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
56 #include <linux/socket.h>
57 #include <linux/file.h>
58 #include <linux/net.h>
59 #include <linux/interrupt.h>
60 #include <linux/thread_info.h>
61 #include <linux/rcupdate.h>
62 #include <linux/netdevice.h>
63 #include <linux/proc_fs.h>
64 #include <linux/seq_file.h>
65 #include <linux/mutex.h>
66 #include <linux/if_bridge.h>
67 #include <linux/if_frad.h>
68 #include <linux/if_vlan.h>
69 #include <linux/ptp_classify.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74 #include <linux/highmem.h>
75 #include <linux/mount.h>
76 #include <linux/pseudo_fs.h>
77 #include <linux/security.h>
78 #include <linux/syscalls.h>
79 #include <linux/compat.h>
80 #include <linux/kmod.h>
81 #include <linux/audit.h>
82 #include <linux/wireless.h>
83 #include <linux/nsproxy.h>
84 #include <linux/magic.h>
85 #include <linux/slab.h>
86 #include <linux/xattr.h>
87 #include <linux/nospec.h>
88 #include <linux/indirect_call_wrapper.h>
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
95 #include <net/cls_cgroup.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/termios.h>
104 #include <linux/sockios.h>
105 #include <net/busy_poll.h>
106 #include <linux/errqueue.h>
108 #ifdef CONFIG_NET_RX_BUSY_POLL
109 unsigned int sysctl_net_busy_read __read_mostly;
110 unsigned int sysctl_net_busy_poll __read_mostly;
113 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
115 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
117 static int sock_close(struct inode *inode, struct file *file);
118 static __poll_t sock_poll(struct file *file,
119 struct poll_table_struct *wait);
120 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
122 static long compat_sock_ioctl(struct file *file,
123 unsigned int cmd, unsigned long arg);
125 static int sock_fasync(int fd, struct file *filp, int on);
126 static ssize_t sock_sendpage(struct file *file, struct page *page,
127 int offset, size_t size, loff_t *ppos, int more);
128 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
129 struct pipe_inode_info *pipe, size_t len,
132 #ifdef CONFIG_PROC_FS
133 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135 struct socket *sock = f->private_data;
137 if (sock->ops->show_fdinfo)
138 sock->ops->show_fdinfo(m, sock);
141 #define sock_show_fdinfo NULL
145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 * in the operation structures but are done directly via the socketcall() multiplexor.
149 static const struct file_operations socket_file_ops = {
150 .owner = THIS_MODULE,
152 .read_iter = sock_read_iter,
153 .write_iter = sock_write_iter,
155 .unlocked_ioctl = sock_ioctl,
157 .compat_ioctl = compat_sock_ioctl,
160 .release = sock_close,
161 .fasync = sock_fasync,
162 .sendpage = sock_sendpage,
163 .splice_write = generic_splice_sendpage,
164 .splice_read = sock_splice_read,
165 .show_fdinfo = sock_show_fdinfo,
169 * The protocol list. Each protocol is registered in here.
172 static DEFINE_SPINLOCK(net_family_lock);
173 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
198 if (copy_from_user(kaddr, uaddr, ulen))
200 return audit_sockaddr(ulen, kaddr);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221 void __user *uaddr, int __user *ulen)
226 BUG_ON(klen > sizeof(struct sockaddr_storage));
227 err = get_user(len, ulen);
235 if (audit_sockaddr(klen, kaddr))
237 if (copy_to_user(uaddr, kaddr, len))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen, ulen);
247 static struct kmem_cache *sock_inode_cachep __ro_after_init;
249 static struct inode *sock_alloc_inode(struct super_block *sb)
251 struct socket_alloc *ei;
253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
256 init_waitqueue_head(&ei->socket.wq.wait);
257 ei->socket.wq.fasync_list = NULL;
258 ei->socket.wq.flags = 0;
260 ei->socket.state = SS_UNCONNECTED;
261 ei->socket.flags = 0;
262 ei->socket.ops = NULL;
263 ei->socket.sk = NULL;
264 ei->socket.file = NULL;
266 return &ei->vfs_inode;
269 static void sock_free_inode(struct inode *inode)
271 struct socket_alloc *ei;
273 ei = container_of(inode, struct socket_alloc, vfs_inode);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static void init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
291 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
293 BUG_ON(sock_inode_cachep == NULL);
296 static const struct super_operations sockfs_ops = {
297 .alloc_inode = sock_alloc_inode,
298 .free_inode = sock_free_inode,
299 .statfs = simple_statfs,
303 * sockfs_dname() is called from d_path().
305 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
307 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
308 d_inode(dentry)->i_ino);
311 static const struct dentry_operations sockfs_dentry_operations = {
312 .d_dname = sockfs_dname,
315 static int sockfs_xattr_get(const struct xattr_handler *handler,
316 struct dentry *dentry, struct inode *inode,
317 const char *suffix, void *value, size_t size)
320 if (dentry->d_name.len + 1 > size)
322 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
324 return dentry->d_name.len + 1;
327 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
331 static const struct xattr_handler sockfs_xattr_handler = {
332 .name = XATTR_NAME_SOCKPROTONAME,
333 .get = sockfs_xattr_get,
336 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337 struct dentry *dentry, struct inode *inode,
338 const char *suffix, const void *value,
339 size_t size, int flags)
341 /* Handled by LSM. */
345 static const struct xattr_handler sockfs_security_xattr_handler = {
346 .prefix = XATTR_SECURITY_PREFIX,
347 .set = sockfs_security_xattr_set,
350 static const struct xattr_handler *sockfs_xattr_handlers[] = {
351 &sockfs_xattr_handler,
352 &sockfs_security_xattr_handler,
356 static int sockfs_init_fs_context(struct fs_context *fc)
358 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
361 ctx->ops = &sockfs_ops;
362 ctx->dops = &sockfs_dentry_operations;
363 ctx->xattr = sockfs_xattr_handlers;
367 static struct vfsmount *sock_mnt __read_mostly;
369 static struct file_system_type sock_fs_type = {
371 .init_fs_context = sockfs_init_fs_context,
372 .kill_sb = kill_anon_super,
376 * Obtains the first available file descriptor and sets it up for use.
378 * These functions create file structures and maps them to fd space
379 * of the current process. On success it returns file descriptor
380 * and file struct implicitly stored in sock->file.
381 * Note that another thread may close file descriptor before we return
382 * from this function. We use the fact that now we do not refer
383 * to socket after mapping. If one day we will need it, this
384 * function will increment ref. count on file by 1.
386 * In any case returned fd MAY BE not valid!
387 * This race condition is unavoidable
388 * with shared fd spaces, we cannot solve it inside kernel,
389 * but we take care of internal coherence yet.
393 * sock_alloc_file - Bind a &socket to a &file
395 * @flags: file status flags
396 * @dname: protocol name
398 * Returns the &file bound with @sock, implicitly storing it
399 * in sock->file. If dname is %NULL, sets to "".
400 * On failure the return is a ERR pointer (see linux/err.h).
401 * This function uses GFP_KERNEL internally.
404 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
409 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
411 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
412 O_RDWR | (flags & O_NONBLOCK),
420 file->private_data = sock;
421 stream_open(SOCK_INODE(sock), file);
424 EXPORT_SYMBOL(sock_alloc_file);
426 static int sock_map_fd(struct socket *sock, int flags)
428 struct file *newfile;
429 int fd = get_unused_fd_flags(flags);
430 if (unlikely(fd < 0)) {
435 newfile = sock_alloc_file(sock, flags, NULL);
436 if (!IS_ERR(newfile)) {
437 fd_install(fd, newfile);
442 return PTR_ERR(newfile);
446 * sock_from_file - Return the &socket bounded to @file.
448 * @err: pointer to an error code return
450 * On failure returns %NULL and assigns -ENOTSOCK to @err.
453 struct socket *sock_from_file(struct file *file, int *err)
455 if (file->f_op == &socket_file_ops)
456 return file->private_data; /* set in sock_map_fd */
461 EXPORT_SYMBOL(sock_from_file);
464 * sockfd_lookup - Go from a file number to its socket slot
466 * @err: pointer to an error code return
468 * The file handle passed in is locked and the socket it is bound
469 * to is returned. If an error occurs the err pointer is overwritten
470 * with a negative errno code and NULL is returned. The function checks
471 * for both invalid handles and passing a handle which is not a socket.
473 * On a success the socket object pointer is returned.
