1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
96 #include <net/cls_cgroup.h>
99 #include <linux/netfilter.h>
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
109 #include <trace/events/sock.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 static int sock_close(struct inode *inode, struct file *file);
121 static __poll_t sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
135 #ifdef CONFIG_PROC_FS
136 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138 struct socket *sock = f->private_data;
140 if (sock->ops->show_fdinfo)
141 sock->ops->show_fdinfo(m, sock);
144 #define sock_show_fdinfo NULL
148 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
149 * in the operation structures but are done directly via the socketcall() multiplexor.
152 static const struct file_operations socket_file_ops = {
153 .owner = THIS_MODULE,
155 .read_iter = sock_read_iter,
156 .write_iter = sock_write_iter,
158 .unlocked_ioctl = sock_ioctl,
160 .compat_ioctl = compat_sock_ioctl,
163 .release = sock_close,
164 .fasync = sock_fasync,
165 .sendpage = sock_sendpage,
166 .splice_write = generic_splice_sendpage,
167 .splice_read = sock_splice_read,
168 .show_fdinfo = sock_show_fdinfo,
171 static const char * const pf_family_names[] = {
172 [PF_UNSPEC] = "PF_UNSPEC",
173 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
174 [PF_INET] = "PF_INET",
175 [PF_AX25] = "PF_AX25",
177 [PF_APPLETALK] = "PF_APPLETALK",
178 [PF_NETROM] = "PF_NETROM",
179 [PF_BRIDGE] = "PF_BRIDGE",
180 [PF_ATMPVC] = "PF_ATMPVC",
182 [PF_INET6] = "PF_INET6",
183 [PF_ROSE] = "PF_ROSE",
184 [PF_DECnet] = "PF_DECnet",
185 [PF_NETBEUI] = "PF_NETBEUI",
186 [PF_SECURITY] = "PF_SECURITY",
188 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
189 [PF_PACKET] = "PF_PACKET",
191 [PF_ECONET] = "PF_ECONET",
192 [PF_ATMSVC] = "PF_ATMSVC",
195 [PF_IRDA] = "PF_IRDA",
196 [PF_PPPOX] = "PF_PPPOX",
197 [PF_WANPIPE] = "PF_WANPIPE",
200 [PF_MPLS] = "PF_MPLS",
202 [PF_TIPC] = "PF_TIPC",
203 [PF_BLUETOOTH] = "PF_BLUETOOTH",
204 [PF_IUCV] = "PF_IUCV",
205 [PF_RXRPC] = "PF_RXRPC",
206 [PF_ISDN] = "PF_ISDN",
207 [PF_PHONET] = "PF_PHONET",
208 [PF_IEEE802154] = "PF_IEEE802154",
209 [PF_CAIF] = "PF_CAIF",
212 [PF_VSOCK] = "PF_VSOCK",
214 [PF_QIPCRTR] = "PF_QIPCRTR",
217 [PF_MCTP] = "PF_MCTP",
221 * The protocol list. Each protocol is registered in here.
224 static DEFINE_SPINLOCK(net_family_lock);
225 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
229 * Move socket addresses back and forth across the kernel/user
230 * divide and look after the messy bits.
234 * move_addr_to_kernel - copy a socket address into kernel space
235 * @uaddr: Address in user space
236 * @kaddr: Address in kernel space
237 * @ulen: Length in user space
239 * The address is copied into kernel space. If the provided address is
240 * too long an error code of -EINVAL is returned. If the copy gives
241 * invalid addresses -EFAULT is returned. On a success 0 is returned.
244 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
246 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
250 if (copy_from_user(kaddr, uaddr, ulen))
252 return audit_sockaddr(ulen, kaddr);
256 * move_addr_to_user - copy an address to user space
257 * @kaddr: kernel space address
258 * @klen: length of address in kernel
259 * @uaddr: user space address
260 * @ulen: pointer to user length field
262 * The value pointed to by ulen on entry is the buffer length available.
263 * This is overwritten with the buffer space used. -EINVAL is returned
264 * if an overlong buffer is specified or a negative buffer size. -EFAULT
265 * is returned if either the buffer or the length field are not
267 * After copying the data up to the limit the user specifies, the true
268 * length of the data is written over the length limit the user
269 * specified. Zero is returned for a success.
272 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
273 void __user *uaddr, int __user *ulen)
278 BUG_ON(klen > sizeof(struct sockaddr_storage));
279 err = get_user(len, ulen);
287 if (audit_sockaddr(klen, kaddr))
289 if (copy_to_user(uaddr, kaddr, len))
293 * "fromlen shall refer to the value before truncation.."
296 return __put_user(klen, ulen);
299 static struct kmem_cache *sock_inode_cachep __ro_after_init;
301 static struct inode *sock_alloc_inode(struct super_block *sb)
303 struct socket_alloc *ei;
305 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
308 init_waitqueue_head(&ei->socket.wq.wait);
309 ei->socket.wq.fasync_list = NULL;
310 ei->socket.wq.flags = 0;
312 ei->socket.state = SS_UNCONNECTED;
313 ei->socket.flags = 0;
314 ei->socket.ops = NULL;
315 ei->socket.sk = NULL;
316 ei->socket.file = NULL;
318 return &ei->vfs_inode;
321 static void sock_free_inode(struct inode *inode)
323 struct socket_alloc *ei;
325 ei = container_of(inode, struct socket_alloc, vfs_inode);
326 kmem_cache_free(sock_inode_cachep, ei);
329 static void init_once(void *foo)
331 struct socket_alloc *ei = (struct socket_alloc *)foo;
333 inode_init_once(&ei->vfs_inode);
336 static void init_inodecache(void)
338 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
339 sizeof(struct socket_alloc),
341 (SLAB_HWCACHE_ALIGN |
342 SLAB_RECLAIM_ACCOUNT |
343 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
345 BUG_ON(sock_inode_cachep == NULL);
348 static const struct super_operations sockfs_ops = {
349 .alloc_inode = sock_alloc_inode,
350 .free_inode = sock_free_inode,
351 .statfs = simple_statfs,
355 * sockfs_dname() is called from d_path().
357 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
359 return dynamic_dname(buffer, buflen, "socket:[%lu]",
360 d_inode(dentry)->i_ino);
363 static const struct dentry_operations sockfs_dentry_operations = {
364 .d_dname = sockfs_dname,
367 static int sockfs_xattr_get(const struct xattr_handler *handler,
368 struct dentry *dentry, struct inode *inode,
369 const char *suffix, void *value, size_t size)
372 if (dentry->d_name.len + 1 > size)
374 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
376 return dentry->d_name.len + 1;
379 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
380 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
381 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
383 static const struct xattr_handler sockfs_xattr_handler = {
384 .name = XATTR_NAME_SOCKPROTONAME,
385 .get = sockfs_xattr_get,
388 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
389 struct mnt_idmap *idmap,
390 struct dentry *dentry, struct inode *inode,
391 const char *suffix, const void *value,
392 size_t size, int flags)
394 /* Handled by LSM. */
398 static const struct xattr_handler sockfs_security_xattr_handler = {
399 .prefix = XATTR_SECURITY_PREFIX,
400 .set = sockfs_security_xattr_set,
403 static const struct xattr_handler *sockfs_xattr_handlers[] = {
404 &sockfs_xattr_handler,
405 &sockfs_security_xattr_handler,
409 static int sockfs_init_fs_context(struct fs_context *fc)
411 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
414 ctx->ops = &sockfs_ops;
415 ctx->dops = &sockfs_dentry_operations;
416 ctx->xattr = sockfs_xattr_handlers;
420 static struct vfsmount *sock_mnt __read_mostly;
422 static struct file_system_type sock_fs_type = {
424 .init_fs_context = sockfs_init_fs_context,
425 .kill_sb = kill_anon_super,
429 * Obtains the first available file descriptor and sets it up for use.
431 * These functions create file structures and maps them to fd space
432 * of the current process. On success it returns file descriptor
433 * and file struct implicitly stored in sock->file.
434 * Note that another thread may close file descriptor before we return
435 * from this function. We use the fact that now we do not refer
436 * to socket after mapping. If one day we will need it, this
437 * function will increment ref. count on file by 1.
439 * In any case returned fd MAY BE not valid!
440 * This race condition is unavoidable
441 * with shared fd spaces, we cannot solve it inside kernel,
442 * but we take care of internal coherence yet.
446 * sock_alloc_file - Bind a &socket to a &file
448 * @flags: file status flags
449 * @dname: protocol name
451 * Returns the &file bound with @sock, implicitly storing it
452 * in sock->file. If dname is %NULL, sets to "".
