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
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,
156 .read_iter = sock_read_iter,
157 .write_iter = sock_write_iter,
159 .unlocked_ioctl = sock_ioctl,
161 .compat_ioctl = compat_sock_ioctl,
163 .uring_cmd = io_uring_cmd_sock,
165 .release = sock_close,
166 .fasync = sock_fasync,
167 .splice_write = splice_to_socket,
168 .splice_read = sock_splice_read,
169 .splice_eof = sock_splice_eof,
170 .show_fdinfo = sock_show_fdinfo,
173 static const char * const pf_family_names[] = {
174 [PF_UNSPEC] = "PF_UNSPEC",
175 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
176 [PF_INET] = "PF_INET",
177 [PF_AX25] = "PF_AX25",
179 [PF_APPLETALK] = "PF_APPLETALK",
180 [PF_NETROM] = "PF_NETROM",
181 [PF_BRIDGE] = "PF_BRIDGE",
182 [PF_ATMPVC] = "PF_ATMPVC",
184 [PF_INET6] = "PF_INET6",
185 [PF_ROSE] = "PF_ROSE",
186 [PF_DECnet] = "PF_DECnet",
187 [PF_NETBEUI] = "PF_NETBEUI",
188 [PF_SECURITY] = "PF_SECURITY",
190 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
191 [PF_PACKET] = "PF_PACKET",
193 [PF_ECONET] = "PF_ECONET",
194 [PF_ATMSVC] = "PF_ATMSVC",
197 [PF_IRDA] = "PF_IRDA",
198 [PF_PPPOX] = "PF_PPPOX",
199 [PF_WANPIPE] = "PF_WANPIPE",
202 [PF_MPLS] = "PF_MPLS",
204 [PF_TIPC] = "PF_TIPC",
205 [PF_BLUETOOTH] = "PF_BLUETOOTH",
206 [PF_IUCV] = "PF_IUCV",
207 [PF_RXRPC] = "PF_RXRPC",
208 [PF_ISDN] = "PF_ISDN",
209 [PF_PHONET] = "PF_PHONET",
210 [PF_IEEE802154] = "PF_IEEE802154",
211 [PF_CAIF] = "PF_CAIF",
214 [PF_VSOCK] = "PF_VSOCK",
216 [PF_QIPCRTR] = "PF_QIPCRTR",
219 [PF_MCTP] = "PF_MCTP",
223 * The protocol list. Each protocol is registered in here.
226 static DEFINE_SPINLOCK(net_family_lock);
227 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
231 * Move socket addresses back and forth across the kernel/user
232 * divide and look after the messy bits.
236 * move_addr_to_kernel - copy a socket address into kernel space
237 * @uaddr: Address in user space
238 * @kaddr: Address in kernel space
239 * @ulen: Length in user space
241 * The address is copied into kernel space. If the provided address is
242 * too long an error code of -EINVAL is returned. If the copy gives
243 * invalid addresses -EFAULT is returned. On a success 0 is returned.
246 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
248 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
252 if (copy_from_user(kaddr, uaddr, ulen))
254 return audit_sockaddr(ulen, kaddr);
258 * move_addr_to_user - copy an address to user space
259 * @kaddr: kernel space address
260 * @klen: length of address in kernel
261 * @uaddr: user space address
262 * @ulen: pointer to user length field
264 * The value pointed to by ulen on entry is the buffer length available.
265 * This is overwritten with the buffer space used. -EINVAL is returned
266 * if an overlong buffer is specified or a negative buffer size. -EFAULT
267 * is returned if either the buffer or the length field are not
269 * After copying the data up to the limit the user specifies, the true
270 * length of the data is written over the length limit the user
271 * specified. Zero is returned for a success.
274 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
275 void __user *uaddr, int __user *ulen)
280 BUG_ON(klen > sizeof(struct sockaddr_storage));
281 err = get_user(len, ulen);
289 if (audit_sockaddr(klen, kaddr))
291 if (copy_to_user(uaddr, kaddr, len))
295 * "fromlen shall refer to the value before truncation.."
298 return __put_user(klen, ulen);
301 static struct kmem_cache *sock_inode_cachep __ro_after_init;
303 static struct inode *sock_alloc_inode(struct super_block *sb)
305 struct socket_alloc *ei;
307 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
310 init_waitqueue_head(&ei->socket.wq.wait);
311 ei->socket.wq.fasync_list = NULL;
312 ei->socket.wq.flags = 0;
314 ei->socket.state = SS_UNCONNECTED;
315 ei->socket.flags = 0;
316 ei->socket.ops = NULL;
317 ei->socket.sk = NULL;
318 ei->socket.file = NULL;
320 return &ei->vfs_inode;
323 static void sock_free_inode(struct inode *inode)
325 struct socket_alloc *ei;
327 ei = container_of(inode, struct socket_alloc, vfs_inode);
328 kmem_cache_free(sock_inode_cachep, ei);
331 static void init_once(void *foo)
333 struct socket_alloc *ei = (struct socket_alloc *)foo;
335 inode_init_once(&ei->vfs_inode);
338 static void init_inodecache(void)
340 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
341 sizeof(struct socket_alloc),
343 (SLAB_HWCACHE_ALIGN |
344 SLAB_RECLAIM_ACCOUNT |
347 BUG_ON(sock_inode_cachep == NULL);
350 static const struct super_operations sockfs_ops = {
351 .alloc_inode = sock_alloc_inode,
352 .free_inode = sock_free_inode,
353 .statfs = simple_statfs,
357 * sockfs_dname() is called from d_path().
359 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
361 return dynamic_dname(buffer, buflen, "socket:[%lu]",
362 d_inode(dentry)->i_ino);
365 static const struct dentry_operations sockfs_dentry_operations = {
366 .d_dname = sockfs_dname,
369 static int sockfs_xattr_get(const struct xattr_handler *handler,
370 struct dentry *dentry, struct inode *inode,
371 const char *suffix, void *value, size_t size)
374 if (dentry->d_name.len + 1 > size)
376 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
378 return dentry->d_name.len + 1;
381 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
382 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
383 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
385 static const struct xattr_handler sockfs_xattr_handler = {
386 .name = XATTR_NAME_SOCKPROTONAME,
387 .get = sockfs_xattr_get,
390 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
391 struct mnt_idmap *idmap,
392 struct dentry *dentry, struct inode *inode,
393 const char *suffix, const void *value,
394 size_t size, int flags)
396 /* Handled by LSM. */
400 static const struct xattr_handler sockfs_security_xattr_handler = {
401 .prefix = XATTR_SECURITY_PREFIX,
402 .set = sockfs_security_xattr_set,
405 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
406 &sockfs_xattr_handler,
407 &sockfs_security_xattr_handler,
411 static int sockfs_init_fs_context(struct fs_context *fc)
413 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
416 ctx->ops = &sockfs_ops;
417 ctx->dops = &sockfs_dentry_operations;
418 ctx->xattr = sockfs_xattr_handlers;
422 static struct vfsmount *sock_mnt __read_mostly;
424 static struct file_system_type sock_fs_type = {
426 .init_fs_context = sockfs_init_fs_context,
427 .kill_sb = kill_anon_super,
431 * Obtains the first available file descriptor and sets it up for use.
433 * These functions create file structures and maps them to fd space
434 * of the current process. On success it returns file descriptor
435 * and file struct implicitly stored in sock->file.
436 * Note that another thread may close file descriptor before we return
437 * from this function. We use the fact that now we do not refer
438 * to socket after mapping. If one day we will need it, this
439 * function will increment ref. count on file by 1.
441 * In any case returned fd MAY BE not valid!
442 * This race condition is unavoidable
443 * with shared fd spaces, we cannot solve it inside kernel,
444 * but we take care of internal coherence yet.
448 * sock_alloc_file - Bind a &socket to a &file
450 * @flags: file status flags
451 * @dname: protocol name
453 * Returns the &file bound with @sock, implicitly storing it
454 * in sock->file. If dname is %NULL, sets to "".
456 * On failure @sock is released, and an ERR pointer is returned.
458 * This function uses GFP_KERNEL internally.
461 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
466 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
468 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
469 O_RDWR | (flags & O_NONBLOCK),
476 file->f_mode |= FMODE_NOWAIT;
478 file->private_data = sock;
479 stream_open(SOCK_INODE(sock), file);
482 EXPORT_SYMBOL(sock_alloc_file);
484 static int sock_map_fd(struct socket *sock, int flags)
486 struct file *newfile;
487 int fd = get_unused_fd_flags(flags);
488 if (unlikely(fd < 0)) {
493 newfile = sock_alloc_file(sock, flags, NULL);
494 if (!IS_ERR(newfile)) {
495 fd_install(fd, newfile);
500 return PTR_ERR(newfile);
504 * sock_from_file - Return the &socket bounded to @file.