476 struct socket *sockfd_lookup(int fd, int *err)
487 sock = sock_from_file(file, err);
492 EXPORT_SYMBOL(sockfd_lookup);
494 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
496 struct fd f = fdget(fd);
501 sock = sock_from_file(f.file, err);
503 *fput_needed = f.flags;
511 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
517 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
527 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
532 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
539 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
541 int err = simple_setattr(dentry, iattr);
543 if (!err && (iattr->ia_valid & ATTR_UID)) {
544 struct socket *sock = SOCKET_I(d_inode(dentry));
547 sock->sk->sk_uid = iattr->ia_uid;
555 static const struct inode_operations sockfs_inode_ops = {
556 .listxattr = sockfs_listxattr,
557 .setattr = sockfs_setattr,
561 * sock_alloc - allocate a socket
563 * Allocate a new inode and socket object. The two are bound together
564 * and initialised. The socket is then returned. If we are out of inodes
565 * NULL is returned. This functions uses GFP_KERNEL internally.
568 struct socket *sock_alloc(void)
573 inode = new_inode_pseudo(sock_mnt->mnt_sb);
577 sock = SOCKET_I(inode);
579 inode->i_ino = get_next_ino();
580 inode->i_mode = S_IFSOCK | S_IRWXUGO;
581 inode->i_uid = current_fsuid();
582 inode->i_gid = current_fsgid();
583 inode->i_op = &sockfs_inode_ops;
587 EXPORT_SYMBOL(sock_alloc);
590 * sock_release - close a socket
591 * @sock: socket to close
593 * The socket is released from the protocol stack if it has a release
594 * callback, and the inode is then released if the socket is bound to
595 * an inode not a file.
598 static void __sock_release(struct socket *sock, struct inode *inode)
601 struct module *owner = sock->ops->owner;
605 sock->ops->release(sock);
613 if (sock->wq.fasync_list)
614 pr_err("%s: fasync list not empty!\n", __func__);
617 iput(SOCK_INODE(sock));
623 void sock_release(struct socket *sock)
625 __sock_release(sock, NULL);
627 EXPORT_SYMBOL(sock_release);
629 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
631 u8 flags = *tx_flags;
633 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
634 flags |= SKBTX_HW_TSTAMP;
636 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
637 flags |= SKBTX_SW_TSTAMP;
639 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
640 flags |= SKBTX_SCHED_TSTAMP;
644 EXPORT_SYMBOL(__sock_tx_timestamp);
646 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
648 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
650 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
652 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
653 inet_sendmsg, sock, msg,
655 BUG_ON(ret == -EIOCBQUEUED);
660 * sock_sendmsg - send a message through @sock
662 * @msg: message to send
664 * Sends @msg through @sock, passing through LSM.
665 * Returns the number of bytes sent, or an error code.
667 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
669 int err = security_socket_sendmsg(sock, msg,
672 return err ?: sock_sendmsg_nosec(sock, msg);
674 EXPORT_SYMBOL(sock_sendmsg);
677 * kernel_sendmsg - send a message through @sock (kernel-space)
679 * @msg: message header
681 * @num: vec array length
682 * @size: total message data size
684 * Builds the message data with @vec and sends it through @sock.
685 * Returns the number of bytes sent, or an error code.
688 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
689 struct kvec *vec, size_t num, size_t size)
691 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
692 return sock_sendmsg(sock, msg);
694 EXPORT_SYMBOL(kernel_sendmsg);
697 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
699 * @msg: message header
700 * @vec: output s/g array
701 * @num: output s/g array length
702 * @size: total message data size
704 * Builds the message data with @vec and sends it through @sock.
705 * Returns the number of bytes sent, or an error code.
706 * Caller must hold @sk.
709 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
710 struct kvec *vec, size_t num, size_t size)
712 struct socket *sock = sk->sk_socket;
714 if (!sock->ops->sendmsg_locked)
715 return sock_no_sendmsg_locked(sk, msg, size);
717 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
719 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
721 EXPORT_SYMBOL(kernel_sendmsg_locked);
723 static bool skb_is_err_queue(const struct sk_buff *skb)
725 /* pkt_type of skbs enqueued on the error queue are set to
726 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
727 * in recvmsg, since skbs received on a local socket will never
728 * have a pkt_type of PACKET_OUTGOING.
730 return skb->pkt_type == PACKET_OUTGOING;
733 /* On transmit, software and hardware timestamps are returned independently.
734 * As the two skb clones share the hardware timestamp, which may be updated
735 * before the software timestamp is received, a hardware TX timestamp may be
736 * returned only if there is no software TX timestamp. Ignore false software
737 * timestamps, which may be made in the __sock_recv_timestamp() call when the
738 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
739 * hardware timestamp.
741 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
743 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
746 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
748 struct scm_ts_pktinfo ts_pktinfo;
749 struct net_device *orig_dev;
751 if (!skb_mac_header_was_set(skb))
754 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
757 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
759 ts_pktinfo.if_index = orig_dev->ifindex;
762 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
763 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
764 sizeof(ts_pktinfo), &ts_pktinfo);
768 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
770 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
773 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
774 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
775 struct scm_timestamping_internal tss;
777 int empty = 1, false_tstamp = 0;
778 struct skb_shared_hwtstamps *shhwtstamps =
781 /* Race occurred between timestamp enabling and packet
782 receiving. Fill in the current time for now. */
783 if (need_software_tstamp && skb->tstamp == 0) {
784 __net_timestamp(skb);
788 if (need_software_tstamp) {
789 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
791 struct __kernel_sock_timeval tv;
793 skb_get_new_timestamp(skb, &tv);
794 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
797 struct __kernel_old_timeval tv;
799 skb_get_timestamp(skb, &tv);
800 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
805 struct __kernel_timespec ts;
807 skb_get_new_timestampns(skb, &ts);
808 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
811 struct __kernel_old_timespec ts;
813 skb_get_timestampns(skb, &ts);
814 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
820 memset(&tss, 0, sizeof(tss));
821 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
822 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
825 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
826 !skb_is_swtx_tstamp(skb, false_tstamp) &&
827 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
829 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
830 !skb_is_err_queue(skb))
831 put_ts_pktinfo(msg, skb);
834 if (sock_flag(sk, SOCK_TSTAMP_NEW))
835 put_cmsg_scm_timestamping64(msg, &tss);
837 put_cmsg_scm_timestamping(msg, &tss);
839 if (skb_is_err_queue(skb) && skb->len &&
840 SKB_EXT_ERR(skb)->opt_stats)
841 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
842 skb->len, skb->data);
845 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
847 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
852 if (!sock_flag(sk, SOCK_WIFI_STATUS))
854 if (!skb->wifi_acked_valid)
857 ack = skb->wifi_acked;
859 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
861 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
863 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
866 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
867 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
868 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
871 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
874 sock_recv_timestamp(msg, sk, skb);
875 sock_recv_drops(msg, sk, skb);
877 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
879 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
881 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
883 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
886 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
887 inet_recvmsg, sock, msg, msg_data_left(msg),
892 * sock_recvmsg - receive a message from @sock
894 * @msg: message to receive
895 * @flags: message flags
897 * Receives @msg from @sock, passing through LSM. Returns the total number
898 * of bytes received, or an error.
900 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
902 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
904 return err ?: sock_recvmsg_nosec(sock, msg, flags);
906 EXPORT_SYMBOL(sock_recvmsg);
909 * kernel_recvmsg - Receive a message from a socket (kernel space)
910 * @sock: The socket to receive the message from
911 * @msg: Received message
912 * @vec: Input s/g array for message data
913 * @num: Size of input s/g array
914 * @size: Number of bytes to read
915 * @flags: Message flags (MSG_DONTWAIT, etc...)
917 * On return the msg structure contains the scatter/gather array passed in the
918 * vec argument. The array is modified so that it consists of the unfilled
919 * portion of the original array.
921 * The returned value is the total number of bytes received, or an error.
924 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
925 struct kvec *vec, size_t num, size_t size, int flags)
927 msg->msg_control_is_user = false;
928 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
929 return sock_recvmsg(sock, msg, flags);
931 EXPORT_SYMBOL(kernel_recvmsg);
933 static ssize_t sock_sendpage(struct file *file, struct page *page,
934 int offset, size_t size, loff_t *ppos, int more)
939 sock = file->private_data;
941 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
942 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
945 return kernel_sendpage(sock, page, offset, size, flags);
948 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
949 struct pipe_inode_info *pipe, size_t len,
952 struct socket *sock = file->private_data;
954 if (unlikely(!sock->ops->splice_read))
955 return generic_file_splice_read(file, ppos, pipe, len, flags);
957 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
960 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
962 struct file *file = iocb->ki_filp;
963 struct socket *sock = file->private_data;
964 struct msghdr msg = {.msg_iter = *to,
968 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
969 msg.msg_flags = MSG_DONTWAIT;
971 if (iocb->ki_pos != 0)
974 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
977 res = sock_recvmsg(sock, &msg, msg.msg_flags);
982 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
984 struct file *file = iocb->ki_filp;
985 struct socket *sock = file->private_data;
986 struct msghdr msg = {.msg_iter = *from,
990 if (iocb->ki_pos != 0)
993 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
994 msg.msg_flags = MSG_DONTWAIT;
996 if (sock->type == SOCK_SEQPACKET)
997 msg.msg_flags |= MSG_EOR;
999 res = sock_sendmsg(sock, &msg);
1000 *from = msg.msg_iter;
1005 * Atomic setting of ioctl hooks to avoid race
1006 * with module unload.
