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 mm_segment_t oldfs = get_fs();
930 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
932 result = sock_recvmsg(sock, msg, flags);
936 EXPORT_SYMBOL(kernel_recvmsg);
938 static ssize_t sock_sendpage(struct file *file, struct page *page,
939 int offset, size_t size, loff_t *ppos, int more)
944 sock = file->private_data;
946 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
947 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
950 return kernel_sendpage(sock, page, offset, size, flags);
953 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
954 struct pipe_inode_info *pipe, size_t len,
957 struct socket *sock = file->private_data;
959 if (unlikely(!sock->ops->splice_read))
960 return generic_file_splice_read(file, ppos, pipe, len, flags);
962 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
965 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
967 struct file *file = iocb->ki_filp;
968 struct socket *sock = file->private_data;
969 struct msghdr msg = {.msg_iter = *to,
973 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
974 msg.msg_flags = MSG_DONTWAIT;
976 if (iocb->ki_pos != 0)
979 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
982 res = sock_recvmsg(sock, &msg, msg.msg_flags);
987 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
989 struct file *file = iocb->ki_filp;
990 struct socket *sock = file->private_data;
991 struct msghdr msg = {.msg_iter = *from,
995 if (iocb->ki_pos != 0)
998 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
999 msg.msg_flags = MSG_DONTWAIT;
1001 if (sock->type == SOCK_SEQPACKET)
1002 msg.msg_flags |= MSG_EOR;
1004 res = sock_sendmsg(sock, &msg);
1005 *from = msg.msg_iter;
1010 * Atomic setting of ioctl hooks to avoid race
1011 * with module unload.
1014 static DEFINE_MUTEX(br_ioctl_mutex);
1015 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1017 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1019 mutex_lock(&br_ioctl_mutex);
1020 br_ioctl_hook = hook;
1021 mutex_unlock(&br_ioctl_mutex);
1023 EXPORT_SYMBOL(brioctl_set);
1025 static DEFINE_MUTEX(vlan_ioctl_mutex);
1026 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1028 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1030 mutex_lock(&vlan_ioctl_mutex);
1031 vlan_ioctl_hook = hook;
1032 mutex_unlock(&vlan_ioctl_mutex);
1034 EXPORT_SYMBOL(vlan_ioctl_set);
1036 static DEFINE_MUTEX(dlci_ioctl_mutex);
1037 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1039 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1041 mutex_lock(&dlci_ioctl_mutex);
1042 dlci_ioctl_hook = hook;
1043 mutex_unlock(&dlci_ioctl_mutex);
1045 EXPORT_SYMBOL(dlci_ioctl_set);
1047 static long sock_do_ioctl(struct net *net, struct socket *sock,
1048 unsigned int cmd, unsigned long arg)
1051 void __user *argp = (void __user *)arg;
1053 err = sock->ops->ioctl(sock, cmd, arg);
1056 * If this ioctl is unknown try to hand it down
1057 * to the NIC driver.
1059 if (err != -ENOIOCTLCMD)
1062 if (cmd == SIOCGIFCONF) {
1064 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1067 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1069 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1074 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1076 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1077 if (!err && need_copyout)
1078 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1085 * With an ioctl, arg may well be a user mode pointer, but we don't know
1086 * what to do with it - that's up to the protocol still.
1090 * get_net_ns - increment the refcount of the network namespace
1091 * @ns: common namespace (net)
1093 * Returns the net's common namespace.
1096 struct ns_common *get_net_ns(struct ns_common *ns)
1098 return &get_net(container_of(ns, struct net, ns))->ns;
1100 EXPORT_SYMBOL_GPL(get_net_ns);
1102 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1104 struct socket *sock;
1106 void __user *argp = (void __user *)arg;
1110 sock = file->private_data;
1113 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1116 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1118 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1119 if (!err && need_copyout)
1120 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1123 #ifdef CONFIG_WEXT_CORE
1124 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1125 err = wext_handle_ioctl(net, cmd, argp);
1132 if (get_user(pid, (int __user *)argp))
1134 err = f_setown(sock->file, pid, 1);
1138 err = put_user(f_getown(sock->file),
1139 (int __user *)argp);
1147 request_module("bridge");
1149 mutex_lock(&br_ioctl_mutex);
1151 err = br_ioctl_hook(net, cmd, argp);
1152 mutex_unlock(&br_ioctl_mutex);
1157 if (!vlan_ioctl_hook)
1158 request_module("8021q");
1160 mutex_lock(&vlan_ioctl_mutex);
1161 if (vlan_ioctl_hook)
1162 err = vlan_ioctl_hook(net, argp);
1163 mutex_unlock(&vlan_ioctl_mutex);
1168 if (!dlci_ioctl_hook)
1169 request_module("dlci");
1171 mutex_lock(&dlci_ioctl_mutex);
1172 if (dlci_ioctl_hook)
1173 err = dlci_ioctl_hook(cmd, argp);
1174 mutex_unlock(&dlci_ioctl_mutex);
1178 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1181 err = open_related_ns(&net->ns, get_net_ns);
1183 case SIOCGSTAMP_OLD:
1184 case SIOCGSTAMPNS_OLD:
1185 if (!sock->ops->gettstamp) {
1189 err = sock->ops->gettstamp(sock, argp,
1190 cmd == SIOCGSTAMP_OLD,
1191 !IS_ENABLED(CONFIG_64BIT));
1193 case SIOCGSTAMP_NEW:
1194 case SIOCGSTAMPNS_NEW:
1195 if (!sock->ops->gettstamp) {
1199 err = sock->ops->gettstamp(sock, argp,
1200 cmd == SIOCGSTAMP_NEW,
1204 err = sock_do_ioctl(net, sock, cmd, arg);
1211 * sock_create_lite - creates a socket
1212 * @family: protocol family (AF_INET, ...)
1213 * @type: communication type (SOCK_STREAM, ...)
1214 * @protocol: protocol (0, ...)
1217 * Creates a new socket and assigns it to @res, passing through LSM.
1218 * The new socket initialization is not complete, see kernel_accept().
1219 * Returns 0 or an error. On failure @res is set to %NULL.
1220 * This function internally uses GFP_KERNEL.
1223 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1226 struct socket *sock = NULL;
1228 err = security_socket_create(family, type, protocol, 1);
1232 sock = sock_alloc();
1239 err = security_socket_post_create(sock, family, type, protocol, 1);
1251 EXPORT_SYMBOL(sock_create_lite);
1253 /* No kernel lock held - perfect */
1254 static __poll_t sock_poll(struct file *file, poll_table *wait)
1256 struct socket *sock = file->private_data;
1257 __poll_t events = poll_requested_events(wait), flag = 0;
1259 if (!sock->ops->poll)
1262 if (sk_can_busy_loop(sock->sk)) {
1263 /* poll once if requested by the syscall */
1264 if (events & POLL_BUSY_LOOP)
1265 sk_busy_loop(sock->sk, 1);
1267 /* if this socket can poll_ll, tell the system call */
1268 flag = POLL_BUSY_LOOP;
1271 return sock->ops->poll(file, sock, wait) | flag;
1274 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1276 struct socket *sock = file->private_data;
1278 return sock->ops->mmap(file, sock, vma);
1281 static int sock_close(struct inode *inode, struct file *filp)
1283 __sock_release(SOCKET_I(inode), inode);
1288 * Update the socket async list
1290 * Fasync_list locking strategy.