507 * On failure returns %NULL.
510 struct socket *sock_from_file(struct file *file)
512 if (file->f_op == &socket_file_ops)
513 return file->private_data; /* set in sock_alloc_file */
517 EXPORT_SYMBOL(sock_from_file);
520 * sockfd_lookup - Go from a file number to its socket slot
522 * @err: pointer to an error code return
524 * The file handle passed in is locked and the socket it is bound
525 * to is returned. If an error occurs the err pointer is overwritten
526 * with a negative errno code and NULL is returned. The function checks
527 * for both invalid handles and passing a handle which is not a socket.
529 * On a success the socket object pointer is returned.
532 struct socket *sockfd_lookup(int fd, int *err)
543 sock = sock_from_file(file);
550 EXPORT_SYMBOL(sockfd_lookup);
552 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
554 struct fd f = fdget(fd);
559 sock = sock_from_file(fd_file(f));
561 *fput_needed = f.word & FDPUT_FPUT;
570 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
576 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
586 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
591 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
598 static int sockfs_setattr(struct mnt_idmap *idmap,
599 struct dentry *dentry, struct iattr *iattr)
601 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
603 if (!err && (iattr->ia_valid & ATTR_UID)) {
604 struct socket *sock = SOCKET_I(d_inode(dentry));
607 sock->sk->sk_uid = iattr->ia_uid;
615 static const struct inode_operations sockfs_inode_ops = {
616 .listxattr = sockfs_listxattr,
617 .setattr = sockfs_setattr,
621 * sock_alloc - allocate a socket
623 * Allocate a new inode and socket object. The two are bound together
624 * and initialised. The socket is then returned. If we are out of inodes
625 * NULL is returned. This functions uses GFP_KERNEL internally.
628 struct socket *sock_alloc(void)
633 inode = new_inode_pseudo(sock_mnt->mnt_sb);
637 sock = SOCKET_I(inode);
639 inode->i_ino = get_next_ino();
640 inode->i_mode = S_IFSOCK | S_IRWXUGO;
641 inode->i_uid = current_fsuid();
642 inode->i_gid = current_fsgid();
643 inode->i_op = &sockfs_inode_ops;
647 EXPORT_SYMBOL(sock_alloc);
649 static void __sock_release(struct socket *sock, struct inode *inode)
651 const struct proto_ops *ops = READ_ONCE(sock->ops);
654 struct module *owner = ops->owner;
666 if (sock->wq.fasync_list)
667 pr_err("%s: fasync list not empty!\n", __func__);
670 iput(SOCK_INODE(sock));
677 * sock_release - close a socket
678 * @sock: socket to close
680 * The socket is released from the protocol stack if it has a release
681 * callback, and the inode is then released if the socket is bound to
682 * an inode not a file.
684 void sock_release(struct socket *sock)
686 __sock_release(sock, NULL);
688 EXPORT_SYMBOL(sock_release);
690 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
692 u8 flags = *tx_flags;
694 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
695 flags |= SKBTX_HW_TSTAMP;
697 /* PTP hardware clocks can provide a free running cycle counter
698 * as a time base for virtual clocks. Tell driver to use the
699 * free running cycle counter for timestamp if socket is bound
702 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
703 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
706 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
707 flags |= SKBTX_SW_TSTAMP;
709 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
710 flags |= SKBTX_SCHED_TSTAMP;
714 EXPORT_SYMBOL(__sock_tx_timestamp);
716 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
718 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
721 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
724 trace_sock_send_length(sk, ret, 0);
727 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
729 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
730 inet_sendmsg, sock, msg,
732 BUG_ON(ret == -EIOCBQUEUED);
734 if (trace_sock_send_length_enabled())
735 call_trace_sock_send_length(sock->sk, ret, 0);
739 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
741 int err = security_socket_sendmsg(sock, msg,
744 return err ?: sock_sendmsg_nosec(sock, msg);
748 * sock_sendmsg - send a message through @sock
750 * @msg: message to send
752 * Sends @msg through @sock, passing through LSM.
753 * Returns the number of bytes sent, or an error code.
755 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
757 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
758 struct sockaddr_storage address;
759 int save_len = msg->msg_namelen;
763 memcpy(&address, msg->msg_name, msg->msg_namelen);
764 msg->msg_name = &address;
767 ret = __sock_sendmsg(sock, msg);
768 msg->msg_name = save_addr;
769 msg->msg_namelen = save_len;
773 EXPORT_SYMBOL(sock_sendmsg);
776 * kernel_sendmsg - send a message through @sock (kernel-space)
778 * @msg: message header
780 * @num: vec array length
781 * @size: total message data size
783 * Builds the message data with @vec and sends it through @sock.
784 * Returns the number of bytes sent, or an error code.
787 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
788 struct kvec *vec, size_t num, size_t size)
790 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
791 return sock_sendmsg(sock, msg);
793 EXPORT_SYMBOL(kernel_sendmsg);
796 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
798 * @msg: message header
799 * @vec: output s/g array
800 * @num: output s/g array length
801 * @size: total message data size
803 * Builds the message data with @vec and sends it through @sock.
804 * Returns the number of bytes sent, or an error code.
805 * Caller must hold @sk.
808 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
809 struct kvec *vec, size_t num, size_t size)
811 struct socket *sock = sk->sk_socket;
812 const struct proto_ops *ops = READ_ONCE(sock->ops);
814 if (!ops->sendmsg_locked)
815 return sock_no_sendmsg_locked(sk, msg, size);
817 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
819 return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
821 EXPORT_SYMBOL(kernel_sendmsg_locked);
823 static bool skb_is_err_queue(const struct sk_buff *skb)
825 /* pkt_type of skbs enqueued on the error queue are set to
826 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
827 * in recvmsg, since skbs received on a local socket will never
828 * have a pkt_type of PACKET_OUTGOING.
830 return skb->pkt_type == PACKET_OUTGOING;
833 /* On transmit, software and hardware timestamps are returned independently.
834 * As the two skb clones share the hardware timestamp, which may be updated
835 * before the software timestamp is received, a hardware TX timestamp may be
836 * returned only if there is no software TX timestamp. Ignore false software
837 * timestamps, which may be made in the __sock_recv_timestamp() call when the
838 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
839 * hardware timestamp.
841 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
843 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
846 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
848 bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
849 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
850 struct net_device *orig_dev;
854 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 *if_index = orig_dev->ifindex;
857 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
859 hwtstamp = shhwtstamps->hwtstamp;
866 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
869 struct scm_ts_pktinfo ts_pktinfo;
870 struct net_device *orig_dev;
872 if (!skb_mac_header_was_set(skb))
875 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
879 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
881 if_index = orig_dev->ifindex;
884 ts_pktinfo.if_index = if_index;
886 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
887 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
888 sizeof(ts_pktinfo), &ts_pktinfo);
892 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
894 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
897 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
898 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
899 struct scm_timestamping_internal tss;
900 int empty = 1, false_tstamp = 0;
901 struct skb_shared_hwtstamps *shhwtstamps =
907 /* Race occurred between timestamp enabling and packet
908 receiving. Fill in the current time for now. */
909 if (need_software_tstamp && skb->tstamp == 0) {
910 __net_timestamp(skb);
914 if (need_software_tstamp) {
915 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
917 struct __kernel_sock_timeval tv;
919 skb_get_new_timestamp(skb, &tv);
920 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
923 struct __kernel_old_timeval tv;
925 skb_get_timestamp(skb, &tv);
926 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
931 struct __kernel_timespec ts;
933 skb_get_new_timestampns(skb, &ts);
934 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
937 struct __kernel_old_timespec ts;
939 skb_get_timestampns(skb, &ts);
940 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
946 memset(&tss, 0, sizeof(tss));
947 tsflags = READ_ONCE(sk->sk_tsflags);
948 if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
949 (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
950 skb_is_err_queue(skb) ||
951 !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
952 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
955 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
956 (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
957 skb_is_err_queue(skb) ||
958 !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
959 !skb_is_swtx_tstamp(skb, false_tstamp)) {
961 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
962 hwtstamp = get_timestamp(sk, skb, &if_index);
964 hwtstamp = shhwtstamps->hwtstamp;
966 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
967 hwtstamp = ptp_convert_timestamp(&hwtstamp,
968 READ_ONCE(sk->sk_bind_phc));
970 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
973 if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
974 !skb_is_err_queue(skb))
975 put_ts_pktinfo(msg, skb, if_index);
979 if (sock_flag(sk, SOCK_TSTAMP_NEW))
980 put_cmsg_scm_timestamping64(msg, &tss);
982 put_cmsg_scm_timestamping(msg, &tss);
984 if (skb_is_err_queue(skb) && skb->len &&
985 SKB_EXT_ERR(skb)->opt_stats)
986 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
987 skb->len, skb->data);
990 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
992 #ifdef CONFIG_WIRELESS
993 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
998 if (!sock_flag(sk, SOCK_WIFI_STATUS))
1000 if (!skb->wifi_acked_valid)
1003 ack = skb->wifi_acked;
1005 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1007 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1010 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1011 struct sk_buff *skb)
1013 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1014 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1015 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1018 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1019 struct sk_buff *skb)
1021 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1022 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1023 __u32 mark = skb->mark;
1025 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1029 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1030 struct sk_buff *skb)
1032 sock_recv_timestamp(msg, sk, skb);
1033 sock_recv_drops(msg, sk, skb);
1034 sock_recv_mark(msg, sk, skb);
1036 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1038 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1040 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1043 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1045 trace_sock_recv_length(sk, ret, flags);
1048 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1051 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1053 inet_recvmsg, sock, msg,
1054 msg_data_left(msg), flags);
1055 if (trace_sock_recv_length_enabled())
1056 call_trace_sock_recv_length(sock->sk, ret, flags);
1061 * sock_recvmsg - receive a message from @sock
1063 * @msg: message to receive
1064 * @flags: message flags
1066 * Receives @msg from @sock, passing through LSM. Returns the total number
1067 * of bytes received, or an error.
