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/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
96 #include <net/cls_cgroup.h>
99 #include <linux/netfilter.h>
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly;
112 unsigned int sysctl_net_busy_poll __read_mostly;
115 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
116 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
117 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119 static int sock_close(struct inode *inode, struct file *file);
120 static __poll_t sock_poll(struct file *file,
121 struct poll_table_struct *wait);
122 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124 static long compat_sock_ioctl(struct file *file,
125 unsigned int cmd, unsigned long arg);
127 static int sock_fasync(int fd, struct file *filp, int on);
128 static ssize_t sock_sendpage(struct file *file, struct page *page,
129 int offset, size_t size, loff_t *ppos, int more);
130 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
131 struct pipe_inode_info *pipe, size_t len,
134 #ifdef CONFIG_PROC_FS
135 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
137 struct socket *sock = f->private_data;
139 if (sock->ops->show_fdinfo)
140 sock->ops->show_fdinfo(m, sock);
143 #define sock_show_fdinfo NULL
147 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
148 * in the operation structures but are done directly via the socketcall() multiplexor.
151 static const struct file_operations socket_file_ops = {
152 .owner = THIS_MODULE,
154 .read_iter = sock_read_iter,
155 .write_iter = sock_write_iter,
157 .unlocked_ioctl = sock_ioctl,
159 .compat_ioctl = compat_sock_ioctl,
162 .release = sock_close,
163 .fasync = sock_fasync,
164 .sendpage = sock_sendpage,
165 .splice_write = generic_splice_sendpage,
166 .splice_read = sock_splice_read,
167 .show_fdinfo = sock_show_fdinfo,
170 static const char * const pf_family_names[] = {
171 [PF_UNSPEC] = "PF_UNSPEC",
172 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
173 [PF_INET] = "PF_INET",
174 [PF_AX25] = "PF_AX25",
176 [PF_APPLETALK] = "PF_APPLETALK",
177 [PF_NETROM] = "PF_NETROM",
178 [PF_BRIDGE] = "PF_BRIDGE",
179 [PF_ATMPVC] = "PF_ATMPVC",
181 [PF_INET6] = "PF_INET6",
182 [PF_ROSE] = "PF_ROSE",
183 [PF_DECnet] = "PF_DECnet",
184 [PF_NETBEUI] = "PF_NETBEUI",
185 [PF_SECURITY] = "PF_SECURITY",
187 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
188 [PF_PACKET] = "PF_PACKET",
190 [PF_ECONET] = "PF_ECONET",
191 [PF_ATMSVC] = "PF_ATMSVC",
194 [PF_IRDA] = "PF_IRDA",
195 [PF_PPPOX] = "PF_PPPOX",
196 [PF_WANPIPE] = "PF_WANPIPE",
199 [PF_MPLS] = "PF_MPLS",
201 [PF_TIPC] = "PF_TIPC",
202 [PF_BLUETOOTH] = "PF_BLUETOOTH",
203 [PF_IUCV] = "PF_IUCV",
204 [PF_RXRPC] = "PF_RXRPC",
205 [PF_ISDN] = "PF_ISDN",
206 [PF_PHONET] = "PF_PHONET",
207 [PF_IEEE802154] = "PF_IEEE802154",
208 [PF_CAIF] = "PF_CAIF",
211 [PF_VSOCK] = "PF_VSOCK",
213 [PF_QIPCRTR] = "PF_QIPCRTR",
216 [PF_MCTP] = "PF_MCTP",
220 * The protocol list. Each protocol is registered in here.
223 static DEFINE_SPINLOCK(net_family_lock);
224 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
228 * Move socket addresses back and forth across the kernel/user
229 * divide and look after the messy bits.
233 * move_addr_to_kernel - copy a socket address into kernel space
234 * @uaddr: Address in user space
235 * @kaddr: Address in kernel space
236 * @ulen: Length in user space
238 * The address is copied into kernel space. If the provided address is
239 * too long an error code of -EINVAL is returned. If the copy gives
240 * invalid addresses -EFAULT is returned. On a success 0 is returned.
243 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
245 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
249 if (copy_from_user(kaddr, uaddr, ulen))
251 return audit_sockaddr(ulen, kaddr);
255 * move_addr_to_user - copy an address to user space
256 * @kaddr: kernel space address
257 * @klen: length of address in kernel
258 * @uaddr: user space address
259 * @ulen: pointer to user length field
261 * The value pointed to by ulen on entry is the buffer length available.
262 * This is overwritten with the buffer space used. -EINVAL is returned
263 * if an overlong buffer is specified or a negative buffer size. -EFAULT
264 * is returned if either the buffer or the length field are not
266 * After copying the data up to the limit the user specifies, the true
267 * length of the data is written over the length limit the user
268 * specified. Zero is returned for a success.
271 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
272 void __user *uaddr, int __user *ulen)
277 BUG_ON(klen > sizeof(struct sockaddr_storage));
278 err = get_user(len, ulen);
286 if (audit_sockaddr(klen, kaddr))
288 if (copy_to_user(uaddr, kaddr, len))
292 * "fromlen shall refer to the value before truncation.."
295 return __put_user(klen, ulen);
298 static struct kmem_cache *sock_inode_cachep __ro_after_init;
300 static struct inode *sock_alloc_inode(struct super_block *sb)
302 struct socket_alloc *ei;
304 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
307 init_waitqueue_head(&ei->socket.wq.wait);
308 ei->socket.wq.fasync_list = NULL;
309 ei->socket.wq.flags = 0;
311 ei->socket.state = SS_UNCONNECTED;
312 ei->socket.flags = 0;
313 ei->socket.ops = NULL;
314 ei->socket.sk = NULL;
315 ei->socket.file = NULL;
317 return &ei->vfs_inode;
320 static void sock_free_inode(struct inode *inode)
322 struct socket_alloc *ei;
324 ei = container_of(inode, struct socket_alloc, vfs_inode);
325 kmem_cache_free(sock_inode_cachep, ei);
328 static void init_once(void *foo)
330 struct socket_alloc *ei = (struct socket_alloc *)foo;
332 inode_init_once(&ei->vfs_inode);
335 static void init_inodecache(void)
337 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
338 sizeof(struct socket_alloc),
340 (SLAB_HWCACHE_ALIGN |
341 SLAB_RECLAIM_ACCOUNT |
342 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
344 BUG_ON(sock_inode_cachep == NULL);
347 static const struct super_operations sockfs_ops = {
348 .alloc_inode = sock_alloc_inode,
349 .free_inode = sock_free_inode,
350 .statfs = simple_statfs,
354 * sockfs_dname() is called from d_path().
356 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
358 return dynamic_dname(buffer, buflen, "socket:[%lu]",
359 d_inode(dentry)->i_ino);
362 static const struct dentry_operations sockfs_dentry_operations = {
363 .d_dname = sockfs_dname,
366 static int sockfs_xattr_get(const struct xattr_handler *handler,
367 struct dentry *dentry, struct inode *inode,
368 const char *suffix, void *value, size_t size)
371 if (dentry->d_name.len + 1 > size)
373 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
375 return dentry->d_name.len + 1;
378 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
379 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
380 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
382 static const struct xattr_handler sockfs_xattr_handler = {
383 .name = XATTR_NAME_SOCKPROTONAME,
384 .get = sockfs_xattr_get,
387 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
388 struct user_namespace *mnt_userns,
389 struct dentry *dentry, struct inode *inode,
390 const char *suffix, const void *value,
391 size_t size, int flags)
393 /* Handled by LSM. */
397 static const struct xattr_handler sockfs_security_xattr_handler = {
398 .prefix = XATTR_SECURITY_PREFIX,
399 .set = sockfs_security_xattr_set,
402 static const struct xattr_handler *sockfs_xattr_handlers[] = {
403 &sockfs_xattr_handler,
404 &sockfs_security_xattr_handler,
408 static int sockfs_init_fs_context(struct fs_context *fc)
410 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
413 ctx->ops = &sockfs_ops;
414 ctx->dops = &sockfs_dentry_operations;
415 ctx->xattr = sockfs_xattr_handlers;
419 static struct vfsmount *sock_mnt __read_mostly;
421 static struct file_system_type sock_fs_type = {
423 .init_fs_context = sockfs_init_fs_context,
424 .kill_sb = kill_anon_super,
428 * Obtains the first available file descriptor and sets it up for use.
430 * These functions create file structures and maps them to fd space
431 * of the current process. On success it returns file descriptor
432 * and file struct implicitly stored in sock->file.
433 * Note that another thread may close file descriptor before we return
434 * from this function. We use the fact that now we do not refer
435 * to socket after mapping. If one day we will need it, this
436 * function will increment ref. count on file by 1.