1292 * 1. fasync_list is modified only under process context socket lock
1293 * i.e. under semaphore.
1294 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1295 * or under socket lock
1298 static int sock_fasync(int fd, struct file *filp, int on)
1300 struct socket *sock = filp->private_data;
1301 struct sock *sk = sock->sk;
1302 struct socket_wq *wq = &sock->wq;
1308 fasync_helper(fd, filp, on, &wq->fasync_list);
1310 if (!wq->fasync_list)
1311 sock_reset_flag(sk, SOCK_FASYNC);
1313 sock_set_flag(sk, SOCK_FASYNC);
1319 /* This function may be called only under rcu_lock */
1321 int sock_wake_async(struct socket_wq *wq, int how, int band)
1323 if (!wq || !wq->fasync_list)
1327 case SOCK_WAKE_WAITD:
1328 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1331 case SOCK_WAKE_SPACE:
1332 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1337 kill_fasync(&wq->fasync_list, SIGIO, band);
1340 kill_fasync(&wq->fasync_list, SIGURG, band);
1345 EXPORT_SYMBOL(sock_wake_async);
1348 * __sock_create - creates a socket
1349 * @net: net namespace
1350 * @family: protocol family (AF_INET, ...)
1351 * @type: communication type (SOCK_STREAM, ...)
1352 * @protocol: protocol (0, ...)
1354 * @kern: boolean for kernel space sockets
1356 * Creates a new socket and assigns it to @res, passing through LSM.
1357 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1358 * be set to true if the socket resides in kernel space.
1359 * This function internally uses GFP_KERNEL.
1362 int __sock_create(struct net *net, int family, int type, int protocol,
1363 struct socket **res, int kern)
1366 struct socket *sock;
1367 const struct net_proto_family *pf;
1370 * Check protocol is in range
1372 if (family < 0 || family >= NPROTO)
1373 return -EAFNOSUPPORT;
1374 if (type < 0 || type >= SOCK_MAX)
1379 This uglymoron is moved from INET layer to here to avoid
1380 deadlock in module load.
1382 if (family == PF_INET && type == SOCK_PACKET) {
1383 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1388 err = security_socket_create(family, type, protocol, kern);
1393 * Allocate the socket and allow the family to set things up. if
1394 * the protocol is 0, the family is instructed to select an appropriate
1397 sock = sock_alloc();
1399 net_warn_ratelimited("socket: no more sockets\n");
1400 return -ENFILE; /* Not exactly a match, but its the
1401 closest posix thing */
1406 #ifdef CONFIG_MODULES
1407 /* Attempt to load a protocol module if the find failed.
1409 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1410 * requested real, full-featured networking support upon configuration.
1411 * Otherwise module support will break!
1413 if (rcu_access_pointer(net_families[family]) == NULL)
1414 request_module("net-pf-%d", family);
1418 pf = rcu_dereference(net_families[family]);
1419 err = -EAFNOSUPPORT;
1424 * We will call the ->create function, that possibly is in a loadable
1425 * module, so we have to bump that loadable module refcnt first.
1427 if (!try_module_get(pf->owner))
1430 /* Now protected by module ref count */
1433 err = pf->create(net, sock, protocol, kern);
1435 goto out_module_put;
1438 * Now to bump the refcnt of the [loadable] module that owns this
1439 * socket at sock_release time we decrement its refcnt.
1441 if (!try_module_get(sock->ops->owner))
1442 goto out_module_busy;
1445 * Now that we're done with the ->create function, the [loadable]
1446 * module can have its refcnt decremented
1448 module_put(pf->owner);
1449 err = security_socket_post_create(sock, family, type, protocol, kern);
1451 goto out_sock_release;
1457 err = -EAFNOSUPPORT;
1460 module_put(pf->owner);
1467 goto out_sock_release;
1469 EXPORT_SYMBOL(__sock_create);
1472 * sock_create - creates a socket
1473 * @family: protocol family (AF_INET, ...)
1474 * @type: communication type (SOCK_STREAM, ...)
1475 * @protocol: protocol (0, ...)
1478 * A wrapper around __sock_create().
1479 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1482 int sock_create(int family, int type, int protocol, struct socket **res)
1484 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1486 EXPORT_SYMBOL(sock_create);
1489 * sock_create_kern - creates a socket (kernel space)
1490 * @net: net namespace
1491 * @family: protocol family (AF_INET, ...)
1492 * @type: communication type (SOCK_STREAM, ...)
1493 * @protocol: protocol (0, ...)
1496 * A wrapper around __sock_create().
1497 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1500 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1502 return __sock_create(net, family, type, protocol, res, 1);
1504 EXPORT_SYMBOL(sock_create_kern);
1506 int __sys_socket(int family, int type, int protocol)
1509 struct socket *sock;
1512 /* Check the SOCK_* constants for consistency. */
1513 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1514 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1515 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1516 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1518 flags = type & ~SOCK_TYPE_MASK;
1519 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1521 type &= SOCK_TYPE_MASK;
1523 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1524 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1526 retval = sock_create(family, type, protocol, &sock);
1530 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1533 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1535 return __sys_socket(family, type, protocol);
1539 * Create a pair of connected sockets.
1542 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1544 struct socket *sock1, *sock2;
1546 struct file *newfile1, *newfile2;
1549 flags = type & ~SOCK_TYPE_MASK;
1550 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1552 type &= SOCK_TYPE_MASK;
1554 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1555 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1558 * reserve descriptors and make sure we won't fail
1559 * to return them to userland.
1561 fd1 = get_unused_fd_flags(flags);
1562 if (unlikely(fd1 < 0))
1565 fd2 = get_unused_fd_flags(flags);
1566 if (unlikely(fd2 < 0)) {
1571 err = put_user(fd1, &usockvec[0]);
1575 err = put_user(fd2, &usockvec[1]);
1580 * Obtain the first socket and check if the underlying protocol
1581 * supports the socketpair call.
