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>
91 #include <linux/io_uring/net.h>
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/termios.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 #include <linux/ptp_clock_kernel.h>
111 #include <trace/events/sock.h>
113 #ifdef CONFIG_NET_RX_BUSY_POLL
114 unsigned int sysctl_net_busy_read __read_mostly;
115 unsigned int sysctl_net_busy_poll __read_mostly;
118 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
119 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
122 static int sock_close(struct inode *inode, struct file *file);
123 static __poll_t sock_poll(struct file *file,
124 struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
127 static long compat_sock_ioctl(struct file *file,
128 unsigned int cmd, unsigned long arg);
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
134 static void sock_splice_eof(struct file *file);
136 #ifdef CONFIG_PROC_FS
137 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
139 struct socket *sock = f->private_data;
140 const struct proto_ops *ops = READ_ONCE(sock->ops);
142 if (ops->show_fdinfo)
143 ops->show_fdinfo(m, sock);
146 #define sock_show_fdinfo NULL
150 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151 * in the operation structures but are done directly via the socketcall() multiplexor.
154 static const struct file_operations socket_file_ops = {
155 .owner = THIS_MODULE,
157 .read_iter = sock_read_iter,
158 .write_iter = sock_write_iter,
160 .unlocked_ioctl = sock_ioctl,
162 .compat_ioctl = compat_sock_ioctl,
164 .uring_cmd = io_uring_cmd_sock,
166 .release = sock_close,
167 .fasync = sock_fasync,
168 .splice_write = splice_to_socket,
169 .splice_read = sock_splice_read,
170 .splice_eof = sock_splice_eof,
171 .show_fdinfo = sock_show_fdinfo,
174 static const char * const pf_family_names[] = {
175 [PF_UNSPEC] = "PF_UNSPEC",
176 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
177 [PF_INET] = "PF_INET",
178 [PF_AX25] = "PF_AX25",
180 [PF_APPLETALK] = "PF_APPLETALK",
181 [PF_NETROM] = "PF_NETROM",
182 [PF_BRIDGE] = "PF_BRIDGE",
183 [PF_ATMPVC] = "PF_ATMPVC",
185 [PF_INET6] = "PF_INET6",
186 [PF_ROSE] = "PF_ROSE",
187 [PF_DECnet] = "PF_DECnet",
188 [PF_NETBEUI] = "PF_NETBEUI",
189 [PF_SECURITY] = "PF_SECURITY",
191 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
192 [PF_PACKET] = "PF_PACKET",
194 [PF_ECONET] = "PF_ECONET",
195 [PF_ATMSVC] = "PF_ATMSVC",
198 [PF_IRDA] = "PF_IRDA",
199 [PF_PPPOX] = "PF_PPPOX",
200 [PF_WANPIPE] = "PF_WANPIPE",
203 [PF_MPLS] = "PF_MPLS",
205 [PF_TIPC] = "PF_TIPC",
206 [PF_BLUETOOTH] = "PF_BLUETOOTH",
207 [PF_IUCV] = "PF_IUCV",
208 [PF_RXRPC] = "PF_RXRPC",
209 [PF_ISDN] = "PF_ISDN",
210 [PF_PHONET] = "PF_PHONET",
211 [PF_IEEE802154] = "PF_IEEE802154",
212 [PF_CAIF] = "PF_CAIF",
215 [PF_VSOCK] = "PF_VSOCK",
217 [PF_QIPCRTR] = "PF_QIPCRTR",
220 [PF_MCTP] = "PF_MCTP",
224 * The protocol list. Each protocol is registered in here.
227 static DEFINE_SPINLOCK(net_family_lock);
228 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
232 * Move socket addresses back and forth across the kernel/user
233 * divide and look after the messy bits.
237 * move_addr_to_kernel - copy a socket address into kernel space
238 * @uaddr: Address in user space
239 * @kaddr: Address in kernel space
240 * @ulen: Length in user space
242 * The address is copied into kernel space. If the provided address is
243 * too long an error code of -EINVAL is returned. If the copy gives
244 * invalid addresses -EFAULT is returned. On a success 0 is returned.
247 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
249 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
253 if (copy_from_user(kaddr, uaddr, ulen))
255 return audit_sockaddr(ulen, kaddr);
259 * move_addr_to_user - copy an address to user space
260 * @kaddr: kernel space address
261 * @klen: length of address in kernel
262 * @uaddr: user space address
263 * @ulen: pointer to user length field
265 * The value pointed to by ulen on entry is the buffer length available.
266 * This is overwritten with the buffer space used. -EINVAL is returned
267 * if an overlong buffer is specified or a negative buffer size. -EFAULT
268 * is returned if either the buffer or the length field are not
270 * After copying the data up to the limit the user specifies, the true
271 * length of the data is written over the length limit the user
272 * specified. Zero is returned for a success.
275 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
276 void __user *uaddr, int __user *ulen)
281 BUG_ON(klen > sizeof(struct sockaddr_storage));
282 err = get_user(len, ulen);
290 if (audit_sockaddr(klen, kaddr))
292 if (copy_to_user(uaddr, kaddr, len))
296 * "fromlen shall refer to the value before truncation.."
299 return __put_user(klen, ulen);
302 static struct kmem_cache *sock_inode_cachep __ro_after_init;
304 static struct inode *sock_alloc_inode(struct super_block *sb)
306 struct socket_alloc *ei;
308 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
311 init_waitqueue_head(&ei->socket.wq.wait);
312 ei->socket.wq.fasync_list = NULL;
313 ei->socket.wq.flags = 0;
315 ei->socket.state = SS_UNCONNECTED;
316 ei->socket.flags = 0;
317 ei->socket.ops = NULL;
318 ei->socket.sk = NULL;
319 ei->socket.file = NULL;
321 return &ei->vfs_inode;
324 static void sock_free_inode(struct inode *inode)
326 struct socket_alloc *ei;
328 ei = container_of(inode, struct socket_alloc, vfs_inode);
329 kmem_cache_free(sock_inode_cachep, ei);
332 static void init_once(void *foo)
334 struct socket_alloc *ei = (struct socket_alloc *)foo;
336 inode_init_once(&ei->vfs_inode);
339 static void init_inodecache(void)
341 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
342 sizeof(struct socket_alloc),
344 (SLAB_HWCACHE_ALIGN |
345 SLAB_RECLAIM_ACCOUNT |
348 BUG_ON(sock_inode_cachep == NULL);
351 static const struct super_operations sockfs_ops = {
352 .alloc_inode = sock_alloc_inode,
353 .free_inode = sock_free_inode,
354 .statfs = simple_statfs,
358 * sockfs_dname() is called from d_path().
360 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
362 return dynamic_dname(buffer, buflen, "socket:[%lu]",
363 d_inode(dentry)->i_ino);
366 static const struct dentry_operations sockfs_dentry_operations = {
367 .d_dname = sockfs_dname,
370 static int sockfs_xattr_get(const struct xattr_handler *handler,
371 struct dentry *dentry, struct inode *inode,
372 const char *suffix, void *value, size_t size)
375 if (dentry->d_name.len + 1 > size)
377 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
379 return dentry->d_name.len + 1;
382 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
383 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
384 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
386 static const struct xattr_handler sockfs_xattr_handler = {
387 .name = XATTR_NAME_SOCKPROTONAME,
388 .get = sockfs_xattr_get,
391 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
392 struct mnt_idmap *idmap,
393 struct dentry *dentry, struct inode *inode,
394 const char *suffix, const void *value,
395 size_t size, int flags)
397 /* Handled by LSM. */
401 static const struct xattr_handler sockfs_security_xattr_handler = {
402 .prefix = XATTR_SECURITY_PREFIX,
403 .set = sockfs_security_xattr_set,
406 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
407 &sockfs_xattr_handler,
408 &sockfs_security_xattr_handler,
412 static int sockfs_init_fs_context(struct fs_context *fc)
414 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
417 ctx->ops = &sockfs_ops;
418 ctx->dops = &sockfs_dentry_operations;
419 ctx->xattr = sockfs_xattr_handlers;
423 static struct vfsmount *sock_mnt __read_mostly;
425 static struct file_system_type sock_fs_type = {
427 .init_fs_context = sockfs_init_fs_context,
428 .kill_sb = kill_anon_super,
432 * Obtains the first available file descriptor and sets it up for use.
434 * These functions create file structures and maps them to fd space
435 * of the current process. On success it returns file descriptor
436 * and file struct implicitly stored in sock->file.
437 * Note that another thread may close file descriptor before we return
438 * from this function. We use the fact that now we do not refer
439 * to socket after mapping. If one day we will need it, this
440 * function will increment ref. count on file by 1.
442 * In any case returned fd MAY BE not valid!