1069 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1071 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1073 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1075 EXPORT_SYMBOL(sock_recvmsg);
1078 * kernel_recvmsg - Receive a message from a socket (kernel space)
1079 * @sock: The socket to receive the message from
1080 * @msg: Received message
1081 * @vec: Input s/g array for message data
1082 * @num: Size of input s/g array
1083 * @size: Number of bytes to read
1084 * @flags: Message flags (MSG_DONTWAIT, etc...)
1086 * On return the msg structure contains the scatter/gather array passed in the
1087 * vec argument. The array is modified so that it consists of the unfilled
1088 * portion of the original array.
1090 * The returned value is the total number of bytes received, or an error.
1093 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1094 struct kvec *vec, size_t num, size_t size, int flags)
1096 msg->msg_control_is_user = false;
1097 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1098 return sock_recvmsg(sock, msg, flags);
1100 EXPORT_SYMBOL(kernel_recvmsg);
1102 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1103 struct pipe_inode_info *pipe, size_t len,
1106 struct socket *sock = file->private_data;
1107 const struct proto_ops *ops;
1109 ops = READ_ONCE(sock->ops);
1110 if (unlikely(!ops->splice_read))
1111 return copy_splice_read(file, ppos, pipe, len, flags);
1113 return ops->splice_read(sock, ppos, pipe, len, flags);
1116 static void sock_splice_eof(struct file *file)
1118 struct socket *sock = file->private_data;
1119 const struct proto_ops *ops;
1121 ops = READ_ONCE(sock->ops);
1122 if (ops->splice_eof)
1123 ops->splice_eof(sock);
1126 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1128 struct file *file = iocb->ki_filp;
1129 struct socket *sock = file->private_data;
1130 struct msghdr msg = {.msg_iter = *to,
1134 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1135 msg.msg_flags = MSG_DONTWAIT;
1137 if (iocb->ki_pos != 0)
1140 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1143 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1148 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1150 struct file *file = iocb->ki_filp;
1151 struct socket *sock = file->private_data;
1152 struct msghdr msg = {.msg_iter = *from,
1156 if (iocb->ki_pos != 0)
1159 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1160 msg.msg_flags = MSG_DONTWAIT;
1162 if (sock->type == SOCK_SEQPACKET)
1163 msg.msg_flags |= MSG_EOR;
1165 res = __sock_sendmsg(sock, &msg);
1166 *from = msg.msg_iter;
1171 * Atomic setting of ioctl hooks to avoid race
1172 * with module unload.
1175 static DEFINE_MUTEX(br_ioctl_mutex);
1176 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1177 unsigned int cmd, struct ifreq *ifr,
1180 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1181 unsigned int cmd, struct ifreq *ifr,
1184 mutex_lock(&br_ioctl_mutex);
1185 br_ioctl_hook = hook;
1186 mutex_unlock(&br_ioctl_mutex);
1188 EXPORT_SYMBOL(brioctl_set);
1190 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1191 struct ifreq *ifr, void __user *uarg)
1196 request_module("bridge");
1198 mutex_lock(&br_ioctl_mutex);
1200 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1201 mutex_unlock(&br_ioctl_mutex);
1206 static DEFINE_MUTEX(vlan_ioctl_mutex);
1207 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1209 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1211 mutex_lock(&vlan_ioctl_mutex);
1212 vlan_ioctl_hook = hook;
1213 mutex_unlock(&vlan_ioctl_mutex);
1215 EXPORT_SYMBOL(vlan_ioctl_set);
1217 static long sock_do_ioctl(struct net *net, struct socket *sock,
1218 unsigned int cmd, unsigned long arg)
1220 const struct proto_ops *ops = READ_ONCE(sock->ops);
1224 void __user *argp = (void __user *)arg;
1227 err = ops->ioctl(sock, cmd, arg);
1230 * If this ioctl is unknown try to hand it down
1231 * to the NIC driver.
1233 if (err != -ENOIOCTLCMD)
1236 if (!is_socket_ioctl_cmd(cmd))
1239 if (get_user_ifreq(&ifr, &data, argp))
1241 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1242 if (!err && need_copyout)
1243 if (put_user_ifreq(&ifr, argp))
1250 * With an ioctl, arg may well be a user mode pointer, but we don't know
1251 * what to do with it - that's up to the protocol still.
1254 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1256 const struct proto_ops *ops;
1257 struct socket *sock;
1259 void __user *argp = (void __user *)arg;
1263 sock = file->private_data;
1264 ops = READ_ONCE(sock->ops);
1267 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1271 if (get_user_ifreq(&ifr, &data, argp))
1273 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1274 if (!err && need_copyout)
1275 if (put_user_ifreq(&ifr, argp))
1278 #ifdef CONFIG_WEXT_CORE
1279 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1280 err = wext_handle_ioctl(net, cmd, argp);
1287 if (get_user(pid, (int __user *)argp))
1289 err = f_setown(sock->file, pid, 1);
1293 err = put_user(f_getown(sock->file),
1294 (int __user *)argp);
1300 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1305 if (!vlan_ioctl_hook)
1306 request_module("8021q");
1308 mutex_lock(&vlan_ioctl_mutex);
1309 if (vlan_ioctl_hook)
1310 err = vlan_ioctl_hook(net, argp);
1311 mutex_unlock(&vlan_ioctl_mutex);
1315 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1318 err = open_related_ns(&net->ns, get_net_ns);
1320 case SIOCGSTAMP_OLD:
1321 case SIOCGSTAMPNS_OLD:
1322 if (!ops->gettstamp) {
1326 err = ops->gettstamp(sock, argp,
1327 cmd == SIOCGSTAMP_OLD,
1328 !IS_ENABLED(CONFIG_64BIT));
1330 case SIOCGSTAMP_NEW:
1331 case SIOCGSTAMPNS_NEW:
1332 if (!ops->gettstamp) {
1336 err = ops->gettstamp(sock, argp,
1337 cmd == SIOCGSTAMP_NEW,
1342 err = dev_ifconf(net, argp);
1346 err = sock_do_ioctl(net, sock, cmd, arg);
1353 * sock_create_lite - creates a socket
1354 * @family: protocol family (AF_INET, ...)
1355 * @type: communication type (SOCK_STREAM, ...)
1356 * @protocol: protocol (0, ...)
1359 * Creates a new socket and assigns it to @res, passing through LSM.
1360 * The new socket initialization is not complete, see kernel_accept().
1361 * Returns 0 or an error. On failure @res is set to %NULL.
1362 * This function internally uses GFP_KERNEL.