1584 err = sock_create(family, type, protocol, &sock1);
1585 if (unlikely(err < 0))
1588 err = sock_create(family, type, protocol, &sock2);
1589 if (unlikely(err < 0)) {
1590 sock_release(sock1);
1594 err = security_socket_socketpair(sock1, sock2);
1595 if (unlikely(err)) {
1596 sock_release(sock2);
1597 sock_release(sock1);
1601 err = sock1->ops->socketpair(sock1, sock2);
1602 if (unlikely(err < 0)) {
1603 sock_release(sock2);
1604 sock_release(sock1);
1608 newfile1 = sock_alloc_file(sock1, flags, NULL);
1609 if (IS_ERR(newfile1)) {
1610 err = PTR_ERR(newfile1);
1611 sock_release(sock2);
1615 newfile2 = sock_alloc_file(sock2, flags, NULL);
1616 if (IS_ERR(newfile2)) {
1617 err = PTR_ERR(newfile2);
1622 audit_fd_pair(fd1, fd2);
1624 fd_install(fd1, newfile1);
1625 fd_install(fd2, newfile2);
1634 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1635 int __user *, usockvec)
1637 return __sys_socketpair(family, type, protocol, usockvec);
1641 * Bind a name to a socket. Nothing much to do here since it's
1642 * the protocol's responsibility to handle the local address.
1644 * We move the socket address to kernel space before we call
1645 * the protocol layer (having also checked the address is ok).
1648 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1650 struct socket *sock;
1651 struct sockaddr_storage address;
1652 int err, fput_needed;
1654 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1656 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1658 err = security_socket_bind(sock,
1659 (struct sockaddr *)&address,
1662 err = sock->ops->bind(sock,
1666 fput_light(sock->file, fput_needed);
1671 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1673 return __sys_bind(fd, umyaddr, addrlen);
1677 * Perform a listen. Basically, we allow the protocol to do anything
1678 * necessary for a listen, and if that works, we mark the socket as
1679 * ready for listening.
1682 int __sys_listen(int fd, int backlog)
1684 struct socket *sock;
1685 int err, fput_needed;
1688 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1691 if ((unsigned int)backlog > somaxconn)
1692 backlog = somaxconn;
1694 err = security_socket_listen(sock, backlog);
1696 err = sock->ops->listen(sock, backlog);
1698 fput_light(sock->file, fput_needed);
1703 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1705 return __sys_listen(fd, backlog);
1708 int __sys_accept4_file(struct file *file, unsigned file_flags,
1709 struct sockaddr __user *upeer_sockaddr,
1710 int __user *upeer_addrlen, int flags,
1711 unsigned long nofile)
1713 struct socket *sock, *newsock;
1714 struct file *newfile;
1715 int err, len, newfd;
1716 struct sockaddr_storage address;
1718 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1721 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1722 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1724 sock = sock_from_file(file, &err);
1729 newsock = sock_alloc();
1733 newsock->type = sock->type;
1734 newsock->ops = sock->ops;
1737 * We don't need try_module_get here, as the listening socket (sock)
1738 * has the protocol module (sock->ops->owner) held.
1740 __module_get(newsock->ops->owner);
1742 newfd = __get_unused_fd_flags(flags, nofile);
1743 if (unlikely(newfd < 0)) {
1745 sock_release(newsock);
1748 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1749 if (IS_ERR(newfile)) {
1750 err = PTR_ERR(newfile);
1751 put_unused_fd(newfd);
1755 err = security_socket_accept(sock, newsock);
1759 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1764 if (upeer_sockaddr) {
1765 len = newsock->ops->getname(newsock,
1766 (struct sockaddr *)&address, 2);
1768 err = -ECONNABORTED;
1771 err = move_addr_to_user(&address,
1772 len, upeer_sockaddr, upeer_addrlen);
1777 /* File flags are not inherited via accept() unlike another OSes. */
1779 fd_install(newfd, newfile);
1785 put_unused_fd(newfd);
1791 * For accept, we attempt to create a new socket, set up the link
1792 * with the client, wake up the client, then return the new
1793 * connected fd. We collect the address of the connector in kernel
1794 * space and move it to user at the very end. This is unclean because
1795 * we open the socket then return an error.
1797 * 1003.1g adds the ability to recvmsg() to query connection pending
1798 * status to recvmsg. We need to add that support in a way thats
1799 * clean when we restructure accept also.
1802 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1803 int __user *upeer_addrlen, int flags)
1810 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1811 upeer_addrlen, flags,
1812 rlimit(RLIMIT_NOFILE));
1820 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1821 int __user *, upeer_addrlen, int, flags)
1823 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1826 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1827 int __user *, upeer_addrlen)
1829 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1833 * Attempt to connect to a socket with the server address. The address
1834 * is in user space so we verify it is OK and move it to kernel space.
1836 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1839 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1840 * other SEQPACKET protocols that take time to connect() as it doesn't
1841 * include the -EINPROGRESS status for such sockets.
1844 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1845 int addrlen, int file_flags)
1847 struct socket *sock;
1850 sock = sock_from_file(file, &err);
1855 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1859 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1860 sock->file->f_flags | file_flags);
1865 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1872 struct sockaddr_storage address;
1874 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1876 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1884 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1887 return __sys_connect(fd, uservaddr, addrlen);
1891 * Get the local address ('name') of a socket object. Move the obtained
1892 * name to user space.
1895 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1896 int __user *usockaddr_len)
1898 struct socket *sock;
1899 struct sockaddr_storage address;
1900 int err, fput_needed;
1902 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1906 err = security_socket_getsockname(sock);
1910 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1913 /* "err" is actually length in this case */
1914 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1917 fput_light(sock->file, fput_needed);
1922 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1923 int __user *, usockaddr_len)
1925 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1929 * Get the remote address ('name') of a socket object. Move the obtained
1930 * name to user space.
1933 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1934 int __user *usockaddr_len)
1936 struct socket *sock;
1937 struct sockaddr_storage address;
1938 int err, fput_needed;
1940 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1942 err = security_socket_getpeername(sock);
1944 fput_light(sock->file, fput_needed);
1948 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1950 /* "err" is actually length in this case */
1951 err = move_addr_to_user(&address, err, usockaddr,
1953 fput_light(sock->file, fput_needed);
1958 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1959 int __user *, usockaddr_len)
1961 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1965 * Send a datagram to a given address. We move the address into kernel
1966 * space and check the user space data area is readable before invoking
1969 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1970 struct sockaddr __user *addr, int addr_len)
1972 struct socket *sock;
1973 struct sockaddr_storage address;
1979 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1982 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1986 msg.msg_name = NULL;
1987 msg.msg_control = NULL;
1988 msg.msg_controllen = 0;
1989 msg.msg_namelen = 0;
1991 err = move_addr_to_kernel(addr, addr_len, &address);
1994 msg.msg_name = (struct sockaddr *)&address;
1995 msg.msg_namelen = addr_len;
1997 if (sock->file->f_flags & O_NONBLOCK)
1998 flags |= MSG_DONTWAIT;
1999 msg.msg_flags = flags;
2000 err = sock_sendmsg(sock, &msg);
2003 fput_light(sock->file, fput_needed);
2008 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2009 unsigned int, flags, struct sockaddr __user *, addr,
2012 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2016 * Send a datagram down a socket.
2019 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2020 unsigned int, flags)
2022 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2026 * Receive a frame from the socket and optionally record the address of the
2027 * sender. We verify the buffers are writable and if needed move the
2028 * sender address from kernel to user space.
