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,
133 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
134 * in the operation structures but are done directly via the socketcall() multiplexor.
137 static const struct file_operations socket_file_ops = {
138 .owner = THIS_MODULE,
140 .read_iter = sock_read_iter,
141 .write_iter = sock_write_iter,
143 .unlocked_ioctl = sock_ioctl,
145 .compat_ioctl = compat_sock_ioctl,
148 .release = sock_close,
149 .fasync = sock_fasync,
150 .sendpage = sock_sendpage,
151 .splice_write = generic_splice_sendpage,
152 .splice_read = sock_splice_read,
156 * The protocol list. Each protocol is registered in here.
159 static DEFINE_SPINLOCK(net_family_lock);
160 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164 * Move socket addresses back and forth across the kernel/user
165 * divide and look after the messy bits.
169 * move_addr_to_kernel - copy a socket address into kernel space
170 * @uaddr: Address in user space
171 * @kaddr: Address in kernel space
172 * @ulen: Length in user space
174 * The address is copied into kernel space. If the provided address is
175 * too long an error code of -EINVAL is returned. If the copy gives
176 * invalid addresses -EFAULT is returned. On a success 0 is returned.
179 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
181 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
185 if (copy_from_user(kaddr, uaddr, ulen))
187 return audit_sockaddr(ulen, kaddr);
191 * move_addr_to_user - copy an address to user space
192 * @kaddr: kernel space address
193 * @klen: length of address in kernel
194 * @uaddr: user space address
195 * @ulen: pointer to user length field
197 * The value pointed to by ulen on entry is the buffer length available.
198 * This is overwritten with the buffer space used. -EINVAL is returned
199 * if an overlong buffer is specified or a negative buffer size. -EFAULT
200 * is returned if either the buffer or the length field are not
202 * After copying the data up to the limit the user specifies, the true
203 * length of the data is written over the length limit the user
204 * specified. Zero is returned for a success.
207 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
208 void __user *uaddr, int __user *ulen)
213 BUG_ON(klen > sizeof(struct sockaddr_storage));
214 err = get_user(len, ulen);
222 if (audit_sockaddr(klen, kaddr))
224 if (copy_to_user(uaddr, kaddr, len))
228 * "fromlen shall refer to the value before truncation.."
231 return __put_user(klen, ulen);
234 static struct kmem_cache *sock_inode_cachep __ro_after_init;
236 static struct inode *sock_alloc_inode(struct super_block *sb)
238 struct socket_alloc *ei;
240 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
243 init_waitqueue_head(&ei->socket.wq.wait);
244 ei->socket.wq.fasync_list = NULL;
245 ei->socket.wq.flags = 0;
247 ei->socket.state = SS_UNCONNECTED;
248 ei->socket.flags = 0;
249 ei->socket.ops = NULL;
250 ei->socket.sk = NULL;
251 ei->socket.file = NULL;
253 return &ei->vfs_inode;
256 static void sock_free_inode(struct inode *inode)
258 struct socket_alloc *ei;
260 ei = container_of(inode, struct socket_alloc, vfs_inode);
261 kmem_cache_free(sock_inode_cachep, ei);
264 static void init_once(void *foo)
266 struct socket_alloc *ei = (struct socket_alloc *)foo;
268 inode_init_once(&ei->vfs_inode);
271 static void init_inodecache(void)
273 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
274 sizeof(struct socket_alloc),
276 (SLAB_HWCACHE_ALIGN |
277 SLAB_RECLAIM_ACCOUNT |
278 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
280 BUG_ON(sock_inode_cachep == NULL);
283 static const struct super_operations sockfs_ops = {
284 .alloc_inode = sock_alloc_inode,
285 .free_inode = sock_free_inode,
286 .statfs = simple_statfs,
290 * sockfs_dname() is called from d_path().