1009 static DEFINE_MUTEX(br_ioctl_mutex);
1010 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1012 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1014 mutex_lock(&br_ioctl_mutex);
1015 br_ioctl_hook = hook;
1016 mutex_unlock(&br_ioctl_mutex);
1018 EXPORT_SYMBOL(brioctl_set);
1020 static DEFINE_MUTEX(vlan_ioctl_mutex);
1021 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1023 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1025 mutex_lock(&vlan_ioctl_mutex);
1026 vlan_ioctl_hook = hook;
1027 mutex_unlock(&vlan_ioctl_mutex);
1029 EXPORT_SYMBOL(vlan_ioctl_set);
1031 static DEFINE_MUTEX(dlci_ioctl_mutex);
1032 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1034 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1036 mutex_lock(&dlci_ioctl_mutex);
1037 dlci_ioctl_hook = hook;
1038 mutex_unlock(&dlci_ioctl_mutex);
1040 EXPORT_SYMBOL(dlci_ioctl_set);
1042 static long sock_do_ioctl(struct net *net, struct socket *sock,
1043 unsigned int cmd, unsigned long arg)
1046 void __user *argp = (void __user *)arg;
1048 err = sock->ops->ioctl(sock, cmd, arg);
1051 * If this ioctl is unknown try to hand it down
1052 * to the NIC driver.
1054 if (err != -ENOIOCTLCMD)
1057 if (cmd == SIOCGIFCONF) {
1059 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1062 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1064 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1069 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1071 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1072 if (!err && need_copyout)
1073 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1080 * With an ioctl, arg may well be a user mode pointer, but we don't know
1081 * what to do with it - that's up to the protocol still.
1085 * get_net_ns - increment the refcount of the network namespace
1086 * @ns: common namespace (net)
1088 * Returns the net's common namespace.
1091 struct ns_common *get_net_ns(struct ns_common *ns)
1093 return &get_net(container_of(ns, struct net, ns))->ns;
1095 EXPORT_SYMBOL_GPL(get_net_ns);
1097 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1099 struct socket *sock;
1101 void __user *argp = (void __user *)arg;
1105 sock = file->private_data;
1108 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1111 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1113 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1114 if (!err && need_copyout)
1115 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1118 #ifdef CONFIG_WEXT_CORE
1119 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1120 err = wext_handle_ioctl(net, cmd, argp);
1127 if (get_user(pid, (int __user *)argp))
1129 err = f_setown(sock->file, pid, 1);
1133 err = put_user(f_getown(sock->file),
1134 (int __user *)argp);
1142 request_module("bridge");
1144 mutex_lock(&br_ioctl_mutex);
1146 err = br_ioctl_hook(net, cmd, argp);
1147 mutex_unlock(&br_ioctl_mutex);
1152 if (!vlan_ioctl_hook)
1153 request_module("8021q");
1155 mutex_lock(&vlan_ioctl_mutex);
1156 if (vlan_ioctl_hook)
1157 err = vlan_ioctl_hook(net, argp);
1158 mutex_unlock(&vlan_ioctl_mutex);
1163 if (!dlci_ioctl_hook)
1164 request_module("dlci");
1166 mutex_lock(&dlci_ioctl_mutex);
1167 if (dlci_ioctl_hook)
1168 err = dlci_ioctl_hook(cmd, argp);
1169 mutex_unlock(&dlci_ioctl_mutex);
1173 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1176 err = open_related_ns(&net->ns, get_net_ns);
1178 case SIOCGSTAMP_OLD:
1179 case SIOCGSTAMPNS_OLD:
1180 if (!sock->ops->gettstamp) {
1184 err = sock->ops->gettstamp(sock, argp,
1185 cmd == SIOCGSTAMP_OLD,
1186 !IS_ENABLED(CONFIG_64BIT));
1188 case SIOCGSTAMP_NEW:
1189 case SIOCGSTAMPNS_NEW:
1190 if (!sock->ops->gettstamp) {
1194 err = sock->ops->gettstamp(sock, argp,
1195 cmd == SIOCGSTAMP_NEW,
1199 err = sock_do_ioctl(net, sock, cmd, arg);
1206 * sock_create_lite - creates a socket
1207 * @family: protocol family (AF_INET, ...)
1208 * @type: communication type (SOCK_STREAM, ...)
1209 * @protocol: protocol (0, ...)
1212 * Creates a new socket and assigns it to @res, passing through LSM.
1213 * The new socket initialization is not complete, see kernel_accept().
1214 * Returns 0 or an error. On failure @res is set to %NULL.
1215 * This function internally uses GFP_KERNEL.
1218 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1221 struct socket *sock = NULL;
1223 err = security_socket_create(family, type, protocol, 1);
1227 sock = sock_alloc();
1234 err = security_socket_post_create(sock, family, type, protocol, 1);
1246 EXPORT_SYMBOL(sock_create_lite);
1248 /* No kernel lock held - perfect */
1249 static __poll_t sock_poll(struct file *file, poll_table *wait)
1251 struct socket *sock = file->private_data;
1252 __poll_t events = poll_requested_events(wait), flag = 0;
1254 if (!sock->ops->poll)
1257 if (sk_can_busy_loop(sock->sk)) {
1258 /* poll once if requested by the syscall */
1259 if (events & POLL_BUSY_LOOP)
1260 sk_busy_loop(sock->sk, 1);
1262 /* if this socket can poll_ll, tell the system call */
1263 flag = POLL_BUSY_LOOP;
1266 return sock->ops->poll(file, sock, wait) | flag;
1269 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1271 struct socket *sock = file->private_data;
1273 return sock->ops->mmap(file, sock, vma);
1276 static int sock_close(struct inode *inode, struct file *filp)
1278 __sock_release(SOCKET_I(inode), inode);
1283 * Update the socket async list
1285 * Fasync_list locking strategy.
1287 * 1. fasync_list is modified only under process context socket lock
1288 * i.e. under semaphore.
1289 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1290 * or under socket lock
1293 static int sock_fasync(int fd, struct file *filp, int on)
1295 struct socket *sock = filp->private_data;
1296 struct sock *sk = sock->sk;
1297 struct socket_wq *wq = &sock->wq;
1303 fasync_helper(fd, filp, on, &wq->fasync_list);
1305 if (!wq->fasync_list)
1306 sock_reset_flag(sk, SOCK_FASYNC);
1308 sock_set_flag(sk, SOCK_FASYNC);
1314 /* This function may be called only under rcu_lock */
1316 int sock_wake_async(struct socket_wq *wq, int how, int band)
1318 if (!wq || !wq->fasync_list)
1322 case SOCK_WAKE_WAITD:
1323 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1326 case SOCK_WAKE_SPACE:
1327 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1332 kill_fasync(&wq->fasync_list, SIGIO, band);
1335 kill_fasync(&wq->fasync_list, SIGURG, band);
1340 EXPORT_SYMBOL(sock_wake_async);
1343 * __sock_create - creates a socket
1344 * @net: net namespace
1345 * @family: protocol family (AF_INET, ...)
1346 * @type: communication type (SOCK_STREAM, ...)
1347 * @protocol: protocol (0, ...)
1349 * @kern: boolean for kernel space sockets
1351 * Creates a new socket and assigns it to @res, passing through LSM.
1352 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1353 * be set to true if the socket resides in kernel space.
1354 * This function internally uses GFP_KERNEL.
1357 int __sock_create(struct net *net, int family, int type, int protocol,
1358 struct socket **res, int kern)
1361 struct socket *sock;
1362 const struct net_proto_family *pf;
1365 * Check protocol is in range
1367 if (family < 0 || family >= NPROTO)
1368 return -EAFNOSUPPORT;
1369 if (type < 0 || type >= SOCK_MAX)
1374 This uglymoron is moved from INET layer to here to avoid
1375 deadlock in module load.
