1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/splice.h>
61 #include <linux/net.h>
62 #include <linux/interrupt.h>
63 #include <linux/thread_info.h>
64 #include <linux/rcupdate.h>
65 #include <linux/netdevice.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68 #include <linux/mutex.h>
69 #include <linux/if_bridge.h>
70 #include <linux/if_vlan.h>
71 #include <linux/ptp_classify.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
77 #include <linux/mount.h>
78 #include <linux/pseudo_fs.h>
79 #include <linux/security.h>
80 #include <linux/syscalls.h>
81 #include <linux/compat.h>
82 #include <linux/kmod.h>
83 #include <linux/audit.h>
84 #include <linux/wireless.h>
85 #include <linux/nsproxy.h>
86 #include <linux/magic.h>
87 #include <linux/slab.h>
88 #include <linux/xattr.h>
89 #include <linux/nospec.h>
90 #include <linux/indirect_call_wrapper.h>
92 #include <linux/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/termios.h>
106 #include <linux/sockios.h>
107 #include <net/busy_poll.h>
108 #include <linux/errqueue.h>
109 #include <linux/ptp_clock_kernel.h>
110 #include <trace/events/sock.h>
112 #ifdef CONFIG_NET_RX_BUSY_POLL
113 unsigned int sysctl_net_busy_read __read_mostly;
114 unsigned int sysctl_net_busy_poll __read_mostly;
117 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
118 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 static int sock_close(struct inode *inode, struct file *file);
122 static __poll_t sock_poll(struct file *file,
123 struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 static long compat_sock_ioctl(struct file *file,
127 unsigned int cmd, unsigned long arg);
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
131 struct pipe_inode_info *pipe, size_t len,
133 static void sock_splice_eof(struct file *file);
135 #ifdef CONFIG_PROC_FS
136 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138 struct socket *sock = f->private_data;
140 if (sock->ops->show_fdinfo)
141 sock->ops->show_fdinfo(m, sock);
144 #define sock_show_fdinfo NULL
148 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
149 * in the operation structures but are done directly via the socketcall() multiplexor.
152 static const struct file_operations socket_file_ops = {
153 .owner = THIS_MODULE,
155 .read_iter = sock_read_iter,
156 .write_iter = sock_write_iter,
158 .unlocked_ioctl = sock_ioctl,
160 .compat_ioctl = compat_sock_ioctl,
163 .release = sock_close,
164 .fasync = sock_fasync,
165 .splice_write = splice_to_socket,
166 .splice_read = sock_splice_read,
167 .splice_eof = sock_splice_eof,
168 .show_fdinfo = sock_show_fdinfo,
171 static const char * const pf_family_names[] = {
172 [PF_UNSPEC] = "PF_UNSPEC",
173 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
174 [PF_INET] = "PF_INET",
175 [PF_AX25] = "PF_AX25",
177 [PF_APPLETALK] = "PF_APPLETALK",
178 [PF_NETROM] = "PF_NETROM",
179 [PF_BRIDGE] = "PF_BRIDGE",
180 [PF_ATMPVC] = "PF_ATMPVC",
182 [PF_INET6] = "PF_INET6",
183 [PF_ROSE] = "PF_ROSE",
184 [PF_DECnet] = "PF_DECnet",
185 [PF_NETBEUI] = "PF_NETBEUI",
186 [PF_SECURITY] = "PF_SECURITY",
188 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
189 [PF_PACKET] = "PF_PACKET",
191 [PF_ECONET] = "PF_ECONET",
192 [PF_ATMSVC] = "PF_ATMSVC",
195 [PF_IRDA] = "PF_IRDA",
196 [PF_PPPOX] = "PF_PPPOX",
197 [PF_WANPIPE] = "PF_WANPIPE",
200 [PF_MPLS] = "PF_MPLS",
202 [PF_TIPC] = "PF_TIPC",
203 [PF_BLUETOOTH] = "PF_BLUETOOTH",
204 [PF_IUCV] = "PF_IUCV",
205 [PF_RXRPC] = "PF_RXRPC",
206 [PF_ISDN] = "PF_ISDN",
207 [PF_PHONET] = "PF_PHONET",
208 [PF_IEEE802154] = "PF_IEEE802154",
209 [PF_CAIF] = "PF_CAIF",
212 [PF_VSOCK] = "PF_VSOCK",
214 [PF_QIPCRTR] = "PF_QIPCRTR",
217 [PF_MCTP] = "PF_MCTP",
221 * The protocol list. Each protocol is registered in here.
224 static DEFINE_SPINLOCK(net_family_lock);
225 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
229 * Move socket addresses back and forth across the kernel/user
230 * divide and look after the messy bits.
234 * move_addr_to_kernel - copy a socket address into kernel space
235 * @uaddr: Address in user space
236 * @kaddr: Address in kernel space
237 * @ulen: Length in user space
239 * The address is copied into kernel space. If the provided address is
240 * too long an error code of -EINVAL is returned. If the copy gives
241 * invalid addresses -EFAULT is returned. On a success 0 is returned.
244 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
246 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
250 if (copy_from_user(kaddr, uaddr, ulen))
252 return audit_sockaddr(ulen, kaddr);
256 * move_addr_to_user - copy an address to user space
257 * @kaddr: kernel space address
258 * @klen: length of address in kernel
259 * @uaddr: user space address
260 * @ulen: pointer to user length field
262 * The value pointed to by ulen on entry is the buffer length available.
263 * This is overwritten with the buffer space used. -EINVAL is returned
264 * if an overlong buffer is specified or a negative buffer size. -EFAULT
265 * is returned if either the buffer or the length field are not
267 * After copying the data up to the limit the user specifies, the true
268 * length of the data is written over the length limit the user
269 * specified. Zero is returned for a success.
272 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
273 void __user *uaddr, int __user *ulen)
278 BUG_ON(klen > sizeof(struct sockaddr_storage));
279 err = get_user(len, ulen);
287 if (audit_sockaddr(klen, kaddr))
289 if (copy_to_user(uaddr, kaddr, len))
293 * "fromlen shall refer to the value before truncation.."
296 return __put_user(klen, ulen);
299 static struct kmem_cache *sock_inode_cachep __ro_after_init;
301 static struct inode *sock_alloc_inode(struct super_block *sb)
303 struct socket_alloc *ei;
305 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
308 init_waitqueue_head(&ei->socket.wq.wait);
309 ei->socket.wq.fasync_list = NULL;
310 ei->socket.wq.flags = 0;
312 ei->socket.state = SS_UNCONNECTED;
313 ei->socket.flags = 0;
314 ei->socket.ops = NULL;
315 ei->socket.sk = NULL;
316 ei->socket.file = NULL;
318 return &ei->vfs_inode;
321 static void sock_free_inode(struct inode *inode)
323 struct socket_alloc *ei;
325 ei = container_of(inode, struct socket_alloc, vfs_inode);
326 kmem_cache_free(sock_inode_cachep, ei);
329 static void init_once(void *foo)
331 struct socket_alloc *ei = (struct socket_alloc *)foo;
333 inode_init_once(&ei->vfs_inode);
336 static void init_inodecache(void)
338 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
339 sizeof(struct socket_alloc),
341 (SLAB_HWCACHE_ALIGN |
342 SLAB_RECLAIM_ACCOUNT |
343 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
345 BUG_ON(sock_inode_cachep == NULL);
348 static const struct super_operations sockfs_ops = {
349 .alloc_inode = sock_alloc_inode,
350 .free_inode = sock_free_inode,
351 .statfs = simple_statfs,
355 * sockfs_dname() is called from d_path().
357 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
359 return dynamic_dname(buffer, buflen, "socket:[%lu]",
360 d_inode(dentry)->i_ino);
363 static const struct dentry_operations sockfs_dentry_operations = {
364 .d_dname = sockfs_dname,
367 static int sockfs_xattr_get(const struct xattr_handler *handler,
368 struct dentry *dentry, struct inode *inode,
369 const char *suffix, void *value, size_t size)
372 if (dentry->d_name.len + 1 > size)
374 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
376 return dentry->d_name.len + 1;
379 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
380 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
381 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
383 static const struct xattr_handler sockfs_xattr_handler = {
384 .name = XATTR_NAME_SOCKPROTONAME,
385 .get = sockfs_xattr_get,
388 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
389 struct mnt_idmap *idmap,
390 struct dentry *dentry, struct inode *inode,
391 const char *suffix, const void *value,
392 size_t size, int flags)
394 /* Handled by LSM. */
398 static const struct xattr_handler sockfs_security_xattr_handler = {
399 .prefix = XATTR_SECURITY_PREFIX,
400 .set = sockfs_security_xattr_set,
403 static const struct xattr_handler *sockfs_xattr_handlers[] = {
404 &sockfs_xattr_handler,
405 &sockfs_security_xattr_handler,
409 static int sockfs_init_fs_context(struct fs_context *fc)
411 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
414 ctx->ops = &sockfs_ops;
415 ctx->dops = &sockfs_dentry_operations;
416 ctx->xattr = sockfs_xattr_handlers;
420 static struct vfsmount *sock_mnt __read_mostly;
422 static struct file_system_type sock_fs_type = {
424 .init_fs_context = sockfs_init_fs_context,
425 .kill_sb = kill_anon_super,
429 * Obtains the first available file descriptor and sets it up for use.