2030 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2031 struct sockaddr __user *addr, int __user *addr_len)
2033 struct socket *sock;
2036 struct sockaddr_storage address;
2040 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2043 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2047 msg.msg_control = NULL;
2048 msg.msg_controllen = 0;
2049 /* Save some cycles and don't copy the address if not needed */
2050 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2051 /* We assume all kernel code knows the size of sockaddr_storage */
2052 msg.msg_namelen = 0;
2053 msg.msg_iocb = NULL;
2055 if (sock->file->f_flags & O_NONBLOCK)
2056 flags |= MSG_DONTWAIT;
2057 err = sock_recvmsg(sock, &msg, flags);
2059 if (err >= 0 && addr != NULL) {
2060 err2 = move_addr_to_user(&address,
2061 msg.msg_namelen, addr, addr_len);
2066 fput_light(sock->file, fput_needed);
2071 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2072 unsigned int, flags, struct sockaddr __user *, addr,
2073 int __user *, addr_len)
2075 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2079 * Receive a datagram from a socket.
2082 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2083 unsigned int, flags)
2085 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2089 * Set a socket option. Because we don't know the option lengths we have
2090 * to pass the user mode parameter for the protocols to sort out.
2093 static int __sys_setsockopt(int fd, int level, int optname,
2094 char __user *optval, int optlen)
2096 mm_segment_t oldfs = get_fs();
2097 char *kernel_optval = NULL;
2098 int err, fput_needed;
2099 struct socket *sock;
2104 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2106 err = security_socket_setsockopt(sock, level, optname);
2110 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2111 &optname, optval, &optlen,
2116 } else if (err > 0) {
2121 if (kernel_optval) {
2123 optval = (char __user __force *)kernel_optval;
2126 if (level == SOL_SOCKET)
2128 sock_setsockopt(sock, level, optname, optval,
2132 sock->ops->setsockopt(sock, level, optname, optval,
2135 if (kernel_optval) {
2137 kfree(kernel_optval);
2140 fput_light(sock->file, fput_needed);
2145 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2146 char __user *, optval, int, optlen)
2148 return __sys_setsockopt(fd, level, optname, optval, optlen);
2152 * Get a socket option. Because we don't know the option lengths we have
2153 * to pass a user mode parameter for the protocols to sort out.
2156 static int __sys_getsockopt(int fd, int level, int optname,
2157 char __user *optval, int __user *optlen)
2159 int err, fput_needed;
2160 struct socket *sock;
2163 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2165 err = security_socket_getsockopt(sock, level, optname);
2169 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2171 if (level == SOL_SOCKET)
2173 sock_getsockopt(sock, level, optname, optval,
2177 sock->ops->getsockopt(sock, level, optname, optval,
2180 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2184 fput_light(sock->file, fput_needed);
2189 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2190 char __user *, optval, int __user *, optlen)
2192 return __sys_getsockopt(fd, level, optname, optval, optlen);
2196 * Shutdown a socket.
2199 int __sys_shutdown(int fd, int how)
2201 int err, fput_needed;
2202 struct socket *sock;
2204 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2206 err = security_socket_shutdown(sock, how);
2208 err = sock->ops->shutdown(sock, how);
2209 fput_light(sock->file, fput_needed);
2214 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2216 return __sys_shutdown(fd, how);
2219 /* A couple of helpful macros for getting the address of the 32/64 bit
2220 * fields which are the same type (int / unsigned) on our platforms.
2222 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2223 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2224 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2226 struct used_address {
2227 struct sockaddr_storage name;
2228 unsigned int name_len;
2231 int __copy_msghdr_from_user(struct msghdr *kmsg,
2232 struct user_msghdr __user *umsg,
2233 struct sockaddr __user **save_addr,
2234 struct iovec __user **uiov, size_t *nsegs)
2236 struct user_msghdr msg;
2239 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2242 kmsg->msg_control = (void __force *)msg.msg_control;
2243 kmsg->msg_controllen = msg.msg_controllen;
2244 kmsg->msg_flags = msg.msg_flags;
2246 kmsg->msg_namelen = msg.msg_namelen;
2248 kmsg->msg_namelen = 0;
2250 if (kmsg->msg_namelen < 0)
2253 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2254 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2257 *save_addr = msg.msg_name;
2259 if (msg.msg_name && kmsg->msg_namelen) {
2261 err = move_addr_to_kernel(msg.msg_name,
2268 kmsg->msg_name = NULL;
2269 kmsg->msg_namelen = 0;
2272 if (msg.msg_iovlen > UIO_MAXIOV)
2275 kmsg->msg_iocb = NULL;
2276 *uiov = msg.msg_iov;
2277 *nsegs = msg.msg_iovlen;
2281 static int copy_msghdr_from_user(struct msghdr *kmsg,
2282 struct user_msghdr __user *umsg,
2283 struct sockaddr __user **save_addr,
2286 struct user_msghdr msg;
2289 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2294 err = import_iovec(save_addr ? READ : WRITE,
2295 msg.msg_iov, msg.msg_iovlen,
2296 UIO_FASTIOV, iov, &kmsg->msg_iter);
2297 return err < 0 ? err : 0;
2300 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2301 unsigned int flags, struct used_address *used_address,
2302 unsigned int allowed_msghdr_flags)
2304 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2305 __aligned(sizeof(__kernel_size_t));
2306 /* 20 is size of ipv6_pktinfo */
2307 unsigned char *ctl_buf = ctl;
2313 if (msg_sys->msg_controllen > INT_MAX)
2315 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2316 ctl_len = msg_sys->msg_controllen;
2317 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2319 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2323 ctl_buf = msg_sys->msg_control;
2324 ctl_len = msg_sys->msg_controllen;
2325 } else if (ctl_len) {
2326 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2327 CMSG_ALIGN(sizeof(struct cmsghdr)));
2328 if (ctl_len > sizeof(ctl)) {
2329 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2330 if (ctl_buf == NULL)
2335 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2336 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2337 * checking falls down on this.
2339 if (copy_from_user(ctl_buf,
2340 (void __user __force *)msg_sys->msg_control,
2343 msg_sys->msg_control = ctl_buf;
2345 msg_sys->msg_flags = flags;
2347 if (sock->file->f_flags & O_NONBLOCK)
2348 msg_sys->msg_flags |= MSG_DONTWAIT;
2350 * If this is sendmmsg() and current destination address is same as
2351 * previously succeeded address, omit asking LSM's decision.
2352 * used_address->name_len is initialized to UINT_MAX so that the first
2353 * destination address never matches.
2355 if (used_address && msg_sys->msg_name &&
2356 used_address->name_len == msg_sys->msg_namelen &&
2357 !memcmp(&used_address->name, msg_sys->msg_name,
2358 used_address->name_len)) {
2359 err = sock_sendmsg_nosec(sock, msg_sys);
2362 err = sock_sendmsg(sock, msg_sys);
2364 * If this is sendmmsg() and sending to current destination address was
2365 * successful, remember it.
