2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92 #include <linux/nospec.h>
94 #include <linux/uaccess.h>
95 #include <asm/unistd.h>
97 #include <net/compat.h>
99 #include <net/cls_cgroup.h>
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 static int sock_close(struct inode *inode, struct file *file);
121 static __poll_t sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
140 static const struct file_operations socket_file_ops = {
141 .owner = THIS_MODULE,
143 .read_iter = sock_read_iter,
144 .write_iter = sock_write_iter,
146 .unlocked_ioctl = sock_ioctl,
148 .compat_ioctl = compat_sock_ioctl,
151 .release = sock_close,
152 .fasync = sock_fasync,
153 .sendpage = sock_sendpage,
154 .splice_write = generic_splice_sendpage,
155 .splice_read = sock_splice_read,
159 * The protocol list. Each protocol is registered in here.
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
172 * move_addr_to_kernel - copy a socket address into kernel space
173 * @uaddr: Address in user space
174 * @kaddr: Address in kernel space
175 * @ulen: Length in user space
177 * The address is copied into kernel space. If the provided address is
178 * too long an error code of -EINVAL is returned. If the copy gives
179 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
184 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
188 if (copy_from_user(kaddr, uaddr, ulen))
190 return audit_sockaddr(ulen, kaddr);
194 * move_addr_to_user - copy an address to user space
195 * @kaddr: kernel space address
196 * @klen: length of address in kernel
197 * @uaddr: user space address
198 * @ulen: pointer to user length field
200 * The value pointed to by ulen on entry is the buffer length available.
201 * This is overwritten with the buffer space used. -EINVAL is returned
202 * if an overlong buffer is specified or a negative buffer size. -EFAULT
203 * is returned if either the buffer or the length field are not
205 * After copying the data up to the limit the user specifies, the true
206 * length of the data is written over the length limit the user
207 * specified. Zero is returned for a success.
210 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
211 void __user *uaddr, int __user *ulen)
216 BUG_ON(klen > sizeof(struct sockaddr_storage));
217 err = get_user(len, ulen);
225 if (audit_sockaddr(klen, kaddr))
227 if (copy_to_user(uaddr, kaddr, len))
231 * "fromlen shall refer to the value before truncation.."
234 return __put_user(klen, ulen);
237 static struct kmem_cache *sock_inode_cachep __ro_after_init;
239 static struct inode *sock_alloc_inode(struct super_block *sb)
241 struct socket_alloc *ei;
242 struct socket_wq *wq;
244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
247 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249 kmem_cache_free(sock_inode_cachep, ei);
252 init_waitqueue_head(&wq->wait);
253 wq->fasync_list = NULL;
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
263 return &ei->vfs_inode;
266 static void sock_destroy_inode(struct inode *inode)
268 struct socket_alloc *ei;
270 ei = container_of(inode, struct socket_alloc, vfs_inode);
271 kfree_rcu(ei->socket.wq, rcu);
272 kmem_cache_free(sock_inode_cachep, ei);
275 static void init_once(void *foo)
277 struct socket_alloc *ei = (struct socket_alloc *)foo;
279 inode_init_once(&ei->vfs_inode);
282 static void init_inodecache(void)
284 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
285 sizeof(struct socket_alloc),
287 (SLAB_HWCACHE_ALIGN |
288 SLAB_RECLAIM_ACCOUNT |
289 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
291 BUG_ON(sock_inode_cachep == NULL);
294 static const struct super_operations sockfs_ops = {
295 .alloc_inode = sock_alloc_inode,
296 .destroy_inode = sock_destroy_inode,
297 .statfs = simple_statfs,
301 * sockfs_dname() is called from d_path().
303 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
305 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
306 d_inode(dentry)->i_ino);
309 static const struct dentry_operations sockfs_dentry_operations = {
310 .d_dname = sockfs_dname,
313 static int sockfs_xattr_get(const struct xattr_handler *handler,
314 struct dentry *dentry, struct inode *inode,
315 const char *suffix, void *value, size_t size)
318 if (dentry->d_name.len + 1 > size)
320 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
322 return dentry->d_name.len + 1;
325 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
326 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
327 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
329 static const struct xattr_handler sockfs_xattr_handler = {
330 .name = XATTR_NAME_SOCKPROTONAME,
331 .get = sockfs_xattr_get,
334 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
335 struct dentry *dentry, struct inode *inode,
336 const char *suffix, const void *value,
337 size_t size, int flags)
339 /* Handled by LSM. */
343 static const struct xattr_handler sockfs_security_xattr_handler = {
344 .prefix = XATTR_SECURITY_PREFIX,
345 .set = sockfs_security_xattr_set,
348 static const struct xattr_handler *sockfs_xattr_handlers[] = {
349 &sockfs_xattr_handler,
350 &sockfs_security_xattr_handler,
354 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
355 int flags, const char *dev_name, void *data)
357 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
358 sockfs_xattr_handlers,
359 &sockfs_dentry_operations, SOCKFS_MAGIC);
362 static struct vfsmount *sock_mnt __read_mostly;
364 static struct file_system_type sock_fs_type = {
366 .mount = sockfs_mount,
367 .kill_sb = kill_anon_super,
371 * Obtains the first available file descriptor and sets it up for use.
373 * These functions create file structures and maps them to fd space
374 * of the current process. On success it returns file descriptor
375 * and file struct implicitly stored in sock->file.
376 * Note that another thread may close file descriptor before we return
377 * from this function. We use the fact that now we do not refer
378 * to socket after mapping. If one day we will need it, this
379 * function will increment ref. count on file by 1.
381 * In any case returned fd MAY BE not valid!
382 * This race condition is unavoidable
383 * with shared fd spaces, we cannot solve it inside kernel,
384 * but we take care of internal coherence yet.
388 * sock_alloc_file - Bind a &socket to a &file
390 * @flags: file status flags
391 * @dname: protocol name
393 * Returns the &file bound with @sock, implicitly storing it
394 * in sock->file. If dname is %NULL, sets to "".
395 * On failure the return is a ERR pointer (see linux/err.h).
396 * This function uses GFP_KERNEL internally.
399 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
404 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
406 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
407 O_RDWR | (flags & O_NONBLOCK),
415 file->private_data = sock;
418 EXPORT_SYMBOL(sock_alloc_file);
420 static int sock_map_fd(struct socket *sock, int flags)
422 struct file *newfile;
423 int fd = get_unused_fd_flags(flags);
424 if (unlikely(fd < 0)) {
429 newfile = sock_alloc_file(sock, flags, NULL);
430 if (likely(!IS_ERR(newfile))) {
431 fd_install(fd, newfile);
436 return PTR_ERR(newfile);
440 * sock_from_file - Return the &socket bounded to @file.
442 * @err: pointer to an error code return
444 * On failure returns %NULL and assigns -ENOTSOCK to @err.
447 struct socket *sock_from_file(struct file *file, int *err)
449 if (file->f_op == &socket_file_ops)
450 return file->private_data; /* set in sock_map_fd */
455 EXPORT_SYMBOL(sock_from_file);
458 * sockfd_lookup - Go from a file number to its socket slot
460 * @err: pointer to an error code return
462 * The file handle passed in is locked and the socket it is bound
463 * to is returned. If an error occurs the err pointer is overwritten
464 * with a negative errno code and NULL is returned. The function checks
465 * for both invalid handles and passing a handle which is not a socket.
467 * On a success the socket object pointer is returned.
470 struct socket *sockfd_lookup(int fd, int *err)
481 sock = sock_from_file(file, err);
486 EXPORT_SYMBOL(sockfd_lookup);
488 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
490 struct fd f = fdget(fd);
495 sock = sock_from_file(f.file, err);
497 *fput_needed = f.flags & FDPUT_FPUT;
505 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
511 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
521 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
526 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
533 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
535 int err = simple_setattr(dentry, iattr);
537 if (!err && (iattr->ia_valid & ATTR_UID)) {
538 struct socket *sock = SOCKET_I(d_inode(dentry));
541 sock->sk->sk_uid = iattr->ia_uid;
549 static const struct inode_operations sockfs_inode_ops = {
550 .listxattr = sockfs_listxattr,
551 .setattr = sockfs_setattr,
555 * sock_alloc - allocate a socket
557 * Allocate a new inode and socket object. The two are bound together
558 * and initialised. The socket is then returned. If we are out of inodes
559 * NULL is returned. This functions uses GFP_KERNEL internally.