1365 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1368 struct socket *sock = NULL;
1370 err = security_socket_create(family, type, protocol, 1);
1374 sock = sock_alloc();
1381 err = security_socket_post_create(sock, family, type, protocol, 1);
1393 EXPORT_SYMBOL(sock_create_lite);
1395 /* No kernel lock held - perfect */
1396 static __poll_t sock_poll(struct file *file, poll_table *wait)
1398 struct socket *sock = file->private_data;
1399 const struct proto_ops *ops = READ_ONCE(sock->ops);
1400 __poll_t events = poll_requested_events(wait), flag = 0;
1405 if (sk_can_busy_loop(sock->sk)) {
1406 /* poll once if requested by the syscall */
1407 if (events & POLL_BUSY_LOOP)
1408 sk_busy_loop(sock->sk, 1);
1410 /* if this socket can poll_ll, tell the system call */
1411 flag = POLL_BUSY_LOOP;
1414 return ops->poll(file, sock, wait) | flag;
1417 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1419 struct socket *sock = file->private_data;
1421 return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1424 static int sock_close(struct inode *inode, struct file *filp)
1426 __sock_release(SOCKET_I(inode), inode);
1431 * Update the socket async list
1433 * Fasync_list locking strategy.
1435 * 1. fasync_list is modified only under process context socket lock
1436 * i.e. under semaphore.
1437 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1438 * or under socket lock
1441 static int sock_fasync(int fd, struct file *filp, int on)
1443 struct socket *sock = filp->private_data;
1444 struct sock *sk = sock->sk;
1445 struct socket_wq *wq = &sock->wq;
1451 fasync_helper(fd, filp, on, &wq->fasync_list);
1453 if (!wq->fasync_list)
1454 sock_reset_flag(sk, SOCK_FASYNC);
1456 sock_set_flag(sk, SOCK_FASYNC);
1462 /* This function may be called only under rcu_lock */
1464 int sock_wake_async(struct socket_wq *wq, int how, int band)
1466 if (!wq || !wq->fasync_list)
1470 case SOCK_WAKE_WAITD:
1471 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1474 case SOCK_WAKE_SPACE:
1475 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1480 kill_fasync(&wq->fasync_list, SIGIO, band);
1483 kill_fasync(&wq->fasync_list, SIGURG, band);
1488 EXPORT_SYMBOL(sock_wake_async);
1491 * __sock_create - creates a socket
1492 * @net: net namespace
1493 * @family: protocol family (AF_INET, ...)
1494 * @type: communication type (SOCK_STREAM, ...)
1495 * @protocol: protocol (0, ...)
1497 * @kern: boolean for kernel space sockets
1499 * Creates a new socket and assigns it to @res, passing through LSM.
1500 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1501 * be set to true if the socket resides in kernel space.
1502 * This function internally uses GFP_KERNEL.
1505 int __sock_create(struct net *net, int family, int type, int protocol,
1506 struct socket **res, int kern)
1509 struct socket *sock;
1510 const struct net_proto_family *pf;
1513 * Check protocol is in range
1515 if (family < 0 || family >= NPROTO)
1516 return -EAFNOSUPPORT;
1517 if (type < 0 || type >= SOCK_MAX)
1522 This uglymoron is moved from INET layer to here to avoid
1523 deadlock in module load.
1525 if (family == PF_INET && type == SOCK_PACKET) {
1526 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1531 err = security_socket_create(family, type, protocol, kern);
1536 * Allocate the socket and allow the family to set things up. if
1537 * the protocol is 0, the family is instructed to select an appropriate
1540 sock = sock_alloc();
1542 net_warn_ratelimited("socket: no more sockets\n");
1543 return -ENFILE; /* Not exactly a match, but its the
1544 closest posix thing */
1549 #ifdef CONFIG_MODULES
1550 /* Attempt to load a protocol module if the find failed.
1552 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1553 * requested real, full-featured networking support upon configuration.
1554 * Otherwise module support will break!
1556 if (rcu_access_pointer(net_families[family]) == NULL)
1557 request_module("net-pf-%d", family);
1561 pf = rcu_dereference(net_families[family]);
1562 err = -EAFNOSUPPORT;
1567 * We will call the ->create function, that possibly is in a loadable
1568 * module, so we have to bump that loadable module refcnt first.
1570 if (!try_module_get(pf->owner))
1573 /* Now protected by module ref count */
1576 err = pf->create(net, sock, protocol, kern);
1578 goto out_module_put;
1581 * Now to bump the refcnt of the [loadable] module that owns this
1582 * socket at sock_release time we decrement its refcnt.
1584 if (!try_module_get(sock->ops->owner))
1585 goto out_module_busy;
1588 * Now that we're done with the ->create function, the [loadable]
1589 * module can have its refcnt decremented
1591 module_put(pf->owner);
1592 err = security_socket_post_create(sock, family, type, protocol, kern);
1594 goto out_sock_release;
1600 err = -EAFNOSUPPORT;
1603 module_put(pf->owner);
1610 goto out_sock_release;
1612 EXPORT_SYMBOL(__sock_create);
1615 * sock_create - creates a socket
1616 * @family: protocol family (AF_INET, ...)
1617 * @type: communication type (SOCK_STREAM, ...)
1618 * @protocol: protocol (0, ...)
1621 * A wrapper around __sock_create().
1622 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1625 int sock_create(int family, int type, int protocol, struct socket **res)
1627 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1629 EXPORT_SYMBOL(sock_create);
1632 * sock_create_kern - creates a socket (kernel space)
1633 * @net: net namespace
1634 * @family: protocol family (AF_INET, ...)
1635 * @type: communication type (SOCK_STREAM, ...)
1636 * @protocol: protocol (0, ...)
1639 * A wrapper around __sock_create().
1640 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1643 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1645 return __sock_create(net, family, type, protocol, res, 1);
1647 EXPORT_SYMBOL(sock_create_kern);
1649 static struct socket *__sys_socket_create(int family, int type, int protocol)
1651 struct socket *sock;
1654 /* Check the SOCK_* constants for consistency. */
1655 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1656 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1657 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1658 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1660 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1661 return ERR_PTR(-EINVAL);
1662 type &= SOCK_TYPE_MASK;
1664 retval = sock_create(family, type, protocol, &sock);
1666 return ERR_PTR(retval);
1671 struct file *__sys_socket_file(int family, int type, int protocol)
1673 struct socket *sock;
1676 sock = __sys_socket_create(family, type, protocol);
1678 return ERR_CAST(sock);
1680 flags = type & ~SOCK_TYPE_MASK;
1681 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1682 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1684 return sock_alloc_file(sock, flags, NULL);
1687 /* A hook for bpf progs to attach to and update socket protocol.
1689 * A static noinline declaration here could cause the compiler to
1690 * optimize away the function. A global noinline declaration will
1691 * keep the definition, but may optimize away the callsite.
1692 * Therefore, __weak is needed to ensure that the call is still
1693 * emitted, by telling the compiler that we don't know what the
1694 * function might eventually be.
1699 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1706 int __sys_socket(int family, int type, int protocol)
1708 struct socket *sock;
1711 sock = __sys_socket_create(family, type,
1712 update_socket_protocol(family, type, protocol));
1714 return PTR_ERR(sock);
1716 flags = type & ~SOCK_TYPE_MASK;
1717 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1718 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1720 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1723 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1725 return __sys_socket(family, type, protocol);
1729 * Create a pair of connected sockets.
1732 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1734 struct socket *sock1, *sock2;
1736 struct file *newfile1, *newfile2;
1739 flags = type & ~SOCK_TYPE_MASK;
1740 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1742 type &= SOCK_TYPE_MASK;
1744 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1745 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1748 * reserve descriptors and make sure we won't fail
1749 * to return them to userland.
1751 fd1 = get_unused_fd_flags(flags);
1752 if (unlikely(fd1 < 0))
1755 fd2 = get_unused_fd_flags(flags);
1756 if (unlikely(fd2 < 0)) {
1761 err = put_user(fd1, &usockvec[0]);
1765 err = put_user(fd2, &usockvec[1]);
1770 * Obtain the first socket and check if the underlying protocol
1771 * supports the socketpair call.
1774 err = sock_create(family, type, protocol, &sock1);
1775 if (unlikely(err < 0))
1778 err = sock_create(family, type, protocol, &sock2);
1779 if (unlikely(err < 0)) {
1780 sock_release(sock1);
1784 err = security_socket_socketpair(sock1, sock2);
1785 if (unlikely(err)) {
1786 sock_release(sock2);
1787 sock_release(sock1);
1791 err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1792 if (unlikely(err < 0)) {
1793 sock_release(sock2);
1794 sock_release(sock1);
1798 newfile1 = sock_alloc_file(sock1, flags, NULL);
1799 if (IS_ERR(newfile1)) {
1800 err = PTR_ERR(newfile1);
1801 sock_release(sock2);
1805 newfile2 = sock_alloc_file(sock2, flags, NULL);
1806 if (IS_ERR(newfile2)) {
1807 err = PTR_ERR(newfile2);
1812 audit_fd_pair(fd1, fd2);
1814 fd_install(fd1, newfile1);
1815 fd_install(fd2, newfile2);
1824 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1825 int __user *, usockvec)
1827 return __sys_socketpair(family, type, protocol, usockvec);
1830 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1835 err = security_socket_bind(sock, (struct sockaddr *)address,
1838 err = READ_ONCE(sock->ops)->bind(sock,
1839 (struct sockaddr *)address,
1845 * Bind a name to a socket. Nothing much to do here since it's
1846 * the protocol's responsibility to handle the local address.