443 * This race condition is unavoidable
444 * with shared fd spaces, we cannot solve it inside kernel,
445 * but we take care of internal coherence yet.
449 * sock_alloc_file - Bind a &socket to a &file
451 * @flags: file status flags
452 * @dname: protocol name
454 * Returns the &file bound with @sock, implicitly storing it
455 * in sock->file. If dname is %NULL, sets to "".
457 * On failure @sock is released, and an ERR pointer is returned.
459 * This function uses GFP_KERNEL internally.
462 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
467 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
469 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
470 O_RDWR | (flags & O_NONBLOCK),
477 file->f_mode |= FMODE_NOWAIT;
479 file->private_data = sock;
480 stream_open(SOCK_INODE(sock), file);
483 EXPORT_SYMBOL(sock_alloc_file);
485 static int sock_map_fd(struct socket *sock, int flags)
487 struct file *newfile;
488 int fd = get_unused_fd_flags(flags);
489 if (unlikely(fd < 0)) {
494 newfile = sock_alloc_file(sock, flags, NULL);
495 if (!IS_ERR(newfile)) {
496 fd_install(fd, newfile);
501 return PTR_ERR(newfile);
505 * sock_from_file - Return the &socket bounded to @file.
508 * On failure returns %NULL.
511 struct socket *sock_from_file(struct file *file)
513 if (file->f_op == &socket_file_ops)
514 return file->private_data; /* set in sock_alloc_file */
518 EXPORT_SYMBOL(sock_from_file);
521 * sockfd_lookup - Go from a file number to its socket slot
523 * @err: pointer to an error code return
525 * The file handle passed in is locked and the socket it is bound
526 * to is returned. If an error occurs the err pointer is overwritten
527 * with a negative errno code and NULL is returned. The function checks
528 * for both invalid handles and passing a handle which is not a socket.
530 * On a success the socket object pointer is returned.
533 struct socket *sockfd_lookup(int fd, int *err)
544 sock = sock_from_file(file);
551 EXPORT_SYMBOL(sockfd_lookup);
553 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
555 struct fd f = fdget(fd);
560 sock = sock_from_file(f.file);
562 *fput_needed = f.flags & FDPUT_FPUT;
571 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
577 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
587 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
592 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
599 static int sockfs_setattr(struct mnt_idmap *idmap,
600 struct dentry *dentry, struct iattr *iattr)
602 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
604 if (!err && (iattr->ia_valid & ATTR_UID)) {
605 struct socket *sock = SOCKET_I(d_inode(dentry));
608 sock->sk->sk_uid = iattr->ia_uid;
616 static const struct inode_operations sockfs_inode_ops = {
617 .listxattr = sockfs_listxattr,
618 .setattr = sockfs_setattr,
622 * sock_alloc - allocate a socket
624 * Allocate a new inode and socket object. The two are bound together
625 * and initialised. The socket is then returned. If we are out of inodes
626 * NULL is returned. This functions uses GFP_KERNEL internally.
629 struct socket *sock_alloc(void)
634 inode = new_inode_pseudo(sock_mnt->mnt_sb);
638 sock = SOCKET_I(inode);
640 inode->i_ino = get_next_ino();
641 inode->i_mode = S_IFSOCK | S_IRWXUGO;
642 inode->i_uid = current_fsuid();
643 inode->i_gid = current_fsgid();
644 inode->i_op = &sockfs_inode_ops;
648 EXPORT_SYMBOL(sock_alloc);
650 static void __sock_release(struct socket *sock, struct inode *inode)
652 const struct proto_ops *ops = READ_ONCE(sock->ops);
655 struct module *owner = ops->owner;
667 if (sock->wq.fasync_list)
668 pr_err("%s: fasync list not empty!\n", __func__);
671 iput(SOCK_INODE(sock));
678 * sock_release - close a socket
679 * @sock: socket to close
681 * The socket is released from the protocol stack if it has a release
682 * callback, and the inode is then released if the socket is bound to
683 * an inode not a file.
685 void sock_release(struct socket *sock)
687 __sock_release(sock, NULL);
689 EXPORT_SYMBOL(sock_release);
691 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
693 u8 flags = *tx_flags;
695 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
696 flags |= SKBTX_HW_TSTAMP;
698 /* PTP hardware clocks can provide a free running cycle counter
699 * as a time base for virtual clocks. Tell driver to use the
700 * free running cycle counter for timestamp if socket is bound
703 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
704 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
707 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
708 flags |= SKBTX_SW_TSTAMP;
710 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
711 flags |= SKBTX_SCHED_TSTAMP;
715 EXPORT_SYMBOL(__sock_tx_timestamp);
717 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
719 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
722 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
725 trace_sock_send_length(sk, ret, 0);
728 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
730 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
731 inet_sendmsg, sock, msg,
733 BUG_ON(ret == -EIOCBQUEUED);
735 if (trace_sock_send_length_enabled())
736 call_trace_sock_send_length(sock->sk, ret, 0);
740 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
742 int err = security_socket_sendmsg(sock, msg,
745 return err ?: sock_sendmsg_nosec(sock, msg);
749 * sock_sendmsg - send a message through @sock
751 * @msg: message to send
753 * Sends @msg through @sock, passing through LSM.
754 * Returns the number of bytes sent, or an error code.
756 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
758 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
759 struct sockaddr_storage address;
760 int save_len = msg->msg_namelen;
764 memcpy(&address, msg->msg_name, msg->msg_namelen);
765 msg->msg_name = &address;
768 ret = __sock_sendmsg(sock, msg);
769 msg->msg_name = save_addr;
770 msg->msg_namelen = save_len;
774 EXPORT_SYMBOL(sock_sendmsg);
777 * kernel_sendmsg - send a message through @sock (kernel-space)
779 * @msg: message header
781 * @num: vec array length
782 * @size: total message data size
784 * Builds the message data with @vec and sends it through @sock.
785 * Returns the number of bytes sent, or an error code.
788 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
789 struct kvec *vec, size_t num, size_t size)
791 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
792 return sock_sendmsg(sock, msg);
794 EXPORT_SYMBOL(kernel_sendmsg);
797 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
799 * @msg: message header
800 * @vec: output s/g array
801 * @num: output s/g array length
802 * @size: total message data size
804 * Builds the message data with @vec and sends it through @sock.
805 * Returns the number of bytes sent, or an error code.
806 * Caller must hold @sk.
809 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
810 struct kvec *vec, size_t num, size_t size)
812 struct socket *sock = sk->sk_socket;
813 const struct proto_ops *ops = READ_ONCE(sock->ops);
815 if (!ops->sendmsg_locked)
816 return sock_no_sendmsg_locked(sk, msg, size);
818 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
820 return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
822 EXPORT_SYMBOL(kernel_sendmsg_locked);
824 static bool skb_is_err_queue(const struct sk_buff *skb)
826 /* pkt_type of skbs enqueued on the error queue are set to
827 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
828 * in recvmsg, since skbs received on a local socket will never
829 * have a pkt_type of PACKET_OUTGOING.
831 return skb->pkt_type == PACKET_OUTGOING;
834 /* On transmit, software and hardware timestamps are returned independently.
835 * As the two skb clones share the hardware timestamp, which may be updated
836 * before the software timestamp is received, a hardware TX timestamp may be
837 * returned only if there is no software TX timestamp. Ignore false software
838 * timestamps, which may be made in the __sock_recv_timestamp() call when the
839 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
840 * hardware timestamp.