431 * These functions create file structures and maps them to fd space
432 * of the current process. On success it returns file descriptor
433 * and file struct implicitly stored in sock->file.
434 * Note that another thread may close file descriptor before we return
435 * from this function. We use the fact that now we do not refer
436 * to socket after mapping. If one day we will need it, this
437 * function will increment ref. count on file by 1.
439 * In any case returned fd MAY BE not valid!
440 * This race condition is unavoidable
441 * with shared fd spaces, we cannot solve it inside kernel,
442 * but we take care of internal coherence yet.
446 * sock_alloc_file - Bind a &socket to a &file
448 * @flags: file status flags
449 * @dname: protocol name
451 * Returns the &file bound with @sock, implicitly storing it
452 * in sock->file. If dname is %NULL, sets to "".
454 * On failure @sock is released, and an ERR pointer is returned.
456 * This function uses GFP_KERNEL internally.
459 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
464 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
466 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
467 O_RDWR | (flags & O_NONBLOCK),
475 file->private_data = sock;
476 stream_open(SOCK_INODE(sock), file);
479 EXPORT_SYMBOL(sock_alloc_file);
481 static int sock_map_fd(struct socket *sock, int flags)
483 struct file *newfile;
484 int fd = get_unused_fd_flags(flags);
485 if (unlikely(fd < 0)) {
490 newfile = sock_alloc_file(sock, flags, NULL);
491 if (!IS_ERR(newfile)) {
492 fd_install(fd, newfile);
497 return PTR_ERR(newfile);
501 * sock_from_file - Return the &socket bounded to @file.
504 * On failure returns %NULL.
507 struct socket *sock_from_file(struct file *file)
509 if (file->f_op == &socket_file_ops)
510 return file->private_data; /* set in sock_alloc_file */
514 EXPORT_SYMBOL(sock_from_file);
517 * sockfd_lookup - Go from a file number to its socket slot
519 * @err: pointer to an error code return
521 * The file handle passed in is locked and the socket it is bound
522 * to is returned. If an error occurs the err pointer is overwritten
523 * with a negative errno code and NULL is returned. The function checks
524 * for both invalid handles and passing a handle which is not a socket.
526 * On a success the socket object pointer is returned.
529 struct socket *sockfd_lookup(int fd, int *err)
540 sock = sock_from_file(file);
547 EXPORT_SYMBOL(sockfd_lookup);
549 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
551 struct fd f = fdget(fd);
556 sock = sock_from_file(f.file);
558 *fput_needed = f.flags & FDPUT_FPUT;
567 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
573 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
583 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
588 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
595 static int sockfs_setattr(struct mnt_idmap *idmap,
596 struct dentry *dentry, struct iattr *iattr)
598 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
600 if (!err && (iattr->ia_valid & ATTR_UID)) {
601 struct socket *sock = SOCKET_I(d_inode(dentry));
604 sock->sk->sk_uid = iattr->ia_uid;
612 static const struct inode_operations sockfs_inode_ops = {
613 .listxattr = sockfs_listxattr,
614 .setattr = sockfs_setattr,
618 * sock_alloc - allocate a socket
620 * Allocate a new inode and socket object. The two are bound together
621 * and initialised. The socket is then returned. If we are out of inodes
622 * NULL is returned. This functions uses GFP_KERNEL internally.
625 struct socket *sock_alloc(void)
630 inode = new_inode_pseudo(sock_mnt->mnt_sb);
634 sock = SOCKET_I(inode);
636 inode->i_ino = get_next_ino();
637 inode->i_mode = S_IFSOCK | S_IRWXUGO;
638 inode->i_uid = current_fsuid();
639 inode->i_gid = current_fsgid();
640 inode->i_op = &sockfs_inode_ops;
644 EXPORT_SYMBOL(sock_alloc);
646 static void __sock_release(struct socket *sock, struct inode *inode)
649 struct module *owner = sock->ops->owner;
653 sock->ops->release(sock);
661 if (sock->wq.fasync_list)
662 pr_err("%s: fasync list not empty!\n", __func__);
665 iput(SOCK_INODE(sock));
672 * sock_release - close a socket
673 * @sock: socket to close
675 * The socket is released from the protocol stack if it has a release
676 * callback, and the inode is then released if the socket is bound to
677 * an inode not a file.
679 void sock_release(struct socket *sock)
681 __sock_release(sock, NULL);
683 EXPORT_SYMBOL(sock_release);
685 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
687 u8 flags = *tx_flags;
689 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
690 flags |= SKBTX_HW_TSTAMP;
692 /* PTP hardware clocks can provide a free running cycle counter
693 * as a time base for virtual clocks. Tell driver to use the
694 * free running cycle counter for timestamp if socket is bound
697 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
698 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
701 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
702 flags |= SKBTX_SW_TSTAMP;
704 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
705 flags |= SKBTX_SCHED_TSTAMP;
709 EXPORT_SYMBOL(__sock_tx_timestamp);
711 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
713 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
716 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
719 trace_sock_send_length(sk, ret, 0);
722 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
724 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
725 inet_sendmsg, sock, msg,
727 BUG_ON(ret == -EIOCBQUEUED);
729 if (trace_sock_send_length_enabled())
730 call_trace_sock_send_length(sock->sk, ret, 0);
735 * sock_sendmsg - send a message through @sock
737 * @msg: message to send
739 * Sends @msg through @sock, passing through LSM.
740 * Returns the number of bytes sent, or an error code.
742 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
744 int err = security_socket_sendmsg(sock, msg,
747 return err ?: sock_sendmsg_nosec(sock, msg);
749 EXPORT_SYMBOL(sock_sendmsg);
752 * kernel_sendmsg - send a message through @sock (kernel-space)
754 * @msg: message header
756 * @num: vec array length
757 * @size: total message data size
759 * Builds the message data with @vec and sends it through @sock.
760 * Returns the number of bytes sent, or an error code.
763 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
764 struct kvec *vec, size_t num, size_t size)
766 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
767 return sock_sendmsg(sock, msg);
769 EXPORT_SYMBOL(kernel_sendmsg);
772 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
774 * @msg: message header
775 * @vec: output s/g array
776 * @num: output s/g array length
777 * @size: total message data size
779 * Builds the message data with @vec and sends it through @sock.
780 * Returns the number of bytes sent, or an error code.
781 * Caller must hold @sk.
784 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
785 struct kvec *vec, size_t num, size_t size)
787 struct socket *sock = sk->sk_socket;
789 if (!sock->ops->sendmsg_locked)
790 return sock_no_sendmsg_locked(sk, msg, size);
792 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
794 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
796 EXPORT_SYMBOL(kernel_sendmsg_locked);
798 static bool skb_is_err_queue(const struct sk_buff *skb)
800 /* pkt_type of skbs enqueued on the error queue are set to
801 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
802 * in recvmsg, since skbs received on a local socket will never
803 * have a pkt_type of PACKET_OUTGOING.
805 return skb->pkt_type == PACKET_OUTGOING;
808 /* On transmit, software and hardware timestamps are returned independently.
809 * As the two skb clones share the hardware timestamp, which may be updated
810 * before the software timestamp is received, a hardware TX timestamp may be
811 * returned only if there is no software TX timestamp. Ignore false software
812 * timestamps, which may be made in the __sock_recv_timestamp() call when the
813 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
814 * hardware timestamp.