438 * In any case returned fd MAY BE not valid!
439 * This race condition is unavoidable
440 * with shared fd spaces, we cannot solve it inside kernel,
441 * but we take care of internal coherence yet.
445 * sock_alloc_file - Bind a &socket to a &file
447 * @flags: file status flags
448 * @dname: protocol name
450 * Returns the &file bound with @sock, implicitly storing it
451 * in sock->file. If dname is %NULL, sets to "".
453 * On failure @sock is released, and an ERR pointer is returned.
455 * This function uses GFP_KERNEL internally.
458 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
463 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
465 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
466 O_RDWR | (flags & O_NONBLOCK),
474 file->private_data = sock;
475 stream_open(SOCK_INODE(sock), file);
478 EXPORT_SYMBOL(sock_alloc_file);
480 static int sock_map_fd(struct socket *sock, int flags)
482 struct file *newfile;
483 int fd = get_unused_fd_flags(flags);
484 if (unlikely(fd < 0)) {
489 newfile = sock_alloc_file(sock, flags, NULL);
490 if (!IS_ERR(newfile)) {
491 fd_install(fd, newfile);
496 return PTR_ERR(newfile);
500 * sock_from_file - Return the &socket bounded to @file.
503 * On failure returns %NULL.
506 struct socket *sock_from_file(struct file *file)
508 if (file->f_op == &socket_file_ops)
509 return file->private_data; /* set in sock_alloc_file */
513 EXPORT_SYMBOL(sock_from_file);
516 * sockfd_lookup - Go from a file number to its socket slot
518 * @err: pointer to an error code return
520 * The file handle passed in is locked and the socket it is bound
521 * to is returned. If an error occurs the err pointer is overwritten
522 * with a negative errno code and NULL is returned. The function checks
523 * for both invalid handles and passing a handle which is not a socket.
525 * On a success the socket object pointer is returned.
528 struct socket *sockfd_lookup(int fd, int *err)
539 sock = sock_from_file(file);
546 EXPORT_SYMBOL(sockfd_lookup);
548 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
550 struct fd f = fdget(fd);
555 sock = sock_from_file(f.file);
557 *fput_needed = f.flags & FDPUT_FPUT;
566 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
572 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
582 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
587 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
594 static int sockfs_setattr(struct user_namespace *mnt_userns,
595 struct dentry *dentry, struct iattr *iattr)
597 int err = simple_setattr(&init_user_ns, dentry, iattr);
599 if (!err && (iattr->ia_valid & ATTR_UID)) {
600 struct socket *sock = SOCKET_I(d_inode(dentry));
603 sock->sk->sk_uid = iattr->ia_uid;
611 static const struct inode_operations sockfs_inode_ops = {
612 .listxattr = sockfs_listxattr,
613 .setattr = sockfs_setattr,
617 * sock_alloc - allocate a socket
619 * Allocate a new inode and socket object. The two are bound together
620 * and initialised. The socket is then returned. If we are out of inodes
621 * NULL is returned. This functions uses GFP_KERNEL internally.
624 struct socket *sock_alloc(void)
629 inode = new_inode_pseudo(sock_mnt->mnt_sb);
633 sock = SOCKET_I(inode);
635 inode->i_ino = get_next_ino();
636 inode->i_mode = S_IFSOCK | S_IRWXUGO;
637 inode->i_uid = current_fsuid();
638 inode->i_gid = current_fsgid();
639 inode->i_op = &sockfs_inode_ops;
643 EXPORT_SYMBOL(sock_alloc);
645 static void __sock_release(struct socket *sock, struct inode *inode)
648 struct module *owner = sock->ops->owner;
652 sock->ops->release(sock);
660 if (sock->wq.fasync_list)
661 pr_err("%s: fasync list not empty!\n", __func__);
664 iput(SOCK_INODE(sock));
671 * sock_release - close a socket
672 * @sock: socket to close
674 * The socket is released from the protocol stack if it has a release
675 * callback, and the inode is then released if the socket is bound to
676 * an inode not a file.
678 void sock_release(struct socket *sock)
680 __sock_release(sock, NULL);
682 EXPORT_SYMBOL(sock_release);
684 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
686 u8 flags = *tx_flags;
688 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
689 flags |= SKBTX_HW_TSTAMP;
691 /* PTP hardware clocks can provide a free running cycle counter
692 * as a time base for virtual clocks. Tell driver to use the
693 * free running cycle counter for timestamp if socket is bound
696 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
697 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
700 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
701 flags |= SKBTX_SW_TSTAMP;
703 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
704 flags |= SKBTX_SCHED_TSTAMP;
708 EXPORT_SYMBOL(__sock_tx_timestamp);
710 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
712 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
714 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
716 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
717 inet_sendmsg, sock, msg,
719 BUG_ON(ret == -EIOCBQUEUED);
723 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
725 int err = security_socket_sendmsg(sock, msg,
728 return err ?: sock_sendmsg_nosec(sock, msg);
732 * sock_sendmsg - send a message through @sock
734 * @msg: message to send
736 * Sends @msg through @sock, passing through LSM.
737 * Returns the number of bytes sent, or an error code.
739 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
741 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
742 struct sockaddr_storage address;
746 memcpy(&address, msg->msg_name, msg->msg_namelen);
747 msg->msg_name = &address;
750 ret = __sock_sendmsg(sock, msg);
751 msg->msg_name = save_addr;
755 EXPORT_SYMBOL(sock_sendmsg);
758 * kernel_sendmsg - send a message through @sock (kernel-space)
760 * @msg: message header
762 * @num: vec array length
763 * @size: total message data size
765 * Builds the message data with @vec and sends it through @sock.
766 * Returns the number of bytes sent, or an error code.
769 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
770 struct kvec *vec, size_t num, size_t size)
772 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
773 return sock_sendmsg(sock, msg);
775 EXPORT_SYMBOL(kernel_sendmsg);
778 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
780 * @msg: message header
781 * @vec: output s/g array
782 * @num: output s/g array length
783 * @size: total message data size
785 * Builds the message data with @vec and sends it through @sock.
786 * Returns the number of bytes sent, or an error code.
787 * Caller must hold @sk.
790 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
791 struct kvec *vec, size_t num, size_t size)
793 struct socket *sock = sk->sk_socket;
795 if (!sock->ops->sendmsg_locked)
796 return sock_no_sendmsg_locked(sk, msg, size);
798 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
800 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
802 EXPORT_SYMBOL(kernel_sendmsg_locked);
804 static bool skb_is_err_queue(const struct sk_buff *skb)
806 /* pkt_type of skbs enqueued on the error queue are set to
807 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
808 * in recvmsg, since skbs received on a local socket will never
809 * have a pkt_type of PACKET_OUTGOING.
811 return skb->pkt_type == PACKET_OUTGOING;
814 /* On transmit, software and hardware timestamps are returned independently.
815 * As the two skb clones share the hardware timestamp, which may be updated
816 * before the software timestamp is received, a hardware TX timestamp may be
817 * returned only if there is no software TX timestamp. Ignore false software
818 * timestamps, which may be made in the __sock_recv_timestamp() call when the
819 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
820 * hardware timestamp.