454 * On failure @sock is released, and an ERR pointer is returned.
456 * This function uses GFP_KERNEL internally.
459 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
464 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
466 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
467 O_RDWR | (flags & O_NONBLOCK),
475 file->private_data = sock;
476 stream_open(SOCK_INODE(sock), file);
479 EXPORT_SYMBOL(sock_alloc_file);
481 static int sock_map_fd(struct socket *sock, int flags)
483 struct file *newfile;
484 int fd = get_unused_fd_flags(flags);
485 if (unlikely(fd < 0)) {
490 newfile = sock_alloc_file(sock, flags, NULL);
491 if (!IS_ERR(newfile)) {
492 fd_install(fd, newfile);
497 return PTR_ERR(newfile);
501 * sock_from_file - Return the &socket bounded to @file.
504 * On failure returns %NULL.
507 struct socket *sock_from_file(struct file *file)
509 if (file->f_op == &socket_file_ops)
510 return file->private_data; /* set in sock_alloc_file */
514 EXPORT_SYMBOL(sock_from_file);
517 * sockfd_lookup - Go from a file number to its socket slot
519 * @err: pointer to an error code return
521 * The file handle passed in is locked and the socket it is bound
522 * to is returned. If an error occurs the err pointer is overwritten
523 * with a negative errno code and NULL is returned. The function checks
524 * for both invalid handles and passing a handle which is not a socket.
526 * On a success the socket object pointer is returned.
529 struct socket *sockfd_lookup(int fd, int *err)
540 sock = sock_from_file(file);
547 EXPORT_SYMBOL(sockfd_lookup);
549 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
551 struct fd f = fdget(fd);
556 sock = sock_from_file(f.file);
558 *fput_needed = f.flags & FDPUT_FPUT;
567 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
573 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
583 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
588 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
595 static int sockfs_setattr(struct mnt_idmap *idmap,
596 struct dentry *dentry, struct iattr *iattr)
598 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
600 if (!err && (iattr->ia_valid & ATTR_UID)) {
601 struct socket *sock = SOCKET_I(d_inode(dentry));
604 sock->sk->sk_uid = iattr->ia_uid;
612 static const struct inode_operations sockfs_inode_ops = {
613 .listxattr = sockfs_listxattr,
614 .setattr = sockfs_setattr,
618 * sock_alloc - allocate a socket
620 * Allocate a new inode and socket object. The two are bound together
621 * and initialised. The socket is then returned. If we are out of inodes
622 * NULL is returned. This functions uses GFP_KERNEL internally.
625 struct socket *sock_alloc(void)
630 inode = new_inode_pseudo(sock_mnt->mnt_sb);
634 sock = SOCKET_I(inode);
636 inode->i_ino = get_next_ino();
637 inode->i_mode = S_IFSOCK | S_IRWXUGO;
638 inode->i_uid = current_fsuid();
639 inode->i_gid = current_fsgid();
640 inode->i_op = &sockfs_inode_ops;
644 EXPORT_SYMBOL(sock_alloc);
646 static void __sock_release(struct socket *sock, struct inode *inode)
649 struct module *owner = sock->ops->owner;
653 sock->ops->release(sock);
661 if (sock->wq.fasync_list)
662 pr_err("%s: fasync list not empty!\n", __func__);
665 iput(SOCK_INODE(sock));
672 * sock_release - close a socket
673 * @sock: socket to close
675 * The socket is released from the protocol stack if it has a release
676 * callback, and the inode is then released if the socket is bound to
677 * an inode not a file.
679 void sock_release(struct socket *sock)
681 __sock_release(sock, NULL);
683 EXPORT_SYMBOL(sock_release);
685 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
687 u8 flags = *tx_flags;
689 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
690 flags |= SKBTX_HW_TSTAMP;
692 /* PTP hardware clocks can provide a free running cycle counter
693 * as a time base for virtual clocks. Tell driver to use the
694 * free running cycle counter for timestamp if socket is bound
697 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
698 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
701 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
702 flags |= SKBTX_SW_TSTAMP;
704 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
705 flags |= SKBTX_SCHED_TSTAMP;
709 EXPORT_SYMBOL(__sock_tx_timestamp);
711 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
713 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
716 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
719 trace_sock_send_length(sk, ret, 0);
722 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
724 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
725 inet_sendmsg, sock, msg,
727 BUG_ON(ret == -EIOCBQUEUED);
729 if (trace_sock_send_length_enabled())
730 call_trace_sock_send_length(sock->sk, ret, 0);
735 * sock_sendmsg - send a message through @sock
737 * @msg: message to send
739 * Sends @msg through @sock, passing through LSM.
740 * Returns the number of bytes sent, or an error code.
742 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
744 int err = security_socket_sendmsg(sock, msg,
747 return err ?: sock_sendmsg_nosec(sock, msg);
749 EXPORT_SYMBOL(sock_sendmsg);
752 * kernel_sendmsg - send a message through @sock (kernel-space)
754 * @msg: message header
756 * @num: vec array length
757 * @size: total message data size
759 * Builds the message data with @vec and sends it through @sock.
760 * Returns the number of bytes sent, or an error code.
763 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
764 struct kvec *vec, size_t num, size_t size)
766 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
767 return sock_sendmsg(sock, msg);
769 EXPORT_SYMBOL(kernel_sendmsg);
772 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
774 * @msg: message header
775 * @vec: output s/g array
776 * @num: output s/g array length
777 * @size: total message data size
779 * Builds the message data with @vec and sends it through @sock.
780 * Returns the number of bytes sent, or an error code.
781 * Caller must hold @sk.
784 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
785 struct kvec *vec, size_t num, size_t size)
787 struct socket *sock = sk->sk_socket;
789 if (!sock->ops->sendmsg_locked)
790 return sock_no_sendmsg_locked(sk, msg, size);
792 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
794 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
796 EXPORT_SYMBOL(kernel_sendmsg_locked);
798 static bool skb_is_err_queue(const struct sk_buff *skb)
800 /* pkt_type of skbs enqueued on the error queue are set to
801 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
802 * in recvmsg, since skbs received on a local socket will never
803 * have a pkt_type of PACKET_OUTGOING.
805 return skb->pkt_type == PACKET_OUTGOING;
808 /* On transmit, software and hardware timestamps are returned independently.
809 * As the two skb clones share the hardware timestamp, which may be updated
810 * before the software timestamp is received, a hardware TX timestamp may be
811 * returned only if there is no software TX timestamp. Ignore false software
812 * timestamps, which may be made in the __sock_recv_timestamp() call when the
813 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
814 * hardware timestamp.
816 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
818 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
821 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
823 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC;
824 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
825 struct net_device *orig_dev;
829 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
831 *if_index = orig_dev->ifindex;
832 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
834 hwtstamp = shhwtstamps->hwtstamp;
841 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
844 struct scm_ts_pktinfo ts_pktinfo;
845 struct net_device *orig_dev;
847 if (!skb_mac_header_was_set(skb))
850 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
854 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 if_index = orig_dev->ifindex;
859 ts_pktinfo.if_index = if_index;
861 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
862 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
863 sizeof(ts_pktinfo), &ts_pktinfo);
867 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
869 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
872 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
873 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
874 struct scm_timestamping_internal tss;
876 int empty = 1, false_tstamp = 0;
877 struct skb_shared_hwtstamps *shhwtstamps =
882 /* Race occurred between timestamp enabling and packet
883 receiving. Fill in the current time for now. */
884 if (need_software_tstamp && skb->tstamp == 0) {
885 __net_timestamp(skb);
889 if (need_software_tstamp) {
890 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
892 struct __kernel_sock_timeval tv;
894 skb_get_new_timestamp(skb, &tv);
895 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
898 struct __kernel_old_timeval tv;
900 skb_get_timestamp(skb, &tv);
901 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
906 struct __kernel_timespec ts;
908 skb_get_new_timestampns(skb, &ts);
909 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
912 struct __kernel_old_timespec ts;
914 skb_get_timestampns(skb, &ts);
915 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
921 memset(&tss, 0, sizeof(tss));
922 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
923 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
926 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
927 !skb_is_swtx_tstamp(skb, false_tstamp)) {
929 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
930 hwtstamp = get_timestamp(sk, skb, &if_index);
932 hwtstamp = shhwtstamps->hwtstamp;
934 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
935 hwtstamp = ptp_convert_timestamp(&hwtstamp,
938 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
941 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
942 !skb_is_err_queue(skb))
943 put_ts_pktinfo(msg, skb, if_index);
947 if (sock_flag(sk, SOCK_TSTAMP_NEW))
948 put_cmsg_scm_timestamping64(msg, &tss);
950 put_cmsg_scm_timestamping(msg, &tss);
952 if (skb_is_err_queue(skb) && skb->len &&
953 SKB_EXT_ERR(skb)->opt_stats)
954 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
955 skb->len, skb->data);
958 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
960 #ifdef CONFIG_WIRELESS
961 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
966 if (!sock_flag(sk, SOCK_WIFI_STATUS))
968 if (!skb->wifi_acked_valid)
971 ack = skb->wifi_acked;
973 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
975 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
978 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
981 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
982 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
983 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
986 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
989 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
990 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
991 __u32 mark = skb->mark;
993 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
997 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1000 sock_recv_timestamp(msg, sk, skb);
1001 sock_recv_drops(msg, sk, skb);
1002 sock_recv_mark(msg, sk, skb);
1004 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1006 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1008 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1011 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1013 trace_sock_recv_length(sk, ret, flags);
1016 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1019 int ret = INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
1020 inet_recvmsg, sock, msg,
1021 msg_data_left(msg), flags);
1022 if (trace_sock_recv_length_enabled())
1023 call_trace_sock_recv_length(sock->sk, ret, flags);
1028 * sock_recvmsg - receive a message from @sock
1030 * @msg: message to receive
1031 * @flags: message flags
1033 * Receives @msg from @sock, passing through LSM. Returns the total number
1034 * of bytes received, or an error.