2367 if (used_address && err >= 0) {
2368 used_address->name_len = msg_sys->msg_namelen;
2369 if (msg_sys->msg_name)
2370 memcpy(&used_address->name, msg_sys->msg_name,
2371 used_address->name_len);
2376 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2381 int sendmsg_copy_msghdr(struct msghdr *msg,
2382 struct user_msghdr __user *umsg, unsigned flags,
2387 if (flags & MSG_CMSG_COMPAT) {
2388 struct compat_msghdr __user *msg_compat;
2390 msg_compat = (struct compat_msghdr __user *) umsg;
2391 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2393 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2401 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2402 struct msghdr *msg_sys, unsigned int flags,
2403 struct used_address *used_address,
2404 unsigned int allowed_msghdr_flags)
2406 struct sockaddr_storage address;
2407 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2410 msg_sys->msg_name = &address;
2412 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2416 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2417 allowed_msghdr_flags);
2423 * BSD sendmsg interface
2425 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2428 /* disallow ancillary data requests from this path */
2429 if (msg->msg_control || msg->msg_controllen)
2432 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2435 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2436 bool forbid_cmsg_compat)
2438 int fput_needed, err;
2439 struct msghdr msg_sys;
2440 struct socket *sock;
2442 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2445 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2449 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2451 fput_light(sock->file, fput_needed);
2456 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2458 return __sys_sendmsg(fd, msg, flags, true);
2462 * Linux sendmmsg interface
2465 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2466 unsigned int flags, bool forbid_cmsg_compat)
2468 int fput_needed, err, datagrams;
2469 struct socket *sock;
2470 struct mmsghdr __user *entry;
2471 struct compat_mmsghdr __user *compat_entry;
2472 struct msghdr msg_sys;
2473 struct used_address used_address;
2474 unsigned int oflags = flags;
2476 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2479 if (vlen > UIO_MAXIOV)
2484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2488 used_address.name_len = UINT_MAX;
2490 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2494 while (datagrams < vlen) {
2495 if (datagrams == vlen - 1)
2498 if (MSG_CMSG_COMPAT & flags) {
2499 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2500 &msg_sys, flags, &used_address, MSG_EOR);
2503 err = __put_user(err, &compat_entry->msg_len);
2506 err = ___sys_sendmsg(sock,
2507 (struct user_msghdr __user *)entry,
2508 &msg_sys, flags, &used_address, MSG_EOR);
2511 err = put_user(err, &entry->msg_len);
2518 if (msg_data_left(&msg_sys))
2523 fput_light(sock->file, fput_needed);
2525 /* We only return an error if no datagrams were able to be sent */
2532 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2533 unsigned int, vlen, unsigned int, flags)
2535 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2538 int recvmsg_copy_msghdr(struct msghdr *msg,
2539 struct user_msghdr __user *umsg, unsigned flags,
2540 struct sockaddr __user **uaddr,
2545 if (MSG_CMSG_COMPAT & flags) {
2546 struct compat_msghdr __user *msg_compat;
2548 msg_compat = (struct compat_msghdr __user *) umsg;
2549 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2551 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2559 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2560 struct user_msghdr __user *msg,
2561 struct sockaddr __user *uaddr,
2562 unsigned int flags, int nosec)
2564 struct compat_msghdr __user *msg_compat =
2565 (struct compat_msghdr __user *) msg;
2566 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2567 struct sockaddr_storage addr;
2568 unsigned long cmsg_ptr;
2572 msg_sys->msg_name = &addr;
2573 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2574 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2576 /* We assume all kernel code knows the size of sockaddr_storage */
2577 msg_sys->msg_namelen = 0;
2579 if (sock->file->f_flags & O_NONBLOCK)
2580 flags |= MSG_DONTWAIT;
2582 if (unlikely(nosec))
2583 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2585 err = sock_recvmsg(sock, msg_sys, flags);
2591 if (uaddr != NULL) {
2592 err = move_addr_to_user(&addr,
2593 msg_sys->msg_namelen, uaddr,
2598 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2602 if (MSG_CMSG_COMPAT & flags)
2603 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2604 &msg_compat->msg_controllen);
2606 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2607 &msg->msg_controllen);
2615 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2616 struct msghdr *msg_sys, unsigned int flags, int nosec)
2618 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2619 /* user mode address pointers */
2620 struct sockaddr __user *uaddr;
2623 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2627 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2633 * BSD recvmsg interface
2636 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2637 struct user_msghdr __user *umsg,
2638 struct sockaddr __user *uaddr, unsigned int flags)
2640 /* disallow ancillary data requests from this path */
2641 if (msg->msg_control || msg->msg_controllen)
2644 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2647 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2648 bool forbid_cmsg_compat)
2650 int fput_needed, err;
2651 struct msghdr msg_sys;
2652 struct socket *sock;
2654 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2657 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2661 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2663 fput_light(sock->file, fput_needed);
2668 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2669 unsigned int, flags)
2671 return __sys_recvmsg(fd, msg, flags, true);
2675 * Linux recvmmsg interface
2678 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2679 unsigned int vlen, unsigned int flags,
2680 struct timespec64 *timeout)
2682 int fput_needed, err, datagrams;
2683 struct socket *sock;
2684 struct mmsghdr __user *entry;
2685 struct compat_mmsghdr __user *compat_entry;
2686 struct msghdr msg_sys;
2687 struct timespec64 end_time;
2688 struct timespec64 timeout64;
2691 poll_select_set_timeout(&end_time, timeout->tv_sec,
2697 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2701 if (likely(!(flags & MSG_ERRQUEUE))) {
2702 err = sock_error(sock->sk);
2710 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2712 while (datagrams < vlen) {
2714 * No need to ask LSM for more than the first datagram.
2716 if (MSG_CMSG_COMPAT & flags) {
2717 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2718 &msg_sys, flags & ~MSG_WAITFORONE,
2722 err = __put_user(err, &compat_entry->msg_len);
2725 err = ___sys_recvmsg(sock,
2726 (struct user_msghdr __user *)entry,
2727 &msg_sys, flags & ~MSG_WAITFORONE,
2731 err = put_user(err, &entry->msg_len);
2739 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2740 if (flags & MSG_WAITFORONE)
2741 flags |= MSG_DONTWAIT;
2744 ktime_get_ts64(&timeout64);
2745 *timeout = timespec64_sub(end_time, timeout64);
2746 if (timeout->tv_sec < 0) {
2747 timeout->tv_sec = timeout->tv_nsec = 0;
2751 /* Timeout, return less than vlen datagrams */
2752 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2756 /* Out of band data, return right away */
2757 if (msg_sys.msg_flags & MSG_OOB)
2765 if (datagrams == 0) {
2771 * We may return less entries than requested (vlen) if the
2772 * sock is non block and there aren't enough datagrams...
2774 if (err != -EAGAIN) {
2776 * ... or if recvmsg returns an error after we
2777 * received some datagrams, where we record the
2778 * error to return on the next call or if the
2779 * app asks about it using getsockopt(SO_ERROR).