816 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
818 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
821 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
823 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC;
824 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
825 struct net_device *orig_dev;
829 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
831 *if_index = orig_dev->ifindex;
832 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
834 hwtstamp = shhwtstamps->hwtstamp;
841 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
844 struct scm_ts_pktinfo ts_pktinfo;
845 struct net_device *orig_dev;
847 if (!skb_mac_header_was_set(skb))
850 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
854 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 if_index = orig_dev->ifindex;
859 ts_pktinfo.if_index = if_index;
861 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
862 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
863 sizeof(ts_pktinfo), &ts_pktinfo);
867 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
869 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
872 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
873 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
874 struct scm_timestamping_internal tss;
876 int empty = 1, false_tstamp = 0;
877 struct skb_shared_hwtstamps *shhwtstamps =
882 /* Race occurred between timestamp enabling and packet
883 receiving. Fill in the current time for now. */
884 if (need_software_tstamp && skb->tstamp == 0) {
885 __net_timestamp(skb);
889 if (need_software_tstamp) {
890 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
892 struct __kernel_sock_timeval tv;
894 skb_get_new_timestamp(skb, &tv);
895 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
898 struct __kernel_old_timeval tv;
900 skb_get_timestamp(skb, &tv);
901 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
906 struct __kernel_timespec ts;
908 skb_get_new_timestampns(skb, &ts);
909 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
912 struct __kernel_old_timespec ts;
914 skb_get_timestampns(skb, &ts);
915 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
921 memset(&tss, 0, sizeof(tss));
922 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
923 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
926 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
927 !skb_is_swtx_tstamp(skb, false_tstamp)) {
929 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
930 hwtstamp = get_timestamp(sk, skb, &if_index);
932 hwtstamp = shhwtstamps->hwtstamp;
934 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
935 hwtstamp = ptp_convert_timestamp(&hwtstamp,
938 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
941 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
942 !skb_is_err_queue(skb))
943 put_ts_pktinfo(msg, skb, if_index);
947 if (sock_flag(sk, SOCK_TSTAMP_NEW))
948 put_cmsg_scm_timestamping64(msg, &tss);
950 put_cmsg_scm_timestamping(msg, &tss);
952 if (skb_is_err_queue(skb) && skb->len &&
953 SKB_EXT_ERR(skb)->opt_stats)
954 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
955 skb->len, skb->data);
958 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
960 #ifdef CONFIG_WIRELESS
961 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
966 if (!sock_flag(sk, SOCK_WIFI_STATUS))
968 if (!skb->wifi_acked_valid)
971 ack = skb->wifi_acked;
973 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
975 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
978 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
981 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
982 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
983 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
986 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
989 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
990 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
991 __u32 mark = skb->mark;
993 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
997 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1000 sock_recv_timestamp(msg, sk, skb);
1001 sock_recv_drops(msg, sk, skb);
1002 sock_recv_mark(msg, sk, skb);
1004 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1006 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1008 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1011 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1013 trace_sock_recv_length(sk, ret, flags);
1016 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1019 int ret = INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
1020 inet_recvmsg, sock, msg,
1021 msg_data_left(msg), flags);
1022 if (trace_sock_recv_length_enabled())
1023 call_trace_sock_recv_length(sock->sk, ret, flags);
1028 * sock_recvmsg - receive a message from @sock
1030 * @msg: message to receive
1031 * @flags: message flags
1033 * Receives @msg from @sock, passing through LSM. Returns the total number
1034 * of bytes received, or an error.
1036 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1038 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1040 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1042 EXPORT_SYMBOL(sock_recvmsg);
1045 * kernel_recvmsg - Receive a message from a socket (kernel space)
1046 * @sock: The socket to receive the message from
1047 * @msg: Received message
1048 * @vec: Input s/g array for message data
1049 * @num: Size of input s/g array
1050 * @size: Number of bytes to read
1051 * @flags: Message flags (MSG_DONTWAIT, etc...)
1053 * On return the msg structure contains the scatter/gather array passed in the
1054 * vec argument. The array is modified so that it consists of the unfilled
1055 * portion of the original array.
1057 * The returned value is the total number of bytes received, or an error.
1060 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1061 struct kvec *vec, size_t num, size_t size, int flags)
1063 msg->msg_control_is_user = false;
1064 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1065 return sock_recvmsg(sock, msg, flags);
1067 EXPORT_SYMBOL(kernel_recvmsg);
1069 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1070 struct pipe_inode_info *pipe, size_t len,
1073 struct socket *sock = file->private_data;
1075 if (unlikely(!sock->ops->splice_read))
1076 return generic_file_splice_read(file, ppos, pipe, len, flags);
1078 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1081 static void sock_splice_eof(struct file *file)
1083 struct socket *sock = file->private_data;
1085 if (sock->ops->splice_eof)
1086 sock->ops->splice_eof(sock);
1089 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1091 struct file *file = iocb->ki_filp;
1092 struct socket *sock = file->private_data;
1093 struct msghdr msg = {.msg_iter = *to,
1097 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1098 msg.msg_flags = MSG_DONTWAIT;
1100 if (iocb->ki_pos != 0)
1103 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1106 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1111 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1113 struct file *file = iocb->ki_filp;
1114 struct socket *sock = file->private_data;
1115 struct msghdr msg = {.msg_iter = *from,
1119 if (iocb->ki_pos != 0)
1122 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1123 msg.msg_flags = MSG_DONTWAIT;
1125 if (sock->type == SOCK_SEQPACKET)
1126 msg.msg_flags |= MSG_EOR;
1128 res = sock_sendmsg(sock, &msg);
1129 *from = msg.msg_iter;
1134 * Atomic setting of ioctl hooks to avoid race
1135 * with module unload.
1138 static DEFINE_MUTEX(br_ioctl_mutex);
1139 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1140 unsigned int cmd, struct ifreq *ifr,
1143 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1144 unsigned int cmd, struct ifreq *ifr,
1147 mutex_lock(&br_ioctl_mutex);
1148 br_ioctl_hook = hook;
1149 mutex_unlock(&br_ioctl_mutex);
1151 EXPORT_SYMBOL(brioctl_set);
1153 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1154 struct ifreq *ifr, void __user *uarg)
1159 request_module("bridge");
1161 mutex_lock(&br_ioctl_mutex);
1163 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1164 mutex_unlock(&br_ioctl_mutex);
1169 static DEFINE_MUTEX(vlan_ioctl_mutex);
1170 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1172 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1174 mutex_lock(&vlan_ioctl_mutex);
1175 vlan_ioctl_hook = hook;
1176 mutex_unlock(&vlan_ioctl_mutex);
1178 EXPORT_SYMBOL(vlan_ioctl_set);
1180 static long sock_do_ioctl(struct net *net, struct socket *sock,
1181 unsigned int cmd, unsigned long arg)
1186 void __user *argp = (void __user *)arg;
1189 err = sock->ops->ioctl(sock, cmd, arg);
1192 * If this ioctl is unknown try to hand it down
1193 * to the NIC driver.
1195 if (err != -ENOIOCTLCMD)
1198 if (!is_socket_ioctl_cmd(cmd))
1201 if (get_user_ifreq(&ifr, &data, argp))
1203 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1204 if (!err && need_copyout)
1205 if (put_user_ifreq(&ifr, argp))
1212 * With an ioctl, arg may well be a user mode pointer, but we don't know
1213 * what to do with it - that's up to the protocol still.
1216 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1218 struct socket *sock;
1220 void __user *argp = (void __user *)arg;
1224 sock = file->private_data;
1227 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1231 if (get_user_ifreq(&ifr, &data, argp))
1233 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1234 if (!err && need_copyout)
1235 if (put_user_ifreq(&ifr, argp))
1238 #ifdef CONFIG_WEXT_CORE
1239 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1240 err = wext_handle_ioctl(net, cmd, argp);
1247 if (get_user(pid, (int __user *)argp))
1249 err = f_setown(sock->file, pid, 1);
1253 err = put_user(f_getown(sock->file),
1254 (int __user *)argp);
1260 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1265 if (!vlan_ioctl_hook)
1266 request_module("8021q");
1268 mutex_lock(&vlan_ioctl_mutex);
1269 if (vlan_ioctl_hook)
1270 err = vlan_ioctl_hook(net, argp);
1271 mutex_unlock(&vlan_ioctl_mutex);
1275 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1278 err = open_related_ns(&net->ns, get_net_ns);
1280 case SIOCGSTAMP_OLD:
1281 case SIOCGSTAMPNS_OLD:
1282 if (!sock->ops->gettstamp) {
1286 err = sock->ops->gettstamp(sock, argp,
1287 cmd == SIOCGSTAMP_OLD,
1288 !IS_ENABLED(CONFIG_64BIT));
1290 case SIOCGSTAMP_NEW:
1291 case SIOCGSTAMPNS_NEW:
1292 if (!sock->ops->gettstamp) {
1296 err = sock->ops->gettstamp(sock, argp,
1297 cmd == SIOCGSTAMP_NEW,
1302 err = dev_ifconf(net, argp);
1306 err = sock_do_ioctl(net, sock, cmd, arg);
1313 * sock_create_lite - creates a socket
1314 * @family: protocol family (AF_INET, ...)
1315 * @type: communication type (SOCK_STREAM, ...)
1316 * @protocol: protocol (0, ...)
1319 * Creates a new socket and assigns it to @res, passing through LSM.