1036 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1038 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1040 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1042 EXPORT_SYMBOL(sock_recvmsg);
1045 * kernel_recvmsg - Receive a message from a socket (kernel space)
1046 * @sock: The socket to receive the message from
1047 * @msg: Received message
1048 * @vec: Input s/g array for message data
1049 * @num: Size of input s/g array
1050 * @size: Number of bytes to read
1051 * @flags: Message flags (MSG_DONTWAIT, etc...)
1053 * On return the msg structure contains the scatter/gather array passed in the
1054 * vec argument. The array is modified so that it consists of the unfilled
1055 * portion of the original array.
1057 * The returned value is the total number of bytes received, or an error.
1060 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1061 struct kvec *vec, size_t num, size_t size, int flags)
1063 msg->msg_control_is_user = false;
1064 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1065 return sock_recvmsg(sock, msg, flags);
1067 EXPORT_SYMBOL(kernel_recvmsg);
1069 static ssize_t sock_sendpage(struct file *file, struct page *page,
1070 int offset, size_t size, loff_t *ppos, int more)
1072 struct socket *sock;
1076 sock = file->private_data;
1078 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1079 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1082 ret = kernel_sendpage(sock, page, offset, size, flags);
1084 if (trace_sock_send_length_enabled())
1085 call_trace_sock_send_length(sock->sk, ret, 0);
1089 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1090 struct pipe_inode_info *pipe, size_t len,
1093 struct socket *sock = file->private_data;
1095 if (unlikely(!sock->ops->splice_read))
1096 return generic_file_splice_read(file, ppos, pipe, len, flags);
1098 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1101 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1103 struct file *file = iocb->ki_filp;
1104 struct socket *sock = file->private_data;
1105 struct msghdr msg = {.msg_iter = *to,
1109 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1110 msg.msg_flags = MSG_DONTWAIT;
1112 if (iocb->ki_pos != 0)
1115 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1118 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1123 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1125 struct file *file = iocb->ki_filp;
1126 struct socket *sock = file->private_data;
1127 struct msghdr msg = {.msg_iter = *from,
1131 if (iocb->ki_pos != 0)
1134 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1135 msg.msg_flags = MSG_DONTWAIT;
1137 if (sock->type == SOCK_SEQPACKET)
1138 msg.msg_flags |= MSG_EOR;
1140 res = sock_sendmsg(sock, &msg);
1141 *from = msg.msg_iter;
1146 * Atomic setting of ioctl hooks to avoid race
1147 * with module unload.
1150 static DEFINE_MUTEX(br_ioctl_mutex);
1151 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1152 unsigned int cmd, struct ifreq *ifr,
1155 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1156 unsigned int cmd, struct ifreq *ifr,
1159 mutex_lock(&br_ioctl_mutex);
1160 br_ioctl_hook = hook;
1161 mutex_unlock(&br_ioctl_mutex);
1163 EXPORT_SYMBOL(brioctl_set);
1165 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1166 struct ifreq *ifr, void __user *uarg)
1171 request_module("bridge");
1173 mutex_lock(&br_ioctl_mutex);
1175 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1176 mutex_unlock(&br_ioctl_mutex);
1181 static DEFINE_MUTEX(vlan_ioctl_mutex);
1182 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1184 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1186 mutex_lock(&vlan_ioctl_mutex);
1187 vlan_ioctl_hook = hook;
1188 mutex_unlock(&vlan_ioctl_mutex);
1190 EXPORT_SYMBOL(vlan_ioctl_set);
1192 static long sock_do_ioctl(struct net *net, struct socket *sock,
1193 unsigned int cmd, unsigned long arg)
1198 void __user *argp = (void __user *)arg;
1201 err = sock->ops->ioctl(sock, cmd, arg);
1204 * If this ioctl is unknown try to hand it down
1205 * to the NIC driver.
1207 if (err != -ENOIOCTLCMD)
1210 if (!is_socket_ioctl_cmd(cmd))
1213 if (get_user_ifreq(&ifr, &data, argp))
1215 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1216 if (!err && need_copyout)
1217 if (put_user_ifreq(&ifr, argp))
1224 * With an ioctl, arg may well be a user mode pointer, but we don't know
1225 * what to do with it - that's up to the protocol still.
1228 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1230 struct socket *sock;
1232 void __user *argp = (void __user *)arg;
1236 sock = file->private_data;
1239 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1243 if (get_user_ifreq(&ifr, &data, argp))
1245 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1246 if (!err && need_copyout)
1247 if (put_user_ifreq(&ifr, argp))
1250 #ifdef CONFIG_WEXT_CORE
1251 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1252 err = wext_handle_ioctl(net, cmd, argp);
1259 if (get_user(pid, (int __user *)argp))
1261 err = f_setown(sock->file, pid, 1);
1265 err = put_user(f_getown(sock->file),
1266 (int __user *)argp);
1272 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1277 if (!vlan_ioctl_hook)
1278 request_module("8021q");
1280 mutex_lock(&vlan_ioctl_mutex);
1281 if (vlan_ioctl_hook)
1282 err = vlan_ioctl_hook(net, argp);
1283 mutex_unlock(&vlan_ioctl_mutex);
1287 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1290 err = open_related_ns(&net->ns, get_net_ns);
1292 case SIOCGSTAMP_OLD:
1293 case SIOCGSTAMPNS_OLD:
1294 if (!sock->ops->gettstamp) {
1298 err = sock->ops->gettstamp(sock, argp,
1299 cmd == SIOCGSTAMP_OLD,
1300 !IS_ENABLED(CONFIG_64BIT));
1302 case SIOCGSTAMP_NEW:
1303 case SIOCGSTAMPNS_NEW:
1304 if (!sock->ops->gettstamp) {
1308 err = sock->ops->gettstamp(sock, argp,
1309 cmd == SIOCGSTAMP_NEW,
1314 err = dev_ifconf(net, argp);
1318 err = sock_do_ioctl(net, sock, cmd, arg);
1325 * sock_create_lite - creates a socket
1326 * @family: protocol family (AF_INET, ...)
1327 * @type: communication type (SOCK_STREAM, ...)
1328 * @protocol: protocol (0, ...)
1331 * Creates a new socket and assigns it to @res, passing through LSM.
1332 * The new socket initialization is not complete, see kernel_accept().
1333 * Returns 0 or an error. On failure @res is set to %NULL.
1334 * This function internally uses GFP_KERNEL.