1377 if (family == PF_INET && type == SOCK_PACKET) {
1378 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1383 err = security_socket_create(family, type, protocol, kern);
1388 * Allocate the socket and allow the family to set things up. if
1389 * the protocol is 0, the family is instructed to select an appropriate
1392 sock = sock_alloc();
1394 net_warn_ratelimited("socket: no more sockets\n");
1395 return -ENFILE; /* Not exactly a match, but its the
1396 closest posix thing */
1401 #ifdef CONFIG_MODULES
1402 /* Attempt to load a protocol module if the find failed.
1404 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1405 * requested real, full-featured networking support upon configuration.
1406 * Otherwise module support will break!
1408 if (rcu_access_pointer(net_families[family]) == NULL)
1409 request_module("net-pf-%d", family);
1413 pf = rcu_dereference(net_families[family]);
1414 err = -EAFNOSUPPORT;
1419 * We will call the ->create function, that possibly is in a loadable
1420 * module, so we have to bump that loadable module refcnt first.
1422 if (!try_module_get(pf->owner))
1425 /* Now protected by module ref count */
1428 err = pf->create(net, sock, protocol, kern);
1430 goto out_module_put;
1433 * Now to bump the refcnt of the [loadable] module that owns this
1434 * socket at sock_release time we decrement its refcnt.
1436 if (!try_module_get(sock->ops->owner))
1437 goto out_module_busy;
1440 * Now that we're done with the ->create function, the [loadable]
1441 * module can have its refcnt decremented
1443 module_put(pf->owner);
1444 err = security_socket_post_create(sock, family, type, protocol, kern);
1446 goto out_sock_release;
1452 err = -EAFNOSUPPORT;
1455 module_put(pf->owner);
1462 goto out_sock_release;
1464 EXPORT_SYMBOL(__sock_create);
1467 * sock_create - creates a socket
1468 * @family: protocol family (AF_INET, ...)
1469 * @type: communication type (SOCK_STREAM, ...)
1470 * @protocol: protocol (0, ...)
1473 * A wrapper around __sock_create().
1474 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1477 int sock_create(int family, int type, int protocol, struct socket **res)
1479 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1481 EXPORT_SYMBOL(sock_create);
1484 * sock_create_kern - creates a socket (kernel space)
1485 * @net: net namespace
1486 * @family: protocol family (AF_INET, ...)
1487 * @type: communication type (SOCK_STREAM, ...)
1488 * @protocol: protocol (0, ...)
1491 * A wrapper around __sock_create().
1492 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1495 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1497 return __sock_create(net, family, type, protocol, res, 1);
1499 EXPORT_SYMBOL(sock_create_kern);
1501 int __sys_socket(int family, int type, int protocol)
1504 struct socket *sock;
1507 /* Check the SOCK_* constants for consistency. */
1508 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1509 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1510 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1511 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1513 flags = type & ~SOCK_TYPE_MASK;
1514 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1516 type &= SOCK_TYPE_MASK;
1518 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1519 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1521 retval = sock_create(family, type, protocol, &sock);
1525 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1528 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1530 return __sys_socket(family, type, protocol);
1534 * Create a pair of connected sockets.
1537 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1539 struct socket *sock1, *sock2;
1541 struct file *newfile1, *newfile2;
1544 flags = type & ~SOCK_TYPE_MASK;
1545 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1547 type &= SOCK_TYPE_MASK;
1549 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1550 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1553 * reserve descriptors and make sure we won't fail
1554 * to return them to userland.
1556 fd1 = get_unused_fd_flags(flags);
1557 if (unlikely(fd1 < 0))
1560 fd2 = get_unused_fd_flags(flags);
1561 if (unlikely(fd2 < 0)) {
1566 err = put_user(fd1, &usockvec[0]);
1570 err = put_user(fd2, &usockvec[1]);
1575 * Obtain the first socket and check if the underlying protocol
1576 * supports the socketpair call.
1579 err = sock_create(family, type, protocol, &sock1);
1580 if (unlikely(err < 0))
1583 err = sock_create(family, type, protocol, &sock2);
1584 if (unlikely(err < 0)) {
1585 sock_release(sock1);
1589 err = security_socket_socketpair(sock1, sock2);
1590 if (unlikely(err)) {
1591 sock_release(sock2);
1592 sock_release(sock1);
1596 err = sock1->ops->socketpair(sock1, sock2);
1597 if (unlikely(err < 0)) {
1598 sock_release(sock2);
1599 sock_release(sock1);
1603 newfile1 = sock_alloc_file(sock1, flags, NULL);
1604 if (IS_ERR(newfile1)) {
1605 err = PTR_ERR(newfile1);
1606 sock_release(sock2);
1610 newfile2 = sock_alloc_file(sock2, flags, NULL);
1611 if (IS_ERR(newfile2)) {
1612 err = PTR_ERR(newfile2);
1617 audit_fd_pair(fd1, fd2);
1619 fd_install(fd1, newfile1);
1620 fd_install(fd2, newfile2);
1629 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1630 int __user *, usockvec)
1632 return __sys_socketpair(family, type, protocol, usockvec);
1636 * Bind a name to a socket. Nothing much to do here since it's
1637 * the protocol's responsibility to handle the local address.
1639 * We move the socket address to kernel space before we call
1640 * the protocol layer (having also checked the address is ok).
1643 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1645 struct socket *sock;
1646 struct sockaddr_storage address;
1647 int err, fput_needed;
1649 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1651 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1653 err = security_socket_bind(sock,
1654 (struct sockaddr *)&address,
1657 err = sock->ops->bind(sock,
1661 fput_light(sock->file, fput_needed);
1666 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1668 return __sys_bind(fd, umyaddr, addrlen);
1672 * Perform a listen. Basically, we allow the protocol to do anything
1673 * necessary for a listen, and if that works, we mark the socket as
1674 * ready for listening.
1677 int __sys_listen(int fd, int backlog)
1679 struct socket *sock;
1680 int err, fput_needed;
1683 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1685 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1686 if ((unsigned int)backlog > somaxconn)
1687 backlog = somaxconn;
1689 err = security_socket_listen(sock, backlog);
1691 err = sock->ops->listen(sock, backlog);
1693 fput_light(sock->file, fput_needed);
1698 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1700 return __sys_listen(fd, backlog);
1703 int __sys_accept4_file(struct file *file, unsigned file_flags,
1704 struct sockaddr __user *upeer_sockaddr,
1705 int __user *upeer_addrlen, int flags,
1706 unsigned long nofile)
1708 struct socket *sock, *newsock;
1709 struct file *newfile;
1710 int err, len, newfd;
1711 struct sockaddr_storage address;
1713 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1716 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1717 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1719 sock = sock_from_file(file, &err);
1724 newsock = sock_alloc();
1728 newsock->type = sock->type;
1729 newsock->ops = sock->ops;
1732 * We don't need try_module_get here, as the listening socket (sock)
1733 * has the protocol module (sock->ops->owner) held.
1735 __module_get(newsock->ops->owner);
1737 newfd = __get_unused_fd_flags(flags, nofile);
1738 if (unlikely(newfd < 0)) {
1740 sock_release(newsock);
1743 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1744 if (IS_ERR(newfile)) {
1745 err = PTR_ERR(newfile);
1746 put_unused_fd(newfd);
1750 err = security_socket_accept(sock, newsock);
1754 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1759 if (upeer_sockaddr) {
1760 len = newsock->ops->getname(newsock,
1761 (struct sockaddr *)&address, 2);
1763 err = -ECONNABORTED;
1766 err = move_addr_to_user(&address,
1767 len, upeer_sockaddr, upeer_addrlen);
1772 /* File flags are not inherited via accept() unlike another OSes. */
1774 fd_install(newfd, newfile);
1780 put_unused_fd(newfd);
1786 * For accept, we attempt to create a new socket, set up the link
1787 * with the client, wake up the client, then return the new
1788 * connected fd. We collect the address of the connector in kernel
1789 * space and move it to user at the very end. This is unclean because
1790 * we open the socket then return an error.
1792 * 1003.1g adds the ability to recvmsg() to query connection pending
1793 * status to recvmsg. We need to add that support in a way thats
1794 * clean when we restructure accept also.
1797 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1798 int __user *upeer_addrlen, int flags)
1805 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1806 upeer_addrlen, flags,
1807 rlimit(RLIMIT_NOFILE));
1815 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1816 int __user *, upeer_addrlen, int, flags)
1818 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1821 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1822 int __user *, upeer_addrlen)
1824 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1828 * Attempt to connect to a socket with the server address. The address
1829 * is in user space so we verify it is OK and move it to kernel space.
1831 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1834 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1835 * other SEQPACKET protocols that take time to connect() as it doesn't
1836 * include the -EINPROGRESS status for such sockets.