842 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
844 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
847 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
849 bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
850 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
851 struct net_device *orig_dev;
855 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
857 *if_index = orig_dev->ifindex;
858 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
860 hwtstamp = shhwtstamps->hwtstamp;
867 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
870 struct scm_ts_pktinfo ts_pktinfo;
871 struct net_device *orig_dev;
873 if (!skb_mac_header_was_set(skb))
876 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
880 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
882 if_index = orig_dev->ifindex;
885 ts_pktinfo.if_index = if_index;
887 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
888 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
889 sizeof(ts_pktinfo), &ts_pktinfo);
893 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
895 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
898 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
899 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
900 struct scm_timestamping_internal tss;
901 int empty = 1, false_tstamp = 0;
902 struct skb_shared_hwtstamps *shhwtstamps =
908 /* Race occurred between timestamp enabling and packet
909 receiving. Fill in the current time for now. */
910 if (need_software_tstamp && skb->tstamp == 0) {
911 __net_timestamp(skb);
915 if (need_software_tstamp) {
916 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
918 struct __kernel_sock_timeval tv;
920 skb_get_new_timestamp(skb, &tv);
921 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
924 struct __kernel_old_timeval tv;
926 skb_get_timestamp(skb, &tv);
927 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
932 struct __kernel_timespec ts;
934 skb_get_new_timestampns(skb, &ts);
935 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
938 struct __kernel_old_timespec ts;
940 skb_get_timestampns(skb, &ts);
941 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
947 memset(&tss, 0, sizeof(tss));
948 tsflags = READ_ONCE(sk->sk_tsflags);
949 if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
950 (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
951 skb_is_err_queue(skb) ||
952 !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
953 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
956 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
957 (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
958 skb_is_err_queue(skb) ||
959 !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
960 !skb_is_swtx_tstamp(skb, false_tstamp)) {
962 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
963 hwtstamp = get_timestamp(sk, skb, &if_index);
965 hwtstamp = shhwtstamps->hwtstamp;
967 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
968 hwtstamp = ptp_convert_timestamp(&hwtstamp,
969 READ_ONCE(sk->sk_bind_phc));
971 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
974 if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
975 !skb_is_err_queue(skb))
976 put_ts_pktinfo(msg, skb, if_index);
980 if (sock_flag(sk, SOCK_TSTAMP_NEW))
981 put_cmsg_scm_timestamping64(msg, &tss);
983 put_cmsg_scm_timestamping(msg, &tss);
985 if (skb_is_err_queue(skb) && skb->len &&
986 SKB_EXT_ERR(skb)->opt_stats)
987 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
988 skb->len, skb->data);
991 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
993 #ifdef CONFIG_WIRELESS
994 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
999 if (!sock_flag(sk, SOCK_WIFI_STATUS))
1001 if (!skb->wifi_acked_valid)
1004 ack = skb->wifi_acked;
1006 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1008 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1011 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1012 struct sk_buff *skb)
1014 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1015 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1016 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1019 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1020 struct sk_buff *skb)
1022 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1023 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1024 __u32 mark = skb->mark;
1026 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1030 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1031 struct sk_buff *skb)
1033 sock_recv_timestamp(msg, sk, skb);
1034 sock_recv_drops(msg, sk, skb);
1035 sock_recv_mark(msg, sk, skb);
1037 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1039 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1041 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1044 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1046 trace_sock_recv_length(sk, ret, flags);
1049 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1052 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1054 inet_recvmsg, sock, msg,
1055 msg_data_left(msg), flags);
1056 if (trace_sock_recv_length_enabled())
1057 call_trace_sock_recv_length(sock->sk, ret, flags);
1062 * sock_recvmsg - receive a message from @sock
1064 * @msg: message to receive
1065 * @flags: message flags
1067 * Receives @msg from @sock, passing through LSM. Returns the total number
1068 * of bytes received, or an error.
1070 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1072 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1074 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1076 EXPORT_SYMBOL(sock_recvmsg);
1079 * kernel_recvmsg - Receive a message from a socket (kernel space)
1080 * @sock: The socket to receive the message from
1081 * @msg: Received message
1082 * @vec: Input s/g array for message data
1083 * @num: Size of input s/g array
1084 * @size: Number of bytes to read
1085 * @flags: Message flags (MSG_DONTWAIT, etc...)
1087 * On return the msg structure contains the scatter/gather array passed in the
1088 * vec argument. The array is modified so that it consists of the unfilled
1089 * portion of the original array.
1091 * The returned value is the total number of bytes received, or an error.
1094 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1095 struct kvec *vec, size_t num, size_t size, int flags)
1097 msg->msg_control_is_user = false;
1098 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1099 return sock_recvmsg(sock, msg, flags);
1101 EXPORT_SYMBOL(kernel_recvmsg);
1103 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1104 struct pipe_inode_info *pipe, size_t len,
1107 struct socket *sock = file->private_data;
1108 const struct proto_ops *ops;
1110 ops = READ_ONCE(sock->ops);
1111 if (unlikely(!ops->splice_read))
1112 return copy_splice_read(file, ppos, pipe, len, flags);
1114 return ops->splice_read(sock, ppos, pipe, len, flags);
1117 static void sock_splice_eof(struct file *file)
1119 struct socket *sock = file->private_data;
1120 const struct proto_ops *ops;
1122 ops = READ_ONCE(sock->ops);
1123 if (ops->splice_eof)
1124 ops->splice_eof(sock);
1127 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1129 struct file *file = iocb->ki_filp;
1130 struct socket *sock = file->private_data;
1131 struct msghdr msg = {.msg_iter = *to,
1135 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1136 msg.msg_flags = MSG_DONTWAIT;
1138 if (iocb->ki_pos != 0)
1141 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1144 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1149 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1151 struct file *file = iocb->ki_filp;
1152 struct socket *sock = file->private_data;
1153 struct msghdr msg = {.msg_iter = *from,
1157 if (iocb->ki_pos != 0)
1160 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1161 msg.msg_flags = MSG_DONTWAIT;
1163 if (sock->type == SOCK_SEQPACKET)
1164 msg.msg_flags |= MSG_EOR;
1166 res = __sock_sendmsg(sock, &msg);
1167 *from = msg.msg_iter;
1172 * Atomic setting of ioctl hooks to avoid race
1173 * with module unload.
1176 static DEFINE_MUTEX(br_ioctl_mutex);
1177 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1178 unsigned int cmd, struct ifreq *ifr,
1181 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1182 unsigned int cmd, struct ifreq *ifr,
1185 mutex_lock(&br_ioctl_mutex);
1186 br_ioctl_hook = hook;
1187 mutex_unlock(&br_ioctl_mutex);
1189 EXPORT_SYMBOL(brioctl_set);
1191 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1192 struct ifreq *ifr, void __user *uarg)
1197 request_module("bridge");
1199 mutex_lock(&br_ioctl_mutex);
1201 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1202 mutex_unlock(&br_ioctl_mutex);
1207 static DEFINE_MUTEX(vlan_ioctl_mutex);
1208 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1210 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1212 mutex_lock(&vlan_ioctl_mutex);
1213 vlan_ioctl_hook = hook;
1214 mutex_unlock(&vlan_ioctl_mutex);
1216 EXPORT_SYMBOL(vlan_ioctl_set);
1218 static long sock_do_ioctl(struct net *net, struct socket *sock,
1219 unsigned int cmd, unsigned long arg)
1221 const struct proto_ops *ops = READ_ONCE(sock->ops);
1225 void __user *argp = (void __user *)arg;
1228 err = ops->ioctl(sock, cmd, arg);
1231 * If this ioctl is unknown try to hand it down
1232 * to the NIC driver.
1234 if (err != -ENOIOCTLCMD)
1237 if (!is_socket_ioctl_cmd(cmd))
1240 if (get_user_ifreq(&ifr, &data, argp))
1242 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1243 if (!err && need_copyout)
1244 if (put_user_ifreq(&ifr, argp))
1251 * With an ioctl, arg may well be a user mode pointer, but we don't know
1252 * what to do with it - that's up to the protocol still.
1255 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1257 const struct proto_ops *ops;
1258 struct socket *sock;
1260 void __user *argp = (void __user *)arg;
1264 sock = file->private_data;
1265 ops = READ_ONCE(sock->ops);
1268 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1272 if (get_user_ifreq(&ifr, &data, argp))
1274 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1275 if (!err && need_copyout)
1276 if (put_user_ifreq(&ifr, argp))
1279 #ifdef CONFIG_WEXT_CORE
1280 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1281 err = wext_handle_ioctl(net, cmd, argp);
1288 if (get_user(pid, (int __user *)argp))
1290 err = f_setown(sock->file, pid, 1);
1294 err = put_user(f_getown(sock->file),
1295 (int __user *)argp);
1301 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1306 if (!vlan_ioctl_hook)
1307 request_module("8021q");
1309 mutex_lock(&vlan_ioctl_mutex);
1310 if (vlan_ioctl_hook)
1311 err = vlan_ioctl_hook(net, argp);
1312 mutex_unlock(&vlan_ioctl_mutex);
1316 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1319 err = open_related_ns(&net->ns, get_net_ns);
1321 case SIOCGSTAMP_OLD:
1322 case SIOCGSTAMPNS_OLD:
1323 if (!ops->gettstamp) {
1327 err = ops->gettstamp(sock, argp,
1328 cmd == SIOCGSTAMP_OLD,
1329 !IS_ENABLED(CONFIG_64BIT));
1331 case SIOCGSTAMP_NEW:
1332 case SIOCGSTAMPNS_NEW:
1333 if (!ops->gettstamp) {
1337 err = ops->gettstamp(sock, argp,
1338 cmd == SIOCGSTAMP_NEW,
1343 err = dev_ifconf(net, argp);
1347 err = sock_do_ioctl(net, sock, cmd, arg);
1354 * sock_create_lite - creates a socket
1355 * @family: protocol family (AF_INET, ...)
1356 * @type: communication type (SOCK_STREAM, ...)
1357 * @protocol: protocol (0, ...)
1360 * Creates a new socket and assigns it to @res, passing through LSM.
1361 * The new socket initialization is not complete, see kernel_accept().