1337 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1340 struct socket *sock = NULL;
1342 err = security_socket_create(family, type, protocol, 1);
1346 sock = sock_alloc();
1353 err = security_socket_post_create(sock, family, type, protocol, 1);
1365 EXPORT_SYMBOL(sock_create_lite);
1367 /* No kernel lock held - perfect */
1368 static __poll_t sock_poll(struct file *file, poll_table *wait)
1370 struct socket *sock = file->private_data;
1371 __poll_t events = poll_requested_events(wait), flag = 0;
1373 if (!sock->ops->poll)
1376 if (sk_can_busy_loop(sock->sk)) {
1377 /* poll once if requested by the syscall */
1378 if (events & POLL_BUSY_LOOP)
1379 sk_busy_loop(sock->sk, 1);
1381 /* if this socket can poll_ll, tell the system call */
1382 flag = POLL_BUSY_LOOP;
1385 return sock->ops->poll(file, sock, wait) | flag;
1388 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1390 struct socket *sock = file->private_data;
1392 return sock->ops->mmap(file, sock, vma);
1395 static int sock_close(struct inode *inode, struct file *filp)
1397 __sock_release(SOCKET_I(inode), inode);
1402 * Update the socket async list
1404 * Fasync_list locking strategy.
1406 * 1. fasync_list is modified only under process context socket lock
1407 * i.e. under semaphore.
1408 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1409 * or under socket lock
1412 static int sock_fasync(int fd, struct file *filp, int on)
1414 struct socket *sock = filp->private_data;
1415 struct sock *sk = sock->sk;
1416 struct socket_wq *wq = &sock->wq;
1422 fasync_helper(fd, filp, on, &wq->fasync_list);
1424 if (!wq->fasync_list)
1425 sock_reset_flag(sk, SOCK_FASYNC);
1427 sock_set_flag(sk, SOCK_FASYNC);
1433 /* This function may be called only under rcu_lock */
1435 int sock_wake_async(struct socket_wq *wq, int how, int band)
1437 if (!wq || !wq->fasync_list)
1441 case SOCK_WAKE_WAITD:
1442 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1445 case SOCK_WAKE_SPACE:
1446 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1451 kill_fasync(&wq->fasync_list, SIGIO, band);
1454 kill_fasync(&wq->fasync_list, SIGURG, band);
1459 EXPORT_SYMBOL(sock_wake_async);
1462 * __sock_create - creates a socket
1463 * @net: net namespace
1464 * @family: protocol family (AF_INET, ...)
1465 * @type: communication type (SOCK_STREAM, ...)
1466 * @protocol: protocol (0, ...)
1468 * @kern: boolean for kernel space sockets
1470 * Creates a new socket and assigns it to @res, passing through LSM.
1471 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1472 * be set to true if the socket resides in kernel space.
1473 * This function internally uses GFP_KERNEL.
1476 int __sock_create(struct net *net, int family, int type, int protocol,
1477 struct socket **res, int kern)
1480 struct socket *sock;
1481 const struct net_proto_family *pf;
1484 * Check protocol is in range
1486 if (family < 0 || family >= NPROTO)
1487 return -EAFNOSUPPORT;
1488 if (type < 0 || type >= SOCK_MAX)
1493 This uglymoron is moved from INET layer to here to avoid
1494 deadlock in module load.
1496 if (family == PF_INET && type == SOCK_PACKET) {
1497 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1502 err = security_socket_create(family, type, protocol, kern);
1507 * Allocate the socket and allow the family to set things up. if
1508 * the protocol is 0, the family is instructed to select an appropriate
1511 sock = sock_alloc();
1513 net_warn_ratelimited("socket: no more sockets\n");
1514 return -ENFILE; /* Not exactly a match, but its the
1515 closest posix thing */
1520 #ifdef CONFIG_MODULES
1521 /* Attempt to load a protocol module if the find failed.
1523 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1524 * requested real, full-featured networking support upon configuration.
1525 * Otherwise module support will break!
1527 if (rcu_access_pointer(net_families[family]) == NULL)
1528 request_module("net-pf-%d", family);
1532 pf = rcu_dereference(net_families[family]);
1533 err = -EAFNOSUPPORT;
1538 * We will call the ->create function, that possibly is in a loadable
1539 * module, so we have to bump that loadable module refcnt first.
1541 if (!try_module_get(pf->owner))
1544 /* Now protected by module ref count */
1547 err = pf->create(net, sock, protocol, kern);
1549 goto out_module_put;
1552 * Now to bump the refcnt of the [loadable] module that owns this
1553 * socket at sock_release time we decrement its refcnt.
1555 if (!try_module_get(sock->ops->owner))
1556 goto out_module_busy;
1559 * Now that we're done with the ->create function, the [loadable]
1560 * module can have its refcnt decremented
1562 module_put(pf->owner);
1563 err = security_socket_post_create(sock, family, type, protocol, kern);
1565 goto out_sock_release;
1571 err = -EAFNOSUPPORT;
1574 module_put(pf->owner);
1581 goto out_sock_release;
1583 EXPORT_SYMBOL(__sock_create);
1586 * sock_create - creates a socket
1587 * @family: protocol family (AF_INET, ...)
1588 * @type: communication type (SOCK_STREAM, ...)
1589 * @protocol: protocol (0, ...)
1592 * A wrapper around __sock_create().
1593 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1596 int sock_create(int family, int type, int protocol, struct socket **res)
1598 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1600 EXPORT_SYMBOL(sock_create);
1603 * sock_create_kern - creates a socket (kernel space)
1604 * @net: net namespace
1605 * @family: protocol family (AF_INET, ...)
1606 * @type: communication type (SOCK_STREAM, ...)
1607 * @protocol: protocol (0, ...)
1610 * A wrapper around __sock_create().
1611 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1614 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1616 return __sock_create(net, family, type, protocol, res, 1);
1618 EXPORT_SYMBOL(sock_create_kern);
1620 static struct socket *__sys_socket_create(int family, int type, int protocol)
1622 struct socket *sock;
1625 /* Check the SOCK_* constants for consistency. */
1626 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1627 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1628 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1629 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1631 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1632 return ERR_PTR(-EINVAL);
1633 type &= SOCK_TYPE_MASK;
1635 retval = sock_create(family, type, protocol, &sock);
1637 return ERR_PTR(retval);
1642 struct file *__sys_socket_file(int family, int type, int protocol)
1644 struct socket *sock;
1647 sock = __sys_socket_create(family, type, protocol);
1649 return ERR_CAST(sock);
1651 flags = type & ~SOCK_TYPE_MASK;
1652 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1653 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1655 return sock_alloc_file(sock, flags, NULL);
1658 int __sys_socket(int family, int type, int protocol)
1660 struct socket *sock;
1663 sock = __sys_socket_create(family, type, protocol);
1665 return PTR_ERR(sock);
1667 flags = type & ~SOCK_TYPE_MASK;
1668 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1669 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1671 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1674 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1676 return __sys_socket(family, type, protocol);
1680 * Create a pair of connected sockets.
1683 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1685 struct socket *sock1, *sock2;
1687 struct file *newfile1, *newfile2;
1690 flags = type & ~SOCK_TYPE_MASK;
1691 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1693 type &= SOCK_TYPE_MASK;
1695 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1696 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1699 * reserve descriptors and make sure we won't fail
1700 * to return them to userland.
1702 fd1 = get_unused_fd_flags(flags);
1703 if (unlikely(fd1 < 0))
1706 fd2 = get_unused_fd_flags(flags);
1707 if (unlikely(fd2 < 0)) {
1712 err = put_user(fd1, &usockvec[0]);
1716 err = put_user(fd2, &usockvec[1]);
1721 * Obtain the first socket and check if the underlying protocol
1722 * supports the socketpair call.
1725 err = sock_create(family, type, protocol, &sock1);
1726 if (unlikely(err < 0))
1729 err = sock_create(family, type, protocol, &sock2);
1730 if (unlikely(err < 0)) {
1731 sock_release(sock1);
1735 err = security_socket_socketpair(sock1, sock2);
1736 if (unlikely(err)) {
1737 sock_release(sock2);
1738 sock_release(sock1);
1742 err = sock1->ops->socketpair(sock1, sock2);
1743 if (unlikely(err < 0)) {
1744 sock_release(sock2);
1745 sock_release(sock1);
1749 newfile1 = sock_alloc_file(sock1, flags, NULL);
1750 if (IS_ERR(newfile1)) {
1751 err = PTR_ERR(newfile1);
1752 sock_release(sock2);
1756 newfile2 = sock_alloc_file(sock2, flags, NULL);
1757 if (IS_ERR(newfile2)) {
1758 err = PTR_ERR(newfile2);
1763 audit_fd_pair(fd1, fd2);
1765 fd_install(fd1, newfile1);
1766 fd_install(fd2, newfile2);
1775 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1776 int __user *, usockvec)
1778 return __sys_socketpair(family, type, protocol, usockvec);
1782 * Bind a name to a socket. Nothing much to do here since it's
1783 * the protocol's responsibility to handle the local address.
1785 * We move the socket address to kernel space before we call
1786 * the protocol layer (having also checked the address is ok).