292 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
294 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
295 d_inode(dentry)->i_ino);
298 static const struct dentry_operations sockfs_dentry_operations = {
299 .d_dname = sockfs_dname,
302 static int sockfs_xattr_get(const struct xattr_handler *handler,
303 struct dentry *dentry, struct inode *inode,
304 const char *suffix, void *value, size_t size)
307 if (dentry->d_name.len + 1 > size)
309 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
311 return dentry->d_name.len + 1;
314 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
315 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
316 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
318 static const struct xattr_handler sockfs_xattr_handler = {
319 .name = XATTR_NAME_SOCKPROTONAME,
320 .get = sockfs_xattr_get,
323 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
324 struct dentry *dentry, struct inode *inode,
325 const char *suffix, const void *value,
326 size_t size, int flags)
328 /* Handled by LSM. */
332 static const struct xattr_handler sockfs_security_xattr_handler = {
333 .prefix = XATTR_SECURITY_PREFIX,
334 .set = sockfs_security_xattr_set,
337 static const struct xattr_handler *sockfs_xattr_handlers[] = {
338 &sockfs_xattr_handler,
339 &sockfs_security_xattr_handler,
343 static int sockfs_init_fs_context(struct fs_context *fc)
345 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
348 ctx->ops = &sockfs_ops;
349 ctx->dops = &sockfs_dentry_operations;
350 ctx->xattr = sockfs_xattr_handlers;
354 static struct vfsmount *sock_mnt __read_mostly;
356 static struct file_system_type sock_fs_type = {
358 .init_fs_context = sockfs_init_fs_context,
359 .kill_sb = kill_anon_super,
363 * Obtains the first available file descriptor and sets it up for use.
365 * These functions create file structures and maps them to fd space
366 * of the current process. On success it returns file descriptor
367 * and file struct implicitly stored in sock->file.
368 * Note that another thread may close file descriptor before we return
369 * from this function. We use the fact that now we do not refer
370 * to socket after mapping. If one day we will need it, this
371 * function will increment ref. count on file by 1.
373 * In any case returned fd MAY BE not valid!
374 * This race condition is unavoidable
375 * with shared fd spaces, we cannot solve it inside kernel,
376 * but we take care of internal coherence yet.
380 * sock_alloc_file - Bind a &socket to a &file
382 * @flags: file status flags
383 * @dname: protocol name
385 * Returns the &file bound with @sock, implicitly storing it
386 * in sock->file. If dname is %NULL, sets to "".
387 * On failure the return is a ERR pointer (see linux/err.h).
388 * This function uses GFP_KERNEL internally.
391 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
396 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
398 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
399 O_RDWR | (flags & O_NONBLOCK),
407 file->private_data = sock;
408 stream_open(SOCK_INODE(sock), file);
411 EXPORT_SYMBOL(sock_alloc_file);
413 static int sock_map_fd(struct socket *sock, int flags)
415 struct file *newfile;
416 int fd = get_unused_fd_flags(flags);
417 if (unlikely(fd < 0)) {
422 newfile = sock_alloc_file(sock, flags, NULL);
423 if (!IS_ERR(newfile)) {
424 fd_install(fd, newfile);
429 return PTR_ERR(newfile);
433 * sock_from_file - Return the &socket bounded to @file.
435 * @err: pointer to an error code return
437 * On failure returns %NULL and assigns -ENOTSOCK to @err.
440 struct socket *sock_from_file(struct file *file, int *err)
442 if (file->f_op == &socket_file_ops)
443 return file->private_data; /* set in sock_map_fd */
448 EXPORT_SYMBOL(sock_from_file);
451 * sockfd_lookup - Go from a file number to its socket slot
453 * @err: pointer to an error code return
455 * The file handle passed in is locked and the socket it is bound
456 * to is returned. If an error occurs the err pointer is overwritten
457 * with a negative errno code and NULL is returned. The function checks
458 * for both invalid handles and passing a handle which is not a socket.
460 * On a success the socket object pointer is returned.
463 struct socket *sockfd_lookup(int fd, int *err)
474 sock = sock_from_file(file, err);
479 EXPORT_SYMBOL(sockfd_lookup);
481 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
483 struct fd f = fdget(fd);
488 sock = sock_from_file(f.file, err);
490 *fput_needed = f.flags;
498 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
504 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
514 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
519 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
526 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
528 int err = simple_setattr(dentry, iattr);
530 if (!err && (iattr->ia_valid & ATTR_UID)) {
531 struct socket *sock = SOCKET_I(d_inode(dentry));
534 sock->sk->sk_uid = iattr->ia_uid;
542 static const struct inode_operations sockfs_inode_ops = {
543 .listxattr = sockfs_listxattr,
544 .setattr = sockfs_setattr,
548 * sock_alloc - allocate a socket
550 * Allocate a new inode and socket object. The two are bound together
551 * and initialised. The socket is then returned. If we are out of inodes
552 * NULL is returned. This functions uses GFP_KERNEL internally.