1320 * The new socket initialization is not complete, see kernel_accept().
1321 * Returns 0 or an error. On failure @res is set to %NULL.
1322 * This function internally uses GFP_KERNEL.
1325 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1328 struct socket *sock = NULL;
1330 err = security_socket_create(family, type, protocol, 1);
1334 sock = sock_alloc();
1341 err = security_socket_post_create(sock, family, type, protocol, 1);
1353 EXPORT_SYMBOL(sock_create_lite);
1355 /* No kernel lock held - perfect */
1356 static __poll_t sock_poll(struct file *file, poll_table *wait)
1358 struct socket *sock = file->private_data;
1359 __poll_t events = poll_requested_events(wait), flag = 0;
1361 if (!sock->ops->poll)
1364 if (sk_can_busy_loop(sock->sk)) {
1365 /* poll once if requested by the syscall */
1366 if (events & POLL_BUSY_LOOP)
1367 sk_busy_loop(sock->sk, 1);
1369 /* if this socket can poll_ll, tell the system call */
1370 flag = POLL_BUSY_LOOP;
1373 return sock->ops->poll(file, sock, wait) | flag;
1376 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1378 struct socket *sock = file->private_data;
1380 return sock->ops->mmap(file, sock, vma);
1383 static int sock_close(struct inode *inode, struct file *filp)
1385 __sock_release(SOCKET_I(inode), inode);
1390 * Update the socket async list
1392 * Fasync_list locking strategy.
1394 * 1. fasync_list is modified only under process context socket lock
1395 * i.e. under semaphore.
1396 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1397 * or under socket lock
1400 static int sock_fasync(int fd, struct file *filp, int on)
1402 struct socket *sock = filp->private_data;
1403 struct sock *sk = sock->sk;
1404 struct socket_wq *wq = &sock->wq;
1410 fasync_helper(fd, filp, on, &wq->fasync_list);
1412 if (!wq->fasync_list)
1413 sock_reset_flag(sk, SOCK_FASYNC);
1415 sock_set_flag(sk, SOCK_FASYNC);
1421 /* This function may be called only under rcu_lock */
1423 int sock_wake_async(struct socket_wq *wq, int how, int band)
1425 if (!wq || !wq->fasync_list)
1429 case SOCK_WAKE_WAITD:
1430 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1433 case SOCK_WAKE_SPACE:
1434 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1439 kill_fasync(&wq->fasync_list, SIGIO, band);
1442 kill_fasync(&wq->fasync_list, SIGURG, band);
1447 EXPORT_SYMBOL(sock_wake_async);
1450 * __sock_create - creates a socket
1451 * @net: net namespace
1452 * @family: protocol family (AF_INET, ...)
1453 * @type: communication type (SOCK_STREAM, ...)
1454 * @protocol: protocol (0, ...)
1456 * @kern: boolean for kernel space sockets
1458 * Creates a new socket and assigns it to @res, passing through LSM.
1459 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1460 * be set to true if the socket resides in kernel space.
1461 * This function internally uses GFP_KERNEL.
1464 int __sock_create(struct net *net, int family, int type, int protocol,
1465 struct socket **res, int kern)
1468 struct socket *sock;
1469 const struct net_proto_family *pf;
1472 * Check protocol is in range
1474 if (family < 0 || family >= NPROTO)
1475 return -EAFNOSUPPORT;
1476 if (type < 0 || type >= SOCK_MAX)
1481 This uglymoron is moved from INET layer to here to avoid
1482 deadlock in module load.
1484 if (family == PF_INET && type == SOCK_PACKET) {
1485 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1490 err = security_socket_create(family, type, protocol, kern);
1495 * Allocate the socket and allow the family to set things up. if
1496 * the protocol is 0, the family is instructed to select an appropriate
1499 sock = sock_alloc();
1501 net_warn_ratelimited("socket: no more sockets\n");
1502 return -ENFILE; /* Not exactly a match, but its the
1503 closest posix thing */
1508 #ifdef CONFIG_MODULES
1509 /* Attempt to load a protocol module if the find failed.
1511 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1512 * requested real, full-featured networking support upon configuration.
1513 * Otherwise module support will break!
1515 if (rcu_access_pointer(net_families[family]) == NULL)
1516 request_module("net-pf-%d", family);
1520 pf = rcu_dereference(net_families[family]);
1521 err = -EAFNOSUPPORT;
1526 * We will call the ->create function, that possibly is in a loadable
1527 * module, so we have to bump that loadable module refcnt first.
1529 if (!try_module_get(pf->owner))
1532 /* Now protected by module ref count */
1535 err = pf->create(net, sock, protocol, kern);
1537 goto out_module_put;
1540 * Now to bump the refcnt of the [loadable] module that owns this
1541 * socket at sock_release time we decrement its refcnt.
1543 if (!try_module_get(sock->ops->owner))
1544 goto out_module_busy;
1547 * Now that we're done with the ->create function, the [loadable]
1548 * module can have its refcnt decremented
1550 module_put(pf->owner);
1551 err = security_socket_post_create(sock, family, type, protocol, kern);
1553 goto out_sock_release;
1559 err = -EAFNOSUPPORT;
1562 module_put(pf->owner);
1569 goto out_sock_release;
1571 EXPORT_SYMBOL(__sock_create);
1574 * sock_create - creates a socket
1575 * @family: protocol family (AF_INET, ...)
1576 * @type: communication type (SOCK_STREAM, ...)
1577 * @protocol: protocol (0, ...)
1580 * A wrapper around __sock_create().
1581 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1584 int sock_create(int family, int type, int protocol, struct socket **res)
1586 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1588 EXPORT_SYMBOL(sock_create);
1591 * sock_create_kern - creates a socket (kernel space)
1592 * @net: net namespace
1593 * @family: protocol family (AF_INET, ...)
1594 * @type: communication type (SOCK_STREAM, ...)
1595 * @protocol: protocol (0, ...)
1598 * A wrapper around __sock_create().
1599 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1602 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1604 return __sock_create(net, family, type, protocol, res, 1);
1606 EXPORT_SYMBOL(sock_create_kern);
1608 static struct socket *__sys_socket_create(int family, int type, int protocol)
1610 struct socket *sock;
1613 /* Check the SOCK_* constants for consistency. */
1614 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1615 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1616 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1617 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1619 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1620 return ERR_PTR(-EINVAL);
1621 type &= SOCK_TYPE_MASK;
1623 retval = sock_create(family, type, protocol, &sock);
1625 return ERR_PTR(retval);
1630 struct file *__sys_socket_file(int family, int type, int protocol)
1632 struct socket *sock;
1635 sock = __sys_socket_create(family, type, protocol);
1637 return ERR_CAST(sock);
1639 flags = type & ~SOCK_TYPE_MASK;
1640 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1641 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1643 return sock_alloc_file(sock, flags, NULL);
1646 int __sys_socket(int family, int type, int protocol)
1648 struct socket *sock;
1651 sock = __sys_socket_create(family, type, protocol);
1653 return PTR_ERR(sock);
1655 flags = type & ~SOCK_TYPE_MASK;
1656 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1657 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1659 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1662 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1664 return __sys_socket(family, type, protocol);
1668 * Create a pair of connected sockets.
1671 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1673 struct socket *sock1, *sock2;
1675 struct file *newfile1, *newfile2;
1678 flags = type & ~SOCK_TYPE_MASK;
1679 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1681 type &= SOCK_TYPE_MASK;
1683 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1684 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1687 * reserve descriptors and make sure we won't fail
1688 * to return them to userland.
1690 fd1 = get_unused_fd_flags(flags);
1691 if (unlikely(fd1 < 0))
1694 fd2 = get_unused_fd_flags(flags);
1695 if (unlikely(fd2 < 0)) {
1700 err = put_user(fd1, &usockvec[0]);
1704 err = put_user(fd2, &usockvec[1]);
1709 * Obtain the first socket and check if the underlying protocol
1710 * supports the socketpair call.
1713 err = sock_create(family, type, protocol, &sock1);
1714 if (unlikely(err < 0))
1717 err = sock_create(family, type, protocol, &sock2);
1718 if (unlikely(err < 0)) {
1719 sock_release(sock1);
1723 err = security_socket_socketpair(sock1, sock2);
1724 if (unlikely(err)) {
1725 sock_release(sock2);
1726 sock_release(sock1);
1730 err = sock1->ops->socketpair(sock1, sock2);
1731 if (unlikely(err < 0)) {
1732 sock_release(sock2);
1733 sock_release(sock1);
1737 newfile1 = sock_alloc_file(sock1, flags, NULL);
1738 if (IS_ERR(newfile1)) {
1739 err = PTR_ERR(newfile1);
1740 sock_release(sock2);
1744 newfile2 = sock_alloc_file(sock2, flags, NULL);
1745 if (IS_ERR(newfile2)) {
1746 err = PTR_ERR(newfile2);
1751 audit_fd_pair(fd1, fd2);
1753 fd_install(fd1, newfile1);
1754 fd_install(fd2, newfile2);
1763 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1764 int __user *, usockvec)
1766 return __sys_socketpair(family, type, protocol, usockvec);
1770 * Bind a name to a socket. Nothing much to do here since it's
1771 * the protocol's responsibility to handle the local address.