562 struct socket *sock_alloc(void)
567 inode = new_inode_pseudo(sock_mnt->mnt_sb);
571 sock = SOCKET_I(inode);
573 inode->i_ino = get_next_ino();
574 inode->i_mode = S_IFSOCK | S_IRWXUGO;
575 inode->i_uid = current_fsuid();
576 inode->i_gid = current_fsgid();
577 inode->i_op = &sockfs_inode_ops;
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 static void __sock_release(struct socket *sock, struct inode *inode)
595 struct module *owner = sock->ops->owner;
599 sock->ops->release(sock);
607 if (sock->wq->fasync_list)
608 pr_err("%s: fasync list not empty!\n", __func__);
611 iput(SOCK_INODE(sock));
617 void sock_release(struct socket *sock)
619 __sock_release(sock, NULL);
621 EXPORT_SYMBOL(sock_release);
623 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
625 u8 flags = *tx_flags;
627 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
628 flags |= SKBTX_HW_TSTAMP;
630 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
631 flags |= SKBTX_SW_TSTAMP;
633 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
634 flags |= SKBTX_SCHED_TSTAMP;
638 EXPORT_SYMBOL(__sock_tx_timestamp);
641 * sock_sendmsg - send a message through @sock
643 * @msg: message to send
645 * Sends @msg through @sock, passing through LSM.
646 * Returns the number of bytes sent, or an error code.
649 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
651 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
652 BUG_ON(ret == -EIOCBQUEUED);
656 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
658 int err = security_socket_sendmsg(sock, msg,
661 return err ?: sock_sendmsg_nosec(sock, msg);
663 EXPORT_SYMBOL(sock_sendmsg);
666 * kernel_sendmsg - send a message through @sock (kernel-space)
668 * @msg: message header
670 * @num: vec array length
671 * @size: total message data size
673 * Builds the message data with @vec and sends it through @sock.
674 * Returns the number of bytes sent, or an error code.
677 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
678 struct kvec *vec, size_t num, size_t size)
680 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
681 return sock_sendmsg(sock, msg);
683 EXPORT_SYMBOL(kernel_sendmsg);
686 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
688 * @msg: message header
689 * @vec: output s/g array
690 * @num: output s/g array length
691 * @size: total message data size
693 * Builds the message data with @vec and sends it through @sock.
694 * Returns the number of bytes sent, or an error code.
695 * Caller must hold @sk.
698 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
699 struct kvec *vec, size_t num, size_t size)
701 struct socket *sock = sk->sk_socket;
703 if (!sock->ops->sendmsg_locked)
704 return sock_no_sendmsg_locked(sk, msg, size);
706 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
708 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
710 EXPORT_SYMBOL(kernel_sendmsg_locked);
712 static bool skb_is_err_queue(const struct sk_buff *skb)
714 /* pkt_type of skbs enqueued on the error queue are set to
715 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
716 * in recvmsg, since skbs received on a local socket will never
717 * have a pkt_type of PACKET_OUTGOING.
719 return skb->pkt_type == PACKET_OUTGOING;
722 /* On transmit, software and hardware timestamps are returned independently.
723 * As the two skb clones share the hardware timestamp, which may be updated
724 * before the software timestamp is received, a hardware TX timestamp may be
725 * returned only if there is no software TX timestamp. Ignore false software
726 * timestamps, which may be made in the __sock_recv_timestamp() call when the
727 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
728 * hardware timestamp.
730 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
732 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
735 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
737 struct scm_ts_pktinfo ts_pktinfo;
738 struct net_device *orig_dev;
740 if (!skb_mac_header_was_set(skb))
743 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
746 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
748 ts_pktinfo.if_index = orig_dev->ifindex;
751 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
752 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
753 sizeof(ts_pktinfo), &ts_pktinfo);
757 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
759 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
762 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
763 struct scm_timestamping tss;
764 int empty = 1, false_tstamp = 0;
765 struct skb_shared_hwtstamps *shhwtstamps =
768 /* Race occurred between timestamp enabling and packet
769 receiving. Fill in the current time for now. */
770 if (need_software_tstamp && skb->tstamp == 0) {
771 __net_timestamp(skb);
775 if (need_software_tstamp) {
776 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
778 skb_get_timestamp(skb, &tv);
779 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
783 skb_get_timestampns(skb, &ts);
784 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
789 memset(&tss, 0, sizeof(tss));
790 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
791 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
794 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
795 !skb_is_swtx_tstamp(skb, false_tstamp) &&
796 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
798 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
799 !skb_is_err_queue(skb))
800 put_ts_pktinfo(msg, skb);
803 put_cmsg(msg, SOL_SOCKET,
804 SCM_TIMESTAMPING, sizeof(tss), &tss);
806 if (skb_is_err_queue(skb) && skb->len &&
807 SKB_EXT_ERR(skb)->opt_stats)
808 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
809 skb->len, skb->data);
812 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
814 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
819 if (!sock_flag(sk, SOCK_WIFI_STATUS))
821 if (!skb->wifi_acked_valid)
824 ack = skb->wifi_acked;
826 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
828 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
830 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
833 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
834 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
835 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
838 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
841 sock_recv_timestamp(msg, sk, skb);
842 sock_recv_drops(msg, sk, skb);
844 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
847 * sock_recvmsg - receive a message from @sock
849 * @msg: message to receive
850 * @flags: message flags
852 * Receives @msg from @sock, passing through LSM. Returns the total number
853 * of bytes received, or an error.
856 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
859 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
862 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
864 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
866 return err ?: sock_recvmsg_nosec(sock, msg, flags);
868 EXPORT_SYMBOL(sock_recvmsg);
871 * kernel_recvmsg - Receive a message from a socket (kernel space)
872 * @sock: The socket to receive the message from
873 * @msg: Received message
874 * @vec: Input s/g array for message data
875 * @num: Size of input s/g array
876 * @size: Number of bytes to read
877 * @flags: Message flags (MSG_DONTWAIT, etc...)
879 * On return the msg structure contains the scatter/gather array passed in the
880 * vec argument. The array is modified so that it consists of the unfilled
881 * portion of the original array.
883 * The returned value is the total number of bytes received, or an error.
886 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
887 struct kvec *vec, size_t num, size_t size, int flags)
889 mm_segment_t oldfs = get_fs();
892 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
894 result = sock_recvmsg(sock, msg, flags);
898 EXPORT_SYMBOL(kernel_recvmsg);
900 static ssize_t sock_sendpage(struct file *file, struct page *page,
901 int offset, size_t size, loff_t *ppos, int more)
906 sock = file->private_data;
908 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
909 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
912 return kernel_sendpage(sock, page, offset, size, flags);
915 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
916 struct pipe_inode_info *pipe, size_t len,
919 struct socket *sock = file->private_data;
921 if (unlikely(!sock->ops->splice_read))
924 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
927 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
929 struct file *file = iocb->ki_filp;
930 struct socket *sock = file->private_data;
931 struct msghdr msg = {.msg_iter = *to,
935 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
936 msg.msg_flags = MSG_DONTWAIT;
938 if (iocb->ki_pos != 0)
941 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
944 res = sock_recvmsg(sock, &msg, msg.msg_flags);
949 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
951 struct file *file = iocb->ki_filp;
952 struct socket *sock = file->private_data;
953 struct msghdr msg = {.msg_iter = *from,
957 if (iocb->ki_pos != 0)
960 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
961 msg.msg_flags = MSG_DONTWAIT;
963 if (sock->type == SOCK_SEQPACKET)
964 msg.msg_flags |= MSG_EOR;
966 res = sock_sendmsg(sock, &msg);
967 *from = msg.msg_iter;
972 * Atomic setting of ioctl hooks to avoid race
973 * with module unload.