1848 * We move the socket address to kernel space before we call
1849 * the protocol layer (having also checked the address is ok).
1852 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1854 struct socket *sock;
1855 struct sockaddr_storage address;
1856 int err, fput_needed;
1858 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1860 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1862 err = __sys_bind_socket(sock, &address, addrlen);
1863 fput_light(sock->file, fput_needed);
1868 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1870 return __sys_bind(fd, umyaddr, addrlen);
1874 * Perform a listen. Basically, we allow the protocol to do anything
1875 * necessary for a listen, and if that works, we mark the socket as
1876 * ready for listening.
1878 int __sys_listen_socket(struct socket *sock, int backlog)
1882 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1883 if ((unsigned int)backlog > somaxconn)
1884 backlog = somaxconn;
1886 err = security_socket_listen(sock, backlog);
1888 err = READ_ONCE(sock->ops)->listen(sock, backlog);
1892 int __sys_listen(int fd, int backlog)
1894 struct socket *sock;
1895 int err, fput_needed;
1897 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1899 err = __sys_listen_socket(sock, backlog);
1900 fput_light(sock->file, fput_needed);
1905 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1907 return __sys_listen(fd, backlog);
1910 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1911 struct sockaddr __user *upeer_sockaddr,
1912 int __user *upeer_addrlen, int flags)
1914 struct socket *sock, *newsock;
1915 struct file *newfile;
1917 struct sockaddr_storage address;
1918 const struct proto_ops *ops;
1920 sock = sock_from_file(file);
1922 return ERR_PTR(-ENOTSOCK);
1924 newsock = sock_alloc();
1926 return ERR_PTR(-ENFILE);
1927 ops = READ_ONCE(sock->ops);
1929 newsock->type = sock->type;
1933 * We don't need try_module_get here, as the listening socket (sock)
1934 * has the protocol module (sock->ops->owner) held.
1936 __module_get(ops->owner);
1938 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1939 if (IS_ERR(newfile))
1942 err = security_socket_accept(sock, newsock);
1946 arg->flags |= sock->file->f_flags;
1947 err = ops->accept(sock, newsock, arg);
1951 if (upeer_sockaddr) {
1952 len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1954 err = -ECONNABORTED;
1957 err = move_addr_to_user(&address,
1958 len, upeer_sockaddr, upeer_addrlen);
1963 /* File flags are not inherited via accept() unlike another OSes. */
1967 return ERR_PTR(err);
1970 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1971 int __user *upeer_addrlen, int flags)
1973 struct proto_accept_arg arg = { };
1974 struct file *newfile;
1977 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1980 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1981 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1983 newfd = get_unused_fd_flags(flags);
1984 if (unlikely(newfd < 0))
1987 newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1989 if (IS_ERR(newfile)) {
1990 put_unused_fd(newfd);
1991 return PTR_ERR(newfile);
1993 fd_install(newfd, newfile);
1998 * For accept, we attempt to create a new socket, set up the link
1999 * with the client, wake up the client, then return the new
2000 * connected fd. We collect the address of the connector in kernel
2001 * space and move it to user at the very end. This is unclean because
2002 * we open the socket then return an error.
2004 * 1003.1g adds the ability to recvmsg() to query connection pending
2005 * status to recvmsg. We need to add that support in a way thats
2006 * clean when we restructure accept also.
2009 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2010 int __user *upeer_addrlen, int flags)
2017 ret = __sys_accept4_file(fd_file(f), upeer_sockaddr,
2018 upeer_addrlen, flags);
2025 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2026 int __user *, upeer_addrlen, int, flags)
2028 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2031 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2032 int __user *, upeer_addrlen)
2034 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2038 * Attempt to connect to a socket with the server address. The address
2039 * is in user space so we verify it is OK and move it to kernel space.
2041 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2044 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2045 * other SEQPACKET protocols that take time to connect() as it doesn't
2046 * include the -EINPROGRESS status for such sockets.
2049 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2050 int addrlen, int file_flags)
2052 struct socket *sock;
2055 sock = sock_from_file(file);
2062 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2066 err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2067 addrlen, sock->file->f_flags | file_flags);
2072 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2079 struct sockaddr_storage address;
2081 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2083 ret = __sys_connect_file(fd_file(f), &address, addrlen, 0);
2090 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2093 return __sys_connect(fd, uservaddr, addrlen);
2097 * Get the local address ('name') of a socket object. Move the obtained
2098 * name to user space.
2101 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2102 int __user *usockaddr_len)
2104 struct socket *sock;
2105 struct sockaddr_storage address;
2106 int err, fput_needed;
2108 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2112 err = security_socket_getsockname(sock);
2116 err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2119 /* "err" is actually length in this case */
2120 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2123 fput_light(sock->file, fput_needed);
2128 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2129 int __user *, usockaddr_len)
2131 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2135 * Get the remote address ('name') of a socket object. Move the obtained
2136 * name to user space.
2139 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2140 int __user *usockaddr_len)
2142 struct socket *sock;
2143 struct sockaddr_storage address;
2144 int err, fput_needed;
2146 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2148 const struct proto_ops *ops = READ_ONCE(sock->ops);
2150 err = security_socket_getpeername(sock);
2152 fput_light(sock->file, fput_needed);
2156 err = ops->getname(sock, (struct sockaddr *)&address, 1);
2158 /* "err" is actually length in this case */
2159 err = move_addr_to_user(&address, err, usockaddr,
2161 fput_light(sock->file, fput_needed);
2166 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2167 int __user *, usockaddr_len)
2169 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2173 * Send a datagram to a given address. We move the address into kernel
2174 * space and check the user space data area is readable before invoking
2177 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2178 struct sockaddr __user *addr, int addr_len)
2180 struct socket *sock;
2181 struct sockaddr_storage address;
2186 err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2189 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2193 msg.msg_name = NULL;
2194 msg.msg_control = NULL;
2195 msg.msg_controllen = 0;
2196 msg.msg_namelen = 0;
2197 msg.msg_ubuf = NULL;
2199 err = move_addr_to_kernel(addr, addr_len, &address);
2202 msg.msg_name = (struct sockaddr *)&address;
2203 msg.msg_namelen = addr_len;
2205 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2206 if (sock->file->f_flags & O_NONBLOCK)
2207 flags |= MSG_DONTWAIT;
2208 msg.msg_flags = flags;
2209 err = __sock_sendmsg(sock, &msg);
2212 fput_light(sock->file, fput_needed);
2217 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2218 unsigned int, flags, struct sockaddr __user *, addr,
2221 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2225 * Send a datagram down a socket.
2228 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2229 unsigned int, flags)
2231 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2235 * Receive a frame from the socket and optionally record the address of the
2236 * sender. We verify the buffers are writable and if needed move the
2237 * sender address from kernel to user space.
2239 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2240 struct sockaddr __user *addr, int __user *addr_len)
2242 struct sockaddr_storage address;
2243 struct msghdr msg = {
2244 /* Save some cycles and don't copy the address if not needed */
2245 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2247 struct socket *sock;
2251 err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2254 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2258 if (sock->file->f_flags & O_NONBLOCK)
2259 flags |= MSG_DONTWAIT;
2260 err = sock_recvmsg(sock, &msg, flags);
2262 if (err >= 0 && addr != NULL) {
2263 err2 = move_addr_to_user(&address,
2264 msg.msg_namelen, addr, addr_len);
2269 fput_light(sock->file, fput_needed);
2274 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2275 unsigned int, flags, struct sockaddr __user *, addr,
2276 int __user *, addr_len)
2278 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2282 * Receive a datagram from a socket.
2285 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2286 unsigned int, flags)
2288 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2291 static bool sock_use_custom_sol_socket(const struct socket *sock)
2293 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2296 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2297 int optname, sockptr_t optval, int optlen)
2299 const struct proto_ops *ops;
2300 char *kernel_optval = NULL;
2306 err = security_socket_setsockopt(sock, level, optname);
2311 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2322 optval = KERNEL_SOCKPTR(kernel_optval);
2323 ops = READ_ONCE(sock->ops);
2324 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2325 err = sock_setsockopt(sock, level, optname, optval, optlen);
2326 else if (unlikely(!ops->setsockopt))
2329 err = ops->setsockopt(sock, level, optname, optval,
2331 kfree(kernel_optval);
2335 EXPORT_SYMBOL(do_sock_setsockopt);
2337 /* Set a socket option. Because we don't know the option lengths we have
2338 * to pass the user mode parameter for the protocols to sort out.