1773 * We move the socket address to kernel space before we call
1774 * the protocol layer (having also checked the address is ok).
1777 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1779 struct socket *sock;
1780 struct sockaddr_storage address;
1781 int err, fput_needed;
1783 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1785 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1787 err = security_socket_bind(sock,
1788 (struct sockaddr *)&address,
1791 err = sock->ops->bind(sock,
1795 fput_light(sock->file, fput_needed);
1800 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1802 return __sys_bind(fd, umyaddr, addrlen);
1806 * Perform a listen. Basically, we allow the protocol to do anything
1807 * necessary for a listen, and if that works, we mark the socket as
1808 * ready for listening.
1811 int __sys_listen(int fd, int backlog)
1813 struct socket *sock;
1814 int err, fput_needed;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1819 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1820 if ((unsigned int)backlog > somaxconn)
1821 backlog = somaxconn;
1823 err = security_socket_listen(sock, backlog);
1825 err = sock->ops->listen(sock, backlog);
1827 fput_light(sock->file, fput_needed);
1832 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1834 return __sys_listen(fd, backlog);
1837 struct file *do_accept(struct file *file, unsigned file_flags,
1838 struct sockaddr __user *upeer_sockaddr,
1839 int __user *upeer_addrlen, int flags)
1841 struct socket *sock, *newsock;
1842 struct file *newfile;
1844 struct sockaddr_storage address;
1846 sock = sock_from_file(file);
1848 return ERR_PTR(-ENOTSOCK);
1850 newsock = sock_alloc();
1852 return ERR_PTR(-ENFILE);
1854 newsock->type = sock->type;
1855 newsock->ops = sock->ops;
1858 * We don't need try_module_get here, as the listening socket (sock)
1859 * has the protocol module (sock->ops->owner) held.
1861 __module_get(newsock->ops->owner);
1863 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1864 if (IS_ERR(newfile))
1867 err = security_socket_accept(sock, newsock);
1871 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1876 if (upeer_sockaddr) {
1877 len = newsock->ops->getname(newsock,
1878 (struct sockaddr *)&address, 2);
1880 err = -ECONNABORTED;
1883 err = move_addr_to_user(&address,
1884 len, upeer_sockaddr, upeer_addrlen);
1889 /* File flags are not inherited via accept() unlike another OSes. */
1893 return ERR_PTR(err);
1896 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1897 int __user *upeer_addrlen, int flags)
1899 struct file *newfile;
1902 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1905 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1906 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1908 newfd = get_unused_fd_flags(flags);
1909 if (unlikely(newfd < 0))
1912 newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen,
1914 if (IS_ERR(newfile)) {
1915 put_unused_fd(newfd);
1916 return PTR_ERR(newfile);
1918 fd_install(newfd, newfile);
1923 * For accept, we attempt to create a new socket, set up the link
1924 * with the client, wake up the client, then return the new
1925 * connected fd. We collect the address of the connector in kernel
1926 * space and move it to user at the very end. This is unclean because
1927 * we open the socket then return an error.
1929 * 1003.1g adds the ability to recvmsg() to query connection pending
1930 * status to recvmsg. We need to add that support in a way thats
1931 * clean when we restructure accept also.
1934 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1935 int __user *upeer_addrlen, int flags)
1942 ret = __sys_accept4_file(f.file, upeer_sockaddr,
1943 upeer_addrlen, flags);
1950 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1951 int __user *, upeer_addrlen, int, flags)
1953 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1956 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1957 int __user *, upeer_addrlen)
1959 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1963 * Attempt to connect to a socket with the server address. The address
1964 * is in user space so we verify it is OK and move it to kernel space.
1966 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1969 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1970 * other SEQPACKET protocols that take time to connect() as it doesn't
1971 * include the -EINPROGRESS status for such sockets.
1974 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1975 int addrlen, int file_flags)
1977 struct socket *sock;
1980 sock = sock_from_file(file);
1987 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1991 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1992 sock->file->f_flags | file_flags);
1997 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2004 struct sockaddr_storage address;
2006 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2008 ret = __sys_connect_file(f.file, &address, addrlen, 0);
2015 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2018 return __sys_connect(fd, uservaddr, addrlen);
2022 * Get the local address ('name') of a socket object. Move the obtained
2023 * name to user space.
2026 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2027 int __user *usockaddr_len)
2029 struct socket *sock;
2030 struct sockaddr_storage address;
2031 int err, fput_needed;
2033 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2037 err = security_socket_getsockname(sock);
2041 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
2044 /* "err" is actually length in this case */
2045 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2048 fput_light(sock->file, fput_needed);
2053 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2054 int __user *, usockaddr_len)
2056 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2060 * Get the remote address ('name') of a socket object. Move the obtained
2061 * name to user space.
2064 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2065 int __user *usockaddr_len)
2067 struct socket *sock;
2068 struct sockaddr_storage address;
2069 int err, fput_needed;
2071 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2073 err = security_socket_getpeername(sock);
2075 fput_light(sock->file, fput_needed);
2079 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
2081 /* "err" is actually length in this case */
2082 err = move_addr_to_user(&address, err, usockaddr,
2084 fput_light(sock->file, fput_needed);
2089 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2090 int __user *, usockaddr_len)
2092 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2096 * Send a datagram to a given address. We move the address into kernel
2097 * space and check the user space data area is readable before invoking
2100 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2101 struct sockaddr __user *addr, int addr_len)
2103 struct socket *sock;
2104 struct sockaddr_storage address;
2110 err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter);
2113 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2117 msg.msg_name = NULL;
2118 msg.msg_control = NULL;
2119 msg.msg_controllen = 0;
2120 msg.msg_namelen = 0;
2121 msg.msg_ubuf = NULL;
2123 err = move_addr_to_kernel(addr, addr_len, &address);
2126 msg.msg_name = (struct sockaddr *)&address;
2127 msg.msg_namelen = addr_len;
2129 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2130 if (sock->file->f_flags & O_NONBLOCK)
2131 flags |= MSG_DONTWAIT;
2132 msg.msg_flags = flags;
2133 err = sock_sendmsg(sock, &msg);
2136 fput_light(sock->file, fput_needed);
2141 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2142 unsigned int, flags, struct sockaddr __user *, addr,
2145 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2149 * Send a datagram down a socket.
2152 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2153 unsigned int, flags)
2155 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2159 * Receive a frame from the socket and optionally record the address of the
2160 * sender. We verify the buffers are writable and if needed move the
2161 * sender address from kernel to user space.
2163 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2164 struct sockaddr __user *addr, int __user *addr_len)
2166 struct sockaddr_storage address;
2167 struct msghdr msg = {
2168 /* Save some cycles and don't copy the address if not needed */
2169 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2171 struct socket *sock;
2176 err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter);
2179 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2183 if (sock->file->f_flags & O_NONBLOCK)
2184 flags |= MSG_DONTWAIT;
2185 err = sock_recvmsg(sock, &msg, flags);
2187 if (err >= 0 && addr != NULL) {
2188 err2 = move_addr_to_user(&address,
2189 msg.msg_namelen, addr, addr_len);
2194 fput_light(sock->file, fput_needed);
2199 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2200 unsigned int, flags, struct sockaddr __user *, addr,
2201 int __user *, addr_len)
2203 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2207 * Receive a datagram from a socket.
2210 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2211 unsigned int, flags)
2213 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2216 static bool sock_use_custom_sol_socket(const struct socket *sock)
2218 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2222 * Set a socket option. Because we don't know the option lengths we have
2223 * to pass the user mode parameter for the protocols to sort out.
2225 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2228 sockptr_t optval = USER_SOCKPTR(user_optval);
2229 char *kernel_optval = NULL;
2230 int err, fput_needed;
2231 struct socket *sock;
2236 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2240 err = security_socket_setsockopt(sock, level, optname);
2244 if (!in_compat_syscall())
2245 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2246 user_optval, &optlen,
2256 optval = KERNEL_SOCKPTR(kernel_optval);
2257 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2258 err = sock_setsockopt(sock, level, optname, optval, optlen);
2259 else if (unlikely(!sock->ops->setsockopt))
2262 err = sock->ops->setsockopt(sock, level, optname, optval,
2264 kfree(kernel_optval);
2266 fput_light(sock->file, fput_needed);
2270 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2271 char __user *, optval, int, optlen)
2273 return __sys_setsockopt(fd, level, optname, optval, optlen);
2276 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2280 * Get a socket option. Because we don't know the option lengths we have
2281 * to pass a user mode parameter for the protocols to sort out.