1362 * Returns 0 or an error. On failure @res is set to %NULL.
1363 * This function internally uses GFP_KERNEL.
1366 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1369 struct socket *sock = NULL;
1371 err = security_socket_create(family, type, protocol, 1);
1375 sock = sock_alloc();
1382 err = security_socket_post_create(sock, family, type, protocol, 1);
1394 EXPORT_SYMBOL(sock_create_lite);
1396 /* No kernel lock held - perfect */
1397 static __poll_t sock_poll(struct file *file, poll_table *wait)
1399 struct socket *sock = file->private_data;
1400 const struct proto_ops *ops = READ_ONCE(sock->ops);
1401 __poll_t events = poll_requested_events(wait), flag = 0;
1406 if (sk_can_busy_loop(sock->sk)) {
1407 /* poll once if requested by the syscall */
1408 if (events & POLL_BUSY_LOOP)
1409 sk_busy_loop(sock->sk, 1);
1411 /* if this socket can poll_ll, tell the system call */
1412 flag = POLL_BUSY_LOOP;
1415 return ops->poll(file, sock, wait) | flag;
1418 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1420 struct socket *sock = file->private_data;
1422 return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1425 static int sock_close(struct inode *inode, struct file *filp)
1427 __sock_release(SOCKET_I(inode), inode);
1432 * Update the socket async list
1434 * Fasync_list locking strategy.
1436 * 1. fasync_list is modified only under process context socket lock
1437 * i.e. under semaphore.
1438 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1439 * or under socket lock
1442 static int sock_fasync(int fd, struct file *filp, int on)
1444 struct socket *sock = filp->private_data;
1445 struct sock *sk = sock->sk;
1446 struct socket_wq *wq = &sock->wq;
1452 fasync_helper(fd, filp, on, &wq->fasync_list);
1454 if (!wq->fasync_list)
1455 sock_reset_flag(sk, SOCK_FASYNC);
1457 sock_set_flag(sk, SOCK_FASYNC);
1463 /* This function may be called only under rcu_lock */
1465 int sock_wake_async(struct socket_wq *wq, int how, int band)
1467 if (!wq || !wq->fasync_list)
1471 case SOCK_WAKE_WAITD:
1472 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1475 case SOCK_WAKE_SPACE:
1476 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1481 kill_fasync(&wq->fasync_list, SIGIO, band);
1484 kill_fasync(&wq->fasync_list, SIGURG, band);
1489 EXPORT_SYMBOL(sock_wake_async);
1492 * __sock_create - creates a socket
1493 * @net: net namespace
1494 * @family: protocol family (AF_INET, ...)
1495 * @type: communication type (SOCK_STREAM, ...)
1496 * @protocol: protocol (0, ...)
1498 * @kern: boolean for kernel space sockets
1500 * Creates a new socket and assigns it to @res, passing through LSM.
1501 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1502 * be set to true if the socket resides in kernel space.
1503 * This function internally uses GFP_KERNEL.
1506 int __sock_create(struct net *net, int family, int type, int protocol,
1507 struct socket **res, int kern)
1510 struct socket *sock;
1511 const struct net_proto_family *pf;
1514 * Check protocol is in range
1516 if (family < 0 || family >= NPROTO)
1517 return -EAFNOSUPPORT;
1518 if (type < 0 || type >= SOCK_MAX)
1523 This uglymoron is moved from INET layer to here to avoid
1524 deadlock in module load.
1526 if (family == PF_INET && type == SOCK_PACKET) {
1527 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1532 err = security_socket_create(family, type, protocol, kern);
1537 * Allocate the socket and allow the family to set things up. if
1538 * the protocol is 0, the family is instructed to select an appropriate
1541 sock = sock_alloc();
1543 net_warn_ratelimited("socket: no more sockets\n");
1544 return -ENFILE; /* Not exactly a match, but its the
1545 closest posix thing */
1550 #ifdef CONFIG_MODULES
1551 /* Attempt to load a protocol module if the find failed.
1553 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1554 * requested real, full-featured networking support upon configuration.
1555 * Otherwise module support will break!
1557 if (rcu_access_pointer(net_families[family]) == NULL)
1558 request_module("net-pf-%d", family);
1562 pf = rcu_dereference(net_families[family]);
1563 err = -EAFNOSUPPORT;
1568 * We will call the ->create function, that possibly is in a loadable
1569 * module, so we have to bump that loadable module refcnt first.
1571 if (!try_module_get(pf->owner))
1574 /* Now protected by module ref count */
1577 err = pf->create(net, sock, protocol, kern);
1579 goto out_module_put;
1582 * Now to bump the refcnt of the [loadable] module that owns this
1583 * socket at sock_release time we decrement its refcnt.
1585 if (!try_module_get(sock->ops->owner))
1586 goto out_module_busy;
1589 * Now that we're done with the ->create function, the [loadable]
1590 * module can have its refcnt decremented
1592 module_put(pf->owner);
1593 err = security_socket_post_create(sock, family, type, protocol, kern);
1595 goto out_sock_release;
1601 err = -EAFNOSUPPORT;
1604 module_put(pf->owner);
1611 goto out_sock_release;
1613 EXPORT_SYMBOL(__sock_create);
1616 * sock_create - creates a socket
1617 * @family: protocol family (AF_INET, ...)
1618 * @type: communication type (SOCK_STREAM, ...)
1619 * @protocol: protocol (0, ...)
1622 * A wrapper around __sock_create().
1623 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1626 int sock_create(int family, int type, int protocol, struct socket **res)
1628 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1630 EXPORT_SYMBOL(sock_create);
1633 * sock_create_kern - creates a socket (kernel space)
1634 * @net: net namespace
1635 * @family: protocol family (AF_INET, ...)
1636 * @type: communication type (SOCK_STREAM, ...)
1637 * @protocol: protocol (0, ...)
1640 * A wrapper around __sock_create().
1641 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1644 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1646 return __sock_create(net, family, type, protocol, res, 1);
1648 EXPORT_SYMBOL(sock_create_kern);
1650 static struct socket *__sys_socket_create(int family, int type, int protocol)
1652 struct socket *sock;
1655 /* Check the SOCK_* constants for consistency. */
1656 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1657 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1658 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1659 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1661 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1662 return ERR_PTR(-EINVAL);
1663 type &= SOCK_TYPE_MASK;
1665 retval = sock_create(family, type, protocol, &sock);
1667 return ERR_PTR(retval);
1672 struct file *__sys_socket_file(int family, int type, int protocol)
1674 struct socket *sock;
1677 sock = __sys_socket_create(family, type, protocol);
1679 return ERR_CAST(sock);
1681 flags = type & ~SOCK_TYPE_MASK;
1682 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1683 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1685 return sock_alloc_file(sock, flags, NULL);
1688 /* A hook for bpf progs to attach to and update socket protocol.
1690 * A static noinline declaration here could cause the compiler to
1691 * optimize away the function. A global noinline declaration will
1692 * keep the definition, but may optimize away the callsite.
1693 * Therefore, __weak is needed to ensure that the call is still
1694 * emitted, by telling the compiler that we don't know what the
1695 * function might eventually be.
1700 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1707 int __sys_socket(int family, int type, int protocol)
1709 struct socket *sock;
1712 sock = __sys_socket_create(family, type,
1713 update_socket_protocol(family, type, protocol));
1715 return PTR_ERR(sock);
1717 flags = type & ~SOCK_TYPE_MASK;
1718 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1719 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1721 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1724 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1726 return __sys_socket(family, type, protocol);
1730 * Create a pair of connected sockets.
1733 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1735 struct socket *sock1, *sock2;
1737 struct file *newfile1, *newfile2;
1740 flags = type & ~SOCK_TYPE_MASK;
1741 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1743 type &= SOCK_TYPE_MASK;
1745 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1746 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1749 * reserve descriptors and make sure we won't fail
1750 * to return them to userland.
1752 fd1 = get_unused_fd_flags(flags);
1753 if (unlikely(fd1 < 0))
1756 fd2 = get_unused_fd_flags(flags);
1757 if (unlikely(fd2 < 0)) {
1762 err = put_user(fd1, &usockvec[0]);
1766 err = put_user(fd2, &usockvec[1]);
1771 * Obtain the first socket and check if the underlying protocol
1772 * supports the socketpair call.