1789 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1791 struct socket *sock;
1792 struct sockaddr_storage address;
1793 int err, fput_needed;
1795 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1797 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1799 err = security_socket_bind(sock,
1800 (struct sockaddr *)&address,
1803 err = sock->ops->bind(sock,
1807 fput_light(sock->file, fput_needed);
1812 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1814 return __sys_bind(fd, umyaddr, addrlen);
1818 * Perform a listen. Basically, we allow the protocol to do anything
1819 * necessary for a listen, and if that works, we mark the socket as
1820 * ready for listening.
1823 int __sys_listen(int fd, int backlog)
1825 struct socket *sock;
1826 int err, fput_needed;
1829 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1831 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1832 if ((unsigned int)backlog > somaxconn)
1833 backlog = somaxconn;
1835 err = security_socket_listen(sock, backlog);
1837 err = sock->ops->listen(sock, backlog);
1839 fput_light(sock->file, fput_needed);
1844 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1846 return __sys_listen(fd, backlog);
1849 struct file *do_accept(struct file *file, unsigned file_flags,
1850 struct sockaddr __user *upeer_sockaddr,
1851 int __user *upeer_addrlen, int flags)
1853 struct socket *sock, *newsock;
1854 struct file *newfile;
1856 struct sockaddr_storage address;
1858 sock = sock_from_file(file);
1860 return ERR_PTR(-ENOTSOCK);
1862 newsock = sock_alloc();
1864 return ERR_PTR(-ENFILE);
1866 newsock->type = sock->type;
1867 newsock->ops = sock->ops;
1870 * We don't need try_module_get here, as the listening socket (sock)
1871 * has the protocol module (sock->ops->owner) held.
1873 __module_get(newsock->ops->owner);
1875 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1876 if (IS_ERR(newfile))
1879 err = security_socket_accept(sock, newsock);
1883 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1888 if (upeer_sockaddr) {
1889 len = newsock->ops->getname(newsock,
1890 (struct sockaddr *)&address, 2);
1892 err = -ECONNABORTED;
1895 err = move_addr_to_user(&address,
1896 len, upeer_sockaddr, upeer_addrlen);
1901 /* File flags are not inherited via accept() unlike another OSes. */
1905 return ERR_PTR(err);
1908 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1909 int __user *upeer_addrlen, int flags)
1911 struct file *newfile;
1914 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1917 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1918 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1920 newfd = get_unused_fd_flags(flags);
1921 if (unlikely(newfd < 0))
1924 newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen,
1926 if (IS_ERR(newfile)) {
1927 put_unused_fd(newfd);
1928 return PTR_ERR(newfile);
1930 fd_install(newfd, newfile);
1935 * For accept, we attempt to create a new socket, set up the link
1936 * with the client, wake up the client, then return the new
1937 * connected fd. We collect the address of the connector in kernel
1938 * space and move it to user at the very end. This is unclean because
1939 * we open the socket then return an error.
1941 * 1003.1g adds the ability to recvmsg() to query connection pending
1942 * status to recvmsg. We need to add that support in a way thats
1943 * clean when we restructure accept also.
1946 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1947 int __user *upeer_addrlen, int flags)
1954 ret = __sys_accept4_file(f.file, upeer_sockaddr,
1955 upeer_addrlen, flags);
1962 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1963 int __user *, upeer_addrlen, int, flags)
1965 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1968 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1969 int __user *, upeer_addrlen)
1971 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1975 * Attempt to connect to a socket with the server address. The address
1976 * is in user space so we verify it is OK and move it to kernel space.
1978 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1981 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1982 * other SEQPACKET protocols that take time to connect() as it doesn't
1983 * include the -EINPROGRESS status for such sockets.
1986 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1987 int addrlen, int file_flags)
1989 struct socket *sock;
1992 sock = sock_from_file(file);
1999 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2003 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
2004 sock->file->f_flags | file_flags);
2009 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2016 struct sockaddr_storage address;
2018 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2020 ret = __sys_connect_file(f.file, &address, addrlen, 0);
2027 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2030 return __sys_connect(fd, uservaddr, addrlen);
2034 * Get the local address ('name') of a socket object. Move the obtained
2035 * name to user space.
2038 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2039 int __user *usockaddr_len)
2041 struct socket *sock;
2042 struct sockaddr_storage address;
2043 int err, fput_needed;
2045 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2049 err = security_socket_getsockname(sock);
2053 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
2056 /* "err" is actually length in this case */
2057 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2060 fput_light(sock->file, fput_needed);
2065 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2066 int __user *, usockaddr_len)
2068 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2072 * Get the remote address ('name') of a socket object. Move the obtained
2073 * name to user space.
2076 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2077 int __user *usockaddr_len)
2079 struct socket *sock;
2080 struct sockaddr_storage address;
2081 int err, fput_needed;
2083 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2085 err = security_socket_getpeername(sock);
2087 fput_light(sock->file, fput_needed);
2091 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
2093 /* "err" is actually length in this case */
2094 err = move_addr_to_user(&address, err, usockaddr,
2096 fput_light(sock->file, fput_needed);
2101 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2102 int __user *, usockaddr_len)
2104 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2108 * Send a datagram to a given address. We move the address into kernel
2109 * space and check the user space data area is readable before invoking
2112 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2113 struct sockaddr __user *addr, int addr_len)
2115 struct socket *sock;
2116 struct sockaddr_storage address;
2122 err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter);
2125 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2129 msg.msg_name = NULL;
2130 msg.msg_control = NULL;
2131 msg.msg_controllen = 0;
2132 msg.msg_namelen = 0;
2133 msg.msg_ubuf = NULL;
2135 err = move_addr_to_kernel(addr, addr_len, &address);
2138 msg.msg_name = (struct sockaddr *)&address;
2139 msg.msg_namelen = addr_len;
2141 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2142 if (sock->file->f_flags & O_NONBLOCK)
2143 flags |= MSG_DONTWAIT;
2144 msg.msg_flags = flags;
2145 err = sock_sendmsg(sock, &msg);
2148 fput_light(sock->file, fput_needed);
2153 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2154 unsigned int, flags, struct sockaddr __user *, addr,
2157 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2161 * Send a datagram down a socket.
2164 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2165 unsigned int, flags)
2167 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2171 * Receive a frame from the socket and optionally record the address of the
2172 * sender. We verify the buffers are writable and if needed move the
2173 * sender address from kernel to user space.
2175 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2176 struct sockaddr __user *addr, int __user *addr_len)
2178 struct sockaddr_storage address;
2179 struct msghdr msg = {
2180 /* Save some cycles and don't copy the address if not needed */
2181 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2183 struct socket *sock;
2188 err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter);
2191 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2195 if (sock->file->f_flags & O_NONBLOCK)
2196 flags |= MSG_DONTWAIT;
2197 err = sock_recvmsg(sock, &msg, flags);
2199 if (err >= 0 && addr != NULL) {
2200 err2 = move_addr_to_user(&address,
2201 msg.msg_namelen, addr, addr_len);
2206 fput_light(sock->file, fput_needed);
2211 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2212 unsigned int, flags, struct sockaddr __user *, addr,
2213 int __user *, addr_len)
2215 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2219 * Receive a datagram from a socket.
2222 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2223 unsigned int, flags)
2225 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2228 static bool sock_use_custom_sol_socket(const struct socket *sock)
2230 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2234 * Set a socket option. Because we don't know the option lengths we have
2235 * to pass the user mode parameter for the protocols to sort out.
2237 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2240 sockptr_t optval = USER_SOCKPTR(user_optval);
2241 char *kernel_optval = NULL;
2242 int err, fput_needed;
2243 struct socket *sock;
2248 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2252 err = security_socket_setsockopt(sock, level, optname);
2256 if (!in_compat_syscall())
2257 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2258 user_optval, &optlen,
2268 optval = KERNEL_SOCKPTR(kernel_optval);
2269 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2270 err = sock_setsockopt(sock, level, optname, optval, optlen);
2271 else if (unlikely(!sock->ops->setsockopt))
2274 err = sock->ops->setsockopt(sock, level, optname, optval,
2276 kfree(kernel_optval);
2278 fput_light(sock->file, fput_needed);
2282 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2283 char __user *, optval, int, optlen)
2285 return __sys_setsockopt(fd, level, optname, optval, optlen);
2288 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2292 * Get a socket option. Because we don't know the option lengths we have
2293 * to pass a user mode parameter for the protocols to sort out.