822 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
824 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
827 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
829 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC;
830 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
831 struct net_device *orig_dev;
835 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
837 *if_index = orig_dev->ifindex;
838 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
840 hwtstamp = shhwtstamps->hwtstamp;
847 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
850 struct scm_ts_pktinfo ts_pktinfo;
851 struct net_device *orig_dev;
853 if (!skb_mac_header_was_set(skb))
856 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
860 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
862 if_index = orig_dev->ifindex;
865 ts_pktinfo.if_index = if_index;
867 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
868 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
869 sizeof(ts_pktinfo), &ts_pktinfo);
873 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
875 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
878 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
879 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
880 struct scm_timestamping_internal tss;
882 int empty = 1, false_tstamp = 0;
883 struct skb_shared_hwtstamps *shhwtstamps =
888 /* Race occurred between timestamp enabling and packet
889 receiving. Fill in the current time for now. */
890 if (need_software_tstamp && skb->tstamp == 0) {
891 __net_timestamp(skb);
895 if (need_software_tstamp) {
896 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
898 struct __kernel_sock_timeval tv;
900 skb_get_new_timestamp(skb, &tv);
901 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
904 struct __kernel_old_timeval tv;
906 skb_get_timestamp(skb, &tv);
907 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
912 struct __kernel_timespec ts;
914 skb_get_new_timestampns(skb, &ts);
915 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
918 struct __kernel_old_timespec ts;
920 skb_get_timestampns(skb, &ts);
921 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
927 memset(&tss, 0, sizeof(tss));
928 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
929 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
932 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
933 !skb_is_swtx_tstamp(skb, false_tstamp)) {
935 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
936 hwtstamp = get_timestamp(sk, skb, &if_index);
938 hwtstamp = shhwtstamps->hwtstamp;
940 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
941 hwtstamp = ptp_convert_timestamp(&hwtstamp,
944 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
947 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
948 !skb_is_err_queue(skb))
949 put_ts_pktinfo(msg, skb, if_index);
953 if (sock_flag(sk, SOCK_TSTAMP_NEW))
954 put_cmsg_scm_timestamping64(msg, &tss);
956 put_cmsg_scm_timestamping(msg, &tss);
958 if (skb_is_err_queue(skb) && skb->len &&
959 SKB_EXT_ERR(skb)->opt_stats)
960 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
961 skb->len, skb->data);
964 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
966 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
971 if (!sock_flag(sk, SOCK_WIFI_STATUS))
973 if (!skb->wifi_acked_valid)
976 ack = skb->wifi_acked;
978 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
980 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
982 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
985 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
986 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
987 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
990 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
993 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
994 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
995 __u32 mark = skb->mark;
997 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1001 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1002 struct sk_buff *skb)
1004 sock_recv_timestamp(msg, sk, skb);
1005 sock_recv_drops(msg, sk, skb);
1006 sock_recv_mark(msg, sk, skb);
1008 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1010 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1012 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1014 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1017 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
1018 inet_recvmsg, sock, msg, msg_data_left(msg),
1023 * sock_recvmsg - receive a message from @sock
1025 * @msg: message to receive
1026 * @flags: message flags
1028 * Receives @msg from @sock, passing through LSM. Returns the total number
1029 * of bytes received, or an error.
1031 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1033 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1035 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1037 EXPORT_SYMBOL(sock_recvmsg);
1040 * kernel_recvmsg - Receive a message from a socket (kernel space)
1041 * @sock: The socket to receive the message from
1042 * @msg: Received message
1043 * @vec: Input s/g array for message data
1044 * @num: Size of input s/g array
1045 * @size: Number of bytes to read
1046 * @flags: Message flags (MSG_DONTWAIT, etc...)
1048 * On return the msg structure contains the scatter/gather array passed in the
1049 * vec argument. The array is modified so that it consists of the unfilled
1050 * portion of the original array.
1052 * The returned value is the total number of bytes received, or an error.
1055 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1056 struct kvec *vec, size_t num, size_t size, int flags)
1058 msg->msg_control_is_user = false;
1059 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1060 return sock_recvmsg(sock, msg, flags);
1062 EXPORT_SYMBOL(kernel_recvmsg);
1064 static ssize_t sock_sendpage(struct file *file, struct page *page,
1065 int offset, size_t size, loff_t *ppos, int more)
1067 struct socket *sock;
1070 sock = file->private_data;
1072 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1073 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1076 return kernel_sendpage(sock, page, offset, size, flags);
1079 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1080 struct pipe_inode_info *pipe, size_t len,
1083 struct socket *sock = file->private_data;
1085 if (unlikely(!sock->ops->splice_read))
1086 return generic_file_splice_read(file, ppos, pipe, len, flags);
1088 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1091 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1093 struct file *file = iocb->ki_filp;
1094 struct socket *sock = file->private_data;
1095 struct msghdr msg = {.msg_iter = *to,
1099 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1100 msg.msg_flags = MSG_DONTWAIT;
1102 if (iocb->ki_pos != 0)
1105 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1108 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1113 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1115 struct file *file = iocb->ki_filp;
1116 struct socket *sock = file->private_data;
1117 struct msghdr msg = {.msg_iter = *from,
1121 if (iocb->ki_pos != 0)
1124 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1125 msg.msg_flags = MSG_DONTWAIT;
1127 if (sock->type == SOCK_SEQPACKET)
1128 msg.msg_flags |= MSG_EOR;
1130 res = __sock_sendmsg(sock, &msg);
1131 *from = msg.msg_iter;
1136 * Atomic setting of ioctl hooks to avoid race
1137 * with module unload.
1140 static DEFINE_MUTEX(br_ioctl_mutex);
1141 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1142 unsigned int cmd, struct ifreq *ifr,
1145 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1146 unsigned int cmd, struct ifreq *ifr,
1149 mutex_lock(&br_ioctl_mutex);
1150 br_ioctl_hook = hook;
1151 mutex_unlock(&br_ioctl_mutex);
1153 EXPORT_SYMBOL(brioctl_set);
1155 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1156 struct ifreq *ifr, void __user *uarg)
1161 request_module("bridge");
1163 mutex_lock(&br_ioctl_mutex);
1165 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1166 mutex_unlock(&br_ioctl_mutex);
1171 static DEFINE_MUTEX(vlan_ioctl_mutex);
1172 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1174 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1176 mutex_lock(&vlan_ioctl_mutex);
1177 vlan_ioctl_hook = hook;
1178 mutex_unlock(&vlan_ioctl_mutex);
1180 EXPORT_SYMBOL(vlan_ioctl_set);
1182 static long sock_do_ioctl(struct net *net, struct socket *sock,
1183 unsigned int cmd, unsigned long arg)
1188 void __user *argp = (void __user *)arg;
1191 err = sock->ops->ioctl(sock, cmd, arg);
1194 * If this ioctl is unknown try to hand it down
1195 * to the NIC driver.
1197 if (err != -ENOIOCTLCMD)
1200 if (!is_socket_ioctl_cmd(cmd))
1203 if (get_user_ifreq(&ifr, &data, argp))
1205 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1206 if (!err && need_copyout)
1207 if (put_user_ifreq(&ifr, argp))
1214 * With an ioctl, arg may well be a user mode pointer, but we don't know
1215 * what to do with it - that's up to the protocol still.
1218 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1220 struct socket *sock;
1222 void __user *argp = (void __user *)arg;
1226 sock = file->private_data;
1229 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1233 if (get_user_ifreq(&ifr, &data, argp))
1235 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1236 if (!err && need_copyout)
1237 if (put_user_ifreq(&ifr, argp))
1240 #ifdef CONFIG_WEXT_CORE
1241 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1242 err = wext_handle_ioctl(net, cmd, argp);
1249 if (get_user(pid, (int __user *)argp))
1251 err = f_setown(sock->file, pid, 1);
1255 err = put_user(f_getown(sock->file),
1256 (int __user *)argp);
1262 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1267 if (!vlan_ioctl_hook)
1268 request_module("8021q");
1270 mutex_lock(&vlan_ioctl_mutex);
1271 if (vlan_ioctl_hook)
1272 err = vlan_ioctl_hook(net, argp);
1273 mutex_unlock(&vlan_ioctl_mutex);
1277 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1280 err = open_related_ns(&net->ns, get_net_ns);
1282 case SIOCGSTAMP_OLD:
1283 case SIOCGSTAMPNS_OLD:
1284 if (!sock->ops->gettstamp) {
1288 err = sock->ops->gettstamp(sock, argp,
1289 cmd == SIOCGSTAMP_OLD,
1290 !IS_ENABLED(CONFIG_64BIT));
1292 case SIOCGSTAMP_NEW:
1293 case SIOCGSTAMPNS_NEW:
1294 if (!sock->ops->gettstamp) {
1298 err = sock->ops->gettstamp(sock, argp,
1299 cmd == SIOCGSTAMP_NEW,
1304 err = dev_ifconf(net, argp);
1308 err = sock_do_ioctl(net, sock, cmd, arg);
1315 * sock_create_lite - creates a socket
1316 * @family: protocol family (AF_INET, ...)
1317 * @type: communication type (SOCK_STREAM, ...)
1318 * @protocol: protocol (0, ...)
1321 * Creates a new socket and assigns it to @res, passing through LSM.
1322 * The new socket initialization is not complete, see kernel_accept().
1323 * Returns 0 or an error. On failure @res is set to %NULL.
1324 * This function internally uses GFP_KERNEL.