2781 sock->sk->sk_err = -err;
2784 fput_light(sock->file, fput_needed);
2789 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2790 unsigned int vlen, unsigned int flags,
2791 struct __kernel_timespec __user *timeout,
2792 struct old_timespec32 __user *timeout32)
2795 struct timespec64 timeout_sys;
2797 if (timeout && get_timespec64(&timeout_sys, timeout))
2800 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2803 if (!timeout && !timeout32)
2804 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2806 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2811 if (timeout && put_timespec64(&timeout_sys, timeout))
2812 datagrams = -EFAULT;
2814 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2815 datagrams = -EFAULT;
2820 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2821 unsigned int, vlen, unsigned int, flags,
2822 struct __kernel_timespec __user *, timeout)
2824 if (flags & MSG_CMSG_COMPAT)
2827 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2830 #ifdef CONFIG_COMPAT_32BIT_TIME
2831 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2832 unsigned int, vlen, unsigned int, flags,
2833 struct old_timespec32 __user *, timeout)
2835 if (flags & MSG_CMSG_COMPAT)
2838 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2842 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2843 /* Argument list sizes for sys_socketcall */
2844 #define AL(x) ((x) * sizeof(unsigned long))
2845 static const unsigned char nargs[21] = {
2846 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2847 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2848 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2855 * System call vectors.
2857 * Argument checking cleaned up. Saved 20% in size.
2858 * This function doesn't need to set the kernel lock because
2859 * it is set by the callees.
2862 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2864 unsigned long a[AUDITSC_ARGS];
2865 unsigned long a0, a1;
2869 if (call < 1 || call > SYS_SENDMMSG)
2871 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2874 if (len > sizeof(a))
2877 /* copy_from_user should be SMP safe. */
2878 if (copy_from_user(a, args, len))
2881 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2890 err = __sys_socket(a0, a1, a[2]);
2893 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2896 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2899 err = __sys_listen(a0, a1);
2902 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2903 (int __user *)a[2], 0);
2905 case SYS_GETSOCKNAME:
2907 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2908 (int __user *)a[2]);
2910 case SYS_GETPEERNAME:
2912 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2913 (int __user *)a[2]);
2915 case SYS_SOCKETPAIR:
2916 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2919 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2923 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2924 (struct sockaddr __user *)a[4], a[5]);
2927 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2931 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2932 (struct sockaddr __user *)a[4],
2933 (int __user *)a[5]);
2936 err = __sys_shutdown(a0, a1);
2938 case SYS_SETSOCKOPT:
2939 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2942 case SYS_GETSOCKOPT:
2944 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2945 (int __user *)a[4]);
2948 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2952 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2956 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2960 if (IS_ENABLED(CONFIG_64BIT))
2961 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2963 (struct __kernel_timespec __user *)a[4],
2966 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2968 (struct old_timespec32 __user *)a[4]);
2971 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2972 (int __user *)a[2], a[3]);
2981 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2984 * sock_register - add a socket protocol handler
2985 * @ops: description of protocol
2987 * This function is called by a protocol handler that wants to
2988 * advertise its address family, and have it linked into the
2989 * socket interface. The value ops->family corresponds to the
2990 * socket system call protocol family.
2992 int sock_register(const struct net_proto_family *ops)
2996 if (ops->family >= NPROTO) {
2997 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3001 spin_lock(&net_family_lock);
3002 if (rcu_dereference_protected(net_families[ops->family],
3003 lockdep_is_held(&net_family_lock)))
3006 rcu_assign_pointer(net_families[ops->family], ops);
3009 spin_unlock(&net_family_lock);
3011 pr_info("NET: Registered protocol family %d\n", ops->family);
3014 EXPORT_SYMBOL(sock_register);
3017 * sock_unregister - remove a protocol handler
3018 * @family: protocol family to remove
3020 * This function is called by a protocol handler that wants to
3021 * remove its address family, and have it unlinked from the
3022 * new socket creation.
3024 * If protocol handler is a module, then it can use module reference
3025 * counts to protect against new references. If protocol handler is not
3026 * a module then it needs to provide its own protection in
3027 * the ops->create routine.
3029 void sock_unregister(int family)
3031 BUG_ON(family < 0 || family >= NPROTO);
3033 spin_lock(&net_family_lock);
3034 RCU_INIT_POINTER(net_families[family], NULL);
3035 spin_unlock(&net_family_lock);
3039 pr_info("NET: Unregistered protocol family %d\n", family);
3041 EXPORT_SYMBOL(sock_unregister);
3043 bool sock_is_registered(int family)
3045 return family < NPROTO && rcu_access_pointer(net_families[family]);
3048 static int __init sock_init(void)
3052 * Initialize the network sysctl infrastructure.
3054 err = net_sysctl_init();
3059 * Initialize skbuff SLAB cache
3064 * Initialize the protocols module.
3069 err = register_filesystem(&sock_fs_type);
3072 sock_mnt = kern_mount(&sock_fs_type);
3073 if (IS_ERR(sock_mnt)) {
3074 err = PTR_ERR(sock_mnt);
3078 /* The real protocol initialization is performed in later initcalls.
3081 #ifdef CONFIG_NETFILTER
3082 err = netfilter_init();
3087 ptp_classifier_init();
3093 unregister_filesystem(&sock_fs_type);
3098 core_initcall(sock_init); /* early initcall */
3100 #ifdef CONFIG_PROC_FS
3101 void socket_seq_show(struct seq_file *seq)
3103 seq_printf(seq, "sockets: used %d\n",
3104 sock_inuse_get(seq->private));
3106 #endif /* CONFIG_PROC_FS */
3108 #ifdef CONFIG_COMPAT
3109 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3111 struct compat_ifconf ifc32;
3115 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3118 ifc.ifc_len = ifc32.ifc_len;
3119 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3122 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3127 ifc32.ifc_len = ifc.ifc_len;
3128 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3134 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3136 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3137 bool convert_in = false, convert_out = false;
3138 size_t buf_size = 0;
3139 struct ethtool_rxnfc __user *rxnfc = NULL;
3141 u32 rule_cnt = 0, actual_rule_cnt;
3146 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3149 compat_rxnfc = compat_ptr(data);
3151 if (get_user(ethcmd, &compat_rxnfc->cmd))
3154 /* Most ethtool structures are defined without padding.
3155 * Unfortunately struct ethtool_rxnfc is an exception.
3160 case ETHTOOL_GRXCLSRLALL:
3161 /* Buffer size is variable */
3162 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3164 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3166 buf_size += rule_cnt * sizeof(u32);
3168 case ETHTOOL_GRXRINGS:
3169 case ETHTOOL_GRXCLSRLCNT:
3170 case ETHTOOL_GRXCLSRULE:
3171 case ETHTOOL_SRXCLSRLINS:
3174 case ETHTOOL_SRXCLSRLDEL:
3175 buf_size += sizeof(struct ethtool_rxnfc);
3177 rxnfc = compat_alloc_user_space(buf_size);
3181 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3184 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3187 /* We expect there to be holes between fs.m_ext and
3188 * fs.ring_cookie and at the end of fs, but nowhere else.