976 static DEFINE_MUTEX(br_ioctl_mutex);
977 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
979 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
981 mutex_lock(&br_ioctl_mutex);
982 br_ioctl_hook = hook;
983 mutex_unlock(&br_ioctl_mutex);
985 EXPORT_SYMBOL(brioctl_set);
987 static DEFINE_MUTEX(vlan_ioctl_mutex);
988 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
990 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
992 mutex_lock(&vlan_ioctl_mutex);
993 vlan_ioctl_hook = hook;
994 mutex_unlock(&vlan_ioctl_mutex);
996 EXPORT_SYMBOL(vlan_ioctl_set);
998 static DEFINE_MUTEX(dlci_ioctl_mutex);
999 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1001 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1003 mutex_lock(&dlci_ioctl_mutex);
1004 dlci_ioctl_hook = hook;
1005 mutex_unlock(&dlci_ioctl_mutex);
1007 EXPORT_SYMBOL(dlci_ioctl_set);
1009 static long sock_do_ioctl(struct net *net, struct socket *sock,
1010 unsigned int cmd, unsigned long arg)
1013 void __user *argp = (void __user *)arg;
1015 err = sock->ops->ioctl(sock, cmd, arg);
1018 * If this ioctl is unknown try to hand it down
1019 * to the NIC driver.
1021 if (err != -ENOIOCTLCMD)
1024 if (cmd == SIOCGIFCONF) {
1026 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1029 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1031 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1033 } else if (is_socket_ioctl_cmd(cmd)) {
1036 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1038 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1039 if (!err && need_copyout)
1040 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1049 * With an ioctl, arg may well be a user mode pointer, but we don't know
1050 * what to do with it - that's up to the protocol still.
1053 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1055 struct socket *sock;
1057 void __user *argp = (void __user *)arg;
1061 sock = file->private_data;
1064 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1067 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1069 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1070 if (!err && need_copyout)
1071 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1074 #ifdef CONFIG_WEXT_CORE
1075 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1076 err = wext_handle_ioctl(net, cmd, argp);
1083 if (get_user(pid, (int __user *)argp))
1085 err = f_setown(sock->file, pid, 1);
1089 err = put_user(f_getown(sock->file),
1090 (int __user *)argp);
1098 request_module("bridge");
1100 mutex_lock(&br_ioctl_mutex);
1102 err = br_ioctl_hook(net, cmd, argp);
1103 mutex_unlock(&br_ioctl_mutex);
1108 if (!vlan_ioctl_hook)
1109 request_module("8021q");
1111 mutex_lock(&vlan_ioctl_mutex);
1112 if (vlan_ioctl_hook)
1113 err = vlan_ioctl_hook(net, argp);
1114 mutex_unlock(&vlan_ioctl_mutex);
1119 if (!dlci_ioctl_hook)
1120 request_module("dlci");
1122 mutex_lock(&dlci_ioctl_mutex);
1123 if (dlci_ioctl_hook)
1124 err = dlci_ioctl_hook(cmd, argp);
1125 mutex_unlock(&dlci_ioctl_mutex);
1129 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1132 err = open_related_ns(&net->ns, get_net_ns);
1135 err = sock_do_ioctl(net, sock, cmd, arg);
1142 * sock_create_lite - creates a socket
1143 * @family: protocol family (AF_INET, ...)
1144 * @type: communication type (SOCK_STREAM, ...)
1145 * @protocol: protocol (0, ...)
1148 * Creates a new socket and assigns it to @res, passing through LSM.
1149 * The new socket initialization is not complete, see kernel_accept().
1150 * Returns 0 or an error. On failure @res is set to %NULL.
1151 * This function internally uses GFP_KERNEL.
1154 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1157 struct socket *sock = NULL;
1159 err = security_socket_create(family, type, protocol, 1);
1163 sock = sock_alloc();
1170 err = security_socket_post_create(sock, family, type, protocol, 1);
1182 EXPORT_SYMBOL(sock_create_lite);
1184 /* No kernel lock held - perfect */
1185 static __poll_t sock_poll(struct file *file, poll_table *wait)
1187 struct socket *sock = file->private_data;
1188 __poll_t events = poll_requested_events(wait), flag = 0;
1190 if (!sock->ops->poll)
1193 if (sk_can_busy_loop(sock->sk)) {
1194 /* poll once if requested by the syscall */
1195 if (events & POLL_BUSY_LOOP)
1196 sk_busy_loop(sock->sk, 1);
1198 /* if this socket can poll_ll, tell the system call */
1199 flag = POLL_BUSY_LOOP;
1202 return sock->ops->poll(file, sock, wait) | flag;
1205 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1207 struct socket *sock = file->private_data;
1209 return sock->ops->mmap(file, sock, vma);
1212 static int sock_close(struct inode *inode, struct file *filp)
1214 __sock_release(SOCKET_I(inode), inode);
1219 * Update the socket async list
1221 * Fasync_list locking strategy.
1223 * 1. fasync_list is modified only under process context socket lock
1224 * i.e. under semaphore.
1225 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1226 * or under socket lock
1229 static int sock_fasync(int fd, struct file *filp, int on)
1231 struct socket *sock = filp->private_data;
1232 struct sock *sk = sock->sk;
1233 struct socket_wq *wq;
1240 fasync_helper(fd, filp, on, &wq->fasync_list);
1242 if (!wq->fasync_list)
1243 sock_reset_flag(sk, SOCK_FASYNC);
1245 sock_set_flag(sk, SOCK_FASYNC);
1251 /* This function may be called only under rcu_lock */
1253 int sock_wake_async(struct socket_wq *wq, int how, int band)
1255 if (!wq || !wq->fasync_list)
1259 case SOCK_WAKE_WAITD:
1260 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1263 case SOCK_WAKE_SPACE:
1264 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1269 kill_fasync(&wq->fasync_list, SIGIO, band);
1272 kill_fasync(&wq->fasync_list, SIGURG, band);
1277 EXPORT_SYMBOL(sock_wake_async);
1280 * __sock_create - creates a socket
1281 * @net: net namespace
1282 * @family: protocol family (AF_INET, ...)
1283 * @type: communication type (SOCK_STREAM, ...)
1284 * @protocol: protocol (0, ...)
1286 * @kern: boolean for kernel space sockets
1288 * Creates a new socket and assigns it to @res, passing through LSM.
1289 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1290 * be set to true if the socket resides in kernel space.
1291 * This function internally uses GFP_KERNEL.
1294 int __sock_create(struct net *net, int family, int type, int protocol,
1295 struct socket **res, int kern)
1298 struct socket *sock;
1299 const struct net_proto_family *pf;
1302 * Check protocol is in range
1304 if (family < 0 || family >= NPROTO)
1305 return -EAFNOSUPPORT;
1306 if (type < 0 || type >= SOCK_MAX)
1311 This uglymoron is moved from INET layer to here to avoid
1312 deadlock in module load.
1314 if (family == PF_INET && type == SOCK_PACKET) {
1315 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1320 err = security_socket_create(family, type, protocol, kern);
1325 * Allocate the socket and allow the family to set things up. if
1326 * the protocol is 0, the family is instructed to select an appropriate
1329 sock = sock_alloc();
1331 net_warn_ratelimited("socket: no more sockets\n");
1332 return -ENFILE; /* Not exactly a match, but its the
1333 closest posix thing */
1338 #ifdef CONFIG_MODULES
1339 /* Attempt to load a protocol module if the find failed.
1341 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1342 * requested real, full-featured networking support upon configuration.
1343 * Otherwise module support will break!
1345 if (rcu_access_pointer(net_families[family]) == NULL)
1346 request_module("net-pf-%d", family);
1350 pf = rcu_dereference(net_families[family]);
1351 err = -EAFNOSUPPORT;
1356 * We will call the ->create function, that possibly is in a loadable
1357 * module, so we have to bump that loadable module refcnt first.
1359 if (!try_module_get(pf->owner))
1362 /* Now protected by module ref count */
1365 err = pf->create(net, sock, protocol, kern);
1367 goto out_module_put;
1370 * Now to bump the refcnt of the [loadable] module that owns this
1371 * socket at sock_release time we decrement its refcnt.
1373 if (!try_module_get(sock->ops->owner))
1374 goto out_module_busy;
1377 * Now that we're done with the ->create function, the [loadable]
1378 * module can have its refcnt decremented
1380 module_put(pf->owner);
1381 err = security_socket_post_create(sock, family, type, protocol, kern);
1383 goto out_sock_release;
1389 err = -EAFNOSUPPORT;
1392 module_put(pf->owner);
1399 goto out_sock_release;
1401 EXPORT_SYMBOL(__sock_create);
1404 * sock_create - creates a socket
1405 * @family: protocol family (AF_INET, ...)
1406 * @type: communication type (SOCK_STREAM, ...)
1407 * @protocol: protocol (0, ...)