1327 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1330 struct socket *sock = NULL;
1332 err = security_socket_create(family, type, protocol, 1);
1336 sock = sock_alloc();
1343 err = security_socket_post_create(sock, family, type, protocol, 1);
1355 EXPORT_SYMBOL(sock_create_lite);
1357 /* No kernel lock held - perfect */
1358 static __poll_t sock_poll(struct file *file, poll_table *wait)
1360 struct socket *sock = file->private_data;
1361 __poll_t events = poll_requested_events(wait), flag = 0;
1363 if (!sock->ops->poll)
1366 if (sk_can_busy_loop(sock->sk)) {
1367 /* poll once if requested by the syscall */
1368 if (events & POLL_BUSY_LOOP)
1369 sk_busy_loop(sock->sk, 1);
1371 /* if this socket can poll_ll, tell the system call */
1372 flag = POLL_BUSY_LOOP;
1375 return sock->ops->poll(file, sock, wait) | flag;
1378 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1380 struct socket *sock = file->private_data;
1382 return sock->ops->mmap(file, sock, vma);
1385 static int sock_close(struct inode *inode, struct file *filp)
1387 __sock_release(SOCKET_I(inode), inode);
1392 * Update the socket async list
1394 * Fasync_list locking strategy.
1396 * 1. fasync_list is modified only under process context socket lock
1397 * i.e. under semaphore.
1398 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1399 * or under socket lock
1402 static int sock_fasync(int fd, struct file *filp, int on)
1404 struct socket *sock = filp->private_data;
1405 struct sock *sk = sock->sk;
1406 struct socket_wq *wq = &sock->wq;
1412 fasync_helper(fd, filp, on, &wq->fasync_list);
1414 if (!wq->fasync_list)
1415 sock_reset_flag(sk, SOCK_FASYNC);
1417 sock_set_flag(sk, SOCK_FASYNC);
1423 /* This function may be called only under rcu_lock */
1425 int sock_wake_async(struct socket_wq *wq, int how, int band)
1427 if (!wq || !wq->fasync_list)
1431 case SOCK_WAKE_WAITD:
1432 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1435 case SOCK_WAKE_SPACE:
1436 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1441 kill_fasync(&wq->fasync_list, SIGIO, band);
1444 kill_fasync(&wq->fasync_list, SIGURG, band);
1449 EXPORT_SYMBOL(sock_wake_async);
1452 * __sock_create - creates a socket
1453 * @net: net namespace
1454 * @family: protocol family (AF_INET, ...)
1455 * @type: communication type (SOCK_STREAM, ...)
1456 * @protocol: protocol (0, ...)
1458 * @kern: boolean for kernel space sockets
1460 * Creates a new socket and assigns it to @res, passing through LSM.
1461 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1462 * be set to true if the socket resides in kernel space.
1463 * This function internally uses GFP_KERNEL.
1466 int __sock_create(struct net *net, int family, int type, int protocol,
1467 struct socket **res, int kern)
1470 struct socket *sock;
1471 const struct net_proto_family *pf;
1474 * Check protocol is in range
1476 if (family < 0 || family >= NPROTO)
1477 return -EAFNOSUPPORT;
1478 if (type < 0 || type >= SOCK_MAX)
1483 This uglymoron is moved from INET layer to here to avoid
1484 deadlock in module load.
1486 if (family == PF_INET && type == SOCK_PACKET) {
1487 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1492 err = security_socket_create(family, type, protocol, kern);
1497 * Allocate the socket and allow the family to set things up. if
1498 * the protocol is 0, the family is instructed to select an appropriate
1501 sock = sock_alloc();
1503 net_warn_ratelimited("socket: no more sockets\n");
1504 return -ENFILE; /* Not exactly a match, but its the
1505 closest posix thing */
1510 #ifdef CONFIG_MODULES
1511 /* Attempt to load a protocol module if the find failed.
1513 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1514 * requested real, full-featured networking support upon configuration.
1515 * Otherwise module support will break!
1517 if (rcu_access_pointer(net_families[family]) == NULL)
1518 request_module("net-pf-%d", family);
1522 pf = rcu_dereference(net_families[family]);
1523 err = -EAFNOSUPPORT;
1528 * We will call the ->create function, that possibly is in a loadable
1529 * module, so we have to bump that loadable module refcnt first.
1531 if (!try_module_get(pf->owner))
1534 /* Now protected by module ref count */
1537 err = pf->create(net, sock, protocol, kern);
1539 goto out_module_put;
1542 * Now to bump the refcnt of the [loadable] module that owns this
1543 * socket at sock_release time we decrement its refcnt.
1545 if (!try_module_get(sock->ops->owner))
1546 goto out_module_busy;
1549 * Now that we're done with the ->create function, the [loadable]
1550 * module can have its refcnt decremented
1552 module_put(pf->owner);
1553 err = security_socket_post_create(sock, family, type, protocol, kern);
1555 goto out_sock_release;
1561 err = -EAFNOSUPPORT;
1564 module_put(pf->owner);
1571 goto out_sock_release;
1573 EXPORT_SYMBOL(__sock_create);
1576 * sock_create - creates a socket
1577 * @family: protocol family (AF_INET, ...)
1578 * @type: communication type (SOCK_STREAM, ...)
1579 * @protocol: protocol (0, ...)
1582 * A wrapper around __sock_create().
1583 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1586 int sock_create(int family, int type, int protocol, struct socket **res)
1588 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1590 EXPORT_SYMBOL(sock_create);
1593 * sock_create_kern - creates a socket (kernel space)
1594 * @net: net namespace
1595 * @family: protocol family (AF_INET, ...)
1596 * @type: communication type (SOCK_STREAM, ...)
1597 * @protocol: protocol (0, ...)
1600 * A wrapper around __sock_create().
1601 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1604 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1606 return __sock_create(net, family, type, protocol, res, 1);
1608 EXPORT_SYMBOL(sock_create_kern);
1610 static struct socket *__sys_socket_create(int family, int type, int protocol)
1612 struct socket *sock;
1615 /* Check the SOCK_* constants for consistency. */
1616 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1617 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1618 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1619 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1621 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1622 return ERR_PTR(-EINVAL);
1623 type &= SOCK_TYPE_MASK;
1625 retval = sock_create(family, type, protocol, &sock);
1627 return ERR_PTR(retval);
1632 struct file *__sys_socket_file(int family, int type, int protocol)
1634 struct socket *sock;
1637 sock = __sys_socket_create(family, type, protocol);
1639 return ERR_CAST(sock);
1641 flags = type & ~SOCK_TYPE_MASK;
1642 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1643 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1645 return sock_alloc_file(sock, flags, NULL);
1648 int __sys_socket(int family, int type, int protocol)
1650 struct socket *sock;
1653 sock = __sys_socket_create(family, type, protocol);
1655 return PTR_ERR(sock);
1657 flags = type & ~SOCK_TYPE_MASK;
1658 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1659 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1661 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1664 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1666 return __sys_socket(family, type, protocol);
1670 * Create a pair of connected sockets.
1673 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1675 struct socket *sock1, *sock2;
1677 struct file *newfile1, *newfile2;
1680 flags = type & ~SOCK_TYPE_MASK;
1681 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1683 type &= SOCK_TYPE_MASK;
1685 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1686 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1689 * reserve descriptors and make sure we won't fail
1690 * to return them to userland.
1692 fd1 = get_unused_fd_flags(flags);
1693 if (unlikely(fd1 < 0))
1696 fd2 = get_unused_fd_flags(flags);
1697 if (unlikely(fd2 < 0)) {
1702 err = put_user(fd1, &usockvec[0]);
1706 err = put_user(fd2, &usockvec[1]);
1711 * Obtain the first socket and check if the underlying protocol
1712 * supports the socketpair call.
1715 err = sock_create(family, type, protocol, &sock1);
1716 if (unlikely(err < 0))
1719 err = sock_create(family, type, protocol, &sock2);
1720 if (unlikely(err < 0)) {
1721 sock_release(sock1);
1725 err = security_socket_socketpair(sock1, sock2);
1726 if (unlikely(err)) {
1727 sock_release(sock2);
1728 sock_release(sock1);
1732 err = sock1->ops->socketpair(sock1, sock2);
1733 if (unlikely(err < 0)) {
1734 sock_release(sock2);
1735 sock_release(sock1);
1739 newfile1 = sock_alloc_file(sock1, flags, NULL);
1740 if (IS_ERR(newfile1)) {
1741 err = PTR_ERR(newfile1);
1742 sock_release(sock2);
1746 newfile2 = sock_alloc_file(sock2, flags, NULL);
1747 if (IS_ERR(newfile2)) {
1748 err = PTR_ERR(newfile2);
1753 audit_fd_pair(fd1, fd2);
1755 fd_install(fd1, newfile1);
1756 fd_install(fd2, newfile2);
1765 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1766 int __user *, usockvec)
1768 return __sys_socketpair(family, type, protocol, usockvec);
1772 * Bind a name to a socket. Nothing much to do here since it's
1773 * the protocol's responsibility to handle the local address.
1775 * We move the socket address to kernel space before we call
1776 * the protocol layer (having also checked the address is ok).