1839 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1840 int addrlen, int file_flags)
1842 struct socket *sock;
1845 sock = sock_from_file(file, &err);
1850 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1854 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1855 sock->file->f_flags | file_flags);
1860 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1867 struct sockaddr_storage address;
1869 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1871 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1879 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1882 return __sys_connect(fd, uservaddr, addrlen);
1886 * Get the local address ('name') of a socket object. Move the obtained
1887 * name to user space.
1890 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1891 int __user *usockaddr_len)
1893 struct socket *sock;
1894 struct sockaddr_storage address;
1895 int err, fput_needed;
1897 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1901 err = security_socket_getsockname(sock);
1905 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1908 /* "err" is actually length in this case */
1909 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1912 fput_light(sock->file, fput_needed);
1917 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1918 int __user *, usockaddr_len)
1920 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1924 * Get the remote address ('name') of a socket object. Move the obtained
1925 * name to user space.
1928 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1929 int __user *usockaddr_len)
1931 struct socket *sock;
1932 struct sockaddr_storage address;
1933 int err, fput_needed;
1935 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1937 err = security_socket_getpeername(sock);
1939 fput_light(sock->file, fput_needed);
1943 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1945 /* "err" is actually length in this case */
1946 err = move_addr_to_user(&address, err, usockaddr,
1948 fput_light(sock->file, fput_needed);
1953 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1954 int __user *, usockaddr_len)
1956 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1960 * Send a datagram to a given address. We move the address into kernel
1961 * space and check the user space data area is readable before invoking
1964 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1965 struct sockaddr __user *addr, int addr_len)
1967 struct socket *sock;
1968 struct sockaddr_storage address;
1974 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1977 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1981 msg.msg_name = NULL;
1982 msg.msg_control = NULL;
1983 msg.msg_controllen = 0;
1984 msg.msg_namelen = 0;
1986 err = move_addr_to_kernel(addr, addr_len, &address);
1989 msg.msg_name = (struct sockaddr *)&address;
1990 msg.msg_namelen = addr_len;
1992 if (sock->file->f_flags & O_NONBLOCK)
1993 flags |= MSG_DONTWAIT;
1994 msg.msg_flags = flags;
1995 err = sock_sendmsg(sock, &msg);
1998 fput_light(sock->file, fput_needed);
2003 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2004 unsigned int, flags, struct sockaddr __user *, addr,
2007 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2011 * Send a datagram down a socket.
2014 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2015 unsigned int, flags)
2017 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2021 * Receive a frame from the socket and optionally record the address of the
2022 * sender. We verify the buffers are writable and if needed move the
2023 * sender address from kernel to user space.
2025 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2026 struct sockaddr __user *addr, int __user *addr_len)
2028 struct socket *sock;
2031 struct sockaddr_storage address;
2035 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2038 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2042 msg.msg_control = NULL;
2043 msg.msg_controllen = 0;
2044 /* Save some cycles and don't copy the address if not needed */
2045 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2046 /* We assume all kernel code knows the size of sockaddr_storage */
2047 msg.msg_namelen = 0;
2048 msg.msg_iocb = NULL;
2050 if (sock->file->f_flags & O_NONBLOCK)
2051 flags |= MSG_DONTWAIT;
2052 err = sock_recvmsg(sock, &msg, flags);
2054 if (err >= 0 && addr != NULL) {
2055 err2 = move_addr_to_user(&address,
2056 msg.msg_namelen, addr, addr_len);
2061 fput_light(sock->file, fput_needed);
2066 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2067 unsigned int, flags, struct sockaddr __user *, addr,
2068 int __user *, addr_len)
2070 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2074 * Receive a datagram from a socket.
2077 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2078 unsigned int, flags)
2080 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2084 * Set a socket option. Because we don't know the option lengths we have
2085 * to pass the user mode parameter for the protocols to sort out.
2088 static int __sys_setsockopt(int fd, int level, int optname,
2089 char __user *optval, int optlen)
2091 mm_segment_t oldfs = get_fs();
2092 char *kernel_optval = NULL;
2093 int err, fput_needed;
2094 struct socket *sock;
2099 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2101 err = security_socket_setsockopt(sock, level, optname);
2105 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2106 &optname, optval, &optlen,
2111 } else if (err > 0) {
2116 if (kernel_optval) {
2118 optval = (char __user __force *)kernel_optval;
2121 if (level == SOL_SOCKET)
2123 sock_setsockopt(sock, level, optname, optval,
2127 sock->ops->setsockopt(sock, level, optname, optval,
2130 if (kernel_optval) {
2132 kfree(kernel_optval);
2135 fput_light(sock->file, fput_needed);
2140 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2141 char __user *, optval, int, optlen)
2143 return __sys_setsockopt(fd, level, optname, optval, optlen);
2147 * Get a socket option. Because we don't know the option lengths we have
2148 * to pass a user mode parameter for the protocols to sort out.
2151 static int __sys_getsockopt(int fd, int level, int optname,
2152 char __user *optval, int __user *optlen)
2154 int err, fput_needed;
2155 struct socket *sock;
2158 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2160 err = security_socket_getsockopt(sock, level, optname);
2164 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2166 if (level == SOL_SOCKET)
2168 sock_getsockopt(sock, level, optname, optval,
2172 sock->ops->getsockopt(sock, level, optname, optval,
2175 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2179 fput_light(sock->file, fput_needed);
2184 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2185 char __user *, optval, int __user *, optlen)
2187 return __sys_getsockopt(fd, level, optname, optval, optlen);
2191 * Shutdown a socket.
2194 int __sys_shutdown(int fd, int how)
2196 int err, fput_needed;
2197 struct socket *sock;
2199 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2201 err = security_socket_shutdown(sock, how);
2203 err = sock->ops->shutdown(sock, how);
2204 fput_light(sock->file, fput_needed);
2209 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2211 return __sys_shutdown(fd, how);
2214 /* A couple of helpful macros for getting the address of the 32/64 bit
2215 * fields which are the same type (int / unsigned) on our platforms.
2217 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2218 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2219 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2221 struct used_address {
2222 struct sockaddr_storage name;
2223 unsigned int name_len;
2226 int __copy_msghdr_from_user(struct msghdr *kmsg,
2227 struct user_msghdr __user *umsg,
2228 struct sockaddr __user **save_addr,
2229 struct iovec __user **uiov, size_t *nsegs)
2231 struct user_msghdr msg;
2234 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2237 kmsg->msg_control_is_user = true;
2238 kmsg->msg_control_user = msg.msg_control;
2239 kmsg->msg_controllen = msg.msg_controllen;
2240 kmsg->msg_flags = msg.msg_flags;
2242 kmsg->msg_namelen = msg.msg_namelen;
2244 kmsg->msg_namelen = 0;
2246 if (kmsg->msg_namelen < 0)
2249 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2250 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2253 *save_addr = msg.msg_name;
2255 if (msg.msg_name && kmsg->msg_namelen) {
2257 err = move_addr_to_kernel(msg.msg_name,
2264 kmsg->msg_name = NULL;
2265 kmsg->msg_namelen = 0;
2268 if (msg.msg_iovlen > UIO_MAXIOV)
2271 kmsg->msg_iocb = NULL;
2272 *uiov = msg.msg_iov;
2273 *nsegs = msg.msg_iovlen;
2277 static int copy_msghdr_from_user(struct msghdr *kmsg,
2278 struct user_msghdr __user *umsg,
2279 struct sockaddr __user **save_addr,
2282 struct user_msghdr msg;
2285 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2290 err = import_iovec(save_addr ? READ : WRITE,
2291 msg.msg_iov, msg.msg_iovlen,
2292 UIO_FASTIOV, iov, &kmsg->msg_iter);
2293 return err < 0 ? err : 0;
2296 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2297 unsigned int flags, struct used_address *used_address,
2298 unsigned int allowed_msghdr_flags)
2300 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2301 __aligned(sizeof(__kernel_size_t));
2302 /* 20 is size of ipv6_pktinfo */
2303 unsigned char *ctl_buf = ctl;
2309 if (msg_sys->msg_controllen > INT_MAX)
2311 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2312 ctl_len = msg_sys->msg_controllen;
2313 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2315 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2319 ctl_buf = msg_sys->msg_control;
2320 ctl_len = msg_sys->msg_controllen;
2321 } else if (ctl_len) {
2322 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2323 CMSG_ALIGN(sizeof(struct cmsghdr)));
2324 if (ctl_len > sizeof(ctl)) {
2325 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2326 if (ctl_buf == NULL)
2330 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2332 msg_sys->msg_control = ctl_buf;
2333 msg_sys->msg_control_is_user = false;
2335 msg_sys->msg_flags = flags;
2337 if (sock->file->f_flags & O_NONBLOCK)
2338 msg_sys->msg_flags |= MSG_DONTWAIT;
2340 * If this is sendmmsg() and current destination address is same as
2341 * previously succeeded address, omit asking LSM's decision.