2283 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2286 int max_optlen __maybe_unused;
2287 int err, fput_needed;
2288 struct socket *sock;
2290 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2294 err = security_socket_getsockopt(sock, level, optname);
2298 if (!in_compat_syscall())
2299 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2301 if (level == SOL_SOCKET)
2302 err = sock_getsockopt(sock, level, optname, optval, optlen);
2303 else if (unlikely(!sock->ops->getsockopt))
2306 err = sock->ops->getsockopt(sock, level, optname, optval,
2309 if (!in_compat_syscall())
2310 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2311 optval, optlen, max_optlen,
2314 fput_light(sock->file, fput_needed);
2318 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2319 char __user *, optval, int __user *, optlen)
2321 return __sys_getsockopt(fd, level, optname, optval, optlen);
2325 * Shutdown a socket.
2328 int __sys_shutdown_sock(struct socket *sock, int how)
2332 err = security_socket_shutdown(sock, how);
2334 err = sock->ops->shutdown(sock, how);
2339 int __sys_shutdown(int fd, int how)
2341 int err, fput_needed;
2342 struct socket *sock;
2344 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2346 err = __sys_shutdown_sock(sock, how);
2347 fput_light(sock->file, fput_needed);
2352 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2354 return __sys_shutdown(fd, how);
2357 /* A couple of helpful macros for getting the address of the 32/64 bit
2358 * fields which are the same type (int / unsigned) on our platforms.
2360 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2361 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2362 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2364 struct used_address {
2365 struct sockaddr_storage name;
2366 unsigned int name_len;
2369 int __copy_msghdr(struct msghdr *kmsg,
2370 struct user_msghdr *msg,
2371 struct sockaddr __user **save_addr)
2375 kmsg->msg_control_is_user = true;
2376 kmsg->msg_get_inq = 0;
2377 kmsg->msg_control_user = msg->msg_control;
2378 kmsg->msg_controllen = msg->msg_controllen;
2379 kmsg->msg_flags = msg->msg_flags;
2381 kmsg->msg_namelen = msg->msg_namelen;
2383 kmsg->msg_namelen = 0;
2385 if (kmsg->msg_namelen < 0)
2388 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2389 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2392 *save_addr = msg->msg_name;
2394 if (msg->msg_name && kmsg->msg_namelen) {
2396 err = move_addr_to_kernel(msg->msg_name,
2403 kmsg->msg_name = NULL;
2404 kmsg->msg_namelen = 0;
2407 if (msg->msg_iovlen > UIO_MAXIOV)
2410 kmsg->msg_iocb = NULL;
2411 kmsg->msg_ubuf = NULL;
2415 static int copy_msghdr_from_user(struct msghdr *kmsg,
2416 struct user_msghdr __user *umsg,
2417 struct sockaddr __user **save_addr,
2420 struct user_msghdr msg;
2423 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2426 err = __copy_msghdr(kmsg, &msg, save_addr);
2430 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2431 msg.msg_iov, msg.msg_iovlen,
2432 UIO_FASTIOV, iov, &kmsg->msg_iter);
2433 return err < 0 ? err : 0;
2436 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2437 unsigned int flags, struct used_address *used_address,
2438 unsigned int allowed_msghdr_flags)
2440 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2441 __aligned(sizeof(__kernel_size_t));
2442 /* 20 is size of ipv6_pktinfo */
2443 unsigned char *ctl_buf = ctl;
2449 if (msg_sys->msg_controllen > INT_MAX)
2451 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2452 ctl_len = msg_sys->msg_controllen;
2453 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2455 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2459 ctl_buf = msg_sys->msg_control;
2460 ctl_len = msg_sys->msg_controllen;
2461 } else if (ctl_len) {
2462 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2463 CMSG_ALIGN(sizeof(struct cmsghdr)));
2464 if (ctl_len > sizeof(ctl)) {
2465 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2466 if (ctl_buf == NULL)
2470 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2472 msg_sys->msg_control = ctl_buf;
2473 msg_sys->msg_control_is_user = false;
2475 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2476 msg_sys->msg_flags = flags;
2478 if (sock->file->f_flags & O_NONBLOCK)
2479 msg_sys->msg_flags |= MSG_DONTWAIT;
2481 * If this is sendmmsg() and current destination address is same as
2482 * previously succeeded address, omit asking LSM's decision.
2483 * used_address->name_len is initialized to UINT_MAX so that the first
2484 * destination address never matches.
2486 if (used_address && msg_sys->msg_name &&
2487 used_address->name_len == msg_sys->msg_namelen &&
2488 !memcmp(&used_address->name, msg_sys->msg_name,
2489 used_address->name_len)) {
2490 err = sock_sendmsg_nosec(sock, msg_sys);
2493 err = sock_sendmsg(sock, msg_sys);
2495 * If this is sendmmsg() and sending to current destination address was
2496 * successful, remember it.
2498 if (used_address && err >= 0) {
2499 used_address->name_len = msg_sys->msg_namelen;
2500 if (msg_sys->msg_name)
2501 memcpy(&used_address->name, msg_sys->msg_name,
2502 used_address->name_len);
2507 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2512 int sendmsg_copy_msghdr(struct msghdr *msg,
2513 struct user_msghdr __user *umsg, unsigned flags,
2518 if (flags & MSG_CMSG_COMPAT) {
2519 struct compat_msghdr __user *msg_compat;
2521 msg_compat = (struct compat_msghdr __user *) umsg;
2522 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2524 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2532 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2533 struct msghdr *msg_sys, unsigned int flags,
2534 struct used_address *used_address,
2535 unsigned int allowed_msghdr_flags)
2537 struct sockaddr_storage address;
2538 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2541 msg_sys->msg_name = &address;
2543 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2547 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2548 allowed_msghdr_flags);
2554 * BSD sendmsg interface
2556 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2559 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2562 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2563 bool forbid_cmsg_compat)
2565 int fput_needed, err;
2566 struct msghdr msg_sys;
2567 struct socket *sock;
2569 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2572 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2576 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2578 fput_light(sock->file, fput_needed);
2583 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2585 return __sys_sendmsg(fd, msg, flags, true);
2589 * Linux sendmmsg interface
2592 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2593 unsigned int flags, bool forbid_cmsg_compat)
2595 int fput_needed, err, datagrams;
2596 struct socket *sock;
2597 struct mmsghdr __user *entry;
2598 struct compat_mmsghdr __user *compat_entry;
2599 struct msghdr msg_sys;
2600 struct used_address used_address;
2601 unsigned int oflags = flags;
2603 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2606 if (vlen > UIO_MAXIOV)
2611 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2615 used_address.name_len = UINT_MAX;
2617 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2621 while (datagrams < vlen) {
2622 if (datagrams == vlen - 1)
2625 if (MSG_CMSG_COMPAT & flags) {
2626 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2627 &msg_sys, flags, &used_address, MSG_EOR);
2630 err = __put_user(err, &compat_entry->msg_len);
2633 err = ___sys_sendmsg(sock,
2634 (struct user_msghdr __user *)entry,
2635 &msg_sys, flags, &used_address, MSG_EOR);
2638 err = put_user(err, &entry->msg_len);
2645 if (msg_data_left(&msg_sys))
2650 fput_light(sock->file, fput_needed);
2652 /* We only return an error if no datagrams were able to be sent */
2659 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2660 unsigned int, vlen, unsigned int, flags)
2662 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2665 int recvmsg_copy_msghdr(struct msghdr *msg,
2666 struct user_msghdr __user *umsg, unsigned flags,
2667 struct sockaddr __user **uaddr,
2672 if (MSG_CMSG_COMPAT & flags) {
2673 struct compat_msghdr __user *msg_compat;
2675 msg_compat = (struct compat_msghdr __user *) umsg;
2676 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2678 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2686 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2687 struct user_msghdr __user *msg,
2688 struct sockaddr __user *uaddr,
2689 unsigned int flags, int nosec)
2691 struct compat_msghdr __user *msg_compat =
2692 (struct compat_msghdr __user *) msg;
2693 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2694 struct sockaddr_storage addr;
2695 unsigned long cmsg_ptr;
2699 msg_sys->msg_name = &addr;
2700 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2701 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2703 /* We assume all kernel code knows the size of sockaddr_storage */
2704 msg_sys->msg_namelen = 0;
2706 if (sock->file->f_flags & O_NONBLOCK)
2707 flags |= MSG_DONTWAIT;
2709 if (unlikely(nosec))
2710 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2712 err = sock_recvmsg(sock, msg_sys, flags);
2718 if (uaddr != NULL) {
2719 err = move_addr_to_user(&addr,
2720 msg_sys->msg_namelen, uaddr,
2725 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2729 if (MSG_CMSG_COMPAT & flags)
2730 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2731 &msg_compat->msg_controllen);
2733 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2734 &msg->msg_controllen);
2742 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2743 struct msghdr *msg_sys, unsigned int flags, int nosec)
2745 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2746 /* user mode address pointers */
2747 struct sockaddr __user *uaddr;
2750 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2754 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2760 * BSD recvmsg interface
2763 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2764 struct user_msghdr __user *umsg,
2765 struct sockaddr __user *uaddr, unsigned int flags)
2767 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2770 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2771 bool forbid_cmsg_compat)
2773 int fput_needed, err;
2774 struct msghdr msg_sys;
2775 struct socket *sock;
2777 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2780 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2784 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2786 fput_light(sock->file, fput_needed);
2791 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2792 unsigned int, flags)
2794 return __sys_recvmsg(fd, msg, flags, true);
2798 * Linux recvmmsg interface
2801 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2802 unsigned int vlen, unsigned int flags,
2803 struct timespec64 *timeout)
2805 int fput_needed, err, datagrams;
2806 struct socket *sock;
2807 struct mmsghdr __user *entry;
2808 struct compat_mmsghdr __user *compat_entry;
2809 struct msghdr msg_sys;
2810 struct timespec64 end_time;
2811 struct timespec64 timeout64;
2814 poll_select_set_timeout(&end_time, timeout->tv_sec,
2820 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2824 if (likely(!(flags & MSG_ERRQUEUE))) {
2825 err = sock_error(sock->sk);
2833 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2835 while (datagrams < vlen) {
2837 * No need to ask LSM for more than the first datagram.