1779 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1781 struct socket *sock;
1782 struct sockaddr_storage address;
1783 int err, fput_needed;
1785 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1787 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1789 err = security_socket_bind(sock,
1790 (struct sockaddr *)&address,
1793 err = sock->ops->bind(sock,
1797 fput_light(sock->file, fput_needed);
1802 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1804 return __sys_bind(fd, umyaddr, addrlen);
1808 * Perform a listen. Basically, we allow the protocol to do anything
1809 * necessary for a listen, and if that works, we mark the socket as
1810 * ready for listening.
1813 int __sys_listen(int fd, int backlog)
1815 struct socket *sock;
1816 int err, fput_needed;
1819 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1821 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1822 if ((unsigned int)backlog > somaxconn)
1823 backlog = somaxconn;
1825 err = security_socket_listen(sock, backlog);
1827 err = sock->ops->listen(sock, backlog);
1829 fput_light(sock->file, fput_needed);
1834 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1836 return __sys_listen(fd, backlog);
1839 struct file *do_accept(struct file *file, unsigned file_flags,
1840 struct sockaddr __user *upeer_sockaddr,
1841 int __user *upeer_addrlen, int flags)
1843 struct socket *sock, *newsock;
1844 struct file *newfile;
1846 struct sockaddr_storage address;
1848 sock = sock_from_file(file);
1850 return ERR_PTR(-ENOTSOCK);
1852 newsock = sock_alloc();
1854 return ERR_PTR(-ENFILE);
1856 newsock->type = sock->type;
1857 newsock->ops = sock->ops;
1860 * We don't need try_module_get here, as the listening socket (sock)
1861 * has the protocol module (sock->ops->owner) held.
1863 __module_get(newsock->ops->owner);
1865 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1866 if (IS_ERR(newfile))
1869 err = security_socket_accept(sock, newsock);
1873 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1878 if (upeer_sockaddr) {
1879 len = newsock->ops->getname(newsock,
1880 (struct sockaddr *)&address, 2);
1882 err = -ECONNABORTED;
1885 err = move_addr_to_user(&address,
1886 len, upeer_sockaddr, upeer_addrlen);
1891 /* File flags are not inherited via accept() unlike another OSes. */
1895 return ERR_PTR(err);
1898 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1899 int __user *upeer_addrlen, int flags)
1901 struct file *newfile;
1904 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1907 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1908 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1910 newfd = get_unused_fd_flags(flags);
1911 if (unlikely(newfd < 0))
1914 newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen,
1916 if (IS_ERR(newfile)) {
1917 put_unused_fd(newfd);
1918 return PTR_ERR(newfile);
1920 fd_install(newfd, newfile);
1925 * For accept, we attempt to create a new socket, set up the link
1926 * with the client, wake up the client, then return the new
1927 * connected fd. We collect the address of the connector in kernel
1928 * space and move it to user at the very end. This is unclean because
1929 * we open the socket then return an error.
1931 * 1003.1g adds the ability to recvmsg() to query connection pending
1932 * status to recvmsg. We need to add that support in a way thats
1933 * clean when we restructure accept also.
1936 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1937 int __user *upeer_addrlen, int flags)
1944 ret = __sys_accept4_file(f.file, upeer_sockaddr,
1945 upeer_addrlen, flags);
1952 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1953 int __user *, upeer_addrlen, int, flags)
1955 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1958 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1959 int __user *, upeer_addrlen)
1961 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1965 * Attempt to connect to a socket with the server address. The address
1966 * is in user space so we verify it is OK and move it to kernel space.
1968 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1971 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1972 * other SEQPACKET protocols that take time to connect() as it doesn't
1973 * include the -EINPROGRESS status for such sockets.
1976 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1977 int addrlen, int file_flags)
1979 struct socket *sock;
1982 sock = sock_from_file(file);
1989 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1993 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1994 sock->file->f_flags | file_flags);
1999 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2006 struct sockaddr_storage address;
2008 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2010 ret = __sys_connect_file(f.file, &address, addrlen, 0);
2017 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2020 return __sys_connect(fd, uservaddr, addrlen);
2024 * Get the local address ('name') of a socket object. Move the obtained
2025 * name to user space.
2028 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2029 int __user *usockaddr_len)
2031 struct socket *sock;
2032 struct sockaddr_storage address;
2033 int err, fput_needed;
2035 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2039 err = security_socket_getsockname(sock);
2043 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
2046 /* "err" is actually length in this case */
2047 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2050 fput_light(sock->file, fput_needed);
2055 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2056 int __user *, usockaddr_len)
2058 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2062 * Get the remote address ('name') of a socket object. Move the obtained
2063 * name to user space.
2066 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2067 int __user *usockaddr_len)
2069 struct socket *sock;
2070 struct sockaddr_storage address;
2071 int err, fput_needed;
2073 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2075 err = security_socket_getpeername(sock);
2077 fput_light(sock->file, fput_needed);
2081 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
2083 /* "err" is actually length in this case */
2084 err = move_addr_to_user(&address, err, usockaddr,
2086 fput_light(sock->file, fput_needed);
2091 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2092 int __user *, usockaddr_len)
2094 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2098 * Send a datagram to a given address. We move the address into kernel
2099 * space and check the user space data area is readable before invoking
2102 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2103 struct sockaddr __user *addr, int addr_len)
2105 struct socket *sock;
2106 struct sockaddr_storage address;
2112 err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter);
2115 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2119 msg.msg_name = NULL;
2120 msg.msg_control = NULL;
2121 msg.msg_controllen = 0;
2122 msg.msg_namelen = 0;
2123 msg.msg_ubuf = NULL;
2125 err = move_addr_to_kernel(addr, addr_len, &address);
2128 msg.msg_name = (struct sockaddr *)&address;
2129 msg.msg_namelen = addr_len;
2131 if (sock->file->f_flags & O_NONBLOCK)
2132 flags |= MSG_DONTWAIT;
2133 msg.msg_flags = flags;
2134 err = __sock_sendmsg(sock, &msg);
2137 fput_light(sock->file, fput_needed);
2142 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2143 unsigned int, flags, struct sockaddr __user *, addr,
2146 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2150 * Send a datagram down a socket.
2153 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2154 unsigned int, flags)
2156 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2160 * Receive a frame from the socket and optionally record the address of the
2161 * sender. We verify the buffers are writable and if needed move the
2162 * sender address from kernel to user space.
2164 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2165 struct sockaddr __user *addr, int __user *addr_len)
2167 struct sockaddr_storage address;
2168 struct msghdr msg = {
2169 /* Save some cycles and don't copy the address if not needed */
2170 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2172 struct socket *sock;
2177 err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter);
2180 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2184 if (sock->file->f_flags & O_NONBLOCK)
2185 flags |= MSG_DONTWAIT;
2186 err = sock_recvmsg(sock, &msg, flags);
2188 if (err >= 0 && addr != NULL) {
2189 err2 = move_addr_to_user(&address,
2190 msg.msg_namelen, addr, addr_len);
2195 fput_light(sock->file, fput_needed);
2200 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2201 unsigned int, flags, struct sockaddr __user *, addr,
2202 int __user *, addr_len)
2204 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2208 * Receive a datagram from a socket.
2211 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2212 unsigned int, flags)
2214 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2217 static bool sock_use_custom_sol_socket(const struct socket *sock)
2219 const struct sock *sk = sock->sk;
2221 /* Use sock->ops->setsockopt() for MPTCP */
2222 return IS_ENABLED(CONFIG_MPTCP) &&
2223 sk->sk_protocol == IPPROTO_MPTCP &&
2224 sk->sk_type == SOCK_STREAM &&
2225 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2229 * Set a socket option. Because we don't know the option lengths we have
2230 * to pass the user mode parameter for the protocols to sort out.
2232 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2235 sockptr_t optval = USER_SOCKPTR(user_optval);
2236 char *kernel_optval = NULL;
2237 int err, fput_needed;
2238 struct socket *sock;
2243 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2247 err = security_socket_setsockopt(sock, level, optname);
2251 if (!in_compat_syscall())
2252 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2253 user_optval, &optlen,
2263 optval = KERNEL_SOCKPTR(kernel_optval);
2264 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2265 err = sock_setsockopt(sock, level, optname, optval, optlen);
2266 else if (unlikely(!sock->ops->setsockopt))
2269 err = sock->ops->setsockopt(sock, level, optname, optval,
2271 kfree(kernel_optval);
2273 fput_light(sock->file, fput_needed);
2277 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2278 char __user *, optval, int, optlen)
2280 return __sys_setsockopt(fd, level, optname, optval, optlen);
2283 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2287 * Get a socket option. Because we don't know the option lengths we have
2288 * to pass a user mode parameter for the protocols to sort out.