2342 * used_address->name_len is initialized to UINT_MAX so that the first
2343 * destination address never matches.
2345 if (used_address && msg_sys->msg_name &&
2346 used_address->name_len == msg_sys->msg_namelen &&
2347 !memcmp(&used_address->name, msg_sys->msg_name,
2348 used_address->name_len)) {
2349 err = sock_sendmsg_nosec(sock, msg_sys);
2352 err = sock_sendmsg(sock, msg_sys);
2354 * If this is sendmmsg() and sending to current destination address was
2355 * successful, remember it.
2357 if (used_address && err >= 0) {
2358 used_address->name_len = msg_sys->msg_namelen;
2359 if (msg_sys->msg_name)
2360 memcpy(&used_address->name, msg_sys->msg_name,
2361 used_address->name_len);
2366 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2371 int sendmsg_copy_msghdr(struct msghdr *msg,
2372 struct user_msghdr __user *umsg, unsigned flags,
2377 if (flags & MSG_CMSG_COMPAT) {
2378 struct compat_msghdr __user *msg_compat;
2380 msg_compat = (struct compat_msghdr __user *) umsg;
2381 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2383 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2391 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2392 struct msghdr *msg_sys, unsigned int flags,
2393 struct used_address *used_address,
2394 unsigned int allowed_msghdr_flags)
2396 struct sockaddr_storage address;
2397 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2400 msg_sys->msg_name = &address;
2402 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2406 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2407 allowed_msghdr_flags);
2413 * BSD sendmsg interface
2415 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2418 /* disallow ancillary data requests from this path */
2419 if (msg->msg_control || msg->msg_controllen)
2422 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2425 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2426 bool forbid_cmsg_compat)
2428 int fput_needed, err;
2429 struct msghdr msg_sys;
2430 struct socket *sock;
2432 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2435 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2439 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2441 fput_light(sock->file, fput_needed);
2446 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2448 return __sys_sendmsg(fd, msg, flags, true);
2452 * Linux sendmmsg interface
2455 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2456 unsigned int flags, bool forbid_cmsg_compat)
2458 int fput_needed, err, datagrams;
2459 struct socket *sock;
2460 struct mmsghdr __user *entry;
2461 struct compat_mmsghdr __user *compat_entry;
2462 struct msghdr msg_sys;
2463 struct used_address used_address;
2464 unsigned int oflags = flags;
2466 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2469 if (vlen > UIO_MAXIOV)
2474 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2478 used_address.name_len = UINT_MAX;
2480 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2484 while (datagrams < vlen) {
2485 if (datagrams == vlen - 1)
2488 if (MSG_CMSG_COMPAT & flags) {
2489 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2490 &msg_sys, flags, &used_address, MSG_EOR);
2493 err = __put_user(err, &compat_entry->msg_len);
2496 err = ___sys_sendmsg(sock,
2497 (struct user_msghdr __user *)entry,
2498 &msg_sys, flags, &used_address, MSG_EOR);
2501 err = put_user(err, &entry->msg_len);
2508 if (msg_data_left(&msg_sys))
2513 fput_light(sock->file, fput_needed);
2515 /* We only return an error if no datagrams were able to be sent */
2522 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2523 unsigned int, vlen, unsigned int, flags)
2525 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2528 int recvmsg_copy_msghdr(struct msghdr *msg,
2529 struct user_msghdr __user *umsg, unsigned flags,
2530 struct sockaddr __user **uaddr,
2535 if (MSG_CMSG_COMPAT & flags) {
2536 struct compat_msghdr __user *msg_compat;
2538 msg_compat = (struct compat_msghdr __user *) umsg;
2539 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2541 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2549 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2550 struct user_msghdr __user *msg,
2551 struct sockaddr __user *uaddr,
2552 unsigned int flags, int nosec)
2554 struct compat_msghdr __user *msg_compat =
2555 (struct compat_msghdr __user *) msg;
2556 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2557 struct sockaddr_storage addr;
2558 unsigned long cmsg_ptr;
2562 msg_sys->msg_name = &addr;
2563 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2564 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2566 /* We assume all kernel code knows the size of sockaddr_storage */
2567 msg_sys->msg_namelen = 0;
2569 if (sock->file->f_flags & O_NONBLOCK)
2570 flags |= MSG_DONTWAIT;
2572 if (unlikely(nosec))
2573 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2575 err = sock_recvmsg(sock, msg_sys, flags);
2581 if (uaddr != NULL) {
2582 err = move_addr_to_user(&addr,
2583 msg_sys->msg_namelen, uaddr,
2588 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2592 if (MSG_CMSG_COMPAT & flags)
2593 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2594 &msg_compat->msg_controllen);
2596 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2597 &msg->msg_controllen);
2605 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2606 struct msghdr *msg_sys, unsigned int flags, int nosec)
2608 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2609 /* user mode address pointers */
2610 struct sockaddr __user *uaddr;
2613 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2617 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2623 * BSD recvmsg interface
2626 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2627 struct user_msghdr __user *umsg,
2628 struct sockaddr __user *uaddr, unsigned int flags)
2630 /* disallow ancillary data requests from this path */
2631 if (msg->msg_control || msg->msg_controllen)
2634 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2637 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2638 bool forbid_cmsg_compat)
2640 int fput_needed, err;
2641 struct msghdr msg_sys;
2642 struct socket *sock;
2644 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2647 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2651 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2653 fput_light(sock->file, fput_needed);
2658 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2659 unsigned int, flags)
2661 return __sys_recvmsg(fd, msg, flags, true);
2665 * Linux recvmmsg interface
2668 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2669 unsigned int vlen, unsigned int flags,
2670 struct timespec64 *timeout)
2672 int fput_needed, err, datagrams;
2673 struct socket *sock;
2674 struct mmsghdr __user *entry;
2675 struct compat_mmsghdr __user *compat_entry;
2676 struct msghdr msg_sys;
2677 struct timespec64 end_time;
2678 struct timespec64 timeout64;
2681 poll_select_set_timeout(&end_time, timeout->tv_sec,
2687 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2691 if (likely(!(flags & MSG_ERRQUEUE))) {
2692 err = sock_error(sock->sk);
2700 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2702 while (datagrams < vlen) {
2704 * No need to ask LSM for more than the first datagram.
2706 if (MSG_CMSG_COMPAT & flags) {
2707 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2708 &msg_sys, flags & ~MSG_WAITFORONE,
2712 err = __put_user(err, &compat_entry->msg_len);
2715 err = ___sys_recvmsg(sock,
2716 (struct user_msghdr __user *)entry,
2717 &msg_sys, flags & ~MSG_WAITFORONE,
2721 err = put_user(err, &entry->msg_len);
2729 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2730 if (flags & MSG_WAITFORONE)
2731 flags |= MSG_DONTWAIT;
2734 ktime_get_ts64(&timeout64);
2735 *timeout = timespec64_sub(end_time, timeout64);
2736 if (timeout->tv_sec < 0) {
2737 timeout->tv_sec = timeout->tv_nsec = 0;
2741 /* Timeout, return less than vlen datagrams */
2742 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2746 /* Out of band data, return right away */
2747 if (msg_sys.msg_flags & MSG_OOB)
2755 if (datagrams == 0) {
2761 * We may return less entries than requested (vlen) if the
2762 * sock is non block and there aren't enough datagrams...
2764 if (err != -EAGAIN) {
2766 * ... or if recvmsg returns an error after we
2767 * received some datagrams, where we record the
2768 * error to return on the next call or if the
2769 * app asks about it using getsockopt(SO_ERROR).
2771 sock->sk->sk_err = -err;
2774 fput_light(sock->file, fput_needed);
2779 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2780 unsigned int vlen, unsigned int flags,
2781 struct __kernel_timespec __user *timeout,
2782 struct old_timespec32 __user *timeout32)
2785 struct timespec64 timeout_sys;
2787 if (timeout && get_timespec64(&timeout_sys, timeout))
2790 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2793 if (!timeout && !timeout32)
2794 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2796 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2801 if (timeout && put_timespec64(&timeout_sys, timeout))
2802 datagrams = -EFAULT;
2804 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2805 datagrams = -EFAULT;
2810 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2811 unsigned int, vlen, unsigned int, flags,
2812 struct __kernel_timespec __user *, timeout)
2814 if (flags & MSG_CMSG_COMPAT)
2817 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2820 #ifdef CONFIG_COMPAT_32BIT_TIME
2821 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2822 unsigned int, vlen, unsigned int, flags,
2823 struct old_timespec32 __user *, timeout)
2825 if (flags & MSG_CMSG_COMPAT)
2828 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2832 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2833 /* Argument list sizes for sys_socketcall */
2834 #define AL(x) ((x) * sizeof(unsigned long))
2835 static const unsigned char nargs[21] = {
2836 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2837 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2838 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2845 * System call vectors.