2839 if (MSG_CMSG_COMPAT & flags) {
2840 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2841 &msg_sys, flags & ~MSG_WAITFORONE,
2845 err = __put_user(err, &compat_entry->msg_len);
2848 err = ___sys_recvmsg(sock,
2849 (struct user_msghdr __user *)entry,
2850 &msg_sys, flags & ~MSG_WAITFORONE,
2854 err = put_user(err, &entry->msg_len);
2862 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2863 if (flags & MSG_WAITFORONE)
2864 flags |= MSG_DONTWAIT;
2867 ktime_get_ts64(&timeout64);
2868 *timeout = timespec64_sub(end_time, timeout64);
2869 if (timeout->tv_sec < 0) {
2870 timeout->tv_sec = timeout->tv_nsec = 0;
2874 /* Timeout, return less than vlen datagrams */
2875 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2879 /* Out of band data, return right away */
2880 if (msg_sys.msg_flags & MSG_OOB)
2888 if (datagrams == 0) {
2894 * We may return less entries than requested (vlen) if the
2895 * sock is non block and there aren't enough datagrams...
2897 if (err != -EAGAIN) {
2899 * ... or if recvmsg returns an error after we
2900 * received some datagrams, where we record the
2901 * error to return on the next call or if the
2902 * app asks about it using getsockopt(SO_ERROR).
2904 WRITE_ONCE(sock->sk->sk_err, -err);
2907 fput_light(sock->file, fput_needed);
2912 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2913 unsigned int vlen, unsigned int flags,
2914 struct __kernel_timespec __user *timeout,
2915 struct old_timespec32 __user *timeout32)
2918 struct timespec64 timeout_sys;
2920 if (timeout && get_timespec64(&timeout_sys, timeout))
2923 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2926 if (!timeout && !timeout32)
2927 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2929 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2934 if (timeout && put_timespec64(&timeout_sys, timeout))
2935 datagrams = -EFAULT;
2937 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2938 datagrams = -EFAULT;
2943 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2944 unsigned int, vlen, unsigned int, flags,
2945 struct __kernel_timespec __user *, timeout)
2947 if (flags & MSG_CMSG_COMPAT)
2950 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2953 #ifdef CONFIG_COMPAT_32BIT_TIME
2954 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2955 unsigned int, vlen, unsigned int, flags,
2956 struct old_timespec32 __user *, timeout)
2958 if (flags & MSG_CMSG_COMPAT)
2961 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2965 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2966 /* Argument list sizes for sys_socketcall */
2967 #define AL(x) ((x) * sizeof(unsigned long))
2968 static const unsigned char nargs[21] = {
2969 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2970 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2971 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2978 * System call vectors.
2980 * Argument checking cleaned up. Saved 20% in size.
2981 * This function doesn't need to set the kernel lock because
2982 * it is set by the callees.
2985 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2987 unsigned long a[AUDITSC_ARGS];
2988 unsigned long a0, a1;
2992 if (call < 1 || call > SYS_SENDMMSG)
2994 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2997 if (len > sizeof(a))
3000 /* copy_from_user should be SMP safe. */
3001 if (copy_from_user(a, args, len))
3004 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3013 err = __sys_socket(a0, a1, a[2]);
3016 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3019 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3022 err = __sys_listen(a0, a1);
3025 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3026 (int __user *)a[2], 0);
3028 case SYS_GETSOCKNAME:
3030 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3031 (int __user *)a[2]);
3033 case SYS_GETPEERNAME:
3035 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3036 (int __user *)a[2]);
3038 case SYS_SOCKETPAIR:
3039 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3042 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3046 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3047 (struct sockaddr __user *)a[4], a[5]);
3050 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3054 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3055 (struct sockaddr __user *)a[4],
3056 (int __user *)a[5]);
3059 err = __sys_shutdown(a0, a1);
3061 case SYS_SETSOCKOPT:
3062 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3065 case SYS_GETSOCKOPT:
3067 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3068 (int __user *)a[4]);
3071 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3075 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3079 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3083 if (IS_ENABLED(CONFIG_64BIT))
3084 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3086 (struct __kernel_timespec __user *)a[4],
3089 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3091 (struct old_timespec32 __user *)a[4]);
3094 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3095 (int __user *)a[2], a[3]);
3104 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3107 * sock_register - add a socket protocol handler
3108 * @ops: description of protocol
3110 * This function is called by a protocol handler that wants to
3111 * advertise its address family, and have it linked into the
3112 * socket interface. The value ops->family corresponds to the
3113 * socket system call protocol family.
3115 int sock_register(const struct net_proto_family *ops)
3119 if (ops->family >= NPROTO) {
3120 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3124 spin_lock(&net_family_lock);
3125 if (rcu_dereference_protected(net_families[ops->family],
3126 lockdep_is_held(&net_family_lock)))
3129 rcu_assign_pointer(net_families[ops->family], ops);
3132 spin_unlock(&net_family_lock);
3134 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3137 EXPORT_SYMBOL(sock_register);
3140 * sock_unregister - remove a protocol handler
3141 * @family: protocol family to remove
3143 * This function is called by a protocol handler that wants to
3144 * remove its address family, and have it unlinked from the
3145 * new socket creation.
3147 * If protocol handler is a module, then it can use module reference
3148 * counts to protect against new references. If protocol handler is not
3149 * a module then it needs to provide its own protection in
3150 * the ops->create routine.
3152 void sock_unregister(int family)
3154 BUG_ON(family < 0 || family >= NPROTO);
3156 spin_lock(&net_family_lock);
3157 RCU_INIT_POINTER(net_families[family], NULL);
3158 spin_unlock(&net_family_lock);
3162 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3164 EXPORT_SYMBOL(sock_unregister);
3166 bool sock_is_registered(int family)
3168 return family < NPROTO && rcu_access_pointer(net_families[family]);
3171 static int __init sock_init(void)
3175 * Initialize the network sysctl infrastructure.
3177 err = net_sysctl_init();
3182 * Initialize skbuff SLAB cache
3187 * Initialize the protocols module.
3192 err = register_filesystem(&sock_fs_type);
3195 sock_mnt = kern_mount(&sock_fs_type);
3196 if (IS_ERR(sock_mnt)) {
3197 err = PTR_ERR(sock_mnt);
3201 /* The real protocol initialization is performed in later initcalls.
