreq);
2725 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2727 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2728 ifr32 = compat_ptr(ifc32.ifcbuf);
2729 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2730 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2736 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2739 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2743 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2747 ifr32 = compat_ptr(ifc32.ifcbuf);
2749 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2750 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2751 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2757 if (ifc32.ifcbuf == 0) {
2758 /* Translate from 64-bit structure multiple to
2762 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2767 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2773 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2775 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2776 bool convert_in = false, convert_out = false;
2777 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2778 struct ethtool_rxnfc __user *rxnfc;
2779 struct ifreq __user *ifr;
2780 u32 rule_cnt = 0, actual_rule_cnt;
2785 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2788 compat_rxnfc = compat_ptr(data);
2790 if (get_user(ethcmd, &compat_rxnfc->cmd))
2793 /* Most ethtool structures are defined without padding.
2794 * Unfortunately struct ethtool_rxnfc is an exception.
2799 case ETHTOOL_GRXCLSRLALL:
2800 /* Buffer size is variable */
2801 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2803 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2805 buf_size += rule_cnt * sizeof(u32);
2807 case ETHTOOL_GRXRINGS:
2808 case ETHTOOL_GRXCLSRLCNT:
2809 case ETHTOOL_GRXCLSRULE:
2810 case ETHTOOL_SRXCLSRLINS:
2813 case ETHTOOL_SRXCLSRLDEL:
2814 buf_size += sizeof(struct ethtool_rxnfc);
2819 ifr = compat_alloc_user_space(buf_size);
2820 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2822 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2825 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2826 &ifr->ifr_ifru.ifru_data))
2830 /* We expect there to be holes between fs.m_ext and
2831 * fs.ring_cookie and at the end of fs, but nowhere else.
2833 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2834 sizeof(compat_rxnfc->fs.m_ext) !=
2835 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2836 sizeof(rxnfc->fs.m_ext));
2838 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2839 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2840 offsetof(struct ethtool_rxnfc, fs.location) -
2841 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2843 if (copy_in_user(rxnfc, compat_rxnfc,
2844 (void __user *)(&rxnfc->fs.m_ext + 1) -
2845 (void __user *)rxnfc) ||
2846 copy_in_user(&rxnfc->fs.ring_cookie,
2847 &compat_rxnfc->fs.ring_cookie,
2848 (void __user *)(&rxnfc->fs.location + 1) -
2849 (void __user *)&rxnfc->fs.ring_cookie) ||
2850 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2851 sizeof(rxnfc->rule_cnt)))
2855 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2860 if (copy_in_user(compat_rxnfc, rxnfc,
2861 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2862 (const void __user *)rxnfc) ||
2863 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2864 &rxnfc->fs.ring_cookie,
2865 (const void __user *)(&rxnfc->fs.location + 1) -
2866 (const void __user *)&rxnfc->fs.ring_cookie) ||
2867 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2868 sizeof(rxnfc->rule_cnt)))
2871 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2872 /* As an optimisation, we only copy the actual
2873 * number of rules that the underlying
2874 * function returned. Since Mallory might
2875 * change the rule count in user memory, we
2876 * check that it is less than the rule count
2877 * originally given (as the user buffer size),
2878 * which has been range-checked.