2340 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2343 sockptr_t optval = USER_SOCKPTR(user_optval);
2344 bool compat = in_compat_syscall();
2345 int err, fput_needed;
2346 struct socket *sock;
2348 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2352 err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2354 fput_light(sock->file, fput_needed);
2358 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2359 char __user *, optval, int, optlen)
2361 return __sys_setsockopt(fd, level, optname, optval, optlen);
2364 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2367 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2368 int optname, sockptr_t optval, sockptr_t optlen)
2370 int max_optlen __maybe_unused = 0;
2371 const struct proto_ops *ops;
2374 err = security_socket_getsockopt(sock, level, optname);
2379 copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2381 ops = READ_ONCE(sock->ops);
2382 if (level == SOL_SOCKET) {
2383 err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2384 } else if (unlikely(!ops->getsockopt)) {
2387 if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2388 "Invalid argument type"))
2391 err = ops->getsockopt(sock, level, optname, optval.user,
2396 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2397 optval, optlen, max_optlen,
2402 EXPORT_SYMBOL(do_sock_getsockopt);
2405 * Get a socket option. Because we don't know the option lengths we have
2406 * to pass a user mode parameter for the protocols to sort out.
2408 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2411 int err, fput_needed;
2412 struct socket *sock;
2415 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2419 compat = in_compat_syscall();
2420 err = do_sock_getsockopt(sock, compat, level, optname,
2421 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2423 fput_light(sock->file, fput_needed);
2427 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2428 char __user *, optval, int __user *, optlen)
2430 return __sys_getsockopt(fd, level, optname, optval, optlen);
2434 * Shutdown a socket.
2437 int __sys_shutdown_sock(struct socket *sock, int how)
2441 err = security_socket_shutdown(sock, how);
2443 err = READ_ONCE(sock->ops)->shutdown(sock, how);
2448 int __sys_shutdown(int fd, int how)
2450 int err, fput_needed;
2451 struct socket *sock;
2453 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2455 err = __sys_shutdown_sock(sock, how);
2456 fput_light(sock->file, fput_needed);
2461 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2463 return __sys_shutdown(fd, how);
2466 /* A couple of helpful macros for getting the address of the 32/64 bit
2467 * fields which are the same type (int / unsigned) on our platforms.
2469 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2470 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2471 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2473 struct used_address {
2474 struct sockaddr_storage name;
2475 unsigned int name_len;
2478 int __copy_msghdr(struct msghdr *kmsg,
2479 struct user_msghdr *msg,
2480 struct sockaddr __user **save_addr)
2484 kmsg->msg_control_is_user = true;
2485 kmsg->msg_get_inq = 0;
2486 kmsg->msg_control_user = msg->msg_control;
2487 kmsg->msg_controllen = msg->msg_controllen;
2488 kmsg->msg_flags = msg->msg_flags;
2490 kmsg->msg_namelen = msg->msg_namelen;
2492 kmsg->msg_namelen = 0;
2494 if (kmsg->msg_namelen < 0)
2497 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2498 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2501 *save_addr = msg->msg_name;
2503 if (msg->msg_name && kmsg->msg_namelen) {
2505 err = move_addr_to_kernel(msg->msg_name,
2512 kmsg->msg_name = NULL;
2513 kmsg->msg_namelen = 0;
2516 if (msg->msg_iovlen > UIO_MAXIOV)
2519 kmsg->msg_iocb = NULL;
2520 kmsg->msg_ubuf = NULL;
2524 static int copy_msghdr_from_user(struct msghdr *kmsg,
2525 struct user_msghdr __user *umsg,
2526 struct sockaddr __user **save_addr,
2529 struct user_msghdr msg;
2532 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2535 err = __copy_msghdr(kmsg, &msg, save_addr);
2539 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2540 msg.msg_iov, msg.msg_iovlen,
2541 UIO_FASTIOV, iov, &kmsg->msg_iter);
2542 return err < 0 ? err : 0;
2545 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2546 unsigned int flags, struct used_address *used_address,
2547 unsigned int allowed_msghdr_flags)
2549 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2550 __aligned(sizeof(__kernel_size_t));
2551 /* 20 is size of ipv6_pktinfo */
2552 unsigned char *ctl_buf = ctl;
2558 if (msg_sys->msg_controllen > INT_MAX)
2560 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2561 ctl_len = msg_sys->msg_controllen;
2562 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2564 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2568 ctl_buf = msg_sys->msg_control;
2569 ctl_len = msg_sys->msg_controllen;
2570 } else if (ctl_len) {
2571 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2572 CMSG_ALIGN(sizeof(struct cmsghdr)));
2573 if (ctl_len > sizeof(ctl)) {
2574 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2575 if (ctl_buf == NULL)
2579 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2581 msg_sys->msg_control = ctl_buf;
2582 msg_sys->msg_control_is_user = false;
2584 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2585 msg_sys->msg_flags = flags;
2587 if (sock->file->f_flags & O_NONBLOCK)
2588 msg_sys->msg_flags |= MSG_DONTWAIT;
2590 * If this is sendmmsg() and current destination address is same as
2591 * previously succeeded address, omit asking LSM's decision.
2592 * used_address->name_len is initialized to UINT_MAX so that the first
2593 * destination address never matches.
2595 if (used_address && msg_sys->msg_name &&
2596 used_address->name_len == msg_sys->msg_namelen &&
2597 !memcmp(&used_address->name, msg_sys->msg_name,
2598 used_address->name_len)) {
2599 err = sock_sendmsg_nosec(sock, msg_sys);
2602 err = __sock_sendmsg(sock, msg_sys);
2604 * If this is sendmmsg() and sending to current destination address was
2605 * successful, remember it.
2607 if (used_address && err >= 0) {
2608 used_address->name_len = msg_sys->msg_namelen;
2609 if (msg_sys->msg_name)
2610 memcpy(&used_address->name, msg_sys->msg_name,
2611 used_address->name_len);
2616 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2621 static int sendmsg_copy_msghdr(struct msghdr *msg,
2622 struct user_msghdr __user *umsg, unsigned flags,
2627 if (flags & MSG_CMSG_COMPAT) {
2628 struct compat_msghdr __user *msg_compat;
2630 msg_compat = (struct compat_msghdr __user *) umsg;
2631 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2633 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2641 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2642 struct msghdr *msg_sys, unsigned int flags,
2643 struct used_address *used_address,
2644 unsigned int allowed_msghdr_flags)
2646 struct sockaddr_storage address;
2647 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2650 msg_sys->msg_name = &address;
2652 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2656 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2657 allowed_msghdr_flags);
2663 * BSD sendmsg interface
2665 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2668 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2671 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2672 bool forbid_cmsg_compat)
2674 int fput_needed, err;
2675 struct msghdr msg_sys;
2676 struct socket *sock;
2678 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2681 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2685 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2687 fput_light(sock->file, fput_needed);
2692 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2694 return __sys_sendmsg(fd, msg, flags, true);
2698 * Linux sendmmsg interface
2701 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2702 unsigned int flags, bool forbid_cmsg_compat)
2704 int fput_needed, err, datagrams;
2705 struct socket *sock;
2706 struct mmsghdr __user *entry;
2707 struct compat_mmsghdr __user *compat_entry;
2708 struct msghdr msg_sys;
2709 struct used_address used_address;
2710 unsigned int oflags = flags;
2712 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2715 if (vlen > UIO_MAXIOV)
2720 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2724 used_address.name_len = UINT_MAX;
2726 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2730 while (datagrams < vlen) {
2731 if (datagrams == vlen - 1)
2734 if (MSG_CMSG_COMPAT & flags) {
2735 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2736 &msg_sys, flags, &used_address, MSG_EOR);
2739 err = __put_user(err, &compat_entry->msg_len);
2742 err = ___sys_sendmsg(sock,
2743 (struct user_msghdr __user *)entry,
2744 &msg_sys, flags, &used_address, MSG_EOR);
2747 err = put_user(err, &entry->msg_len);
2754 if (msg_data_left(&msg_sys))
2759 fput_light(sock->file, fput_needed);
2761 /* We only return an error if no datagrams were able to be sent */
2768 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2769 unsigned int, vlen, unsigned int, flags)
2771 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2774 static int recvmsg_copy_msghdr(struct msghdr *msg,
2775 struct user_msghdr __user *umsg, unsigned flags,
2776 struct sockaddr __user **uaddr,
2781 if (MSG_CMSG_COMPAT & flags) {
2782 struct compat_msghdr __user *msg_compat;
2784 msg_compat = (struct compat_msghdr __user *) umsg;
2785 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2787 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2795 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2796 struct user_msghdr __user *msg,
2797 struct sockaddr __user *uaddr,
2798 unsigned int flags, int nosec)
2800 struct compat_msghdr __user *msg_compat =