555 struct socket *sock_alloc(void)
560 inode = new_inode_pseudo(sock_mnt->mnt_sb);
564 sock = SOCKET_I(inode);
566 inode->i_ino = get_next_ino();
567 inode->i_mode = S_IFSOCK | S_IRWXUGO;
568 inode->i_uid = current_fsuid();
569 inode->i_gid = current_fsgid();
570 inode->i_op = &sockfs_inode_ops;
574 EXPORT_SYMBOL(sock_alloc);
577 * sock_release - close a socket
578 * @sock: socket to close
580 * The socket is released from the protocol stack if it has a release
581 * callback, and the inode is then released if the socket is bound to
582 * an inode not a file.
585 static void __sock_release(struct socket *sock, struct inode *inode)
588 struct module *owner = sock->ops->owner;
592 sock->ops->release(sock);
600 if (sock->wq.fasync_list)
601 pr_err("%s: fasync list not empty!\n", __func__);
604 iput(SOCK_INODE(sock));
610 void sock_release(struct socket *sock)
612 __sock_release(sock, NULL);
614 EXPORT_SYMBOL(sock_release);
616 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
618 u8 flags = *tx_flags;
620 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
621 flags |= SKBTX_HW_TSTAMP;
623 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
624 flags |= SKBTX_SW_TSTAMP;
626 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
627 flags |= SKBTX_SCHED_TSTAMP;
631 EXPORT_SYMBOL(__sock_tx_timestamp);
633 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
635 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
637 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
639 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
640 inet_sendmsg, sock, msg,
642 BUG_ON(ret == -EIOCBQUEUED);
647 * sock_sendmsg - send a message through @sock
649 * @msg: message to send
651 * Sends @msg through @sock, passing through LSM.
652 * Returns the number of bytes sent, or an error code.
654 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
656 int err = security_socket_sendmsg(sock, msg,
659 return err ?: sock_sendmsg_nosec(sock, msg);
661 EXPORT_SYMBOL(sock_sendmsg);
664 * kernel_sendmsg - send a message through @sock (kernel-space)
666 * @msg: message header
668 * @num: vec array length
669 * @size: total message data size
671 * Builds the message data with @vec and sends it through @sock.
672 * Returns the number of bytes sent, or an error code.
675 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
676 struct kvec *vec, size_t num, size_t size)
678 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
679 return sock_sendmsg(sock, msg);
681 EXPORT_SYMBOL(kernel_sendmsg);
684 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
686 * @msg: message header
687 * @vec: output s/g array
688 * @num: output s/g array length
689 * @size: total message data size
691 * Builds the message data with @vec and sends it through @sock.
692 * Returns the number of bytes sent, or an error code.
693 * Caller must hold @sk.
696 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
697 struct kvec *vec, size_t num, size_t size)
699 struct socket *sock = sk->sk_socket;
701 if (!sock->ops->sendmsg_locked)
702 return sock_no_sendmsg_locked(sk, msg, size);
704 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
706 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
708 EXPORT_SYMBOL(kernel_sendmsg_locked);
710 static bool skb_is_err_queue(const struct sk_buff *skb)
712 /* pkt_type of skbs enqueued on the error queue are set to
713 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
714 * in recvmsg, since skbs received on a local socket will never
715 * have a pkt_type of PACKET_OUTGOING.
717 return skb->pkt_type == PACKET_OUTGOING;
720 /* On transmit, software and hardware timestamps are returned independently.
721 * As the two skb clones share the hardware timestamp, which may be updated
722 * before the software timestamp is received, a hardware TX timestamp may be
723 * returned only if there is no software TX timestamp. Ignore false software
724 * timestamps, which may be made in the __sock_recv_timestamp() call when the
725 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
726 * hardware timestamp.