2880 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2882 if (actual_rule_cnt < rule_cnt)
2883 rule_cnt = actual_rule_cnt;
2884 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2885 &rxnfc->rule_locs[0],
2886 rule_cnt * sizeof(u32)))
2894 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2897 compat_uptr_t uptr32;
2898 struct ifreq __user *uifr;
2900 uifr = compat_alloc_user_space(sizeof(*uifr));
2901 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2904 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2907 uptr = compat_ptr(uptr32);
2909 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2912 return dev_ioctl(net, SIOCWANDEV, uifr);
2915 static int bond_ioctl(struct net *net, unsigned int cmd,
2916 struct compat_ifreq __user *ifr32)
2919 mm_segment_t old_fs;
2923 case SIOCBONDENSLAVE:
2924 case SIOCBONDRELEASE:
2925 case SIOCBONDSETHWADDR:
2926 case SIOCBONDCHANGEACTIVE:
2927 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2932 err = dev_ioctl(net, cmd,
2933 (struct ifreq __user __force *) &kifr);
2938 return -ENOIOCTLCMD;
2942 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2943 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2944 struct compat_ifreq __user *u_ifreq32)
2946 struct ifreq __user *u_ifreq64;
2947 char tmp_buf[IFNAMSIZ];
2948 void __user *data64;
2951 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2954 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2956 data64 = compat_ptr(data32);
2958 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2960 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2963 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2966 return dev_ioctl(net, cmd, u_ifreq64);
2969 static int dev_ifsioc(struct net *net, struct socket *sock,
2970 unsigned int cmd, struct compat_ifreq __user *uifr32)
2972 struct ifreq __user *uifr;
2975 uifr = compat_alloc_user_space(sizeof(*uifr));
2976 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2979 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2990 case SIOCGIFBRDADDR:
2991 case SIOCGIFDSTADDR:
2992 case SIOCGIFNETMASK:
2997 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3005 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3006 struct compat_ifreq __user *uifr32)
3009 struct compat_ifmap __user *uifmap32;
3010 mm_segment_t old_fs;
3013 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3014 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3015 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3016 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3017 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3018 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3019 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3020 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3026 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3029 if (cmd == SIOCGIFMAP && !err) {
3030 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3031 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3032 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3033 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3034 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3035 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3036 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3045 struct sockaddr rt_dst; /* target address */
3046 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3047 struct sockaddr rt_genmask; /* target network mask (IP) */
3048 unsigned short rt_flags;
3051 unsigned char rt_tos;
3052 unsigned char rt_class;
3054 short rt_metric; /* +1 for binary compatibility! */
3055 /* char * */ u32 rt_dev; /* forcing the device at add */
3056 u32 rt_mtu; /* per route MTU/Window */
3057 u32 rt_window; /* Window clamping */
3058 unsigned short rt_irtt; /* Initial RTT */
3061 struct in6_rtmsg32 {
3062 struct in6_addr rtmsg_dst;
3063 struct in6_addr rtmsg_src;
3064 struct in6_addr rtmsg_gateway;
3074 static int routing_ioctl(struct net *net, struct socket *sock,
3075 unsigned int cmd, void __user *argp)
3079 struct in6_rtmsg r6;
3083 mm_segment_t old_fs = get_fs();
3085 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3086 struct in6_rtmsg32 __user *ur6 = argp;
3087 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3088 3 * sizeof(struct in6_addr));
3089 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3090 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3091 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3092 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3093 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3094 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3095 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3099 struct rtentry32 __user *ur4 = argp;
3100 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3101 3 * sizeof(struct sockaddr));
3102 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3103 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3104 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3105 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3106 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3107 ret |= get_user(rtdev, &(ur4->rt_dev));
3109 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3110 r4.rt_dev = (char __user __force *)devname;
3124 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3131 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3132 * for some operations; this forces use of the newer bridge-utils that