2290 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2293 int err, fput_needed;
2294 struct socket *sock;
2297 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2301 err = security_socket_getsockopt(sock, level, optname);
2305 if (!in_compat_syscall())
2306 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2308 if (level == SOL_SOCKET)
2309 err = sock_getsockopt(sock, level, optname, optval, optlen);
2310 else if (unlikely(!sock->ops->getsockopt))
2313 err = sock->ops->getsockopt(sock, level, optname, optval,
2316 if (!in_compat_syscall())
2317 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2318 optval, optlen, max_optlen,
2321 fput_light(sock->file, fput_needed);
2325 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2326 char __user *, optval, int __user *, optlen)
2328 return __sys_getsockopt(fd, level, optname, optval, optlen);
2332 * Shutdown a socket.
2335 int __sys_shutdown_sock(struct socket *sock, int how)
2339 err = security_socket_shutdown(sock, how);
2341 err = sock->ops->shutdown(sock, how);
2346 int __sys_shutdown(int fd, int how)
2348 int err, fput_needed;
2349 struct socket *sock;
2351 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2353 err = __sys_shutdown_sock(sock, how);
2354 fput_light(sock->file, fput_needed);
2359 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2361 return __sys_shutdown(fd, how);
2364 /* A couple of helpful macros for getting the address of the 32/64 bit
2365 * fields which are the same type (int / unsigned) on our platforms.
2367 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2368 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2369 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2371 struct used_address {
2372 struct sockaddr_storage name;
2373 unsigned int name_len;
2376 int __copy_msghdr(struct msghdr *kmsg,
2377 struct user_msghdr *msg,
2378 struct sockaddr __user **save_addr)
2382 kmsg->msg_control_is_user = true;
2383 kmsg->msg_get_inq = 0;
2384 kmsg->msg_control_user = msg->msg_control;
2385 kmsg->msg_controllen = msg->msg_controllen;
2386 kmsg->msg_flags = msg->msg_flags;
2388 kmsg->msg_namelen = msg->msg_namelen;
2390 kmsg->msg_namelen = 0;
2392 if (kmsg->msg_namelen < 0)
2395 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2396 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2399 *save_addr = msg->msg_name;
2401 if (msg->msg_name && kmsg->msg_namelen) {
2403 err = move_addr_to_kernel(msg->msg_name,
2410 kmsg->msg_name = NULL;
2411 kmsg->msg_namelen = 0;
2414 if (msg->msg_iovlen > UIO_MAXIOV)
2417 kmsg->msg_iocb = NULL;
2418 kmsg->msg_ubuf = NULL;
2422 static int copy_msghdr_from_user(struct msghdr *kmsg,
2423 struct user_msghdr __user *umsg,
2424 struct sockaddr __user **save_addr,
2427 struct user_msghdr msg;
2430 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2433 err = __copy_msghdr(kmsg, &msg, save_addr);
2437 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2438 msg.msg_iov, msg.msg_iovlen,
2439 UIO_FASTIOV, iov, &kmsg->msg_iter);
2440 return err < 0 ? err : 0;
2443 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2444 unsigned int flags, struct used_address *used_address,
2445 unsigned int allowed_msghdr_flags)
2447 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2448 __aligned(sizeof(__kernel_size_t));
2449 /* 20 is size of ipv6_pktinfo */
2450 unsigned char *ctl_buf = ctl;
2456 if (msg_sys->msg_controllen > INT_MAX)
2458 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2459 ctl_len = msg_sys->msg_controllen;
2460 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2462 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2466 ctl_buf = msg_sys->msg_control;
2467 ctl_len = msg_sys->msg_controllen;
2468 } else if (ctl_len) {
2469 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2470 CMSG_ALIGN(sizeof(struct cmsghdr)));
2471 if (ctl_len > sizeof(ctl)) {
2472 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2473 if (ctl_buf == NULL)
2477 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2479 msg_sys->msg_control = ctl_buf;
2480 msg_sys->msg_control_is_user = false;
2482 msg_sys->msg_flags = flags;
2484 if (sock->file->f_flags & O_NONBLOCK)
2485 msg_sys->msg_flags |= MSG_DONTWAIT;
2487 * If this is sendmmsg() and current destination address is same as
2488 * previously succeeded address, omit asking LSM's decision.
2489 * used_address->name_len is initialized to UINT_MAX so that the first
2490 * destination address never matches.
2492 if (used_address && msg_sys->msg_name &&
2493 used_address->name_len == msg_sys->msg_namelen &&
2494 !memcmp(&used_address->name, msg_sys->msg_name,
2495 used_address->name_len)) {
2496 err = sock_sendmsg_nosec(sock, msg_sys);
2499 err = __sock_sendmsg(sock, msg_sys);
2501 * If this is sendmmsg() and sending to current destination address was
2502 * successful, remember it.
2504 if (used_address && err >= 0) {
2505 used_address->name_len = msg_sys->msg_namelen;
2506 if (msg_sys->msg_name)
2507 memcpy(&used_address->name, msg_sys->msg_name,
2508 used_address->name_len);
2513 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2518 int sendmsg_copy_msghdr(struct msghdr *msg,
2519 struct user_msghdr __user *umsg, unsigned flags,
2524 if (flags & MSG_CMSG_COMPAT) {
2525 struct compat_msghdr __user *msg_compat;
2527 msg_compat = (struct compat_msghdr __user *) umsg;
2528 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2530 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2538 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2539 struct msghdr *msg_sys, unsigned int flags,
2540 struct used_address *used_address,
2541 unsigned int allowed_msghdr_flags)
2543 struct sockaddr_storage address;
2544 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2547 msg_sys->msg_name = &address;
2549 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2553 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2554 allowed_msghdr_flags);
2560 * BSD sendmsg interface
2562 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2565 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2568 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2569 bool forbid_cmsg_compat)
2571 int fput_needed, err;
2572 struct msghdr msg_sys;
2573 struct socket *sock;
2575 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2578 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2582 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2584 fput_light(sock->file, fput_needed);
2589 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2591 return __sys_sendmsg(fd, msg, flags, true);
2595 * Linux sendmmsg interface
2598 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2599 unsigned int flags, bool forbid_cmsg_compat)
2601 int fput_needed, err, datagrams;
2602 struct socket *sock;
2603 struct mmsghdr __user *entry;
2604 struct compat_mmsghdr __user *compat_entry;
2605 struct msghdr msg_sys;
2606 struct used_address used_address;
2607 unsigned int oflags = flags;
2609 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2612 if (vlen > UIO_MAXIOV)
2617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2621 used_address.name_len = UINT_MAX;
2623 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2627 while (datagrams < vlen) {
2628 if (datagrams == vlen - 1)
2631 if (MSG_CMSG_COMPAT & flags) {
2632 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2633 &msg_sys, flags, &used_address, MSG_EOR);
2636 err = __put_user(err, &compat_entry->msg_len);
2639 err = ___sys_sendmsg(sock,
2640 (struct user_msghdr __user *)entry,
2641 &msg_sys, flags, &used_address, MSG_EOR);
2644 err = put_user(err, &entry->msg_len);
2651 if (msg_data_left(&msg_sys))
2656 fput_light(sock->file, fput_needed);
2658 /* We only return an error if no datagrams were able to be sent */
2665 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2666 unsigned int, vlen, unsigned int, flags)
2668 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2671 int recvmsg_copy_msghdr(struct msghdr *msg,
2672 struct user_msghdr __user *umsg, unsigned flags,
2673 struct sockaddr __user **uaddr,
2678 if (MSG_CMSG_COMPAT & flags) {
2679 struct compat_msghdr __user *msg_compat;
2681 msg_compat = (struct compat_msghdr __user *) umsg;
2682 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2684 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2692 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2693 struct user_msghdr __user *msg,
2694 struct sockaddr __user *uaddr,
2695 unsigned int flags, int nosec)
2697 struct compat_msghdr __user *msg_compat =
2698 (struct compat_msghdr __user *) msg;
2699 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2700 struct sockaddr_storage addr;
2701 unsigned long cmsg_ptr;
2705 msg_sys->msg_name = &addr;
2706 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2707 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2709 /* We assume all kernel code knows the size of sockaddr_storage */
2710 msg_sys->msg_namelen = 0;
2712 if (sock->file->f_flags & O_NONBLOCK)
2713 flags |= MSG_DONTWAIT;
2715 if (unlikely(nosec))
2716 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2718 err = sock_recvmsg(sock, msg_sys, flags);
2724 if (uaddr != NULL) {
2725 err = move_addr_to_user(&addr,
2726 msg_sys->msg_namelen, uaddr,
2731 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2735 if (MSG_CMSG_COMPAT & flags)
2736 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2737 &msg_compat->msg_controllen);
2739 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2740 &msg->msg_controllen);
2748 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2749 struct msghdr *msg_sys, unsigned int flags, int nosec)
2751 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2752 /* user mode address pointers */
2753 struct sockaddr __user *uaddr;
2756 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2760 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2766 * BSD recvmsg interface
2769 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2770 struct user_msghdr __user *umsg,
2771 struct sockaddr __user *uaddr, unsigned int flags)
2773 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2776 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2777 bool forbid_cmsg_compat)
2779 int fput_needed, err;
2780 struct msghdr msg_sys;
2781 struct socket *sock;
2783 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2790 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2792 fput_light(sock->file, fput_needed);
2797 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2798 unsigned int, flags)
2800 return __sys_recvmsg(fd, msg, flags, true);
2804 * Linux recvmmsg interface
2807 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2808 unsigned int vlen, unsigned int flags,
2809 struct timespec64 *timeout)
2811 int fput_needed, err, datagrams;
2812 struct socket *sock;
2813 struct mmsghdr __user *entry;
2814 struct compat_mmsghdr __user *compat_entry;
2815 struct msghdr msg_sys;
2816 struct timespec64 end_time;
2817 struct timespec64 timeout64;
2820 poll_select_set_timeout(&end_time, timeout->tv_sec,
2826 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2830 if (likely(!(flags & MSG_ERRQUEUE))) {
2831 err = sock_error(sock->sk);
2839 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2841 while (datagrams < vlen) {
2843 * No need to ask LSM for more than the first datagram.