3204 #ifdef CONFIG_NETFILTER
3205 err = netfilter_init();
3210 ptp_classifier_init();
3216 unregister_filesystem(&sock_fs_type);
3220 core_initcall(sock_init); /* early initcall */
3222 #ifdef CONFIG_PROC_FS
3223 void socket_seq_show(struct seq_file *seq)
3225 seq_printf(seq, "sockets: used %d\n",
3226 sock_inuse_get(seq->private));
3228 #endif /* CONFIG_PROC_FS */
3230 /* Handle the fact that while struct ifreq has the same *layout* on
3231 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3232 * which are handled elsewhere, it still has different *size* due to
3233 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3234 * resulting in struct ifreq being 32 and 40 bytes respectively).
3235 * As a result, if the struct happens to be at the end of a page and
3236 * the next page isn't readable/writable, we get a fault. To prevent
3237 * that, copy back and forth to the full size.
3239 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3241 if (in_compat_syscall()) {
3242 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3244 memset(ifr, 0, sizeof(*ifr));
3245 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3249 *ifrdata = compat_ptr(ifr32->ifr_data);
3254 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3258 *ifrdata = ifr->ifr_data;
3262 EXPORT_SYMBOL(get_user_ifreq);
3264 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3266 size_t size = sizeof(*ifr);
3268 if (in_compat_syscall())
3269 size = sizeof(struct compat_ifreq);
3271 if (copy_to_user(arg, ifr, size))
3276 EXPORT_SYMBOL(put_user_ifreq);
3278 #ifdef CONFIG_COMPAT
3279 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3281 compat_uptr_t uptr32;
3286 if (get_user_ifreq(&ifr, NULL, uifr32))
3289 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3292 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3293 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3295 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3297 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3298 if (put_user_ifreq(&ifr, uifr32))
3304 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3305 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3306 struct compat_ifreq __user *u_ifreq32)
3311 if (!is_socket_ioctl_cmd(cmd))
3313 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3315 ifreq.ifr_data = data;
3317 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3320 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3321 unsigned int cmd, unsigned long arg)
3323 void __user *argp = compat_ptr(arg);
3324 struct sock *sk = sock->sk;
3325 struct net *net = sock_net(sk);
3327 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3328 return sock_ioctl(file, cmd, (unsigned long)argp);
3332 return compat_siocwandev(net, argp);
3333 case SIOCGSTAMP_OLD:
3334 case SIOCGSTAMPNS_OLD:
3335 if (!sock->ops->gettstamp)
3336 return -ENOIOCTLCMD;
3337 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3338 !COMPAT_USE_64BIT_TIME);
3341 case SIOCBONDSLAVEINFOQUERY:
3342 case SIOCBONDINFOQUERY:
3345 return compat_ifr_data_ioctl(net, cmd, argp);
3356 case SIOCGSTAMP_NEW:
3357 case SIOCGSTAMPNS_NEW:
3361 return sock_ioctl(file, cmd, arg);
3380 case SIOCSIFHWBROADCAST:
3382 case SIOCGIFBRDADDR:
3383 case SIOCSIFBRDADDR:
3384 case SIOCGIFDSTADDR:
3385 case SIOCSIFDSTADDR:
3386 case SIOCGIFNETMASK:
3387 case SIOCSIFNETMASK:
3399 case SIOCBONDENSLAVE:
3400 case SIOCBONDRELEASE:
3401 case SIOCBONDSETHWADDR:
3402 case SIOCBONDCHANGEACTIVE:
3409 return sock_do_ioctl(net, sock, cmd, arg);
3412 return -ENOIOCTLCMD;
3415 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3418 struct socket *sock = file->private_data;
3419 int ret = -ENOIOCTLCMD;
3426 if (sock->ops->compat_ioctl)
3427 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3429 if (ret == -ENOIOCTLCMD &&
3430 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3431 ret = compat_wext_handle_ioctl(net, cmd, arg);
3433 if (ret == -ENOIOCTLCMD)
3434 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3441 * kernel_bind - bind an address to a socket (kernel space)
3444 * @addrlen: length of address
3446 * Returns 0 or an error.
3449 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3451 return sock->ops->bind(sock, addr, addrlen);
3453 EXPORT_SYMBOL(kernel_bind);
3456 * kernel_listen - move socket to listening state (kernel space)
3458 * @backlog: pending connections queue size
3460 * Returns 0 or an error.
3463 int kernel_listen(struct socket *sock, int backlog)
3465 return sock->ops->listen(sock, backlog);
3467 EXPORT_SYMBOL(kernel_listen);
3470 * kernel_accept - accept a connection (kernel space)
3471 * @sock: listening socket
3472 * @newsock: new connected socket
3475 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3476 * If it fails, @newsock is guaranteed to be %NULL.
3477 * Returns 0 or an error.
3480 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3482 struct sock *sk = sock->sk;
3485 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3490 err = sock->ops->accept(sock, *newsock, flags, true);
3492 sock_release(*newsock);
3497 (*newsock)->ops = sock->ops;
3498 __module_get((*newsock)->ops->owner);
3503 EXPORT_SYMBOL(kernel_accept);
3506 * kernel_connect - connect a socket (kernel space)
3509 * @addrlen: address length
3510 * @flags: flags (O_NONBLOCK, ...)
3512 * For datagram sockets, @addr is the address to which datagrams are sent
3513 * by default, and the only address from which datagrams are received.
3514 * For stream sockets, attempts to connect to @addr.
3515 * Returns 0 or an error code.
3518 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3521 return sock->ops->connect(sock, addr, addrlen, flags);
3523 EXPORT_SYMBOL(kernel_connect);
3526 * kernel_getsockname - get the address which the socket is bound (kernel space)
3528 * @addr: address holder
3530 * Fills the @addr pointer with the address which the socket is bound.
3531 * Returns the length of the address in bytes or an error code.
3534 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3536 return sock->ops->getname(sock, addr, 0);
3538 EXPORT_SYMBOL(kernel_getsockname);
3541 * kernel_getpeername - get the address which the socket is connected (kernel space)
3543 * @addr: address holder
3545 * Fills the @addr pointer with the address which the socket is connected.
3546 * Returns the length of the address in bytes or an error code.
3549 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3551 return sock->ops->getname(sock, addr, 1);
3553 EXPORT_SYMBOL(kernel_getpeername);
3556 * kernel_sendpage - send a &page through a socket (kernel space)
3559 * @offset: page offset
3560 * @size: total size in bytes
3561 * @flags: flags (MSG_DONTWAIT, ...)
3563 * Returns the total amount sent in bytes or an error.
3566 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3567 size_t size, int flags)
3569 if (sock->ops->sendpage) {
3570 /* Warn in case the improper page to zero-copy send */
3571 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3572 return sock->ops->sendpage(sock, page, offset, size, flags);
3574 return sock_no_sendpage(sock, page, offset, size, flags);
3576 EXPORT_SYMBOL(kernel_sendpage);
3579 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3582 * @offset: page offset
3583 * @size: total size in bytes
3584 * @flags: flags (MSG_DONTWAIT, ...)
3586 * Returns the total amount sent in bytes or an error.
3587 * Caller must hold @sk.
3590 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3591 size_t size, int flags)
3593 struct socket *sock = sk->sk_socket;
3595 if (sock->ops->sendpage_locked)
3596 return sock->ops->sendpage_locked(sk, page, offset, size,
3599 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3601 EXPORT_SYMBOL(kernel_sendpage_locked);
3604 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3606 * @how: connection part
3608 * Returns 0 or an error.
3611 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3613 return sock->ops->shutdown(sock, how);
3615 EXPORT_SYMBOL(kernel_sock_shutdown);
3618 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3621 * This routine returns the IP overhead imposed by a socket i.e.
3622 * the length of the underlying IP header, depending on whether
3623 * this is an IPv4 or IPv6 socket and the length from IP options turned
3624 * on at the socket. Assumes that the caller has a lock on the socket.
3627 u32 kernel_sock_ip_overhead(struct sock *sk)
3629 struct inet_sock *inet;
3630 struct ip_options_rcu *opt;
3632 #if IS_ENABLED(CONFIG_IPV6)
3633 struct ipv6_pinfo *np;
3634 struct ipv6_txoptions *optv6 = NULL;
3635 #endif /* IS_ENABLED(CONFIG_IPV6) */
3640 switch (sk->sk_family) {
3643 overhead += sizeof(struct iphdr);
3644 opt = rcu_dereference_protected(inet->inet_opt,
3645 sock_owned_by_user(sk));
3647 overhead += opt->opt.optlen;
3649 #if IS_ENABLED(CONFIG_IPV6)
3652 overhead += sizeof(struct ipv6hdr);
3654 optv6 = rcu_dereference_protected(np->opt,
3655 sock_owned_by_user(sk));
3657 overhead += (optv6->opt_flen + optv6->opt_nflen);
3659 #endif /* IS_ENABLED(CONFIG_IPV6) */
3660 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3664 EXPORT_SYMBOL(kernel_sock_ip_overhead);