1775 err = sock_create(family, type, protocol, &sock1);
1776 if (unlikely(err < 0))
1779 err = sock_create(family, type, protocol, &sock2);
1780 if (unlikely(err < 0)) {
1781 sock_release(sock1);
1785 err = security_socket_socketpair(sock1, sock2);
1786 if (unlikely(err)) {
1787 sock_release(sock2);
1788 sock_release(sock1);
1792 err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1793 if (unlikely(err < 0)) {
1794 sock_release(sock2);
1795 sock_release(sock1);
1799 newfile1 = sock_alloc_file(sock1, flags, NULL);
1800 if (IS_ERR(newfile1)) {
1801 err = PTR_ERR(newfile1);
1802 sock_release(sock2);
1806 newfile2 = sock_alloc_file(sock2, flags, NULL);
1807 if (IS_ERR(newfile2)) {
1808 err = PTR_ERR(newfile2);
1813 audit_fd_pair(fd1, fd2);
1815 fd_install(fd1, newfile1);
1816 fd_install(fd2, newfile2);
1825 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1826 int __user *, usockvec)
1828 return __sys_socketpair(family, type, protocol, usockvec);
1831 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1836 err = security_socket_bind(sock, (struct sockaddr *)address,
1839 err = READ_ONCE(sock->ops)->bind(sock,
1840 (struct sockaddr *)address,
1846 * Bind a name to a socket. Nothing much to do here since it's
1847 * the protocol's responsibility to handle the local address.
1849 * We move the socket address to kernel space before we call
1850 * the protocol layer (having also checked the address is ok).
1853 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1855 struct socket *sock;
1856 struct sockaddr_storage address;
1857 int err, fput_needed;
1859 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1861 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1863 err = __sys_bind_socket(sock, &address, addrlen);
1864 fput_light(sock->file, fput_needed);
1869 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1871 return __sys_bind(fd, umyaddr, addrlen);
1875 * Perform a listen. Basically, we allow the protocol to do anything
1876 * necessary for a listen, and if that works, we mark the socket as
1877 * ready for listening.
1879 int __sys_listen_socket(struct socket *sock, int backlog)
1883 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1884 if ((unsigned int)backlog > somaxconn)
1885 backlog = somaxconn;
1887 err = security_socket_listen(sock, backlog);
1889 err = READ_ONCE(sock->ops)->listen(sock, backlog);
1893 int __sys_listen(int fd, int backlog)
1895 struct socket *sock;
1896 int err, fput_needed;
1898 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1900 err = __sys_listen_socket(sock, backlog);
1901 fput_light(sock->file, fput_needed);
1906 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1908 return __sys_listen(fd, backlog);
1911 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1912 struct sockaddr __user *upeer_sockaddr,
1913 int __user *upeer_addrlen, int flags)
1915 struct socket *sock, *newsock;
1916 struct file *newfile;
1918 struct sockaddr_storage address;
1919 const struct proto_ops *ops;
1921 sock = sock_from_file(file);
1923 return ERR_PTR(-ENOTSOCK);
1925 newsock = sock_alloc();
1927 return ERR_PTR(-ENFILE);
1928 ops = READ_ONCE(sock->ops);
1930 newsock->type = sock->type;
1934 * We don't need try_module_get here, as the listening socket (sock)
1935 * has the protocol module (sock->ops->owner) held.
1937 __module_get(ops->owner);
1939 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1940 if (IS_ERR(newfile))
1943 err = security_socket_accept(sock, newsock);
1947 arg->flags |= sock->file->f_flags;
1948 err = ops->accept(sock, newsock, arg);
1952 if (upeer_sockaddr) {
1953 len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1955 err = -ECONNABORTED;
1958 err = move_addr_to_user(&address,
1959 len, upeer_sockaddr, upeer_addrlen);
1964 /* File flags are not inherited via accept() unlike another OSes. */
1968 return ERR_PTR(err);
1971 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1972 int __user *upeer_addrlen, int flags)
1974 struct proto_accept_arg arg = { };
1975 struct file *newfile;
1978 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1981 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1982 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1984 newfd = get_unused_fd_flags(flags);
1985 if (unlikely(newfd < 0))
1988 newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1990 if (IS_ERR(newfile)) {
1991 put_unused_fd(newfd);
1992 return PTR_ERR(newfile);
1994 fd_install(newfd, newfile);
1999 * For accept, we attempt to create a new socket, set up the link
2000 * with the client, wake up the client, then return the new
2001 * connected fd. We collect the address of the connector in kernel
2002 * space and move it to user at the very end. This is unclean because
2003 * we open the socket then return an error.
2005 * 1003.1g adds the ability to recvmsg() to query connection pending
2006 * status to recvmsg. We need to add that support in a way thats
2007 * clean when we restructure accept also.
2010 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2011 int __user *upeer_addrlen, int flags)
2018 ret = __sys_accept4_file(f.file, upeer_sockaddr,
2019 upeer_addrlen, flags);
2026 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2027 int __user *, upeer_addrlen, int, flags)
2029 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2032 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2033 int __user *, upeer_addrlen)
2035 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2039 * Attempt to connect to a socket with the server address. The address
2040 * is in user space so we verify it is OK and move it to kernel space.
2042 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2045 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2046 * other SEQPACKET protocols that take time to connect() as it doesn't
2047 * include the -EINPROGRESS status for such sockets.
2050 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2051 int addrlen, int file_flags)
2053 struct socket *sock;
2056 sock = sock_from_file(file);
2063 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2067 err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2068 addrlen, sock->file->f_flags | file_flags);
2073 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2080 struct sockaddr_storage address;
2082 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2084 ret = __sys_connect_file(f.file, &address, addrlen, 0);
2091 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2094 return __sys_connect(fd, uservaddr, addrlen);
2098 * Get the local address ('name') of a socket object. Move the obtained
2099 * name to user space.
2102 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2103 int __user *usockaddr_len)
2105 struct socket *sock;
2106 struct sockaddr_storage address;
2107 int err, fput_needed;
2109 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2113 err = security_socket_getsockname(sock);
2117 err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2120 /* "err" is actually length in this case */
2121 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2124 fput_light(sock->file, fput_needed);
2129 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2130 int __user *, usockaddr_len)
2132 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2136 * Get the remote address ('name') of a socket object. Move the obtained
2137 * name to user space.
2140 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2141 int __user *usockaddr_len)
2143 struct socket *sock;
2144 struct sockaddr_storage address;
2145 int err, fput_needed;
2147 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2149 const struct proto_ops *ops = READ_ONCE(sock->ops);
2151 err = security_socket_getpeername(sock);
2153 fput_light(sock->file, fput_needed);
2157 err = ops->getname(sock, (struct sockaddr *)&address, 1);
2159 /* "err" is actually length in this case */
2160 err = move_addr_to_user(&address, err, usockaddr,
2162 fput_light(sock->file, fput_needed);
2167 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2168 int __user *, usockaddr_len)
2170 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2174 * Send a datagram to a given address. We move the address into kernel
2175 * space and check the user space data area is readable before invoking
2178 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2179 struct sockaddr __user *addr, int addr_len)
2181 struct socket *sock;
2182 struct sockaddr_storage address;
2187 err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2190 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2194 msg.msg_name = NULL;
2195 msg.msg_control = NULL;
2196 msg.msg_controllen = 0;
2197 msg.msg_namelen = 0;
2198 msg.msg_ubuf = NULL;
2200 err = move_addr_to_kernel(addr, addr_len, &address);
2203 msg.msg_name = (struct sockaddr *)&address;
2204 msg.msg_namelen = addr_len;
2206 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2207 if (sock->file->f_flags & O_NONBLOCK)
2208 flags |= MSG_DONTWAIT;
2209 msg.msg_flags = flags;
2210 err = __sock_sendmsg(sock, &msg);
2213 fput_light(sock->file, fput_needed);
2218 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2219 unsigned int, flags, struct sockaddr __user *, addr,
2222 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2226 * Send a datagram down a socket.
2229 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2230 unsigned int, flags)
2232 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2236 * Receive a frame from the socket and optionally record the address of the
2237 * sender. We verify the buffers are writable and if needed move the
2238 * sender address from kernel to user space.
2240 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2241 struct sockaddr __user *addr, int __user *addr_len)
2243 struct sockaddr_storage address;
2244 struct msghdr msg = {
2245 /* Save some cycles and don't copy the address if not needed */
2246 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2248 struct socket *sock;
2252 err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2255 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2259 if (sock->file->f_flags & O_NONBLOCK)
2260 flags |= MSG_DONTWAIT;
2261 err = sock_recvmsg(sock, &msg, flags);
2263 if (err >= 0 && addr != NULL) {
2264 err2 = move_addr_to_user(&address,
2265 msg.msg_namelen, addr, addr_len);
2270 fput_light(sock->file, fput_needed);
2275 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2276 unsigned int, flags, struct sockaddr __user *, addr,
2277 int __user *, addr_len)
2279 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2283 * Receive a datagram from a socket.