1410 * A wrapper around __sock_create().
1411 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1414 int sock_create(int family, int type, int protocol, struct socket **res)
1416 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1418 EXPORT_SYMBOL(sock_create);
1421 * sock_create_kern - creates a socket (kernel space)
1422 * @net: net namespace
1423 * @family: protocol family (AF_INET, ...)
1424 * @type: communication type (SOCK_STREAM, ...)
1425 * @protocol: protocol (0, ...)
1428 * A wrapper around __sock_create().
1429 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1432 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1434 return __sock_create(net, family, type, protocol, res, 1);
1436 EXPORT_SYMBOL(sock_create_kern);
1438 int __sys_socket(int family, int type, int protocol)
1441 struct socket *sock;
1444 /* Check the SOCK_* constants for consistency. */
1445 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1446 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1447 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1448 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1450 flags = type & ~SOCK_TYPE_MASK;
1451 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1453 type &= SOCK_TYPE_MASK;
1455 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1456 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1458 retval = sock_create(family, type, protocol, &sock);
1462 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1465 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1467 return __sys_socket(family, type, protocol);
1471 * Create a pair of connected sockets.
1474 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1476 struct socket *sock1, *sock2;
1478 struct file *newfile1, *newfile2;
1481 flags = type & ~SOCK_TYPE_MASK;
1482 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1484 type &= SOCK_TYPE_MASK;
1486 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1487 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1490 * reserve descriptors and make sure we won't fail
1491 * to return them to userland.
1493 fd1 = get_unused_fd_flags(flags);
1494 if (unlikely(fd1 < 0))
1497 fd2 = get_unused_fd_flags(flags);
1498 if (unlikely(fd2 < 0)) {
1503 err = put_user(fd1, &usockvec[0]);
1507 err = put_user(fd2, &usockvec[1]);
1512 * Obtain the first socket and check if the underlying protocol
1513 * supports the socketpair call.
1516 err = sock_create(family, type, protocol, &sock1);
1517 if (unlikely(err < 0))
1520 err = sock_create(family, type, protocol, &sock2);
1521 if (unlikely(err < 0)) {
1522 sock_release(sock1);
1526 err = security_socket_socketpair(sock1, sock2);
1527 if (unlikely(err)) {
1528 sock_release(sock2);
1529 sock_release(sock1);
1533 err = sock1->ops->socketpair(sock1, sock2);
1534 if (unlikely(err < 0)) {
1535 sock_release(sock2);
1536 sock_release(sock1);
1540 newfile1 = sock_alloc_file(sock1, flags, NULL);
1541 if (IS_ERR(newfile1)) {
1542 err = PTR_ERR(newfile1);
1543 sock_release(sock2);
1547 newfile2 = sock_alloc_file(sock2, flags, NULL);
1548 if (IS_ERR(newfile2)) {
1549 err = PTR_ERR(newfile2);
1554 audit_fd_pair(fd1, fd2);
1556 fd_install(fd1, newfile1);
1557 fd_install(fd2, newfile2);
1566 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1567 int __user *, usockvec)
1569 return __sys_socketpair(family, type, protocol, usockvec);
1573 * Bind a name to a socket. Nothing much to do here since it's
1574 * the protocol's responsibility to handle the local address.
1576 * We move the socket address to kernel space before we call
1577 * the protocol layer (having also checked the address is ok).
1580 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1582 struct socket *sock;
1583 struct sockaddr_storage address;
1584 int err, fput_needed;
1586 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1588 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1590 err = security_socket_bind(sock,
1591 (struct sockaddr *)&address,
1594 err = sock->ops->bind(sock,
1598 fput_light(sock->file, fput_needed);
1603 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1605 return __sys_bind(fd, umyaddr, addrlen);
1609 * Perform a listen. Basically, we allow the protocol to do anything
1610 * necessary for a listen, and if that works, we mark the socket as
1611 * ready for listening.
1614 int __sys_listen(int fd, int backlog)
1616 struct socket *sock;
1617 int err, fput_needed;
1620 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1622 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1623 if ((unsigned int)backlog > somaxconn)
1624 backlog = somaxconn;
1626 err = security_socket_listen(sock, backlog);
1628 err = sock->ops->listen(sock, backlog);
1630 fput_light(sock->file, fput_needed);
1635 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1637 return __sys_listen(fd, backlog);
1641 * For accept, we attempt to create a new socket, set up the link
1642 * with the client, wake up the client, then return the new
1643 * connected fd. We collect the address of the connector in kernel
1644 * space and move it to user at the very end. This is unclean because
1645 * we open the socket then return an error.
1647 * 1003.1g adds the ability to recvmsg() to query connection pending
1648 * status to recvmsg. We need to add that support in a way thats
1649 * clean when we restructure accept also.
1652 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1653 int __user *upeer_addrlen, int flags)
1655 struct socket *sock, *newsock;
1656 struct file *newfile;
1657 int err, len, newfd, fput_needed;
1658 struct sockaddr_storage address;
1660 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1663 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1664 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1666 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1671 newsock = sock_alloc();
1675 newsock->type = sock->type;
1676 newsock->ops = sock->ops;
1679 * We don't need try_module_get here, as the listening socket (sock)
1680 * has the protocol module (sock->ops->owner) held.
1682 __module_get(newsock->ops->owner);
1684 newfd = get_unused_fd_flags(flags);
1685 if (unlikely(newfd < 0)) {
1687 sock_release(newsock);
1690 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1691 if (IS_ERR(newfile)) {
1692 err = PTR_ERR(newfile);
1693 put_unused_fd(newfd);
1697 err = security_socket_accept(sock, newsock);
1701 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1705 if (upeer_sockaddr) {
1706 len = newsock->ops->getname(newsock,
1707 (struct sockaddr *)&address, 2);
1709 err = -ECONNABORTED;
1712 err = move_addr_to_user(&address,
1713 len, upeer_sockaddr, upeer_addrlen);
1718 /* File flags are not inherited via accept() unlike another OSes. */
1720 fd_install(newfd, newfile);
1724 fput_light(sock->file, fput_needed);
1729 put_unused_fd(newfd);
1733 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1734 int __user *, upeer_addrlen, int, flags)
1736 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1739 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1740 int __user *, upeer_addrlen)
1742 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1746 * Attempt to connect to a socket with the server address. The address
1747 * is in user space so we verify it is OK and move it to kernel space.
1749 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1752 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1753 * other SEQPACKET protocols that take time to connect() as it doesn't
1754 * include the -EINPROGRESS status for such sockets.
1757 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1759 struct socket *sock;
1760 struct sockaddr_storage address;
1761 int err, fput_needed;
1763 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1766 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1771 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1775 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1776 sock->file->f_flags);
1778 fput_light(sock->file, fput_needed);
1783 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1786 return __sys_connect(fd, uservaddr, addrlen);
1790 * Get the local address ('name') of a socket object. Move the obtained
1791 * name to user space.
1794 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1795 int __user *usockaddr_len)
1797 struct socket *sock;
1798 struct sockaddr_storage address;
1799 int err, fput_needed;
1801 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1805 err = security_socket_getsockname(sock);
1809 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1812 /* "err" is actually length in this case */
1813 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1816 fput_light(sock->file, fput_needed);
1821 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1822 int __user *, usockaddr_len)
1824 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1828 * Get the remote address ('name') of a socket object. Move the obtained
1829 * name to user space.
1832 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1833 int __user *usockaddr_len)
1835 struct socket *sock;
1836 struct sockaddr_storage address;
1837 int err, fput_needed;
1839 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1841 err = security_socket_getpeername(sock);
1843 fput_light(sock->file, fput_needed);
1847 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1849 /* "err" is actually length in this case */
1850 err = move_addr_to_user(&address, err, usockaddr,
1852 fput_light(sock->file, fput_needed);
1857 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1858 int __user *, usockaddr_len)
1860 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1864 * Send a datagram to a given address. We move the address into kernel
1865 * space and check the user space data area is readable before invoking
1868 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1869 struct sockaddr __user *addr, int addr_len)
1871 struct socket *sock;
1872 struct sockaddr_storage address;
1878 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1881 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1885 msg.msg_name = NULL;
1886 msg.msg_control = NULL;
1887 msg.msg_controllen = 0;
1888 msg.msg_namelen = 0;
1890 err = move_addr_to_kernel(addr, addr_len, &address);
1893 msg.msg_name = (struct sockaddr *)&address;
1894 msg.msg_namelen = addr_len;
1896 if (sock->file->f_flags & O_NONBLOCK)
1897 flags |= MSG_DONTWAIT;
1898 msg.msg_flags = flags;
1899 err = sock_sendmsg(sock, &msg);
1902 fput_light(sock->file, fput_needed);
1907 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1908 unsigned int, flags, struct sockaddr __user *, addr,
1911 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1915 * Send a datagram down a socket.