2845 if (MSG_CMSG_COMPAT & flags) {
2846 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2847 &msg_sys, flags & ~MSG_WAITFORONE,
2851 err = __put_user(err, &compat_entry->msg_len);
2854 err = ___sys_recvmsg(sock,
2855 (struct user_msghdr __user *)entry,
2856 &msg_sys, flags & ~MSG_WAITFORONE,
2860 err = put_user(err, &entry->msg_len);
2868 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2869 if (flags & MSG_WAITFORONE)
2870 flags |= MSG_DONTWAIT;
2873 ktime_get_ts64(&timeout64);
2874 *timeout = timespec64_sub(end_time, timeout64);
2875 if (timeout->tv_sec < 0) {
2876 timeout->tv_sec = timeout->tv_nsec = 0;
2880 /* Timeout, return less than vlen datagrams */
2881 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2885 /* Out of band data, return right away */
2886 if (msg_sys.msg_flags & MSG_OOB)
2894 if (datagrams == 0) {
2900 * We may return less entries than requested (vlen) if the
2901 * sock is non block and there aren't enough datagrams...
2903 if (err != -EAGAIN) {
2905 * ... or if recvmsg returns an error after we
2906 * received some datagrams, where we record the
2907 * error to return on the next call or if the
2908 * app asks about it using getsockopt(SO_ERROR).
2910 WRITE_ONCE(sock->sk->sk_err, -err);
2913 fput_light(sock->file, fput_needed);
2918 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2919 unsigned int vlen, unsigned int flags,
2920 struct __kernel_timespec __user *timeout,
2921 struct old_timespec32 __user *timeout32)
2924 struct timespec64 timeout_sys;
2926 if (timeout && get_timespec64(&timeout_sys, timeout))
2929 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2932 if (!timeout && !timeout32)
2933 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2935 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2940 if (timeout && put_timespec64(&timeout_sys, timeout))
2941 datagrams = -EFAULT;
2943 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2944 datagrams = -EFAULT;
2949 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2950 unsigned int, vlen, unsigned int, flags,
2951 struct __kernel_timespec __user *, timeout)
2953 if (flags & MSG_CMSG_COMPAT)
2956 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2959 #ifdef CONFIG_COMPAT_32BIT_TIME
2960 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2961 unsigned int, vlen, unsigned int, flags,
2962 struct old_timespec32 __user *, timeout)
2964 if (flags & MSG_CMSG_COMPAT)
2967 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2971 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2972 /* Argument list sizes for sys_socketcall */
2973 #define AL(x) ((x) * sizeof(unsigned long))
2974 static const unsigned char nargs[21] = {
2975 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2976 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2977 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2984 * System call vectors.
2986 * Argument checking cleaned up. Saved 20% in size.
2987 * This function doesn't need to set the kernel lock because
2988 * it is set by the callees.
2991 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2993 unsigned long a[AUDITSC_ARGS];
2994 unsigned long a0, a1;
2998 if (call < 1 || call > SYS_SENDMMSG)
3000 call = array_index_nospec(call, SYS_SENDMMSG + 1);
3003 if (len > sizeof(a))
3006 /* copy_from_user should be SMP safe. */
3007 if (copy_from_user(a, args, len))
3010 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3019 err = __sys_socket(a0, a1, a[2]);
3022 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3025 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3028 err = __sys_listen(a0, a1);
3031 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3032 (int __user *)a[2], 0);
3034 case SYS_GETSOCKNAME:
3036 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3037 (int __user *)a[2]);
3039 case SYS_GETPEERNAME:
3041 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3042 (int __user *)a[2]);
3044 case SYS_SOCKETPAIR:
3045 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3048 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3052 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3053 (struct sockaddr __user *)a[4], a[5]);
3056 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3060 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3061 (struct sockaddr __user *)a[4],
3062 (int __user *)a[5]);
3065 err = __sys_shutdown(a0, a1);
3067 case SYS_SETSOCKOPT:
3068 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3071 case SYS_GETSOCKOPT:
3073 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3074 (int __user *)a[4]);
3077 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3081 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3085 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3089 if (IS_ENABLED(CONFIG_64BIT))
3090 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3092 (struct __kernel_timespec __user *)a[4],
3095 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3097 (struct old_timespec32 __user *)a[4]);
3100 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3101 (int __user *)a[2], a[3]);
3110 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3113 * sock_register - add a socket protocol handler
3114 * @ops: description of protocol
3116 * This function is called by a protocol handler that wants to
3117 * advertise its address family, and have it linked into the
3118 * socket interface. The value ops->family corresponds to the
3119 * socket system call protocol family.
3121 int sock_register(const struct net_proto_family *ops)
3125 if (ops->family >= NPROTO) {
3126 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3130 spin_lock(&net_family_lock);
3131 if (rcu_dereference_protected(net_families[ops->family],
3132 lockdep_is_held(&net_family_lock)))
3135 rcu_assign_pointer(net_families[ops->family], ops);
3138 spin_unlock(&net_family_lock);
3140 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3143 EXPORT_SYMBOL(sock_register);
3146 * sock_unregister - remove a protocol handler
3147 * @family: protocol family to remove
3149 * This function is called by a protocol handler that wants to
3150 * remove its address family, and have it unlinked from the
3151 * new socket creation.
3153 * If protocol handler is a module, then it can use module reference
3154 * counts to protect against new references. If protocol handler is not
3155 * a module then it needs to provide its own protection in
3156 * the ops->create routine.
3158 void sock_unregister(int family)
3160 BUG_ON(family < 0 || family >= NPROTO);
3162 spin_lock(&net_family_lock);
3163 RCU_INIT_POINTER(net_families[family], NULL);
3164 spin_unlock(&net_family_lock);
3168 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3170 EXPORT_SYMBOL(sock_unregister);
3172 bool sock_is_registered(int family)
3174 return family < NPROTO && rcu_access_pointer(net_families[family]);
3177 static int __init sock_init(void)
3181 * Initialize the network sysctl infrastructure.
3183 err = net_sysctl_init();
3188 * Initialize skbuff SLAB cache
3193 * Initialize the protocols module.
3198 err = register_filesystem(&sock_fs_type);
3201 sock_mnt = kern_mount(&sock_fs_type);
3202 if (IS_ERR(sock_mnt)) {
3203 err = PTR_ERR(sock_mnt);
3207 /* The real protocol initialization is performed in later initcalls.
3210 #ifdef CONFIG_NETFILTER
3211 err = netfilter_init();
3216 ptp_classifier_init();
3222 unregister_filesystem(&sock_fs_type);
3226 core_initcall(sock_init); /* early initcall */
3228 #ifdef CONFIG_PROC_FS
3229 void socket_seq_show(struct seq_file *seq)
3231 seq_printf(seq, "sockets: used %d\n",
3232 sock_inuse_get(seq->private));
3234 #endif /* CONFIG_PROC_FS */
3236 /* Handle the fact that while struct ifreq has the same *layout* on
3237 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3238 * which are handled elsewhere, it still has different *size* due to
3239 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3240 * resulting in struct ifreq being 32 and 40 bytes respectively).