2801 (struct compat_msghdr __user *) msg;
2802 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2803 struct sockaddr_storage addr;
2804 unsigned long cmsg_ptr;
2808 msg_sys->msg_name = &addr;
2809 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2810 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2812 /* We assume all kernel code knows the size of sockaddr_storage */
2813 msg_sys->msg_namelen = 0;
2815 if (sock->file->f_flags & O_NONBLOCK)
2816 flags |= MSG_DONTWAIT;
2818 if (unlikely(nosec))
2819 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2821 err = sock_recvmsg(sock, msg_sys, flags);
2827 if (uaddr != NULL) {
2828 err = move_addr_to_user(&addr,
2829 msg_sys->msg_namelen, uaddr,
2834 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2838 if (MSG_CMSG_COMPAT & flags)
2839 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2840 &msg_compat->msg_controllen);
2842 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2843 &msg->msg_controllen);
2851 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2852 struct msghdr *msg_sys, unsigned int flags, int nosec)
2854 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2855 /* user mode address pointers */
2856 struct sockaddr __user *uaddr;
2859 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2863 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2869 * BSD recvmsg interface
2872 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2873 struct user_msghdr __user *umsg,
2874 struct sockaddr __user *uaddr, unsigned int flags)
2876 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2879 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2880 bool forbid_cmsg_compat)
2882 int fput_needed, err;
2883 struct msghdr msg_sys;
2884 struct socket *sock;
2886 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2889 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2893 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2895 fput_light(sock->file, fput_needed);
2900 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2901 unsigned int, flags)
2903 return __sys_recvmsg(fd, msg, flags, true);
2907 * Linux recvmmsg interface
2910 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2911 unsigned int vlen, unsigned int flags,
2912 struct timespec64 *timeout)
2914 int fput_needed, err, datagrams;
2915 struct socket *sock;
2916 struct mmsghdr __user *entry;
2917 struct compat_mmsghdr __user *compat_entry;
2918 struct msghdr msg_sys;
2919 struct timespec64 end_time;
2920 struct timespec64 timeout64;
2923 poll_select_set_timeout(&end_time, timeout->tv_sec,
2929 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2933 if (likely(!(flags & MSG_ERRQUEUE))) {
2934 err = sock_error(sock->sk);
2942 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2944 while (datagrams < vlen) {
2946 * No need to ask LSM for more than the first datagram.
2948 if (MSG_CMSG_COMPAT & flags) {
2949 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2950 &msg_sys, flags & ~MSG_WAITFORONE,
2954 err = __put_user(err, &compat_entry->msg_len);
2957 err = ___sys_recvmsg(sock,
2958 (struct user_msghdr __user *)entry,
2959 &msg_sys, flags & ~MSG_WAITFORONE,
2963 err = put_user(err, &entry->msg_len);
2971 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2972 if (flags & MSG_WAITFORONE)
2973 flags |= MSG_DONTWAIT;
2976 ktime_get_ts64(&timeout64);
2977 *timeout = timespec64_sub(end_time, timeout64);
2978 if (timeout->tv_sec < 0) {
2979 timeout->tv_sec = timeout->tv_nsec = 0;
2983 /* Timeout, return less than vlen datagrams */
2984 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2988 /* Out of band data, return right away */
2989 if (msg_sys.msg_flags & MSG_OOB)
2997 if (datagrams == 0) {
3003 * We may return less entries than requested (vlen) if the
3004 * sock is non block and there aren't enough datagrams...
3006 if (err != -EAGAIN) {
3008 * ... or if recvmsg returns an error after we
3009 * received some datagrams, where we record the
3010 * error to return on the next call or if the
3011 * app asks about it using getsockopt(SO_ERROR).
3013 WRITE_ONCE(sock->sk->sk_err, -err);
3016 fput_light(sock->file, fput_needed);
3021 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3022 unsigned int vlen, unsigned int flags,
3023 struct __kernel_timespec __user *timeout,
3024 struct old_timespec32 __user *timeout32)
3027 struct timespec64 timeout_sys;
3029 if (timeout && get_timespec64(&timeout_sys, timeout))
3032 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3035 if (!timeout && !timeout32)
3036 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3038 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3043 if (timeout && put_timespec64(&timeout_sys, timeout))
3044 datagrams = -EFAULT;
3046 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3047 datagrams = -EFAULT;
3052 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3053 unsigned int, vlen, unsigned int, flags,
3054 struct __kernel_timespec __user *, timeout)
3056 if (flags & MSG_CMSG_COMPAT)
3059 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3062 #ifdef CONFIG_COMPAT_32BIT_TIME
3063 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3064 unsigned int, vlen, unsigned int, flags,
3065 struct old_timespec32 __user *, timeout)
3067 if (flags & MSG_CMSG_COMPAT)
3070 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3074 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3075 /* Argument list sizes for sys_socketcall */
3076 #define AL(x) ((x) * sizeof(unsigned long))
3077 static const unsigned char nargs[21] = {
3078 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3079 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3080 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3087 * System call vectors.
3089 * Argument checking cleaned up. Saved 20% in size.
3090 * This function doesn't need to set the kernel lock because
3091 * it is set by the callees.
3094 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3096 unsigned long a[AUDITSC_ARGS];
3097 unsigned long a0, a1;
3101 if (call < 1 || call > SYS_SENDMMSG)
3103 call = array_index_nospec(call, SYS_SENDMMSG + 1);
3106 if (len > sizeof(a))
3109 /* copy_from_user should be SMP safe. */
3110 if (copy_from_user(a, args, len))
3113 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3122 err = __sys_socket(a0, a1, a[2]);
3125 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3128 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3131 err = __sys_listen(a0, a1);
3134 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3135 (int __user *)a[2], 0);
3137 case SYS_GETSOCKNAME:
3139 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3140 (int __user *)a[2]);
3142 case SYS_GETPEERNAME:
3144 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3145 (int __user *)a[2]);
3147 case SYS_SOCKETPAIR:
3148 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3151 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3155 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3156 (struct sockaddr __user *)a[4], a[5]);
3159 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3163 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3164 (struct sockaddr __user *)a[4],
3165 (int __user *)a[5]);
3168 err = __sys_shutdown(a0, a1);
3170 case SYS_SETSOCKOPT:
3171 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3174 case SYS_GETSOCKOPT:
3176 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3177 (int __user *)a[4]);
3180 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3184 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3188 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3192 if (IS_ENABLED(CONFIG_64BIT))
3193 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3195 (struct __kernel_timespec __user *)a[4],
3198 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3200 (struct old_timespec32 __user *)a[4]);
3203 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3204 (int __user *)a[2], a[3]);
3213 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3216 * sock_register - add a socket protocol handler
3217 * @ops: description of protocol
3219 * This function is called by a protocol handler that wants to
3220 * advertise its address family, and have it linked into the
3221 * socket interface. The value ops->family corresponds to the
3222 * socket system call protocol family.
3224 int sock_register(const struct net_proto_family *ops)
3228 if (ops->family >= NPROTO) {
3229 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3233 spin_lock(&net_family_lock);
3234 if (rcu_dereference_protected(net_families[ops->family],
3235 lockdep_is_held(&net_family_lock)))
3238 rcu_assign_pointer(net_families[ops->family], ops);
3241 spin_unlock(&net_family_lock);
3243 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3246 EXPORT_SYMBOL(sock_register);
3249 * sock_unregister - remove a protocol handler
3250 * @family: protocol family to remove
3252 * This function is called by a protocol handler that wants to
3253 * remove its address family, and have it unlinked from the
3254 * new socket creation.
3256 * If protocol handler is a module, then it can use module reference
3257 * counts to protect against new references. If protocol handler is not
3258 * a module then it needs to provide its own protection in
3259 * the ops->create routine.
3261 void sock_unregister(int family)
3263 BUG_ON(family < 0 || family >= NPROTO);
3265 spin_lock(&net_family_lock);
3266 RCU_INIT_POINTER(net_families[family], NULL);
3267 spin_unlock(&net_family_lock);
3271 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3273 EXPORT_SYMBOL(sock_unregister);
3275 bool sock_is_registered(int family)
3277 return family < NPROTO && rcu_access_pointer(net_families[family]);
3280 static int __init sock_init(void)
3284 * Initialize the network sysctl infrastructure.