1918 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1919 unsigned int, flags)
1921 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1925 * Receive a frame from the socket and optionally record the address of the
1926 * sender. We verify the buffers are writable and if needed move the
1927 * sender address from kernel to user space.
1929 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1930 struct sockaddr __user *addr, int __user *addr_len)
1932 struct socket *sock;
1935 struct sockaddr_storage address;
1939 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1942 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1946 msg.msg_control = NULL;
1947 msg.msg_controllen = 0;
1948 /* Save some cycles and don't copy the address if not needed */
1949 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1950 /* We assume all kernel code knows the size of sockaddr_storage */
1951 msg.msg_namelen = 0;
1952 msg.msg_iocb = NULL;
1954 if (sock->file->f_flags & O_NONBLOCK)
1955 flags |= MSG_DONTWAIT;
1956 err = sock_recvmsg(sock, &msg, flags);
1958 if (err >= 0 && addr != NULL) {
1959 err2 = move_addr_to_user(&address,
1960 msg.msg_namelen, addr, addr_len);
1965 fput_light(sock->file, fput_needed);
1970 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1971 unsigned int, flags, struct sockaddr __user *, addr,
1972 int __user *, addr_len)
1974 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1978 * Receive a datagram from a socket.
1981 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1982 unsigned int, flags)
1984 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1988 * Set a socket option. Because we don't know the option lengths we have
1989 * to pass the user mode parameter for the protocols to sort out.
1992 static int __sys_setsockopt(int fd, int level, int optname,
1993 char __user *optval, int optlen)
1995 int err, fput_needed;
1996 struct socket *sock;
2001 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2003 err = security_socket_setsockopt(sock, level, optname);
2007 if (level == SOL_SOCKET)
2009 sock_setsockopt(sock, level, optname, optval,
2013 sock->ops->setsockopt(sock, level, optname, optval,
2016 fput_light(sock->file, fput_needed);
2021 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2022 char __user *, optval, int, optlen)
2024 return __sys_setsockopt(fd, level, optname, optval, optlen);
2028 * Get a socket option. Because we don't know the option lengths we have
2029 * to pass a user mode parameter for the protocols to sort out.
2032 static int __sys_getsockopt(int fd, int level, int optname,
2033 char __user *optval, int __user *optlen)
2035 int err, fput_needed;
2036 struct socket *sock;
2038 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2040 err = security_socket_getsockopt(sock, level, optname);
2044 if (level == SOL_SOCKET)
2046 sock_getsockopt(sock, level, optname, optval,
2050 sock->ops->getsockopt(sock, level, optname, optval,
2053 fput_light(sock->file, fput_needed);
2058 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2059 char __user *, optval, int __user *, optlen)
2061 return __sys_getsockopt(fd, level, optname, optval, optlen);
2065 * Shutdown a socket.
2068 int __sys_shutdown(int fd, int how)
2070 int err, fput_needed;
2071 struct socket *sock;
2073 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2075 err = security_socket_shutdown(sock, how);
2077 err = sock->ops->shutdown(sock, how);
2078 fput_light(sock->file, fput_needed);
2083 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2085 return __sys_shutdown(fd, how);
2088 /* A couple of helpful macros for getting the address of the 32/64 bit
2089 * fields which are the same type (int / unsigned) on our platforms.
2091 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2092 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2093 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2095 struct used_address {
2096 struct sockaddr_storage name;
2097 unsigned int name_len;
2100 static int copy_msghdr_from_user(struct msghdr *kmsg,
2101 struct user_msghdr __user *umsg,
2102 struct sockaddr __user **save_addr,
2105 struct user_msghdr msg;
2108 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2111 kmsg->msg_control = (void __force *)msg.msg_control;
2112 kmsg->msg_controllen = msg.msg_controllen;
2113 kmsg->msg_flags = msg.msg_flags;
2115 kmsg->msg_namelen = msg.msg_namelen;
2117 kmsg->msg_namelen = 0;
2119 if (kmsg->msg_namelen < 0)
2122 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2123 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2126 *save_addr = msg.msg_name;
2128 if (msg.msg_name && kmsg->msg_namelen) {
2130 err = move_addr_to_kernel(msg.msg_name,
2137 kmsg->msg_name = NULL;
2138 kmsg->msg_namelen = 0;
2141 if (msg.msg_iovlen > UIO_MAXIOV)
2144 kmsg->msg_iocb = NULL;
2146 return import_iovec(save_addr ? READ : WRITE,
2147 msg.msg_iov, msg.msg_iovlen,
2148 UIO_FASTIOV, iov, &kmsg->msg_iter);
2151 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2152 struct msghdr *msg_sys, unsigned int flags,
2153 struct used_address *used_address,
2154 unsigned int allowed_msghdr_flags)
2156 struct compat_msghdr __user *msg_compat =
2157 (struct compat_msghdr __user *)msg;
2158 struct sockaddr_storage address;
2159 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2160 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2161 __aligned(sizeof(__kernel_size_t));
2162 /* 20 is size of ipv6_pktinfo */
2163 unsigned char *ctl_buf = ctl;
2167 msg_sys->msg_name = &address;
2169 if (MSG_CMSG_COMPAT & flags)
2170 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2172 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2178 if (msg_sys->msg_controllen > INT_MAX)
2180 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2181 ctl_len = msg_sys->msg_controllen;
2182 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2184 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2188 ctl_buf = msg_sys->msg_control;
2189 ctl_len = msg_sys->msg_controllen;
2190 } else if (ctl_len) {
2191 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2192 CMSG_ALIGN(sizeof(struct cmsghdr)));
2193 if (ctl_len > sizeof(ctl)) {
2194 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2195 if (ctl_buf == NULL)
2200 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2201 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2202 * checking falls down on this.
2204 if (copy_from_user(ctl_buf,
2205 (void __user __force *)msg_sys->msg_control,
2208 msg_sys->msg_control = ctl_buf;
2210 msg_sys->msg_flags = flags;
2212 if (sock->file->f_flags & O_NONBLOCK)
2213 msg_sys->msg_flags |= MSG_DONTWAIT;
2215 * If this is sendmmsg() and current destination address is same as
2216 * previously succeeded address, omit asking LSM's decision.
2217 * used_address->name_len is initialized to UINT_MAX so that the first
2218 * destination address never matches.
2220 if (used_address && msg_sys->msg_name &&
2221 used_address->name_len == msg_sys->msg_namelen &&
2222 !memcmp(&used_address->name, msg_sys->msg_name,
2223 used_address->name_len)) {
2224 err = sock_sendmsg_nosec(sock, msg_sys);
2227 err = sock_sendmsg(sock, msg_sys);
2229 * If this is sendmmsg() and sending to current destination address was
2230 * successful, remember it.