2286 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2287 unsigned int, flags)
2289 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2292 static bool sock_use_custom_sol_socket(const struct socket *sock)
2294 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2297 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2298 int optname, sockptr_t optval, int optlen)
2300 const struct proto_ops *ops;
2301 char *kernel_optval = NULL;
2307 err = security_socket_setsockopt(sock, level, optname);
2312 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2323 optval = KERNEL_SOCKPTR(kernel_optval);
2324 ops = READ_ONCE(sock->ops);
2325 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2326 err = sock_setsockopt(sock, level, optname, optval, optlen);
2327 else if (unlikely(!ops->setsockopt))
2330 err = ops->setsockopt(sock, level, optname, optval,
2332 kfree(kernel_optval);
2336 EXPORT_SYMBOL(do_sock_setsockopt);
2338 /* Set a socket option. Because we don't know the option lengths we have
2339 * to pass the user mode parameter for the protocols to sort out.
2341 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2344 sockptr_t optval = USER_SOCKPTR(user_optval);
2345 bool compat = in_compat_syscall();
2346 int err, fput_needed;
2347 struct socket *sock;
2349 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2353 err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2355 fput_light(sock->file, fput_needed);
2359 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2360 char __user *, optval, int, optlen)
2362 return __sys_setsockopt(fd, level, optname, optval, optlen);
2365 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2368 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2369 int optname, sockptr_t optval, sockptr_t optlen)
2371 int max_optlen __maybe_unused = 0;
2372 const struct proto_ops *ops;
2375 err = security_socket_getsockopt(sock, level, optname);
2380 copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2382 ops = READ_ONCE(sock->ops);
2383 if (level == SOL_SOCKET) {
2384 err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2385 } else if (unlikely(!ops->getsockopt)) {
2388 if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2389 "Invalid argument type"))
2392 err = ops->getsockopt(sock, level, optname, optval.user,
2397 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2398 optval, optlen, max_optlen,
2403 EXPORT_SYMBOL(do_sock_getsockopt);
2406 * Get a socket option. Because we don't know the option lengths we have
2407 * to pass a user mode parameter for the protocols to sort out.
2409 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2412 int err, fput_needed;
2413 struct socket *sock;
2416 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2420 compat = in_compat_syscall();
2421 err = do_sock_getsockopt(sock, compat, level, optname,
2422 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2424 fput_light(sock->file, fput_needed);
2428 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2429 char __user *, optval, int __user *, optlen)
2431 return __sys_getsockopt(fd, level, optname, optval, optlen);
2435 * Shutdown a socket.
2438 int __sys_shutdown_sock(struct socket *sock, int how)
2442 err = security_socket_shutdown(sock, how);
2444 err = READ_ONCE(sock->ops)->shutdown(sock, how);
2449 int __sys_shutdown(int fd, int how)
2451 int err, fput_needed;
2452 struct socket *sock;
2454 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2456 err = __sys_shutdown_sock(sock, how);
2457 fput_light(sock->file, fput_needed);
2462 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2464 return __sys_shutdown(fd, how);
2467 /* A couple of helpful macros for getting the address of the 32/64 bit
2468 * fields which are the same type (int / unsigned) on our platforms.
2470 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2471 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2472 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2474 struct used_address {
2475 struct sockaddr_storage name;
2476 unsigned int name_len;
2479 int __copy_msghdr(struct msghdr *kmsg,
2480 struct user_msghdr *msg,
2481 struct sockaddr __user **save_addr)
2485 kmsg->msg_control_is_user = true;
2486 kmsg->msg_get_inq = 0;
2487 kmsg->msg_control_user = msg->msg_control;
2488 kmsg->msg_controllen = msg->msg_controllen;
2489 kmsg->msg_flags = msg->msg_flags;
2491 kmsg->msg_namelen = msg->msg_namelen;
2493 kmsg->msg_namelen = 0;
2495 if (kmsg->msg_namelen < 0)
2498 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2499 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2502 *save_addr = msg->msg_name;
2504 if (msg->msg_name && kmsg->msg_namelen) {
2506 err = move_addr_to_kernel(msg->msg_name,
2513 kmsg->msg_name = NULL;
2514 kmsg->msg_namelen = 0;
2517 if (msg->msg_iovlen > UIO_MAXIOV)
2520 kmsg->msg_iocb = NULL;
2521 kmsg->msg_ubuf = NULL;
2525 static int copy_msghdr_from_user(struct msghdr *kmsg,
2526 struct user_msghdr __user *umsg,
2527 struct sockaddr __user **save_addr,
2530 struct user_msghdr msg;
2533 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2536 err = __copy_msghdr(kmsg, &msg, save_addr);
2540 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2541 msg.msg_iov, msg.msg_iovlen,
2542 UIO_FASTIOV, iov, &kmsg->msg_iter);
2543 return err < 0 ? err : 0;
2546 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2547 unsigned int flags, struct used_address *used_address,
2548 unsigned int allowed_msghdr_flags)
2550 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2551 __aligned(sizeof(__kernel_size_t));
2552 /* 20 is size of ipv6_pktinfo */
2553 unsigned char *ctl_buf = ctl;
2559 if (msg_sys->msg_controllen > INT_MAX)
2561 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2562 ctl_len = msg_sys->msg_controllen;
2563 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2565 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2569 ctl_buf = msg_sys->msg_control;
2570 ctl_len = msg_sys->msg_controllen;
2571 } else if (ctl_len) {
2572 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2573 CMSG_ALIGN(sizeof(struct cmsghdr)));
2574 if (ctl_len > sizeof(ctl)) {
2575 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2576 if (ctl_buf == NULL)
2580 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2582 msg_sys->msg_control = ctl_buf;
2583 msg_sys->msg_control_is_user = false;
2585 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2586 msg_sys->msg_flags = flags;
2588 if (sock->file->f_flags & O_NONBLOCK)
2589 msg_sys->msg_flags |= MSG_DONTWAIT;
2591 * If this is sendmmsg() and current destination address is same as
2592 * previously succeeded address, omit asking LSM's decision.
2593 * used_address->name_len is initialized to UINT_MAX so that the first
2594 * destination address never matches.
2596 if (used_address && msg_sys->msg_name &&
2597 used_address->name_len == msg_sys->msg_namelen &&
2598 !memcmp(&used_address->name, msg_sys->msg_name,
2599 used_address->name_len)) {
2600 err = sock_sendmsg_nosec(sock, msg_sys);
2603 err = __sock_sendmsg(sock, msg_sys);
2605 * If this is sendmmsg() and sending to current destination address was
2606 * successful, remember it.