2847 * Argument checking cleaned up. Saved 20% in size.
2848 * This function doesn't need to set the kernel lock because
2849 * it is set by the callees.
2852 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2854 unsigned long a[AUDITSC_ARGS];
2855 unsigned long a0, a1;
2859 if (call < 1 || call > SYS_SENDMMSG)
2861 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2864 if (len > sizeof(a))
2867 /* copy_from_user should be SMP safe. */
2868 if (copy_from_user(a, args, len))
2871 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2880 err = __sys_socket(a0, a1, a[2]);
2883 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2886 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2889 err = __sys_listen(a0, a1);
2892 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2893 (int __user *)a[2], 0);
2895 case SYS_GETSOCKNAME:
2897 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2898 (int __user *)a[2]);
2900 case SYS_GETPEERNAME:
2902 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2903 (int __user *)a[2]);
2905 case SYS_SOCKETPAIR:
2906 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2909 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2913 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2914 (struct sockaddr __user *)a[4], a[5]);
2917 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2921 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2922 (struct sockaddr __user *)a[4],
2923 (int __user *)a[5]);
2926 err = __sys_shutdown(a0, a1);
2928 case SYS_SETSOCKOPT:
2929 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2932 case SYS_GETSOCKOPT:
2934 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2935 (int __user *)a[4]);
2938 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2942 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2946 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2950 if (IS_ENABLED(CONFIG_64BIT))
2951 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2953 (struct __kernel_timespec __user *)a[4],
2956 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2958 (struct old_timespec32 __user *)a[4]);
2961 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2962 (int __user *)a[2], a[3]);
2971 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2974 * sock_register - add a socket protocol handler
2975 * @ops: description of protocol
2977 * This function is called by a protocol handler that wants to
2978 * advertise its address family, and have it linked into the
2979 * socket interface. The value ops->family corresponds to the
2980 * socket system call protocol family.
2982 int sock_register(const struct net_proto_family *ops)
2986 if (ops->family >= NPROTO) {
2987 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2991 spin_lock(&net_family_lock);
2992 if (rcu_dereference_protected(net_families[ops->family],
2993 lockdep_is_held(&net_family_lock)))
2996 rcu_assign_pointer(net_families[ops->family], ops);
2999 spin_unlock(&net_family_lock);
3001 pr_info("NET: Registered protocol family %d\n", ops->family);
3004 EXPORT_SYMBOL(sock_register);
3007 * sock_unregister - remove a protocol handler
3008 * @family: protocol family to remove
3010 * This function is called by a protocol handler that wants to
3011 * remove its address family, and have it unlinked from the
3012 * new socket creation.
3014 * If protocol handler is a module, then it can use module reference
3015 * counts to protect against new references. If protocol handler is not
3016 * a module then it needs to provide its own protection in
3017 * the ops->create routine.
3019 void sock_unregister(int family)
3021 BUG_ON(family < 0 || family >= NPROTO);
3023 spin_lock(&net_family_lock);
3024 RCU_INIT_POINTER(net_families[family], NULL);
3025 spin_unlock(&net_family_lock);
3029 pr_info("NET: Unregistered protocol family %d\n", family);
3031 EXPORT_SYMBOL(sock_unregister);
3033 bool sock_is_registered(int family)
3035 return family < NPROTO && rcu_access_pointer(net_families[family]);
3038 static int __init sock_init(void)
3042 * Initialize the network sysctl infrastructure.
3044 err = net_sysctl_init();
3049 * Initialize skbuff SLAB cache
3054 * Initialize the protocols module.
3059 err = register_filesystem(&sock_fs_type);
3062 sock_mnt = kern_mount(&sock_fs_type);
3063 if (IS_ERR(sock_mnt)) {
3064 err = PTR_ERR(sock_mnt);
3068 /* The real protocol initialization is performed in later initcalls.
3071 #ifdef CONFIG_NETFILTER
3072 err = netfilter_init();
3077 ptp_classifier_init();
3083 unregister_filesystem(&sock_fs_type);
3088 core_initcall(sock_init); /* early initcall */
3090 #ifdef CONFIG_PROC_FS
3091 void socket_seq_show(struct seq_file *seq)
3093 seq_printf(seq, "sockets: used %d\n",
3094 sock_inuse_get(seq->private));
3096 #endif /* CONFIG_PROC_FS */
3098 #ifdef CONFIG_COMPAT
3099 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3101 struct compat_ifconf ifc32;
3105 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3108 ifc.ifc_len = ifc32.ifc_len;
3109 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3112 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3117 ifc32.ifc_len = ifc.ifc_len;
3118 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3124 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3126 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3127 bool convert_in = false, convert_out = false;
3128 size_t buf_size = 0;
3129 struct ethtool_rxnfc __user *rxnfc = NULL;
3131 u32 rule_cnt = 0, actual_rule_cnt;
3136 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3139 compat_rxnfc = compat_ptr(data);
3141 if (get_user(ethcmd, &compat_rxnfc->cmd))
3144 /* Most ethtool structures are defined without padding.
3145 * Unfortunately struct ethtool_rxnfc is an exception.
3150 case ETHTOOL_GRXCLSRLALL:
3151 /* Buffer size is variable */
3152 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3154 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3156 buf_size += rule_cnt * sizeof(u32);
3158 case ETHTOOL_GRXRINGS:
3159 case ETHTOOL_GRXCLSRLCNT:
3160 case ETHTOOL_GRXCLSRULE:
3161 case ETHTOOL_SRXCLSRLINS:
3164 case ETHTOOL_SRXCLSRLDEL:
3165 buf_size += sizeof(struct ethtool_rxnfc);
3167 rxnfc = compat_alloc_user_space(buf_size);
3171 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3174 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3177 /* We expect there to be holes between fs.m_ext and
3178 * fs.ring_cookie and at the end of fs, but nowhere else.
3180 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3181 sizeof(compat_rxnfc->fs.m_ext) !=
3182 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3183 sizeof(rxnfc->fs.m_ext));
3185 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3186 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3187 offsetof(struct ethtool_rxnfc, fs.location) -
3188 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3190 if (copy_in_user(rxnfc, compat_rxnfc,
3191 (void __user *)(&rxnfc->fs.m_ext + 1) -
3192 (void __user *)rxnfc) ||
3193 copy_in_user(&rxnfc->fs.ring_cookie,
3194 &compat_rxnfc->fs.ring_cookie,
3195 (void __user *)(&rxnfc->fs.location + 1) -
3196 (void __user *)&rxnfc->fs.ring_cookie))
3198 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3199 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3201 } else if (copy_in_user(&rxnfc->rule_cnt,
3202 &compat_rxnfc->rule_cnt,
3203 sizeof(rxnfc->rule_cnt)))
3207 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3212 if (copy_in_user(compat_rxnfc, rxnfc,
3213 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3214 (const void __user *)rxnfc) ||
3215 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3216 &rxnfc->fs.ring_cookie,
3217 (const void __user *)(&rxnfc->fs.location + 1) -
3218 (const void __user *)&rxnfc->fs.ring_cookie) ||
3219 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3220 sizeof(rxnfc->rule_cnt)))
3223 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3224 /* As an optimisation, we only copy the actual
3225 * number of rules that the underlying
3226 * function returned. Since Mallory might
3227 * change the rule count in user memory, we
3228 * check that it is less than the rule count
3229 * originally given (as the user buffer size),
3230 * which has been range-checked.
3232 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3234 if (actual_rule_cnt < rule_cnt)
3235 rule_cnt = actual_rule_cnt;
3236 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3237 &rxnfc->rule_locs[0],
3238 rule_cnt * sizeof(u32)))
3246 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3248 compat_uptr_t uptr32;
3253 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3256 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3259 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3260 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3262 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3264 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3265 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3271 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3272 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3273 struct compat_ifreq __user *u_ifreq32)
3278 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3280 if (get_user(data32, &u_ifreq32->ifr_data))
3282 ifreq.ifr_data = compat_ptr(data32);
3284 return dev_ioctl(net, cmd, &ifreq, NULL);
3287 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3289 struct compat_ifreq __user *uifr32)
3291 struct ifreq __user *uifr;
3294 /* Handle the fact that while struct ifreq has the same *layout* on
3295 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3296 * which are handled elsewhere, it still has different *size* due to
3297 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3298 * resulting in struct ifreq being 32 and 40 bytes respectively).