3190 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3191 sizeof(compat_rxnfc->fs.m_ext) !=
3192 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3193 sizeof(rxnfc->fs.m_ext));
3195 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3196 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3197 offsetof(struct ethtool_rxnfc, fs.location) -
3198 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3200 if (copy_in_user(rxnfc, compat_rxnfc,
3201 (void __user *)(&rxnfc->fs.m_ext + 1) -
3202 (void __user *)rxnfc) ||
3203 copy_in_user(&rxnfc->fs.ring_cookie,
3204 &compat_rxnfc->fs.ring_cookie,
3205 (void __user *)(&rxnfc->fs.location + 1) -
3206 (void __user *)&rxnfc->fs.ring_cookie))
3208 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3209 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3211 } else if (copy_in_user(&rxnfc->rule_cnt,
3212 &compat_rxnfc->rule_cnt,
3213 sizeof(rxnfc->rule_cnt)))
3217 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3222 if (copy_in_user(compat_rxnfc, rxnfc,
3223 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3224 (const void __user *)rxnfc) ||
3225 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3226 &rxnfc->fs.ring_cookie,
3227 (const void __user *)(&rxnfc->fs.location + 1) -
3228 (const void __user *)&rxnfc->fs.ring_cookie) ||
3229 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3230 sizeof(rxnfc->rule_cnt)))
3233 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3234 /* As an optimisation, we only copy the actual
3235 * number of rules that the underlying
3236 * function returned. Since Mallory might
3237 * change the rule count in user memory, we
3238 * check that it is less than the rule count
3239 * originally given (as the user buffer size),
3240 * which has been range-checked.
3242 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3244 if (actual_rule_cnt < rule_cnt)
3245 rule_cnt = actual_rule_cnt;
3246 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3247 &rxnfc->rule_locs[0],
3248 rule_cnt * sizeof(u32)))
3256 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3258 compat_uptr_t uptr32;
3263 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3266 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3269 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3270 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3272 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3274 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3275 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3281 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3282 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3283 struct compat_ifreq __user *u_ifreq32)
3288 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3290 if (get_user(data32, &u_ifreq32->ifr_data))
3292 ifreq.ifr_data = compat_ptr(data32);
3294 return dev_ioctl(net, cmd, &ifreq, NULL);
3297 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3299 struct compat_ifreq __user *uifr32)
3301 struct ifreq __user *uifr;
3304 /* Handle the fact that while struct ifreq has the same *layout* on
3305 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3306 * which are handled elsewhere, it still has different *size* due to
3307 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3308 * resulting in struct ifreq being 32 and 40 bytes respectively).
3309 * As a result, if the struct happens to be at the end of a page and
3310 * the next page isn't readable/writable, we get a fault. To prevent
3311 * that, copy back and forth to the full size.
3314 uifr = compat_alloc_user_space(sizeof(*uifr));
3315 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3318 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3329 case SIOCGIFBRDADDR:
3330 case SIOCGIFDSTADDR:
3331 case SIOCGIFNETMASK:
3337 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3345 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3346 struct compat_ifreq __user *uifr32)
3349 struct compat_ifmap __user *uifmap32;
3352 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3353 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3354 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3355 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3356 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3357 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3358 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3359 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3363 err = dev_ioctl(net, cmd, &ifr, NULL);
3365 if (cmd == SIOCGIFMAP && !err) {
3366 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3367 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3368 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3369 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3370 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3371 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3372 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3381 struct sockaddr rt_dst; /* target address */
3382 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3383 struct sockaddr rt_genmask; /* target network mask (IP) */
3384 unsigned short rt_flags;
3387 unsigned char rt_tos;
3388 unsigned char rt_class;
3390 short rt_metric; /* +1 for binary compatibility! */
3391 /* char * */ u32 rt_dev; /* forcing the device at add */
3392 u32 rt_mtu; /* per route MTU/Window */
3393 u32 rt_window; /* Window clamping */
3394 unsigned short rt_irtt; /* Initial RTT */
3397 struct in6_rtmsg32 {
3398 struct in6_addr rtmsg_dst;
3399 struct in6_addr rtmsg_src;
3400 struct in6_addr rtmsg_gateway;
3410 static int routing_ioctl(struct net *net, struct socket *sock,
3411 unsigned int cmd, void __user *argp)
3415 struct in6_rtmsg r6;
3419 mm_segment_t old_fs = get_fs();
3421 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3422 struct in6_rtmsg32 __user *ur6 = argp;
3423 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3424 3 * sizeof(struct in6_addr));
3425 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3426 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3427 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3428 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3429 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3430 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3431 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3435 struct rtentry32 __user *ur4 = argp;
3436 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3437 3 * sizeof(struct sockaddr));
3438 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3439 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3440 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3441 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3442 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3443 ret |= get_user(rtdev, &(ur4->rt_dev));
3445 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3446 r4.rt_dev = (char __user __force *)devname;
3460 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3467 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3468 * for some operations; this forces use of the newer bridge-utils that
3469 * use compatible ioctls
3471 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3475 if (get_user(tmp, argp))
3477 if (tmp == BRCTL_GET_VERSION)
3478 return BRCTL_VERSION + 1;
3482 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3483 unsigned int cmd, unsigned long arg)
3485 void __user *argp = compat_ptr(arg);
3486 struct sock *sk = sock->sk;
3487 struct net *net = sock_net(sk);
3489 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3490 return compat_ifr_data_ioctl(net, cmd, argp);
3495 return old_bridge_ioctl(argp);
3497 return compat_dev_ifconf(net, argp);
3499 return ethtool_ioctl(net, argp);
3501 return compat_siocwandev(net, argp);
3504 return compat_sioc_ifmap(net, cmd, argp);
3507 return routing_ioctl(net, sock, cmd, argp);
3508 case SIOCGSTAMP_OLD:
3509 case SIOCGSTAMPNS_OLD:
3510 if (!sock->ops->gettstamp)
3511 return -ENOIOCTLCMD;
3512 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3513 !COMPAT_USE_64BIT_TIME);
3515 case SIOCBONDSLAVEINFOQUERY:
3516 case SIOCBONDINFOQUERY:
3519 return compat_ifr_data_ioctl(net, cmd, argp);
3532 case SIOCGSTAMP_NEW:
3533 case SIOCGSTAMPNS_NEW:
3534 return sock_ioctl(file, cmd, arg);
3551 case SIOCSIFHWBROADCAST:
3553 case SIOCGIFBRDADDR:
3554 case SIOCSIFBRDADDR:
3555 case SIOCGIFDSTADDR:
3556 case SIOCSIFDSTADDR:
3557 case SIOCGIFNETMASK:
3558 case SIOCSIFNETMASK:
3570 case SIOCBONDENSLAVE:
3571 case SIOCBONDRELEASE:
3572 case SIOCBONDSETHWADDR:
3573 case SIOCBONDCHANGEACTIVE:
3574 return compat_ifreq_ioctl(net, sock, cmd, argp);
3582 return sock_do_ioctl(net, sock, cmd, arg);
3585 return -ENOIOCTLCMD;
3588 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3591 struct socket *sock = file->private_data;
3592 int ret = -ENOIOCTLCMD;
3599 if (sock->ops->compat_ioctl)
3600 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3602 if (ret == -ENOIOCTLCMD &&
3603 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3604 ret = compat_wext_handle_ioctl(net, cmd, arg);
3606 if (ret == -ENOIOCTLCMD)
3607 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3614 * kernel_bind - bind an address to a socket (kernel space)
3617 * @addrlen: length of address
3619 * Returns 0 or an error.