2608 if (used_address && err >= 0) {
2609 used_address->name_len = msg_sys->msg_namelen;
2610 if (msg_sys->msg_name)
2611 memcpy(&used_address->name, msg_sys->msg_name,
2612 used_address->name_len);
2617 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2622 static int sendmsg_copy_msghdr(struct msghdr *msg,
2623 struct user_msghdr __user *umsg, unsigned flags,
2628 if (flags & MSG_CMSG_COMPAT) {
2629 struct compat_msghdr __user *msg_compat;
2631 msg_compat = (struct compat_msghdr __user *) umsg;
2632 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2634 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2642 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2643 struct msghdr *msg_sys, unsigned int flags,
2644 struct used_address *used_address,
2645 unsigned int allowed_msghdr_flags)
2647 struct sockaddr_storage address;
2648 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2651 msg_sys->msg_name = &address;
2653 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2657 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2658 allowed_msghdr_flags);
2664 * BSD sendmsg interface
2666 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2669 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2672 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2673 bool forbid_cmsg_compat)
2675 int fput_needed, err;
2676 struct msghdr msg_sys;
2677 struct socket *sock;
2679 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2682 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2686 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2688 fput_light(sock->file, fput_needed);
2693 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2695 return __sys_sendmsg(fd, msg, flags, true);
2699 * Linux sendmmsg interface
2702 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2703 unsigned int flags, bool forbid_cmsg_compat)
2705 int fput_needed, err, datagrams;
2706 struct socket *sock;
2707 struct mmsghdr __user *entry;
2708 struct compat_mmsghdr __user *compat_entry;
2709 struct msghdr msg_sys;
2710 struct used_address used_address;
2711 unsigned int oflags = flags;
2713 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2716 if (vlen > UIO_MAXIOV)
2721 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2725 used_address.name_len = UINT_MAX;
2727 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2731 while (datagrams < vlen) {
2732 if (datagrams == vlen - 1)
2735 if (MSG_CMSG_COMPAT & flags) {
2736 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2737 &msg_sys, flags, &used_address, MSG_EOR);
2740 err = __put_user(err, &compat_entry->msg_len);
2743 err = ___sys_sendmsg(sock,
2744 (struct user_msghdr __user *)entry,
2745 &msg_sys, flags, &used_address, MSG_EOR);
2748 err = put_user(err, &entry->msg_len);
2755 if (msg_data_left(&msg_sys))
2760 fput_light(sock->file, fput_needed);
2762 /* We only return an error if no datagrams were able to be sent */
2769 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2770 unsigned int, vlen, unsigned int, flags)
2772 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2775 static int recvmsg_copy_msghdr(struct msghdr *msg,
2776 struct user_msghdr __user *umsg, unsigned flags,
2777 struct sockaddr __user **uaddr,
2782 if (MSG_CMSG_COMPAT & flags) {
2783 struct compat_msghdr __user *msg_compat;
2785 msg_compat = (struct compat_msghdr __user *) umsg;
2786 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2788 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2796 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2797 struct user_msghdr __user *msg,
2798 struct sockaddr __user *uaddr,
2799 unsigned int flags, int nosec)
2801 struct compat_msghdr __user *msg_compat =
2802 (struct compat_msghdr __user *) msg;
2803 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2804 struct sockaddr_storage addr;
2805 unsigned long cmsg_ptr;
2809 msg_sys->msg_name = &addr;
2810 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2811 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2813 /* We assume all kernel code knows the size of sockaddr_storage */
2814 msg_sys->msg_namelen = 0;
2816 if (sock->file->f_flags & O_NONBLOCK)
2817 flags |= MSG_DONTWAIT;
2819 if (unlikely(nosec))
2820 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2822 err = sock_recvmsg(sock, msg_sys, flags);
2828 if (uaddr != NULL) {
2829 err = move_addr_to_user(&addr,
2830 msg_sys->msg_namelen, uaddr,
2835 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2839 if (MSG_CMSG_COMPAT & flags)
2840 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2841 &msg_compat->msg_controllen);
2843 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2844 &msg->msg_controllen);
2852 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2853 struct msghdr *msg_sys, unsigned int flags, int nosec)
2855 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2856 /* user mode address pointers */
2857 struct sockaddr __user *uaddr;
2860 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2864 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2870 * BSD recvmsg interface
2873 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2874 struct user_msghdr __user *umsg,
2875 struct sockaddr __user *uaddr, unsigned int flags)
2877 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2880 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2881 bool forbid_cmsg_compat)
2883 int fput_needed, err;
2884 struct msghdr msg_sys;
2885 struct socket *sock;
2887 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2890 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2894 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2896 fput_light(sock->file, fput_needed);
2901 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2902 unsigned int, flags)
2904 return __sys_recvmsg(fd, msg, flags, true);
2908 * Linux recvmmsg interface
2911 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2912 unsigned int vlen, unsigned int flags,
2913 struct timespec64 *timeout)
2915 int fput_needed, err, datagrams;
2916 struct socket *sock;
2917 struct mmsghdr __user *entry;
2918 struct compat_mmsghdr __user *compat_entry;
2919 struct msghdr msg_sys;
2920 struct timespec64 end_time;
2921 struct timespec64 timeout64;
2924 poll_select_set_timeout(&end_time, timeout->tv_sec,
2930 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2934 if (likely(!(flags & MSG_ERRQUEUE))) {
2935 err = sock_error(sock->sk);
2943 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2945 while (datagrams < vlen) {
2947 * No need to ask LSM for more than the first datagram.
2949 if (MSG_CMSG_COMPAT & flags) {
2950 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2951 &msg_sys, flags & ~MSG_WAITFORONE,
2955 err = __put_user(err, &compat_entry->msg_len);
2958 err = ___sys_recvmsg(sock,
2959 (struct user_msghdr __user *)entry,
2960 &msg_sys, flags & ~MSG_WAITFORONE,
2964 err = put_user(err, &entry->msg_len);
2972 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2973 if (flags & MSG_WAITFORONE)
2974 flags |= MSG_DONTWAIT;
2977 ktime_get_ts64(&timeout64);
2978 *timeout = timespec64_sub(end_time, timeout64);
2979 if (timeout->tv_sec < 0) {
2980 timeout->tv_sec = timeout->tv_nsec = 0;
2984 /* Timeout, return less than vlen datagrams */
2985 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2989 /* Out of band data, return right away */
2990 if (msg_sys.msg_flags & MSG_OOB)
2998 if (datagrams == 0) {
3004 * We may return less entries than requested (vlen) if the
3005 * sock is non block and there aren't enough datagrams...
3007 if (err != -EAGAIN) {
3009 * ... or if recvmsg returns an error after we
3010 * received some datagrams, where we record the
3011 * error to return on the next call or if the
3012 * app asks about it using getsockopt(SO_ERROR).
3014 WRITE_ONCE(sock->sk->sk_err, -err);
3017 fput_light(sock->file, fput_needed);
3022 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3023 unsigned int vlen, unsigned int flags,
3024 struct __kernel_timespec __user *timeout,
3025 struct old_timespec32 __user *timeout32)
3028 struct timespec64 timeout_sys;
3030 if (timeout && get_timespec64(&timeout_sys, timeout))
3033 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3036 if (!timeout && !timeout32)
3037 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3039 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3044 if (timeout && put_timespec64(&timeout_sys, timeout))
3045 datagrams = -EFAULT;
3047 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3048 datagrams = -EFAULT;
3053 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3054 unsigned int, vlen, unsigned int, flags,
3055 struct __kernel_timespec __user *, timeout)
3057 if (flags & MSG_CMSG_COMPAT)
3060 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3063 #ifdef CONFIG_COMPAT_32BIT_TIME
3064 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3065 unsigned int, vlen, unsigned int, flags,
3066 struct old_timespec32 __user *, timeout)
3068 if (flags & MSG_CMSG_COMPAT)
3071 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3075 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3076 /* Argument list sizes for sys_socketcall */
3077 #define AL(x) ((x) * sizeof(unsigned long))
3078 static const unsigned char nargs[21] = {
3079 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3080 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3081 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3088 * System call vectors.
3090 * Argument checking cleaned up. Saved 20% in size.
3091 * This function doesn't need to set the kernel lock because
3092 * it is set by the callees.
3095 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3097 unsigned long a[AUDITSC_ARGS];
3098 unsigned long a0, a1;
3102 if (call < 1 || call > SYS_SENDMMSG)
3104 call = array_index_nospec(call, SYS_SENDMMSG + 1);
3107 if (len > sizeof(a))
3110 /* copy_from_user should be SMP safe. */
3111 if (copy_from_user(a, args, len))
3114 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3123 err = __sys_socket(a0, a1, a[2]);
3126 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3129 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3132 err = __sys_listen(a0, a1);
3135 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3136 (int __user *)a[2], 0);
3138 case SYS_GETSOCKNAME:
3140 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3141 (int __user *)a[2]);
3143 case SYS_GETPEERNAME:
3145 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3146 (int __user *)a[2]);
3148 case SYS_SOCKETPAIR:
3149 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3152 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3156 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3157 (struct sockaddr __user *)a[4], a[5]);
3160 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3164 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3165 (struct sockaddr __user *)a[4],
3166 (int __user *)a[5]);
3169 err = __sys_shutdown(a0, a1);
3171 case SYS_SETSOCKOPT:
3172 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3175 case SYS_GETSOCKOPT:
3177 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3178 (int __user *)a[4]);
3181 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3185 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3189 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3193 if (IS_ENABLED(CONFIG_64BIT))
3194 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3196 (struct __kernel_timespec __user *)a[4],
3199 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3201 (struct old_timespec32 __user *)a[4]);
3204 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3205 (int __user *)a[2], a[3]);
3214 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3217 * sock_register - add a socket protocol handler
3218 * @ops: description of protocol
3220 * This function is called by a protocol handler that wants to
3221 * advertise its address family, and have it linked into the
3222 * socket interface. The value ops->family corresponds to the
3223 * socket system call protocol family.
3225 int sock_register(const struct net_proto_family *ops)
3229 if (ops->family >= NPROTO) {
3230 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3234 spin_lock(&net_family_lock);
3235 if (rcu_dereference_protected(net_families[ops->family],
3236 lockdep_is_held(&net_family_lock)))
3239 rcu_assign_pointer(net_families[ops->family], ops);
3242 spin_unlock(&net_family_lock);
3244 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3247 EXPORT_SYMBOL(sock_register);
3250 * sock_unregister - remove a protocol handler
3251 * @family: protocol family to remove
3253 * This function is called by a protocol handler that wants to
3254 * remove its address family, and have it unlinked from the
3255 * new socket creation.
3257 * If protocol handler is a module, then it can use module reference
3258 * counts to protect against new references. If protocol handler is not
3259 * a module then it needs to provide its own protection in
3260 * the ops->create routine.
3262 void sock_unregister(int family)
3264 BUG_ON(family < 0 || family >= NPROTO);
3266 spin_lock(&net_family_lock);
3267 RCU_INIT_POINTER(net_families[family], NULL);
3268 spin_unlock(&net_family_lock);
3272 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3274 EXPORT_SYMBOL(sock_unregister);
3276 bool sock_is_registered(int family)
3278 return family < NPROTO && rcu_access_pointer(net_families[family]);
3281 static int __init sock_init(void)
3285 * Initialize the network sysctl infrastructure.