2295 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2298 int max_optlen __maybe_unused;
2299 int err, fput_needed;
2300 struct socket *sock;
2302 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2306 err = security_socket_getsockopt(sock, level, optname);
2310 if (!in_compat_syscall())
2311 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2313 if (level == SOL_SOCKET)
2314 err = sock_getsockopt(sock, level, optname, optval, optlen);
2315 else if (unlikely(!sock->ops->getsockopt))
2318 err = sock->ops->getsockopt(sock, level, optname, optval,
2321 if (!in_compat_syscall())
2322 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2323 optval, optlen, max_optlen,
2326 fput_light(sock->file, fput_needed);
2330 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2331 char __user *, optval, int __user *, optlen)
2333 return __sys_getsockopt(fd, level, optname, optval, optlen);
2337 * Shutdown a socket.
2340 int __sys_shutdown_sock(struct socket *sock, int how)
2344 err = security_socket_shutdown(sock, how);
2346 err = sock->ops->shutdown(sock, how);
2351 int __sys_shutdown(int fd, int how)
2353 int err, fput_needed;
2354 struct socket *sock;
2356 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2358 err = __sys_shutdown_sock(sock, how);
2359 fput_light(sock->file, fput_needed);
2364 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2366 return __sys_shutdown(fd, how);
2369 /* A couple of helpful macros for getting the address of the 32/64 bit
2370 * fields which are the same type (int / unsigned) on our platforms.
2372 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2373 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2374 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2376 struct used_address {
2377 struct sockaddr_storage name;
2378 unsigned int name_len;
2381 int __copy_msghdr(struct msghdr *kmsg,
2382 struct user_msghdr *msg,
2383 struct sockaddr __user **save_addr)
2387 kmsg->msg_control_is_user = true;
2388 kmsg->msg_get_inq = 0;
2389 kmsg->msg_control_user = msg->msg_control;
2390 kmsg->msg_controllen = msg->msg_controllen;
2391 kmsg->msg_flags = msg->msg_flags;
2393 kmsg->msg_namelen = msg->msg_namelen;
2395 kmsg->msg_namelen = 0;
2397 if (kmsg->msg_namelen < 0)
2400 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2401 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2404 *save_addr = msg->msg_name;
2406 if (msg->msg_name && kmsg->msg_namelen) {
2408 err = move_addr_to_kernel(msg->msg_name,
2415 kmsg->msg_name = NULL;
2416 kmsg->msg_namelen = 0;
2419 if (msg->msg_iovlen > UIO_MAXIOV)
2422 kmsg->msg_iocb = NULL;
2423 kmsg->msg_ubuf = NULL;
2427 static int copy_msghdr_from_user(struct msghdr *kmsg,
2428 struct user_msghdr __user *umsg,
2429 struct sockaddr __user **save_addr,
2432 struct user_msghdr msg;
2435 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2438 err = __copy_msghdr(kmsg, &msg, save_addr);
2442 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2443 msg.msg_iov, msg.msg_iovlen,
2444 UIO_FASTIOV, iov, &kmsg->msg_iter);
2445 return err < 0 ? err : 0;
2448 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2449 unsigned int flags, struct used_address *used_address,
2450 unsigned int allowed_msghdr_flags)
2452 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2453 __aligned(sizeof(__kernel_size_t));
2454 /* 20 is size of ipv6_pktinfo */
2455 unsigned char *ctl_buf = ctl;
2461 if (msg_sys->msg_controllen > INT_MAX)
2463 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2464 ctl_len = msg_sys->msg_controllen;
2465 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2467 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2471 ctl_buf = msg_sys->msg_control;
2472 ctl_len = msg_sys->msg_controllen;
2473 } else if (ctl_len) {
2474 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2475 CMSG_ALIGN(sizeof(struct cmsghdr)));
2476 if (ctl_len > sizeof(ctl)) {
2477 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2478 if (ctl_buf == NULL)
2482 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2484 msg_sys->msg_control = ctl_buf;
2485 msg_sys->msg_control_is_user = false;
2487 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2488 msg_sys->msg_flags = flags;
2490 if (sock->file->f_flags & O_NONBLOCK)
2491 msg_sys->msg_flags |= MSG_DONTWAIT;
2493 * If this is sendmmsg() and current destination address is same as
2494 * previously succeeded address, omit asking LSM's decision.
2495 * used_address->name_len is initialized to UINT_MAX so that the first
2496 * destination address never matches.
2498 if (used_address && msg_sys->msg_name &&
2499 used_address->name_len == msg_sys->msg_namelen &&
2500 !memcmp(&used_address->name, msg_sys->msg_name,
2501 used_address->name_len)) {
2502 err = sock_sendmsg_nosec(sock, msg_sys);
2505 err = sock_sendmsg(sock, msg_sys);
2507 * If this is sendmmsg() and sending to current destination address was
2508 * successful, remember it.
2510 if (used_address && err >= 0) {
2511 used_address->name_len = msg_sys->msg_namelen;
2512 if (msg_sys->msg_name)
2513 memcpy(&used_address->name, msg_sys->msg_name,
2514 used_address->name_len);
2519 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2524 int sendmsg_copy_msghdr(struct msghdr *msg,
2525 struct user_msghdr __user *umsg, unsigned flags,
2530 if (flags & MSG_CMSG_COMPAT) {
2531 struct compat_msghdr __user *msg_compat;
2533 msg_compat = (struct compat_msghdr __user *) umsg;
2534 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2536 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2544 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2545 struct msghdr *msg_sys, unsigned int flags,
2546 struct used_address *used_address,
2547 unsigned int allowed_msghdr_flags)
2549 struct sockaddr_storage address;
2550 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2553 msg_sys->msg_name = &address;
2555 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2559 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2560 allowed_msghdr_flags);
2566 * BSD sendmsg interface
2568 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2571 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2574 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2575 bool forbid_cmsg_compat)
2577 int fput_needed, err;
2578 struct msghdr msg_sys;
2579 struct socket *sock;
2581 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2584 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2588 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2590 fput_light(sock->file, fput_needed);
2595 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2597 return __sys_sendmsg(fd, msg, flags, true);
2601 * Linux sendmmsg interface
2604 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2605 unsigned int flags, bool forbid_cmsg_compat)
2607 int fput_needed, err, datagrams;
2608 struct socket *sock;
2609 struct mmsghdr __user *entry;
2610 struct compat_mmsghdr __user *compat_entry;
2611 struct msghdr msg_sys;
2612 struct used_address used_address;
2613 unsigned int oflags = flags;
2615 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2618 if (vlen > UIO_MAXIOV)
2623 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2627 used_address.name_len = UINT_MAX;
2629 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2633 while (datagrams < vlen) {
2634 if (datagrams == vlen - 1)
2637 if (MSG_CMSG_COMPAT & flags) {
2638 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2639 &msg_sys, flags, &used_address, MSG_EOR);
2642 err = __put_user(err, &compat_entry->msg_len);
2645 err = ___sys_sendmsg(sock,
2646 (struct user_msghdr __user *)entry,
2647 &msg_sys, flags, &used_address, MSG_EOR);
2650 err = put_user(err, &entry->msg_len);
2657 if (msg_data_left(&msg_sys))
2662 fput_light(sock->file, fput_needed);
2664 /* We only return an error if no datagrams were able to be sent */
2671 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2672 unsigned int, vlen, unsigned int, flags)
2674 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2677 int recvmsg_copy_msghdr(struct msghdr *msg,
2678 struct user_msghdr __user *umsg, unsigned flags,
2679 struct sockaddr __user **uaddr,
2684 if (MSG_CMSG_COMPAT & flags) {
2685 struct compat_msghdr __user *msg_compat;
2687 msg_compat = (struct compat_msghdr __user *) umsg;
2688 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2690 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2698 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2699 struct user_msghdr __user *msg,
2700 struct sockaddr __user *uaddr,
2701 unsigned int flags, int nosec)
2703 struct compat_msghdr __user *msg_compat =
2704 (struct compat_msghdr __user *) msg;
2705 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2706 struct sockaddr_storage addr;
2707 unsigned long cmsg_ptr;
2711 msg_sys->msg_name = &addr;
2712 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2713 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2715 /* We assume all kernel code knows the size of sockaddr_storage */
2716 msg_sys->msg_namelen = 0;
2718 if (sock->file->f_flags & O_NONBLOCK)
2719 flags |= MSG_DONTWAIT;
2721 if (unlikely(nosec))
2722 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2724 err = sock_recvmsg(sock, msg_sys, flags);
2730 if (uaddr != NULL) {
2731 err = move_addr_to_user(&addr,
2732 msg_sys->msg_namelen, uaddr,
2737 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2741 if (MSG_CMSG_COMPAT & flags)
2742 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2743 &msg_compat->msg_controllen);
2745 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2746 &msg->msg_controllen);
2754 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2755 struct msghdr *msg_sys, unsigned int flags, int nosec)
2757 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2758 /* user mode address pointers */
2759 struct sockaddr __user *uaddr;
2762 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2766 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2772 * BSD recvmsg interface
2775 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2776 struct user_msghdr __user *umsg,
2777 struct sockaddr __user *uaddr, unsigned int flags)
2779 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2782 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2783 bool forbid_cmsg_compat)
2785 int fput_needed, err;
2786 struct msghdr msg_sys;
2787 struct socket *sock;
2789 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2792 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2796 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2798 fput_light(sock->file, fput_needed);
2803 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2804 unsigned int, flags)
2806 return __sys_recvmsg(fd, msg, flags, true);
2810 * Linux recvmmsg interface
2813 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2814 unsigned int vlen, unsigned int flags,
2815 struct timespec64 *timeout)
2817 int fput_needed, err, datagrams;
2818 struct socket *sock;
2819 struct mmsghdr __user *entry;
2820 struct compat_mmsghdr __user *compat_entry;
2821 struct msghdr msg_sys;
2822 struct timespec64 end_time;
2823 struct timespec64 timeout64;
2826 poll_select_set_timeout(&end_time, timeout->tv_sec,
2832 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2836 if (likely(!(flags & MSG_ERRQUEUE))) {
2837 err = sock_error(sock->sk);
2845 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2847 while (datagrams < vlen) {
2849 * No need to ask LSM for more than the first datagram.