3299 * As a result, if the struct happens to be at the end of a page and
3300 * the next page isn't readable/writable, we get a fault. To prevent
3301 * that, copy back and forth to the full size.
3304 uifr = compat_alloc_user_space(sizeof(*uifr));
3305 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3308 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3319 case SIOCGIFBRDADDR:
3320 case SIOCGIFDSTADDR:
3321 case SIOCGIFNETMASK:
3327 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3335 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3336 struct compat_ifreq __user *uifr32)
3339 struct compat_ifmap __user *uifmap32;
3342 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3343 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3344 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3345 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3346 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3347 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3348 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3349 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3353 err = dev_ioctl(net, cmd, &ifr, NULL);
3355 if (cmd == SIOCGIFMAP && !err) {
3356 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3357 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3358 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3359 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3360 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3361 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3362 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3369 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3370 * for some operations; this forces use of the newer bridge-utils that
3371 * use compatible ioctls
3373 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3377 if (get_user(tmp, argp))
3379 if (tmp == BRCTL_GET_VERSION)
3380 return BRCTL_VERSION + 1;
3384 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3385 unsigned int cmd, unsigned long arg)
3387 void __user *argp = compat_ptr(arg);
3388 struct sock *sk = sock->sk;
3389 struct net *net = sock_net(sk);
3391 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3392 return compat_ifr_data_ioctl(net, cmd, argp);
3397 return old_bridge_ioctl(argp);
3399 return compat_dev_ifconf(net, argp);
3401 return ethtool_ioctl(net, argp);
3403 return compat_siocwandev(net, argp);
3406 return compat_sioc_ifmap(net, cmd, argp);
3407 case SIOCGSTAMP_OLD:
3408 case SIOCGSTAMPNS_OLD:
3409 if (!sock->ops->gettstamp)
3410 return -ENOIOCTLCMD;
3411 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3412 !COMPAT_USE_64BIT_TIME);
3414 case SIOCBONDSLAVEINFOQUERY:
3415 case SIOCBONDINFOQUERY:
3418 return compat_ifr_data_ioctl(net, cmd, argp);
3431 case SIOCGSTAMP_NEW:
3432 case SIOCGSTAMPNS_NEW:
3433 return sock_ioctl(file, cmd, arg);
3450 case SIOCSIFHWBROADCAST:
3452 case SIOCGIFBRDADDR:
3453 case SIOCSIFBRDADDR:
3454 case SIOCGIFDSTADDR:
3455 case SIOCSIFDSTADDR:
3456 case SIOCGIFNETMASK:
3457 case SIOCSIFNETMASK:
3469 case SIOCBONDENSLAVE:
3470 case SIOCBONDRELEASE:
3471 case SIOCBONDSETHWADDR:
3472 case SIOCBONDCHANGEACTIVE:
3473 return compat_ifreq_ioctl(net, sock, cmd, argp);
3481 return sock_do_ioctl(net, sock, cmd, arg);
3484 return -ENOIOCTLCMD;
3487 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3490 struct socket *sock = file->private_data;
3491 int ret = -ENOIOCTLCMD;
3498 if (sock->ops->compat_ioctl)
3499 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3501 if (ret == -ENOIOCTLCMD &&
3502 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3503 ret = compat_wext_handle_ioctl(net, cmd, arg);
3505 if (ret == -ENOIOCTLCMD)
3506 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3513 * kernel_bind - bind an address to a socket (kernel space)
3516 * @addrlen: length of address
3518 * Returns 0 or an error.
3521 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3523 return sock->ops->bind(sock, addr, addrlen);
3525 EXPORT_SYMBOL(kernel_bind);
3528 * kernel_listen - move socket to listening state (kernel space)
3530 * @backlog: pending connections queue size
3532 * Returns 0 or an error.
3535 int kernel_listen(struct socket *sock, int backlog)
3537 return sock->ops->listen(sock, backlog);
3539 EXPORT_SYMBOL(kernel_listen);
3542 * kernel_accept - accept a connection (kernel space)
3543 * @sock: listening socket
3544 * @newsock: new connected socket
3547 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3548 * If it fails, @newsock is guaranteed to be %NULL.
3549 * Returns 0 or an error.
3552 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3554 struct sock *sk = sock->sk;
3557 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3562 err = sock->ops->accept(sock, *newsock, flags, true);
3564 sock_release(*newsock);
3569 (*newsock)->ops = sock->ops;
3570 __module_get((*newsock)->ops->owner);
3575 EXPORT_SYMBOL(kernel_accept);
3578 * kernel_connect - connect a socket (kernel space)
3581 * @addrlen: address length
3582 * @flags: flags (O_NONBLOCK, ...)
3584 * For datagram sockets, @addr is the addres to which datagrams are sent
3585 * by default, and the only address from which datagrams are received.
3586 * For stream sockets, attempts to connect to @addr.
3587 * Returns 0 or an error code.
3590 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3593 return sock->ops->connect(sock, addr, addrlen, flags);
3595 EXPORT_SYMBOL(kernel_connect);
3598 * kernel_getsockname - get the address which the socket is bound (kernel space)
3600 * @addr: address holder
3602 * Fills the @addr pointer with the address which the socket is bound.
3603 * Returns 0 or an error code.
3606 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3608 return sock->ops->getname(sock, addr, 0);
3610 EXPORT_SYMBOL(kernel_getsockname);
3613 * kernel_peername - get the address which the socket is connected (kernel space)
3615 * @addr: address holder
3617 * Fills the @addr pointer with the address which the socket is connected.
3618 * Returns 0 or an error code.
3621 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3623 return sock->ops->getname(sock, addr, 1);
3625 EXPORT_SYMBOL(kernel_getpeername);
3628 * kernel_sendpage - send a &page through a socket (kernel space)
3631 * @offset: page offset
3632 * @size: total size in bytes
3633 * @flags: flags (MSG_DONTWAIT, ...)
3635 * Returns the total amount sent in bytes or an error.
3638 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3639 size_t size, int flags)
3641 if (sock->ops->sendpage)
3642 return sock->ops->sendpage(sock, page, offset, size, flags);
3644 return sock_no_sendpage(sock, page, offset, size, flags);
3646 EXPORT_SYMBOL(kernel_sendpage);
3649 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3652 * @offset: page offset
3653 * @size: total size in bytes
3654 * @flags: flags (MSG_DONTWAIT, ...)
3656 * Returns the total amount sent in bytes or an error.
3657 * Caller must hold @sk.
3660 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3661 size_t size, int flags)
3663 struct socket *sock = sk->sk_socket;
3665 if (sock->ops->sendpage_locked)
3666 return sock->ops->sendpage_locked(sk, page, offset, size,
3669 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3671 EXPORT_SYMBOL(kernel_sendpage_locked);
3674 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3676 * @how: connection part
3678 * Returns 0 or an error.
3681 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3683 return sock->ops->shutdown(sock, how);
3685 EXPORT_SYMBOL(kernel_sock_shutdown);
3688 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3691 * This routine returns the IP overhead imposed by a socket i.e.
3692 * the length of the underlying IP header, depending on whether
3693 * this is an IPv4 or IPv6 socket and the length from IP options turned
3694 * on at the socket. Assumes that the caller has a lock on the socket.
3697 u32 kernel_sock_ip_overhead(struct sock *sk)
3699 struct inet_sock *inet;
3700 struct ip_options_rcu *opt;
3702 #if IS_ENABLED(CONFIG_IPV6)
3703 struct ipv6_pinfo *np;
3704 struct ipv6_txoptions *optv6 = NULL;
3705 #endif /* IS_ENABLED(CONFIG_IPV6) */
3710 switch (sk->sk_family) {
3713 overhead += sizeof(struct iphdr);
3714 opt = rcu_dereference_protected(inet->inet_opt,
3715 sock_owned_by_user(sk));
3717 overhead += opt->opt.optlen;
3719 #if IS_ENABLED(CONFIG_IPV6)
3722 overhead += sizeof(struct ipv6hdr);
3724 optv6 = rcu_dereference_protected(np->opt,
3725 sock_owned_by_user(sk));
3727 overhead += (optv6->opt_flen + optv6->opt_nflen);
3729 #endif /* IS_ENABLED(CONFIG_IPV6) */
3730 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3734 EXPORT_SYMBOL(kernel_sock_ip_overhead);