3622 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3624 return sock->ops->bind(sock, addr, addrlen);
3626 EXPORT_SYMBOL(kernel_bind);
3629 * kernel_listen - move socket to listening state (kernel space)
3631 * @backlog: pending connections queue size
3633 * Returns 0 or an error.
3636 int kernel_listen(struct socket *sock, int backlog)
3638 return sock->ops->listen(sock, backlog);
3640 EXPORT_SYMBOL(kernel_listen);
3643 * kernel_accept - accept a connection (kernel space)
3644 * @sock: listening socket
3645 * @newsock: new connected socket
3648 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3649 * If it fails, @newsock is guaranteed to be %NULL.
3650 * Returns 0 or an error.
3653 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3655 struct sock *sk = sock->sk;
3658 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3663 err = sock->ops->accept(sock, *newsock, flags, true);
3665 sock_release(*newsock);
3670 (*newsock)->ops = sock->ops;
3671 __module_get((*newsock)->ops->owner);
3676 EXPORT_SYMBOL(kernel_accept);
3679 * kernel_connect - connect a socket (kernel space)
3682 * @addrlen: address length
3683 * @flags: flags (O_NONBLOCK, ...)
3685 * For datagram sockets, @addr is the addres to which datagrams are sent
3686 * by default, and the only address from which datagrams are received.
3687 * For stream sockets, attempts to connect to @addr.
3688 * Returns 0 or an error code.
3691 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3694 return sock->ops->connect(sock, addr, addrlen, flags);
3696 EXPORT_SYMBOL(kernel_connect);
3699 * kernel_getsockname - get the address which the socket is bound (kernel space)
3701 * @addr: address holder
3703 * Fills the @addr pointer with the address which the socket is bound.
3704 * Returns 0 or an error code.
3707 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3709 return sock->ops->getname(sock, addr, 0);
3711 EXPORT_SYMBOL(kernel_getsockname);
3714 * kernel_peername - get the address which the socket is connected (kernel space)
3716 * @addr: address holder
3718 * Fills the @addr pointer with the address which the socket is connected.
3719 * Returns 0 or an error code.
3722 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3724 return sock->ops->getname(sock, addr, 1);
3726 EXPORT_SYMBOL(kernel_getpeername);
3729 * kernel_getsockopt - get a socket option (kernel space)
3731 * @level: API level (SOL_SOCKET, ...)
3732 * @optname: option tag
3733 * @optval: option value
3734 * @optlen: option length
3736 * Assigns the option length to @optlen.
3737 * Returns 0 or an error.
3740 int kernel_getsockopt(struct socket *sock, int level, int optname,
3741 char *optval, int *optlen)
3743 mm_segment_t oldfs = get_fs();
3744 char __user *uoptval;
3745 int __user *uoptlen;
3748 uoptval = (char __user __force *) optval;
3749 uoptlen = (int __user __force *) optlen;
3752 if (level == SOL_SOCKET)
3753 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3755 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3760 EXPORT_SYMBOL(kernel_getsockopt);
3763 * kernel_setsockopt - set a socket option (kernel space)
3765 * @level: API level (SOL_SOCKET, ...)
3766 * @optname: option tag
3767 * @optval: option value
3768 * @optlen: option length
3770 * Returns 0 or an error.
3773 int kernel_setsockopt(struct socket *sock, int level, int optname,
3774 char *optval, unsigned int optlen)
3776 mm_segment_t oldfs = get_fs();
3777 char __user *uoptval;
3780 uoptval = (char __user __force *) optval;
3783 if (level == SOL_SOCKET)
3784 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3786 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3791 EXPORT_SYMBOL(kernel_setsockopt);
3794 * kernel_sendpage - send a &page through a socket (kernel space)
3797 * @offset: page offset
3798 * @size: total size in bytes
3799 * @flags: flags (MSG_DONTWAIT, ...)
3801 * Returns the total amount sent in bytes or an error.
3804 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3805 size_t size, int flags)
3807 if (sock->ops->sendpage)
3808 return sock->ops->sendpage(sock, page, offset, size, flags);
3810 return sock_no_sendpage(sock, page, offset, size, flags);
3812 EXPORT_SYMBOL(kernel_sendpage);
3815 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3818 * @offset: page offset
3819 * @size: total size in bytes
3820 * @flags: flags (MSG_DONTWAIT, ...)
3822 * Returns the total amount sent in bytes or an error.
3823 * Caller must hold @sk.
3826 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3827 size_t size, int flags)
3829 struct socket *sock = sk->sk_socket;
3831 if (sock->ops->sendpage_locked)
3832 return sock->ops->sendpage_locked(sk, page, offset, size,
3835 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3837 EXPORT_SYMBOL(kernel_sendpage_locked);
3840 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3842 * @how: connection part
3844 * Returns 0 or an error.
3847 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3849 return sock->ops->shutdown(sock, how);
3851 EXPORT_SYMBOL(kernel_sock_shutdown);
3854 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3857 * This routine returns the IP overhead imposed by a socket i.e.
3858 * the length of the underlying IP header, depending on whether
3859 * this is an IPv4 or IPv6 socket and the length from IP options turned
3860 * on at the socket. Assumes that the caller has a lock on the socket.
3863 u32 kernel_sock_ip_overhead(struct sock *sk)
3865 struct inet_sock *inet;
3866 struct ip_options_rcu *opt;
3868 #if IS_ENABLED(CONFIG_IPV6)
3869 struct ipv6_pinfo *np;
3870 struct ipv6_txoptions *optv6 = NULL;
3871 #endif /* IS_ENABLED(CONFIG_IPV6) */
3876 switch (sk->sk_family) {
3879 overhead += sizeof(struct iphdr);
3880 opt = rcu_dereference_protected(inet->inet_opt,
3881 sock_owned_by_user(sk));
3883 overhead += opt->opt.optlen;
3885 #if IS_ENABLED(CONFIG_IPV6)
3888 overhead += sizeof(struct ipv6hdr);
3890 optv6 = rcu_dereference_protected(np->opt,
3891 sock_owned_by_user(sk));
3893 overhead += (optv6->opt_flen + optv6->opt_nflen);
3895 #endif /* IS_ENABLED(CONFIG_IPV6) */
3896 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3900 EXPORT_SYMBOL(kernel_sock_ip_overhead);