3241 * As a result, if the struct happens to be at the end of a page and
3242 * the next page isn't readable/writable, we get a fault. To prevent
3243 * that, copy back and forth to the full size.
3245 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3247 if (in_compat_syscall()) {
3248 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3250 memset(ifr, 0, sizeof(*ifr));
3251 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3255 *ifrdata = compat_ptr(ifr32->ifr_data);
3260 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3264 *ifrdata = ifr->ifr_data;
3268 EXPORT_SYMBOL(get_user_ifreq);
3270 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3272 size_t size = sizeof(*ifr);
3274 if (in_compat_syscall())
3275 size = sizeof(struct compat_ifreq);
3277 if (copy_to_user(arg, ifr, size))
3282 EXPORT_SYMBOL(put_user_ifreq);
3284 #ifdef CONFIG_COMPAT
3285 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3287 compat_uptr_t uptr32;
3292 if (get_user_ifreq(&ifr, NULL, uifr32))
3295 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3298 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3299 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3301 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3303 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3304 if (put_user_ifreq(&ifr, uifr32))
3310 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3311 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3312 struct compat_ifreq __user *u_ifreq32)
3317 if (!is_socket_ioctl_cmd(cmd))
3319 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3321 ifreq.ifr_data = data;
3323 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3326 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3327 unsigned int cmd, unsigned long arg)
3329 void __user *argp = compat_ptr(arg);
3330 struct sock *sk = sock->sk;
3331 struct net *net = sock_net(sk);
3333 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3334 return sock_ioctl(file, cmd, (unsigned long)argp);
3338 return compat_siocwandev(net, argp);
3339 case SIOCGSTAMP_OLD:
3340 case SIOCGSTAMPNS_OLD:
3341 if (!sock->ops->gettstamp)
3342 return -ENOIOCTLCMD;
3343 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3344 !COMPAT_USE_64BIT_TIME);
3347 case SIOCBONDSLAVEINFOQUERY:
3348 case SIOCBONDINFOQUERY:
3351 return compat_ifr_data_ioctl(net, cmd, argp);
3362 case SIOCGSTAMP_NEW:
3363 case SIOCGSTAMPNS_NEW:
3367 return sock_ioctl(file, cmd, arg);
3386 case SIOCSIFHWBROADCAST:
3388 case SIOCGIFBRDADDR:
3389 case SIOCSIFBRDADDR:
3390 case SIOCGIFDSTADDR:
3391 case SIOCSIFDSTADDR:
3392 case SIOCGIFNETMASK:
3393 case SIOCSIFNETMASK:
3405 case SIOCBONDENSLAVE:
3406 case SIOCBONDRELEASE:
3407 case SIOCBONDSETHWADDR:
3408 case SIOCBONDCHANGEACTIVE:
3415 return sock_do_ioctl(net, sock, cmd, arg);
3418 return -ENOIOCTLCMD;
3421 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3424 struct socket *sock = file->private_data;
3425 int ret = -ENOIOCTLCMD;
3432 if (sock->ops->compat_ioctl)
3433 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3435 if (ret == -ENOIOCTLCMD &&
3436 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3437 ret = compat_wext_handle_ioctl(net, cmd, arg);
3439 if (ret == -ENOIOCTLCMD)
3440 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3447 * kernel_bind - bind an address to a socket (kernel space)
3450 * @addrlen: length of address
3452 * Returns 0 or an error.
3455 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3457 struct sockaddr_storage address;
3459 memcpy(&address, addr, addrlen);
3461 return sock->ops->bind(sock, (struct sockaddr *)&address, addrlen);
3463 EXPORT_SYMBOL(kernel_bind);
3466 * kernel_listen - move socket to listening state (kernel space)
3468 * @backlog: pending connections queue size
3470 * Returns 0 or an error.
3473 int kernel_listen(struct socket *sock, int backlog)
3475 return sock->ops->listen(sock, backlog);
3477 EXPORT_SYMBOL(kernel_listen);
3480 * kernel_accept - accept a connection (kernel space)
3481 * @sock: listening socket
3482 * @newsock: new connected socket
3485 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3486 * If it fails, @newsock is guaranteed to be %NULL.
3487 * Returns 0 or an error.
3490 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3492 struct sock *sk = sock->sk;
3495 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3500 err = sock->ops->accept(sock, *newsock, flags, true);
3502 sock_release(*newsock);
3507 (*newsock)->ops = sock->ops;
3508 __module_get((*newsock)->ops->owner);
3513 EXPORT_SYMBOL(kernel_accept);
3516 * kernel_connect - connect a socket (kernel space)
3519 * @addrlen: address length
3520 * @flags: flags (O_NONBLOCK, ...)
3522 * For datagram sockets, @addr is the address to which datagrams are sent
3523 * by default, and the only address from which datagrams are received.
3524 * For stream sockets, attempts to connect to @addr.
3525 * Returns 0 or an error code.
3528 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3531 struct sockaddr_storage address;
3533 memcpy(&address, addr, addrlen);
3535 return sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, flags);
3537 EXPORT_SYMBOL(kernel_connect);
3540 * kernel_getsockname - get the address which the socket is bound (kernel space)
3542 * @addr: address holder
3544 * Fills the @addr pointer with the address which the socket is bound.
3545 * Returns the length of the address in bytes or an error code.
3548 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3550 return sock->ops->getname(sock, addr, 0);
3552 EXPORT_SYMBOL(kernel_getsockname);
3555 * kernel_getpeername - get the address which the socket is connected (kernel space)
3557 * @addr: address holder
3559 * Fills the @addr pointer with the address which the socket is connected.
3560 * Returns the length of the address in bytes or an error code.
3563 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3565 return sock->ops->getname(sock, addr, 1);
3567 EXPORT_SYMBOL(kernel_getpeername);
3570 * kernel_sendpage - send a &page through a socket (kernel space)
3573 * @offset: page offset
3574 * @size: total size in bytes
3575 * @flags: flags (MSG_DONTWAIT, ...)
3577 * Returns the total amount sent in bytes or an error.
3580 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3581 size_t size, int flags)
3583 if (sock->ops->sendpage) {
3584 /* Warn in case the improper page to zero-copy send */
3585 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3586 return sock->ops->sendpage(sock, page, offset, size, flags);
3588 return sock_no_sendpage(sock, page, offset, size, flags);
3590 EXPORT_SYMBOL(kernel_sendpage);
3593 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3596 * @offset: page offset
3597 * @size: total size in bytes
3598 * @flags: flags (MSG_DONTWAIT, ...)
3600 * Returns the total amount sent in bytes or an error.
3601 * Caller must hold @sk.
3604 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3605 size_t size, int flags)
3607 struct socket *sock = sk->sk_socket;
3609 if (sock->ops->sendpage_locked)
3610 return sock->ops->sendpage_locked(sk, page, offset, size,
3613 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3615 EXPORT_SYMBOL(kernel_sendpage_locked);
3618 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3620 * @how: connection part
3622 * Returns 0 or an error.
3625 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3627 return sock->ops->shutdown(sock, how);
3629 EXPORT_SYMBOL(kernel_sock_shutdown);
3632 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3635 * This routine returns the IP overhead imposed by a socket i.e.
3636 * the length of the underlying IP header, depending on whether
3637 * this is an IPv4 or IPv6 socket and the length from IP options turned
3638 * on at the socket. Assumes that the caller has a lock on the socket.
3641 u32 kernel_sock_ip_overhead(struct sock *sk)
3643 struct inet_sock *inet;
3644 struct ip_options_rcu *opt;
3646 #if IS_ENABLED(CONFIG_IPV6)
3647 struct ipv6_pinfo *np;
3648 struct ipv6_txoptions *optv6 = NULL;
3649 #endif /* IS_ENABLED(CONFIG_IPV6) */
3654 switch (sk->sk_family) {
3657 overhead += sizeof(struct iphdr);
3658 opt = rcu_dereference_protected(inet->inet_opt,
3659 sock_owned_by_user(sk));
3661 overhead += opt->opt.optlen;
3663 #if IS_ENABLED(CONFIG_IPV6)
3666 overhead += sizeof(struct ipv6hdr);
3668 optv6 = rcu_dereference_protected(np->opt,
3669 sock_owned_by_user(sk));
3671 overhead += (optv6->opt_flen + optv6->opt_nflen);
3673 #endif /* IS_ENABLED(CONFIG_IPV6) */
3674 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3678 EXPORT_SYMBOL(kernel_sock_ip_overhead);