2232 if (used_address && err >= 0) {
2233 used_address->name_len = msg_sys->msg_namelen;
2234 if (msg_sys->msg_name)
2235 memcpy(&used_address->name, msg_sys->msg_name,
2236 used_address->name_len);
2241 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2248 * BSD sendmsg interface
2251 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2252 bool forbid_cmsg_compat)
2254 int fput_needed, err;
2255 struct msghdr msg_sys;
2256 struct socket *sock;
2258 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2261 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2265 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2267 fput_light(sock->file, fput_needed);
2272 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2274 return __sys_sendmsg(fd, msg, flags, true);
2278 * Linux sendmmsg interface
2281 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2282 unsigned int flags, bool forbid_cmsg_compat)
2284 int fput_needed, err, datagrams;
2285 struct socket *sock;
2286 struct mmsghdr __user *entry;
2287 struct compat_mmsghdr __user *compat_entry;
2288 struct msghdr msg_sys;
2289 struct used_address used_address;
2290 unsigned int oflags = flags;
2292 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2295 if (vlen > UIO_MAXIOV)
2300 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2304 used_address.name_len = UINT_MAX;
2306 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2310 while (datagrams < vlen) {
2311 if (datagrams == vlen - 1)
2314 if (MSG_CMSG_COMPAT & flags) {
2315 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2316 &msg_sys, flags, &used_address, MSG_EOR);
2319 err = __put_user(err, &compat_entry->msg_len);
2322 err = ___sys_sendmsg(sock,
2323 (struct user_msghdr __user *)entry,
2324 &msg_sys, flags, &used_address, MSG_EOR);
2327 err = put_user(err, &entry->msg_len);
2334 if (msg_data_left(&msg_sys))
2339 fput_light(sock->file, fput_needed);
2341 /* We only return an error if no datagrams were able to be sent */
2348 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2349 unsigned int, vlen, unsigned int, flags)
2351 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2354 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2355 struct msghdr *msg_sys, unsigned int flags, int nosec)
2357 struct compat_msghdr __user *msg_compat =
2358 (struct compat_msghdr __user *)msg;
2359 struct iovec iovstack[UIO_FASTIOV];
2360 struct iovec *iov = iovstack;
2361 unsigned long cmsg_ptr;
2365 /* kernel mode address */
2366 struct sockaddr_storage addr;
2368 /* user mode address pointers */
2369 struct sockaddr __user *uaddr;
2370 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2372 msg_sys->msg_name = &addr;
2374 if (MSG_CMSG_COMPAT & flags)
2375 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2377 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2381 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2382 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2384 /* We assume all kernel code knows the size of sockaddr_storage */
2385 msg_sys->msg_namelen = 0;
2387 if (sock->file->f_flags & O_NONBLOCK)
2388 flags |= MSG_DONTWAIT;
2389 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2394 if (uaddr != NULL) {
2395 err = move_addr_to_user(&addr,
2396 msg_sys->msg_namelen, uaddr,
2401 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2405 if (MSG_CMSG_COMPAT & flags)
2406 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2407 &msg_compat->msg_controllen);
2409 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2410 &msg->msg_controllen);
2421 * BSD recvmsg interface
2424 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2425 bool forbid_cmsg_compat)
2427 int fput_needed, err;
2428 struct msghdr msg_sys;
2429 struct socket *sock;
2431 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2434 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2438 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2440 fput_light(sock->file, fput_needed);
2445 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2446 unsigned int, flags)
2448 return __sys_recvmsg(fd, msg, flags, true);
2452 * Linux recvmmsg interface
2455 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2456 unsigned int flags, struct timespec *timeout)
2458 int fput_needed, err, datagrams;
2459 struct socket *sock;
2460 struct mmsghdr __user *entry;
2461 struct compat_mmsghdr __user *compat_entry;
2462 struct msghdr msg_sys;
2463 struct timespec64 end_time;
2464 struct timespec64 timeout64;
2467 poll_select_set_timeout(&end_time, timeout->tv_sec,
2473 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2477 if (likely(!(flags & MSG_ERRQUEUE))) {
2478 err = sock_error(sock->sk);
2486 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2488 while (datagrams < vlen) {
2490 * No need to ask LSM for more than the first datagram.
2492 if (MSG_CMSG_COMPAT & flags) {
2493 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2494 &msg_sys, flags & ~MSG_WAITFORONE,
2498 err = __put_user(err, &compat_entry->msg_len);
2501 err = ___sys_recvmsg(sock,
2502 (struct user_msghdr __user *)entry,
2503 &msg_sys, flags & ~MSG_WAITFORONE,
2507 err = put_user(err, &entry->msg_len);
2515 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2516 if (flags & MSG_WAITFORONE)
2517 flags |= MSG_DONTWAIT;
2520 ktime_get_ts64(&timeout64);
2521 *timeout = timespec64_to_timespec(
2522 timespec64_sub(end_time, timeout64));
2523 if (timeout->tv_sec < 0) {
2524 timeout->tv_sec = timeout->tv_nsec = 0;
2528 /* Timeout, return less than vlen datagrams */
2529 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2533 /* Out of band data, return right away */
2534 if (msg_sys.msg_flags & MSG_OOB)
2542 if (datagrams == 0) {
2548 * We may return less entries than requested (vlen) if the
2549 * sock is non block and there aren't enough datagrams...
2551 if (err != -EAGAIN) {
2553 * ... or if recvmsg returns an error after we
2554 * received some datagrams, where we record the
2555 * error to return on the next call or if the
2556 * app asks about it using getsockopt(SO_ERROR).
2558 sock->sk->sk_err = -err;
2561 fput_light(sock->file, fput_needed);
2566 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2567 unsigned int vlen, unsigned int flags,
2568 struct timespec __user *timeout)
2571 struct timespec timeout_sys;
2573 if (flags & MSG_CMSG_COMPAT)
2577 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2579 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2582 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2584 if (datagrams > 0 &&
2585 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2586 datagrams = -EFAULT;
2591 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2592 unsigned int, vlen, unsigned int, flags,
2593 struct timespec __user *, timeout)
2595 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2598 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2599 /* Argument list sizes for sys_socketcall */
2600 #define AL(x) ((x) * sizeof(unsigned long))
2601 static const unsigned char nargs[21] = {
2602 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2603 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2604 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2611 * System call vectors.
2613 * Argument checking cleaned up. Saved 20% in size.
2614 * This function doesn't need to set the kernel lock because
2615 * it is set by the callees.
2618 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2620 unsigned long a[AUDITSC_ARGS];
2621 unsigned long a0, a1;
2625 if (call < 1 || call > SYS_SENDMMSG)
2627 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2630 if (len > sizeof(a))
2633 /* copy_from_user should be SMP safe. */
2634 if (copy_from_user(a, args, len))
2637 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2646 err = __sys_socket(a0, a1, a[2]);
2649 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2652 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2655 err = __sys_listen(a0, a1);
2658 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2659 (int __user *)a[2], 0);
2661 case SYS_GETSOCKNAME:
2663 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2664 (int __user *)a[2]);
2666 case SYS_GETPEERNAME:
2668 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2669 (int __user *)a[2]);
2671 case SYS_SOCKETPAIR:
2672 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2675 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2679 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2680 (struct sockaddr __user *)a[4], a[5]);
2683 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2687 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2688 (struct sockaddr __user *)a[4],
2689 (int __user *)a[5]);
2692 err = __sys_shutdown(a0, a1);
2694 case SYS_SETSOCKOPT:
2695 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2698 case SYS_GETSOCKOPT:
2700 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2701 (int __user *)a[4]);
2704 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2708 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2712 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2716 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2717 a[3], (struct timespec __user *)a[4]);
2720 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2721 (int __user *)a[2], a[3]);
2730 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2733 * sock_register - add a socket protocol handler
2734 * @ops: description of protocol
2736 * This function is called by a protocol handler that wants to
2737 * advertise its address family, and have it linked into the
2738 * socket interface. The value ops->family corresponds to the
2739 * socket system call protocol family.
2741 int sock_register(const struct net_proto_family *ops)
2745 if (ops->family >= NPROTO) {
2746 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2750 spin_lock(&net_family_lock);
2751 if (rcu_dereference_protected(net_families[ops->family],
2752 lockdep_is_held(&net_family_lock)))
2755 rcu_assign_pointer(net_families[ops->family], ops);
2758 spin_unlock(&net_family_lock);
2760 pr_info("NET: Registered protocol family %d\n", ops->family);
2763 EXPORT_SYMBOL(sock_register);
2766 * sock_unregister - remove a protocol handler
2767 * @family: protocol family to remove
2769 * This function is called by a protocol handler that wants to
2770 * remove its address family, and have it unlinked from the
2771 * new socket creation.
2773 * If protocol handler is a module, then it can use module reference
2774 * counts to protect against new references. If protocol handler is not
2775 * a module then it needs to provide its own protection in
2776 * the ops->create routine.
2778 void sock_unregister(int family)
2780 BUG_ON(family < 0 || family >= NPROTO);
2782 spin_lock(&net_family_lock);
2783 RCU_INIT_POINTER(net_families[family], NULL);
2784 spin_unlock(&net_family_lock);
2788 pr_info("NET: Unregistered protocol family %d\n", family);
2790 EXPORT_SYMBOL(sock_unregister);
2792 bool sock_is_registered(int family)
2794 return family < NPROTO && rcu_access_pointer(net_families[family]);
2797 static int __init sock_init(void)
2801 * Initialize the network sysctl infrastructure.