3287 err = net_sysctl_init();
3292 * Initialize skbuff SLAB cache
3297 * Initialize the protocols module.
3302 err = register_filesystem(&sock_fs_type);
3305 sock_mnt = kern_mount(&sock_fs_type);
3306 if (IS_ERR(sock_mnt)) {
3307 err = PTR_ERR(sock_mnt);
3311 /* The real protocol initialization is performed in later initcalls.
3314 #ifdef CONFIG_NETFILTER
3315 err = netfilter_init();
3320 ptp_classifier_init();
3326 unregister_filesystem(&sock_fs_type);
3330 core_initcall(sock_init); /* early initcall */
3332 #ifdef CONFIG_PROC_FS
3333 void socket_seq_show(struct seq_file *seq)
3335 seq_printf(seq, "sockets: used %d\n",
3336 sock_inuse_get(seq->private));
3338 #endif /* CONFIG_PROC_FS */
3340 /* Handle the fact that while struct ifreq has the same *layout* on
3341 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3342 * which are handled elsewhere, it still has different *size* due to
3343 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3344 * resulting in struct ifreq being 32 and 40 bytes respectively).
3345 * As a result, if the struct happens to be at the end of a page and
3346 * the next page isn't readable/writable, we get a fault. To prevent
3347 * that, copy back and forth to the full size.
3349 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3351 if (in_compat_syscall()) {
3352 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3354 memset(ifr, 0, sizeof(*ifr));
3355 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3359 *ifrdata = compat_ptr(ifr32->ifr_data);
3364 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3368 *ifrdata = ifr->ifr_data;
3372 EXPORT_SYMBOL(get_user_ifreq);
3374 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3376 size_t size = sizeof(*ifr);
3378 if (in_compat_syscall())
3379 size = sizeof(struct compat_ifreq);
3381 if (copy_to_user(arg, ifr, size))
3386 EXPORT_SYMBOL(put_user_ifreq);
3388 #ifdef CONFIG_COMPAT
3389 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3391 compat_uptr_t uptr32;
3396 if (get_user_ifreq(&ifr, NULL, uifr32))
3399 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3402 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3403 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3405 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3407 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3408 if (put_user_ifreq(&ifr, uifr32))
3414 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3415 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3416 struct compat_ifreq __user *u_ifreq32)
3421 if (!is_socket_ioctl_cmd(cmd))
3423 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3425 ifreq.ifr_data = data;
3427 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3430 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3431 unsigned int cmd, unsigned long arg)
3433 void __user *argp = compat_ptr(arg);
3434 struct sock *sk = sock->sk;
3435 struct net *net = sock_net(sk);
3436 const struct proto_ops *ops;
3438 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3439 return sock_ioctl(file, cmd, (unsigned long)argp);
3443 return compat_siocwandev(net, argp);
3444 case SIOCGSTAMP_OLD:
3445 case SIOCGSTAMPNS_OLD:
3446 ops = READ_ONCE(sock->ops);
3447 if (!ops->gettstamp)
3448 return -ENOIOCTLCMD;
3449 return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3450 !COMPAT_USE_64BIT_TIME);
3453 case SIOCBONDSLAVEINFOQUERY:
3454 case SIOCBONDINFOQUERY:
3457 return compat_ifr_data_ioctl(net, cmd, argp);
3468 case SIOCGSTAMP_NEW:
3469 case SIOCGSTAMPNS_NEW:
3473 return sock_ioctl(file, cmd, arg);
3492 case SIOCSIFHWBROADCAST:
3494 case SIOCGIFBRDADDR:
3495 case SIOCSIFBRDADDR:
3496 case SIOCGIFDSTADDR:
3497 case SIOCSIFDSTADDR:
3498 case SIOCGIFNETMASK:
3499 case SIOCSIFNETMASK:
3511 case SIOCBONDENSLAVE:
3512 case SIOCBONDRELEASE:
3513 case SIOCBONDSETHWADDR:
3514 case SIOCBONDCHANGEACTIVE:
3521 return sock_do_ioctl(net, sock, cmd, arg);
3524 return -ENOIOCTLCMD;
3527 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3530 struct socket *sock = file->private_data;
3531 const struct proto_ops *ops = READ_ONCE(sock->ops);
3532 int ret = -ENOIOCTLCMD;
3539 if (ops->compat_ioctl)
3540 ret = ops->compat_ioctl(sock, cmd, arg);
3542 if (ret == -ENOIOCTLCMD &&
3543 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3544 ret = compat_wext_handle_ioctl(net, cmd, arg);
3546 if (ret == -ENOIOCTLCMD)
3547 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3554 * kernel_bind - bind an address to a socket (kernel space)
3557 * @addrlen: length of address
3559 * Returns 0 or an error.
3562 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3564 struct sockaddr_storage address;
3566 memcpy(&address, addr, addrlen);
3568 return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3571 EXPORT_SYMBOL(kernel_bind);
3574 * kernel_listen - move socket to listening state (kernel space)
3576 * @backlog: pending connections queue size
3578 * Returns 0 or an error.
3581 int kernel_listen(struct socket *sock, int backlog)
3583 return READ_ONCE(sock->ops)->listen(sock, backlog);
3585 EXPORT_SYMBOL(kernel_listen);
3588 * kernel_accept - accept a connection (kernel space)
3589 * @sock: listening socket
3590 * @newsock: new connected socket
3593 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3594 * If it fails, @newsock is guaranteed to be %NULL.
3595 * Returns 0 or an error.
3598 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3600 struct sock *sk = sock->sk;
3601 const struct proto_ops *ops = READ_ONCE(sock->ops);
3602 struct proto_accept_arg arg = {
3608 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3613 err = ops->accept(sock, *newsock, &arg);
3615 sock_release(*newsock);
3620 (*newsock)->ops = ops;
3621 __module_get(ops->owner);
3626 EXPORT_SYMBOL(kernel_accept);
3629 * kernel_connect - connect a socket (kernel space)
3632 * @addrlen: address length
3633 * @flags: flags (O_NONBLOCK, ...)
3635 * For datagram sockets, @addr is the address to which datagrams are sent
3636 * by default, and the only address from which datagrams are received.
3637 * For stream sockets, attempts to connect to @addr.
3638 * Returns 0 or an error code.
3641 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3644 struct sockaddr_storage address;
3646 memcpy(&address, addr, addrlen);
3648 return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3651 EXPORT_SYMBOL(kernel_connect);
3654 * kernel_getsockname - get the address which the socket is bound (kernel space)
3656 * @addr: address holder
3658 * Fills the @addr pointer with the address which the socket is bound.
3659 * Returns the length of the address in bytes or an error code.
3662 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3664 return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3666 EXPORT_SYMBOL(kernel_getsockname);
3669 * kernel_getpeername - get the address which the socket is connected (kernel space)
3671 * @addr: address holder
3673 * Fills the @addr pointer with the address which the socket is connected.
3674 * Returns the length of the address in bytes or an error code.
3677 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3679 return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3681 EXPORT_SYMBOL(kernel_getpeername);
3684 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3686 * @how: connection part
3688 * Returns 0 or an error.
3691 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3693 return READ_ONCE(sock->ops)->shutdown(sock, how);
3695 EXPORT_SYMBOL(kernel_sock_shutdown);
3698 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3701 * This routine returns the IP overhead imposed by a socket i.e.
3702 * the length of the underlying IP header, depending on whether
3703 * this is an IPv4 or IPv6 socket and the length from IP options turned
3704 * on at the socket. Assumes that the caller has a lock on the socket.
3707 u32 kernel_sock_ip_overhead(struct sock *sk)
3709 struct inet_sock *inet;
3710 struct ip_options_rcu *opt;
3712 #if IS_ENABLED(CONFIG_IPV6)
3713 struct ipv6_pinfo *np;
3714 struct ipv6_txoptions *optv6 = NULL;
3715 #endif /* IS_ENABLED(CONFIG_IPV6) */
3720 switch (sk->sk_family) {
3723 overhead += sizeof(struct iphdr);
3724 opt = rcu_dereference_protected(inet->inet_opt,
3725 sock_owned_by_user(sk));
3727 overhead += opt->opt.optlen;
3729 #if IS_ENABLED(CONFIG_IPV6)
3732 overhead += sizeof(struct ipv6hdr);
3734 optv6 = rcu_dereference_protected(np->opt,
3735 sock_owned_by_user(sk));
3737 overhead += (optv6->opt_flen + optv6->opt_nflen);
3739 #endif /* IS_ENABLED(CONFIG_IPV6) */
3740 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3744 EXPORT_SYMBOL(kernel_sock_ip_overhead);