3286 err = net_sysctl_init();
3291 * Initialize skbuff SLAB cache
3296 * Initialize the protocols module.
3301 err = register_filesystem(&sock_fs_type);
3304 sock_mnt = kern_mount(&sock_fs_type);
3305 if (IS_ERR(sock_mnt)) {
3306 err = PTR_ERR(sock_mnt);
3310 /* The real protocol initialization is performed in later initcalls.
3313 #ifdef CONFIG_NETFILTER
3314 err = netfilter_init();
3319 ptp_classifier_init();
3325 unregister_filesystem(&sock_fs_type);
3329 core_initcall(sock_init); /* early initcall */
3331 #ifdef CONFIG_PROC_FS
3332 void socket_seq_show(struct seq_file *seq)
3334 seq_printf(seq, "sockets: used %d\n",
3335 sock_inuse_get(seq->private));
3337 #endif /* CONFIG_PROC_FS */
3339 /* Handle the fact that while struct ifreq has the same *layout* on
3340 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3341 * which are handled elsewhere, it still has different *size* due to
3342 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3343 * resulting in struct ifreq being 32 and 40 bytes respectively).
3344 * As a result, if the struct happens to be at the end of a page and
3345 * the next page isn't readable/writable, we get a fault. To prevent
3346 * that, copy back and forth to the full size.
3348 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3350 if (in_compat_syscall()) {
3351 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3353 memset(ifr, 0, sizeof(*ifr));
3354 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3358 *ifrdata = compat_ptr(ifr32->ifr_data);
3363 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3367 *ifrdata = ifr->ifr_data;
3371 EXPORT_SYMBOL(get_user_ifreq);
3373 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3375 size_t size = sizeof(*ifr);
3377 if (in_compat_syscall())
3378 size = sizeof(struct compat_ifreq);
3380 if (copy_to_user(arg, ifr, size))
3385 EXPORT_SYMBOL(put_user_ifreq);
3387 #ifdef CONFIG_COMPAT
3388 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3390 compat_uptr_t uptr32;
3395 if (get_user_ifreq(&ifr, NULL, uifr32))
3398 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3401 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3402 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3404 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3406 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3407 if (put_user_ifreq(&ifr, uifr32))
3413 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3414 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3415 struct compat_ifreq __user *u_ifreq32)
3420 if (!is_socket_ioctl_cmd(cmd))
3422 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3424 ifreq.ifr_data = data;
3426 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3429 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3430 unsigned int cmd, unsigned long arg)
3432 void __user *argp = compat_ptr(arg);
3433 struct sock *sk = sock->sk;
3434 struct net *net = sock_net(sk);
3435 const struct proto_ops *ops;
3437 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3438 return sock_ioctl(file, cmd, (unsigned long)argp);
3442 return compat_siocwandev(net, argp);
3443 case SIOCGSTAMP_OLD:
3444 case SIOCGSTAMPNS_OLD:
3445 ops = READ_ONCE(sock->ops);
3446 if (!ops->gettstamp)
3447 return -ENOIOCTLCMD;
3448 return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3449 !COMPAT_USE_64BIT_TIME);
3452 case SIOCBONDSLAVEINFOQUERY:
3453 case SIOCBONDINFOQUERY:
3456 return compat_ifr_data_ioctl(net, cmd, argp);
3467 case SIOCGSTAMP_NEW:
3468 case SIOCGSTAMPNS_NEW:
3472 return sock_ioctl(file, cmd, arg);
3491 case SIOCSIFHWBROADCAST:
3493 case SIOCGIFBRDADDR:
3494 case SIOCSIFBRDADDR:
3495 case SIOCGIFDSTADDR:
3496 case SIOCSIFDSTADDR:
3497 case SIOCGIFNETMASK:
3498 case SIOCSIFNETMASK:
3510 case SIOCBONDENSLAVE:
3511 case SIOCBONDRELEASE:
3512 case SIOCBONDSETHWADDR:
3513 case SIOCBONDCHANGEACTIVE:
3520 return sock_do_ioctl(net, sock, cmd, arg);
3523 return -ENOIOCTLCMD;
3526 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3529 struct socket *sock = file->private_data;
3530 const struct proto_ops *ops = READ_ONCE(sock->ops);
3531 int ret = -ENOIOCTLCMD;
3538 if (ops->compat_ioctl)
3539 ret = ops->compat_ioctl(sock, cmd, arg);
3541 if (ret == -ENOIOCTLCMD &&
3542 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3543 ret = compat_wext_handle_ioctl(net, cmd, arg);
3545 if (ret == -ENOIOCTLCMD)
3546 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3553 * kernel_bind - bind an address to a socket (kernel space)
3556 * @addrlen: length of address
3558 * Returns 0 or an error.
3561 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3563 struct sockaddr_storage address;
3565 memcpy(&address, addr, addrlen);
3567 return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3570 EXPORT_SYMBOL(kernel_bind);
3573 * kernel_listen - move socket to listening state (kernel space)
3575 * @backlog: pending connections queue size
3577 * Returns 0 or an error.
3580 int kernel_listen(struct socket *sock, int backlog)
3582 return READ_ONCE(sock->ops)->listen(sock, backlog);
3584 EXPORT_SYMBOL(kernel_listen);
3587 * kernel_accept - accept a connection (kernel space)
3588 * @sock: listening socket
3589 * @newsock: new connected socket
3592 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3593 * If it fails, @newsock is guaranteed to be %NULL.
3594 * Returns 0 or an error.
3597 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3599 struct sock *sk = sock->sk;
3600 const struct proto_ops *ops = READ_ONCE(sock->ops);
3601 struct proto_accept_arg arg = {
3607 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3612 err = ops->accept(sock, *newsock, &arg);
3614 sock_release(*newsock);
3619 (*newsock)->ops = ops;
3620 __module_get(ops->owner);
3625 EXPORT_SYMBOL(kernel_accept);
3628 * kernel_connect - connect a socket (kernel space)
3631 * @addrlen: address length
3632 * @flags: flags (O_NONBLOCK, ...)
3634 * For datagram sockets, @addr is the address to which datagrams are sent
3635 * by default, and the only address from which datagrams are received.
3636 * For stream sockets, attempts to connect to @addr.
3637 * Returns 0 or an error code.
3640 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3643 struct sockaddr_storage address;
3645 memcpy(&address, addr, addrlen);
3647 return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3650 EXPORT_SYMBOL(kernel_connect);
3653 * kernel_getsockname - get the address which the socket is bound (kernel space)
3655 * @addr: address holder
3657 * Fills the @addr pointer with the address which the socket is bound.
3658 * Returns the length of the address in bytes or an error code.
3661 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3663 return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3665 EXPORT_SYMBOL(kernel_getsockname);
3668 * kernel_getpeername - get the address which the socket is connected (kernel space)
3670 * @addr: address holder
3672 * Fills the @addr pointer with the address which the socket is connected.
3673 * Returns the length of the address in bytes or an error code.
3676 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3678 return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3680 EXPORT_SYMBOL(kernel_getpeername);
3683 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3685 * @how: connection part
3687 * Returns 0 or an error.
3690 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3692 return READ_ONCE(sock->ops)->shutdown(sock, how);
3694 EXPORT_SYMBOL(kernel_sock_shutdown);
3697 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3700 * This routine returns the IP overhead imposed by a socket i.e.
3701 * the length of the underlying IP header, depending on whether
3702 * this is an IPv4 or IPv6 socket and the length from IP options turned
3703 * on at the socket. Assumes that the caller has a lock on the socket.
3706 u32 kernel_sock_ip_overhead(struct sock *sk)
3708 struct inet_sock *inet;
3709 struct ip_options_rcu *opt;
3711 #if IS_ENABLED(CONFIG_IPV6)
3712 struct ipv6_pinfo *np;
3713 struct ipv6_txoptions *optv6 = NULL;
3714 #endif /* IS_ENABLED(CONFIG_IPV6) */
3719 switch (sk->sk_family) {
3722 overhead += sizeof(struct iphdr);
3723 opt = rcu_dereference_protected(inet->inet_opt,
3724 sock_owned_by_user(sk));
3726 overhead += opt->opt.optlen;
3728 #if IS_ENABLED(CONFIG_IPV6)
3731 overhead += sizeof(struct ipv6hdr);
3733 optv6 = rcu_dereference_protected(np->opt,
3734 sock_owned_by_user(sk));
3736 overhead += (optv6->opt_flen + optv6->opt_nflen);
3738 #endif /* IS_ENABLED(CONFIG_IPV6) */
3739 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3743 EXPORT_SYMBOL(kernel_sock_ip_overhead);