2803 err = net_sysctl_init();
2808 * Initialize skbuff SLAB cache
2813 * Initialize the protocols module.
2818 err = register_filesystem(&sock_fs_type);
2821 sock_mnt = kern_mount(&sock_fs_type);
2822 if (IS_ERR(sock_mnt)) {
2823 err = PTR_ERR(sock_mnt);
2827 /* The real protocol initialization is performed in later initcalls.
2830 #ifdef CONFIG_NETFILTER
2831 err = netfilter_init();
2836 ptp_classifier_init();
2842 unregister_filesystem(&sock_fs_type);
2847 core_initcall(sock_init); /* early initcall */
2849 #ifdef CONFIG_PROC_FS
2850 void socket_seq_show(struct seq_file *seq)
2852 seq_printf(seq, "sockets: used %d\n",
2853 sock_inuse_get(seq->private));
2855 #endif /* CONFIG_PROC_FS */
2857 #ifdef CONFIG_COMPAT
2858 static int do_siocgstamp(struct net *net, struct socket *sock,
2859 unsigned int cmd, void __user *up)
2861 mm_segment_t old_fs = get_fs();
2866 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2869 err = compat_put_timeval(&ktv, up);
2874 static int do_siocgstampns(struct net *net, struct socket *sock,
2875 unsigned int cmd, void __user *up)
2877 mm_segment_t old_fs = get_fs();
2878 struct timespec kts;
2882 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2885 err = compat_put_timespec(&kts, up);
2890 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2892 struct compat_ifconf ifc32;
2896 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2899 ifc.ifc_len = ifc32.ifc_len;
2900 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2903 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2908 ifc32.ifc_len = ifc.ifc_len;
2909 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2915 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2917 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2918 bool convert_in = false, convert_out = false;
2919 size_t buf_size = 0;
2920 struct ethtool_rxnfc __user *rxnfc = NULL;
2922 u32 rule_cnt = 0, actual_rule_cnt;
2927 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2930 compat_rxnfc = compat_ptr(data);
2932 if (get_user(ethcmd, &compat_rxnfc->cmd))
2935 /* Most ethtool structures are defined without padding.
2936 * Unfortunately struct ethtool_rxnfc is an exception.
2941 case ETHTOOL_GRXCLSRLALL:
2942 /* Buffer size is variable */
2943 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2945 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2947 buf_size += rule_cnt * sizeof(u32);
2949 case ETHTOOL_GRXRINGS:
2950 case ETHTOOL_GRXCLSRLCNT:
2951 case ETHTOOL_GRXCLSRULE:
2952 case ETHTOOL_SRXCLSRLINS:
2955 case ETHTOOL_SRXCLSRLDEL:
2956 buf_size += sizeof(struct ethtool_rxnfc);
2958 rxnfc = compat_alloc_user_space(buf_size);
2962 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2965 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2968 /* We expect there to be holes between fs.m_ext and
2969 * fs.ring_cookie and at the end of fs, but nowhere else.
2971 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2972 sizeof(compat_rxnfc->fs.m_ext) !=
2973 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2974 sizeof(rxnfc->fs.m_ext));
2976 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2977 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2978 offsetof(struct ethtool_rxnfc, fs.location) -
2979 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2981 if (copy_in_user(rxnfc, compat_rxnfc,
2982 (void __user *)(&rxnfc->fs.m_ext + 1) -
2983 (void __user *)rxnfc) ||
2984 copy_in_user(&rxnfc->fs.ring_cookie,
2985 &compat_rxnfc->fs.ring_cookie,
2986 (void __user *)(&rxnfc->fs.location + 1) -
2987 (void __user *)&rxnfc->fs.ring_cookie))
2989 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2990 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2992 } else if (copy_in_user(&rxnfc->rule_cnt,
2993 &compat_rxnfc->rule_cnt,
2994 sizeof(rxnfc->rule_cnt)))
2998 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3003 if (copy_in_user(compat_rxnfc, rxnfc,
3004 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3005 (const void __user *)rxnfc) ||
3006 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3007 &rxnfc->fs.ring_cookie,
3008 (const void __user *)(&rxnfc->fs.location + 1) -
3009 (const void __user *)&rxnfc->fs.ring_cookie) ||
3010 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3011 sizeof(rxnfc->rule_cnt)))
3014 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3015 /* As an optimisation, we only copy the actual
3016 * number of rules that the underlying
3017 * function returned. Since Mallory might
3018 * change the rule count in user memory, we
3019 * check that it is less than the rule count
3020 * originally given (as the user buffer size),
3021 * which has been range-checked.
3023 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3025 if (actual_rule_cnt < rule_cnt)
3026 rule_cnt = actual_rule_cnt;
3027 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3028 &rxnfc->rule_locs[0],
3029 rule_cnt * sizeof(u32)))
3037 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3039 compat_uptr_t uptr32;
3044 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3047 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3050 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3051 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3053 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3055 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3056 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3062 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3063 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3064 struct compat_ifreq __user *u_ifreq32)
3069 if (!is_socket_ioctl_cmd(cmd))
3071 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3073 if (get_user(data32, &u_ifreq32->ifr_data))
3075 ifreq.ifr_data = compat_ptr(data32);
3077 return dev_ioctl(net, cmd, &ifreq, NULL);
3080 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3082 struct compat_ifreq __user *uifr32)
3084 struct ifreq __user *uifr;
3087 /* Handle the fact that while struct ifreq has the same *layout* on
3088 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3089 * which are handled elsewhere, it still has different *size* due to
3090 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3091 * resulting in struct ifreq being 32 and 40 bytes respectively).
3092 * As a result, if the struct happens to be at the end of a page and
3093 * the next page isn't readable/writable, we get a fault. To prevent
3094 * that, copy back and forth to the full size.
3097 uifr = compat_alloc_user_space(sizeof(*uifr));
3098 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3101 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3112 case SIOCGIFBRDADDR:
3113 case SIOCGIFDSTADDR:
3114 case SIOCGIFNETMASK:
3120 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3128 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3129 struct compat_ifreq __user *uifr32)
3132 struct compat_ifmap __user *uifmap32;
3135 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3136 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3137 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3138 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3139 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3140 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3141 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3142 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3146 err = dev_ioctl(net, cmd, &ifr, NULL);
3148 if (cmd == SIOCGIFMAP && !err) {
3149 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3150 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3151 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3152 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3153 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3154 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3155 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3164 struct sockaddr rt_dst; /* target address */
3165 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3166 struct sockaddr rt_genmask; /* target network mask (IP) */
3167 unsigned short rt_flags;
3170 unsigned char rt_tos;
3171 unsigned char rt_class;
3173 short rt_metric; /* +1 for binary compatibility! */
3174 /* char * */ u32 rt_dev; /* forcing the device at add */
3175 u32 rt_mtu; /* per route MTU/Window */
3176 u32 rt_window; /* Window clamping */
3177 unsigned short rt_irtt; /* Initial RTT */
3180 struct in6_rtmsg32 {
3181 struct in6_addr rtmsg_dst;
3182 struct in6_addr rtmsg_src;
3183 struct in6_addr rtmsg_gateway;
3193 static int routing_ioctl(struct net *net, struct socket *sock,
3194 unsigned int cmd, void __user *argp)
3198 struct in6_rtmsg r6;
3202 mm_segment_t old_fs = get_fs();
3204 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3205 struct in6_rtmsg32 __user *ur6 = argp;
3206 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3207 3 * sizeof(struct in6_addr));
3208 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3209 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3210 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3211 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3212 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3213 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3214 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3218 struct rtentry32 __user *ur4 = argp;
3219 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3220 3 * sizeof(struct sockaddr));
3221 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3222 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3223 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3224 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3225 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3226 ret |= get_user(rtdev, &(ur4->rt_dev));
3228 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3229 r4.rt_dev = (char __user __force *)devname;
3243 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3250 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3251 * for some operations; this forces use of the newer bridge-utils that
3252 * use compatible ioctls
3254 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3258 if (get_user(tmp, argp))
3260 if (tmp == BRCTL_GET_VERSION)
3261 return BRCTL_VERSION + 1;
3265 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3266 unsigned int cmd, unsigned long arg)
3268 void __user *argp = compat_ptr(arg);
3269 struct sock *sk = sock->sk;
3270 struct net *net = sock_net(sk);
3272 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3273 return compat_ifr_data_ioctl(net, cmd, argp);
3278 return old_bridge_ioctl(argp);
3280 return compat_dev_ifconf(net, argp);
3282 return ethtool_ioctl(net, argp);
3284 return compat_siocwandev(net, argp);
3287 return compat_sioc_ifmap(net, cmd, argp);
3290 return routing_ioctl(net, sock, cmd, argp);
3292 return do_siocgstamp(net, sock, cmd, argp);
3294 return do_siocgstampns(net, sock, cmd, argp);
3295 case SIOCBONDSLAVEINFOQUERY:
3296 case SIOCBONDINFOQUERY:
3299 return compat_ifr_data_ioctl(net, cmd, argp);
3312 return sock_ioctl(file, cmd, arg);
3329 case SIOCSIFHWBROADCAST:
3331 case SIOCGIFBRDADDR:
3332 case SIOCSIFBRDADDR:
3333 case SIOCGIFDSTADDR:
3334 case SIOCSIFDSTADDR:
3335 case SIOCGIFNETMASK:
3336 case SIOCSIFNETMASK:
3348 case SIOCBONDENSLAVE:
3349 case SIOCBONDRELEASE:
3350 case SIOCBONDSETHWADDR:
3351 case SIOCBONDCHANGEACTIVE:
3352 return compat_ifreq_ioctl(net, sock, cmd, argp);
3359 return sock_do_ioctl(net, sock, cmd, arg);
3362 return -ENOIOCTLCMD;
3365 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3368 struct socket *sock = file->private_data;
3369 int ret = -ENOIOCTLCMD;
3376 if (sock->ops->compat_ioctl)
3377 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3379 if (ret == -ENOIOCTLCMD &&
3380 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3381 ret = compat_wext_handle_ioctl(net, cmd, arg);
3383 if (ret == -ENOIOCTLCMD)
3384 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3391 * kernel_bind - bind an address to a socket (kernel space)
3394 * @addrlen: length of address
3396 * Returns 0 or an error.