2851 if (MSG_CMSG_COMPAT & flags) {
2852 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2853 &msg_sys, flags & ~MSG_WAITFORONE,
2857 err = __put_user(err, &compat_entry->msg_len);
2860 err = ___sys_recvmsg(sock,
2861 (struct user_msghdr __user *)entry,
2862 &msg_sys, flags & ~MSG_WAITFORONE,
2866 err = put_user(err, &entry->msg_len);
2874 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2875 if (flags & MSG_WAITFORONE)
2876 flags |= MSG_DONTWAIT;
2879 ktime_get_ts64(&timeout64);
2880 *timeout = timespec64_sub(end_time, timeout64);
2881 if (timeout->tv_sec < 0) {
2882 timeout->tv_sec = timeout->tv_nsec = 0;
2886 /* Timeout, return less than vlen datagrams */
2887 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2891 /* Out of band data, return right away */
2892 if (msg_sys.msg_flags & MSG_OOB)
2900 if (datagrams == 0) {
2906 * We may return less entries than requested (vlen) if the
2907 * sock is non block and there aren't enough datagrams...
2909 if (err != -EAGAIN) {
2911 * ... or if recvmsg returns an error after we
2912 * received some datagrams, where we record the
2913 * error to return on the next call or if the
2914 * app asks about it using getsockopt(SO_ERROR).
2916 WRITE_ONCE(sock->sk->sk_err, -err);
2919 fput_light(sock->file, fput_needed);
2924 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2925 unsigned int vlen, unsigned int flags,
2926 struct __kernel_timespec __user *timeout,
2927 struct old_timespec32 __user *timeout32)
2930 struct timespec64 timeout_sys;
2932 if (timeout && get_timespec64(&timeout_sys, timeout))
2935 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2938 if (!timeout && !timeout32)
2939 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2941 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2946 if (timeout && put_timespec64(&timeout_sys, timeout))
2947 datagrams = -EFAULT;
2949 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2950 datagrams = -EFAULT;
2955 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2956 unsigned int, vlen, unsigned int, flags,
2957 struct __kernel_timespec __user *, timeout)
2959 if (flags & MSG_CMSG_COMPAT)
2962 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2965 #ifdef CONFIG_COMPAT_32BIT_TIME
2966 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2967 unsigned int, vlen, unsigned int, flags,
2968 struct old_timespec32 __user *, timeout)
2970 if (flags & MSG_CMSG_COMPAT)
2973 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2977 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2978 /* Argument list sizes for sys_socketcall */
2979 #define AL(x) ((x) * sizeof(unsigned long))
2980 static const unsigned char nargs[21] = {
2981 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2982 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2983 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2990 * System call vectors.
2992 * Argument checking cleaned up. Saved 20% in size.
2993 * This function doesn't need to set the kernel lock because
2994 * it is set by the callees.
2997 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2999 unsigned long a[AUDITSC_ARGS];
3000 unsigned long a0, a1;
3004 if (call < 1 || call > SYS_SENDMMSG)
3006 call = array_index_nospec(call, SYS_SENDMMSG + 1);
3009 if (len > sizeof(a))
3012 /* copy_from_user should be SMP safe. */
3013 if (copy_from_user(a, args, len))
3016 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3025 err = __sys_socket(a0, a1, a[2]);
3028 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3031 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3034 err = __sys_listen(a0, a1);
3037 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3038 (int __user *)a[2], 0);
3040 case SYS_GETSOCKNAME:
3042 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3043 (int __user *)a[2]);
3045 case SYS_GETPEERNAME:
3047 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3048 (int __user *)a[2]);
3050 case SYS_SOCKETPAIR:
3051 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3054 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3058 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3059 (struct sockaddr __user *)a[4], a[5]);
3062 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3066 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3067 (struct sockaddr __user *)a[4],
3068 (int __user *)a[5]);
3071 err = __sys_shutdown(a0, a1);
3073 case SYS_SETSOCKOPT:
3074 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3077 case SYS_GETSOCKOPT:
3079 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3080 (int __user *)a[4]);
3083 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3087 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3091 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3095 if (IS_ENABLED(CONFIG_64BIT))
3096 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3098 (struct __kernel_timespec __user *)a[4],
3101 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3103 (struct old_timespec32 __user *)a[4]);
3106 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3107 (int __user *)a[2], a[3]);
3116 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3119 * sock_register - add a socket protocol handler
3120 * @ops: description of protocol
3122 * This function is called by a protocol handler that wants to
3123 * advertise its address family, and have it linked into the
3124 * socket interface. The value ops->family corresponds to the
3125 * socket system call protocol family.
3127 int sock_register(const struct net_proto_family *ops)
3131 if (ops->family >= NPROTO) {
3132 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3136 spin_lock(&net_family_lock);
3137 if (rcu_dereference_protected(net_families[ops->family],
3138 lockdep_is_held(&net_family_lock)))
3141 rcu_assign_pointer(net_families[ops->family], ops);
3144 spin_unlock(&net_family_lock);
3146 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3149 EXPORT_SYMBOL(sock_register);
3152 * sock_unregister - remove a protocol handler
3153 * @family: protocol family to remove
3155 * This function is called by a protocol handler that wants to
3156 * remove its address family, and have it unlinked from the
3157 * new socket creation.
3159 * If protocol handler is a module, then it can use module reference
3160 * counts to protect against new references. If protocol handler is not
3161 * a module then it needs to provide its own protection in
3162 * the ops->create routine.
3164 void sock_unregister(int family)
3166 BUG_ON(family < 0 || family >= NPROTO);
3168 spin_lock(&net_family_lock);
3169 RCU_INIT_POINTER(net_families[family], NULL);
3170 spin_unlock(&net_family_lock);
3174 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3176 EXPORT_SYMBOL(sock_unregister);
3178 bool sock_is_registered(int family)
3180 return family < NPROTO && rcu_access_pointer(net_families[family]);
3183 static int __init sock_init(void)
3187 * Initialize the network sysctl infrastructure.
3189 err = net_sysctl_init();
3194 * Initialize skbuff SLAB cache
3199 * Initialize the protocols module.
3204 err = register_filesystem(&sock_fs_type);
3207 sock_mnt = kern_mount(&sock_fs_type);
3208 if (IS_ERR(sock_mnt)) {
3209 err = PTR_ERR(sock_mnt);
3213 /* The real protocol initialization is performed in later initcalls.
3216 #ifdef CONFIG_NETFILTER
3217 err = netfilter_init();
3222 ptp_classifier_init();
3228 unregister_filesystem(&sock_fs_type);
3232 core_initcall(sock_init); /* early initcall */
3234 #ifdef CONFIG_PROC_FS
3235 void socket_seq_show(struct seq_file *seq)
3237 seq_printf(seq, "sockets: used %d\n",
3238 sock_inuse_get(seq->private));
3240 #endif /* CONFIG_PROC_FS */
3242 /* Handle the fact that while struct ifreq has the same *layout* on
3243 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3244 * which are handled elsewhere, it still has different *size* due to
3245 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3246 * resulting in struct ifreq being 32 and 40 bytes respectively).