3133 * use compatible ioctls
3135 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3139 if (get_user(tmp, argp))
3141 if (tmp == BRCTL_GET_VERSION)
3142 return BRCTL_VERSION + 1;
3146 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3147 unsigned int cmd, unsigned long arg)
3149 void __user *argp = compat_ptr(arg);
3150 struct sock *sk = sock->sk;
3151 struct net *net = sock_net(sk);
3153 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3154 return compat_ifr_data_ioctl(net, cmd, argp);
3159 return old_bridge_ioctl(argp);
3161 return dev_ifname32(net, argp);
3163 return dev_ifconf(net, argp);
3165 return ethtool_ioctl(net, argp);
3167 return compat_siocwandev(net, argp);
3170 return compat_sioc_ifmap(net, cmd, argp);
3171 case SIOCBONDENSLAVE:
3172 case SIOCBONDRELEASE:
3173 case SIOCBONDSETHWADDR:
3174 case SIOCBONDCHANGEACTIVE:
3175 return bond_ioctl(net, cmd, argp);
3178 return routing_ioctl(net, sock, cmd, argp);
3180 return do_siocgstamp(net, sock, cmd, argp);
3182 return do_siocgstampns(net, sock, cmd, argp);
3183 case SIOCBONDSLAVEINFOQUERY:
3184 case SIOCBONDINFOQUERY:
3187 return compat_ifr_data_ioctl(net, cmd, argp);
3200 return sock_ioctl(file, cmd, arg);
3217 case SIOCSIFHWBROADCAST:
3219 case SIOCGIFBRDADDR:
3220 case SIOCSIFBRDADDR:
3221 case SIOCGIFDSTADDR:
3222 case SIOCSIFDSTADDR:
3223 case SIOCGIFNETMASK:
3224 case SIOCSIFNETMASK:
3235 return dev_ifsioc(net, sock, cmd, argp);
3241 return sock_do_ioctl(net, sock, cmd, arg);
3244 return -ENOIOCTLCMD;
3247 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3250 struct socket *sock = file->private_data;
3251 int ret = -ENOIOCTLCMD;
3258 if (sock->ops->compat_ioctl)
3259 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3261 if (ret == -ENOIOCTLCMD &&
3262 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3263 ret = compat_wext_handle_ioctl(net, cmd, arg);
3265 if (ret == -ENOIOCTLCMD)
3266 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3272 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3274 return sock->ops->bind(sock, addr, addrlen);
3276 EXPORT_SYMBOL(kernel_bind);
3278 int kernel_listen(struct socket *sock, int backlog)
3280 return sock->ops->listen(sock, backlog);
3282 EXPORT_SYMBOL(kernel_listen);
3284 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3286 struct sock *sk = sock->sk;
3289 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3294 err = sock->ops->accept(sock, *newsock, flags, true);
3296 sock_release(*newsock);
3301 (*newsock)->ops = sock->ops;
3302 __module_get((*newsock)->ops->owner);
3307 EXPORT_SYMBOL(kernel_accept);
3309 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3312 return sock->ops->connect(sock, addr, addrlen, flags);
3314 EXPORT_SYMBOL(kernel_connect);
3316 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3319 return sock->ops->getname(sock, addr, addrlen, 0);
3321 EXPORT_SYMBOL(kernel_getsockname);
3323 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3326 return sock->ops->getname(sock, addr, addrlen, 1);
3328 EXPORT_SYMBOL(kernel_getpeername);
3330 int kernel_getsockopt(struct socket *sock, int level, int optname,
3331 char *optval, int *optlen)
3333 mm_segment_t oldfs = get_fs();
3334 char __user *uoptval;
3335 int __user *uoptlen;
3338 uoptval = (char __user __force *) optval;
3339 uoptlen = (int __user __force *) optlen;
3342 if (level == SOL_SOCKET)
3343 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3345 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3350 EXPORT_SYMBOL(kernel_getsockopt);
3352 int kernel_setsockopt(struct socket *sock, int level, int optname,
3353 char *optval, unsigned int optlen)
3355 mm_segment_t oldfs = get_fs();
3356 char __user *uoptval;
3359 uoptval = (char __user __force *) optval;
3362 if (level == SOL_SOCKET)
3363 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3365 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3370 EXPORT_SYMBOL(kernel_setsockopt);
3372 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3373 size_t size, int flags)
3375 if (sock->ops->sendpage)
3376 return sock->ops->sendpage(sock, page, offset, size, flags);
3378 return sock_no_sendpage(sock, page, offset, size, flags);
3380 EXPORT_SYMBOL(kernel_sendpage);
3382 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3384 mm_segment_t oldfs = get_fs();
3388 err = sock->ops->ioctl(sock, cmd, arg);
3393 EXPORT_SYMBOL(kernel_sock_ioctl);
3395 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3397 return sock->ops->shutdown(sock, how);
3399 EXPORT_SYMBOL(kernel_sock_shutdown);
3401 /* This routine returns the IP overhead imposed by a socket i.e.
3402 * the length of the underlying IP header, depending on whether
3403 * this is an IPv4 or IPv6 socket and the length from IP options turned
3404 * on at the socket. Assumes that the caller has a lock on the socket.
3406 u32 kernel_sock_ip_overhead(struct sock *sk)
3408 struct inet_sock *inet;
3409 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 owned_by_user = sock_owned_by_user(sk);
3421 switch (sk->sk_family) {
3424 overhead += sizeof(struct iphdr);
3425 opt = rcu_dereference_protected(inet->inet_opt,
3428 overhead += opt->opt.optlen;
3430 #if IS_ENABLED(CONFIG_IPV6)
3433 overhead += sizeof(struct ipv6hdr);
3435 optv6 = rcu_dereference_protected(np->opt,
3438 overhead += (optv6->opt_flen + optv6->opt_nflen);
3440 #endif /* IS_ENABLED(CONFIG_IPV6) */
3441 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3445 EXPORT_SYMBOL(kernel_sock_ip_overhead);