3399 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3401 return sock->ops->bind(sock, addr, addrlen);
3403 EXPORT_SYMBOL(kernel_bind);
3406 * kernel_listen - move socket to listening state (kernel space)
3408 * @backlog: pending connections queue size
3410 * Returns 0 or an error.
3413 int kernel_listen(struct socket *sock, int backlog)
3415 return sock->ops->listen(sock, backlog);
3417 EXPORT_SYMBOL(kernel_listen);
3420 * kernel_accept - accept a connection (kernel space)
3421 * @sock: listening socket
3422 * @newsock: new connected socket
3425 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3426 * If it fails, @newsock is guaranteed to be %NULL.
3427 * Returns 0 or an error.
3430 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3432 struct sock *sk = sock->sk;
3435 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3440 err = sock->ops->accept(sock, *newsock, flags, true);
3442 sock_release(*newsock);
3447 (*newsock)->ops = sock->ops;
3448 __module_get((*newsock)->ops->owner);
3453 EXPORT_SYMBOL(kernel_accept);
3456 * kernel_connect - connect a socket (kernel space)
3459 * @addrlen: address length
3460 * @flags: flags (O_NONBLOCK, ...)
3462 * For datagram sockets, @addr is the addres to which datagrams are sent
3463 * by default, and the only address from which datagrams are received.
3464 * For stream sockets, attempts to connect to @addr.
3465 * Returns 0 or an error code.
3468 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3471 return sock->ops->connect(sock, addr, addrlen, flags);
3473 EXPORT_SYMBOL(kernel_connect);
3476 * kernel_getsockname - get the address which the socket is bound (kernel space)
3478 * @addr: address holder
3480 * Fills the @addr pointer with the address which the socket is bound.
3481 * Returns 0 or an error code.
3484 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3486 return sock->ops->getname(sock, addr, 0);
3488 EXPORT_SYMBOL(kernel_getsockname);
3491 * kernel_peername - get the address which the socket is connected (kernel space)
3493 * @addr: address holder
3495 * Fills the @addr pointer with the address which the socket is connected.
3496 * Returns 0 or an error code.
3499 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3501 return sock->ops->getname(sock, addr, 1);
3503 EXPORT_SYMBOL(kernel_getpeername);
3506 * kernel_getsockopt - get a socket option (kernel space)
3508 * @level: API level (SOL_SOCKET, ...)
3509 * @optname: option tag
3510 * @optval: option value
3511 * @optlen: option length
3513 * Assigns the option length to @optlen.
3514 * Returns 0 or an error.
3517 int kernel_getsockopt(struct socket *sock, int level, int optname,
3518 char *optval, int *optlen)
3520 mm_segment_t oldfs = get_fs();
3521 char __user *uoptval;
3522 int __user *uoptlen;
3525 uoptval = (char __user __force *) optval;
3526 uoptlen = (int __user __force *) optlen;
3529 if (level == SOL_SOCKET)
3530 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3532 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3537 EXPORT_SYMBOL(kernel_getsockopt);
3540 * kernel_setsockopt - set a socket option (kernel space)
3542 * @level: API level (SOL_SOCKET, ...)
3543 * @optname: option tag
3544 * @optval: option value
3545 * @optlen: option length
3547 * Returns 0 or an error.
3550 int kernel_setsockopt(struct socket *sock, int level, int optname,
3551 char *optval, unsigned int optlen)
3553 mm_segment_t oldfs = get_fs();
3554 char __user *uoptval;
3557 uoptval = (char __user __force *) optval;
3560 if (level == SOL_SOCKET)
3561 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3563 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3568 EXPORT_SYMBOL(kernel_setsockopt);
3571 * kernel_sendpage - send a &page through a socket (kernel space)
3574 * @offset: page offset
3575 * @size: total size in bytes
3576 * @flags: flags (MSG_DONTWAIT, ...)
3578 * Returns the total amount sent in bytes or an error.
3581 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3582 size_t size, int flags)
3584 if (sock->ops->sendpage)
3585 return sock->ops->sendpage(sock, page, offset, size, flags);
3587 return sock_no_sendpage(sock, page, offset, size, flags);
3589 EXPORT_SYMBOL(kernel_sendpage);
3592 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3595 * @offset: page offset
3596 * @size: total size in bytes
3597 * @flags: flags (MSG_DONTWAIT, ...)
3599 * Returns the total amount sent in bytes or an error.
3600 * Caller must hold @sk.
3603 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3604 size_t size, int flags)
3606 struct socket *sock = sk->sk_socket;
3608 if (sock->ops->sendpage_locked)
3609 return sock->ops->sendpage_locked(sk, page, offset, size,
3612 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3614 EXPORT_SYMBOL(kernel_sendpage_locked);
3617 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3619 * @how: connection part
3621 * Returns 0 or an error.
3624 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3626 return sock->ops->shutdown(sock, how);
3628 EXPORT_SYMBOL(kernel_sock_shutdown);
3631 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3634 * This routine returns the IP overhead imposed by a socket i.e.
3635 * the length of the underlying IP header, depending on whether
3636 * this is an IPv4 or IPv6 socket and the length from IP options turned
3637 * on at the socket. Assumes that the caller has a lock on the socket.
3640 u32 kernel_sock_ip_overhead(struct sock *sk)
3642 struct inet_sock *inet;
3643 struct ip_options_rcu *opt;
3645 #if IS_ENABLED(CONFIG_IPV6)
3646 struct ipv6_pinfo *np;
3647 struct ipv6_txoptions *optv6 = NULL;
3648 #endif /* IS_ENABLED(CONFIG_IPV6) */
3653 switch (sk->sk_family) {
3656 overhead += sizeof(struct iphdr);
3657 opt = rcu_dereference_protected(inet->inet_opt,
3658 sock_owned_by_user(sk));
3660 overhead += opt->opt.optlen;
3662 #if IS_ENABLED(CONFIG_IPV6)
3665 overhead += sizeof(struct ipv6hdr);
3667 optv6 = rcu_dereference_protected(np->opt,
3668 sock_owned_by_user(sk));
3670 overhead += (optv6->opt_flen + optv6->opt_nflen);
3672 #endif /* IS_ENABLED(CONFIG_IPV6) */
3673 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3677 EXPORT_SYMBOL(kernel_sock_ip_overhead);