3247 * As a result, if the struct happens to be at the end of a page and
3248 * the next page isn't readable/writable, we get a fault. To prevent
3249 * that, copy back and forth to the full size.
3251 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3253 if (in_compat_syscall()) {
3254 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3256 memset(ifr, 0, sizeof(*ifr));
3257 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3261 *ifrdata = compat_ptr(ifr32->ifr_data);
3266 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3270 *ifrdata = ifr->ifr_data;
3274 EXPORT_SYMBOL(get_user_ifreq);
3276 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3278 size_t size = sizeof(*ifr);
3280 if (in_compat_syscall())
3281 size = sizeof(struct compat_ifreq);
3283 if (copy_to_user(arg, ifr, size))
3288 EXPORT_SYMBOL(put_user_ifreq);
3290 #ifdef CONFIG_COMPAT
3291 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3293 compat_uptr_t uptr32;
3298 if (get_user_ifreq(&ifr, NULL, uifr32))
3301 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3304 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3305 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3307 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3309 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3310 if (put_user_ifreq(&ifr, uifr32))
3316 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3317 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3318 struct compat_ifreq __user *u_ifreq32)
3323 if (!is_socket_ioctl_cmd(cmd))
3325 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3327 ifreq.ifr_data = data;
3329 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3332 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3333 unsigned int cmd, unsigned long arg)
3335 void __user *argp = compat_ptr(arg);
3336 struct sock *sk = sock->sk;
3337 struct net *net = sock_net(sk);
3339 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3340 return sock_ioctl(file, cmd, (unsigned long)argp);
3344 return compat_siocwandev(net, argp);
3345 case SIOCGSTAMP_OLD:
3346 case SIOCGSTAMPNS_OLD:
3347 if (!sock->ops->gettstamp)
3348 return -ENOIOCTLCMD;
3349 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3350 !COMPAT_USE_64BIT_TIME);
3353 case SIOCBONDSLAVEINFOQUERY:
3354 case SIOCBONDINFOQUERY:
3357 return compat_ifr_data_ioctl(net, cmd, argp);
3368 case SIOCGSTAMP_NEW:
3369 case SIOCGSTAMPNS_NEW:
3373 return sock_ioctl(file, cmd, arg);
3392 case SIOCSIFHWBROADCAST:
3394 case SIOCGIFBRDADDR:
3395 case SIOCSIFBRDADDR:
3396 case SIOCGIFDSTADDR:
3397 case SIOCSIFDSTADDR:
3398 case SIOCGIFNETMASK:
3399 case SIOCSIFNETMASK:
3411 case SIOCBONDENSLAVE:
3412 case SIOCBONDRELEASE:
3413 case SIOCBONDSETHWADDR:
3414 case SIOCBONDCHANGEACTIVE:
3421 return sock_do_ioctl(net, sock, cmd, arg);
3424 return -ENOIOCTLCMD;
3427 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3430 struct socket *sock = file->private_data;
3431 int ret = -ENOIOCTLCMD;
3438 if (sock->ops->compat_ioctl)
3439 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3441 if (ret == -ENOIOCTLCMD &&
3442 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3443 ret = compat_wext_handle_ioctl(net, cmd, arg);
3445 if (ret == -ENOIOCTLCMD)
3446 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3453 * kernel_bind - bind an address to a socket (kernel space)
3456 * @addrlen: length of address
3458 * Returns 0 or an error.
3461 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3463 return sock->ops->bind(sock, addr, addrlen);
3465 EXPORT_SYMBOL(kernel_bind);
3468 * kernel_listen - move socket to listening state (kernel space)
3470 * @backlog: pending connections queue size
3472 * Returns 0 or an error.
3475 int kernel_listen(struct socket *sock, int backlog)
3477 return sock->ops->listen(sock, backlog);
3479 EXPORT_SYMBOL(kernel_listen);
3482 * kernel_accept - accept a connection (kernel space)
3483 * @sock: listening socket
3484 * @newsock: new connected socket
3487 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3488 * If it fails, @newsock is guaranteed to be %NULL.
3489 * Returns 0 or an error.
3492 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3494 struct sock *sk = sock->sk;
3497 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3502 err = sock->ops->accept(sock, *newsock, flags, true);
3504 sock_release(*newsock);
3509 (*newsock)->ops = sock->ops;
3510 __module_get((*newsock)->ops->owner);
3515 EXPORT_SYMBOL(kernel_accept);
3518 * kernel_connect - connect a socket (kernel space)
3521 * @addrlen: address length
3522 * @flags: flags (O_NONBLOCK, ...)
3524 * For datagram sockets, @addr is the address to which datagrams are sent
3525 * by default, and the only address from which datagrams are received.
3526 * For stream sockets, attempts to connect to @addr.
3527 * Returns 0 or an error code.
3530 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3533 return sock->ops->connect(sock, addr, addrlen, flags);
3535 EXPORT_SYMBOL(kernel_connect);
3538 * kernel_getsockname - get the address which the socket is bound (kernel space)
3540 * @addr: address holder
3542 * Fills the @addr pointer with the address which the socket is bound.
3543 * Returns the length of the address in bytes or an error code.
3546 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3548 return sock->ops->getname(sock, addr, 0);
3550 EXPORT_SYMBOL(kernel_getsockname);
3553 * kernel_getpeername - get the address which the socket is connected (kernel space)
3555 * @addr: address holder
3557 * Fills the @addr pointer with the address which the socket is connected.
3558 * Returns the length of the address in bytes or an error code.
3561 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3563 return sock->ops->getname(sock, addr, 1);
3565 EXPORT_SYMBOL(kernel_getpeername);
3568 * kernel_sendpage - send a &page through a socket (kernel space)
3571 * @offset: page offset
3572 * @size: total size in bytes
3573 * @flags: flags (MSG_DONTWAIT, ...)
3575 * Returns the total amount sent in bytes or an error.
3578 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3579 size_t size, int flags)
3581 if (sock->ops->sendpage) {
3582 /* Warn in case the improper page to zero-copy send */
3583 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3584 return sock->ops->sendpage(sock, page, offset, size, flags);
3586 return sock_no_sendpage(sock, page, offset, size, flags);
3588 EXPORT_SYMBOL(kernel_sendpage);
3591 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3594 * @offset: page offset
3595 * @size: total size in bytes
3596 * @flags: flags (MSG_DONTWAIT, ...)
3598 * Returns the total amount sent in bytes or an error.
3599 * Caller must hold @sk.
3602 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3603 size_t size, int flags)
3605 struct socket *sock = sk->sk_socket;
3607 if (sock->ops->sendpage_locked)
3608 return sock->ops->sendpage_locked(sk, page, offset, size,
3611 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3613 EXPORT_SYMBOL(kernel_sendpage_locked);
3616 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3618 * @how: connection part
3620 * Returns 0 or an error.
3623 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3625 return sock->ops->shutdown(sock, how);
3627 EXPORT_SYMBOL(kernel_sock_shutdown);
3630 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3633 * This routine returns the IP overhead imposed by a socket i.e.
3634 * the length of the underlying IP header, depending on whether
3635 * this is an IPv4 or IPv6 socket and the length from IP options turned
3636 * on at the socket. Assumes that the caller has a lock on the socket.
3639 u32 kernel_sock_ip_overhead(struct sock *sk)
3641 struct inet_sock *inet;
3642 struct ip_options_rcu *opt;
3644 #if IS_ENABLED(CONFIG_IPV6)
3645 struct ipv6_pinfo *np;
3646 struct ipv6_txoptions *optv6 = NULL;
3647 #endif /* IS_ENABLED(CONFIG_IPV6) */
3652 switch (sk->sk_family) {
3655 overhead += sizeof(struct iphdr);
3656 opt = rcu_dereference_protected(inet->inet_opt,
3657 sock_owned_by_user(sk));
3659 overhead += opt->opt.optlen;
3661 #if IS_ENABLED(CONFIG_IPV6)
3664 overhead += sizeof(struct ipv6hdr);
3666 optv6 = rcu_dereference_protected(np->opt,
3667 sock_owned_by_user(sk));
3669 overhead += (optv6->opt_flen + optv6->opt_nflen);
3671 #endif /* IS_ENABLED(CONFIG_IPV6) */
3672 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3676 EXPORT_SYMBOL(kernel_sock_ip_overhead);