728 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
730 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
733 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
735 struct scm_ts_pktinfo ts_pktinfo;
736 struct net_device *orig_dev;
738 if (!skb_mac_header_was_set(skb))
741 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
744 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
746 ts_pktinfo.if_index = orig_dev->ifindex;
749 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
750 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
751 sizeof(ts_pktinfo), &ts_pktinfo);
755 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
757 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
760 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
761 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
762 struct scm_timestamping_internal tss;
764 int empty = 1, false_tstamp = 0;
765 struct skb_shared_hwtstamps *shhwtstamps =
768 /* Race occurred between timestamp enabling and packet
769 receiving. Fill in the current time for now. */
770 if (need_software_tstamp && skb->tstamp == 0) {
771 __net_timestamp(skb);
775 if (need_software_tstamp) {
776 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
778 struct __kernel_sock_timeval tv;
780 skb_get_new_timestamp(skb, &tv);
781 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
784 struct __kernel_old_timeval tv;
786 skb_get_timestamp(skb, &tv);
787 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
792 struct __kernel_timespec ts;
794 skb_get_new_timestampns(skb, &ts);
795 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
798 struct __kernel_old_timespec ts;
800 skb_get_timestampns(skb, &ts);
801 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
807 memset(&tss, 0, sizeof(tss));
808 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
809 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
812 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
813 !skb_is_swtx_tstamp(skb, false_tstamp) &&
814 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
816 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
817 !skb_is_err_queue(skb))
818 put_ts_pktinfo(msg, skb);
821 if (sock_flag(sk, SOCK_TSTAMP_NEW))
822 put_cmsg_scm_timestamping64(msg, &tss);
824 put_cmsg_scm_timestamping(msg, &tss);
826 if (skb_is_err_queue(skb) && skb->len &&
827 SKB_EXT_ERR(skb)->opt_stats)
828 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
829 skb->len, skb->data);
832 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
834 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
839 if (!sock_flag(sk, SOCK_WIFI_STATUS))
841 if (!skb->wifi_acked_valid)
844 ack = skb->wifi_acked;
846 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
848 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
850 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
853 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
854 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
855 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
858 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
861 sock_recv_timestamp(msg, sk, skb);
862 sock_recv_drops(msg, sk, skb);
864 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
866 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
868 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
870 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
873 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
874 inet_recvmsg, sock, msg, msg_data_left(msg),
879 * sock_recvmsg - receive a message from @sock
881 * @msg: message to receive
882 * @flags: message flags
884 * Receives @msg from @sock, passing through LSM. Returns the total number
885 * of bytes received, or an error.
887 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
889 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
891 return err ?: sock_recvmsg_nosec(sock, msg, flags);
893 EXPORT_SYMBOL(sock_recvmsg);
896 * kernel_recvmsg - Receive a message from a socket (kernel space)
897 * @sock: The socket to receive the message from
898 * @msg: Received message
899 * @vec: Input s/g array for message data
900 * @num: Size of input s/g array
901 * @size: Number of bytes to read
902 * @flags: Message flags (MSG_DONTWAIT, etc...)
904 * On return the msg structure contains the scatter/gather array passed in the
905 * vec argument. The array is modified so that it consists of the unfilled
906 * portion of the original array.
908 * The returned value is the total number of bytes received, or an error.
911 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
912 struct kvec *vec, size_t num, size_t size, int flags)
914 mm_segment_t oldfs = get_fs();
917 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
919 result = sock_recvmsg(sock, msg, flags);
923 EXPORT_SYMBOL(kernel_recvmsg);
925 static ssize_t sock_sendpage(struct file *file, struct page *page,
926 int offset, size_t size, loff_t *ppos, int more)
931 sock = file->private_data;
933 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
934 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
937 return kernel_sendpage(sock, page, offset, size, flags);
940 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
941 struct pipe_inode_info *pipe, size_t len,
944 struct socket *sock = file->private_data;
946 if (unlikely(!sock->ops->splice_read))
947 return generic_file_splice_read(file, ppos, pipe, len, flags);
949 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
952 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
954 struct file *file = iocb->ki_filp;
955 struct socket *sock = file->private_data;
956 struct msghdr msg = {.msg_iter = *to,
960 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
961 msg.msg_flags = MSG_DONTWAIT;
963 if (iocb->ki_pos != 0)
966 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
969 res = sock_recvmsg(sock, &msg, msg.msg_flags);
974 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
976 struct file *file = iocb->ki_filp;
977 struct socket *sock = file->private_data;
978 struct msghdr msg = {.msg_iter = *from,
982 if (iocb->ki_pos != 0)
985 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
986 msg.msg_flags = MSG_DONTWAIT;
988 if (sock->type == SOCK_SEQPACKET)
989 msg.msg_flags |= MSG_EOR;
991 res = sock_sendmsg(sock, &msg);
992 *from = msg.msg_iter;
997 * Atomic setting of ioctl hooks to avoid race
998 * with module unload.
1001 static DEFINE_MUTEX(br_ioctl_mutex);
1002 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1004 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1006 mutex_lock(&br_ioctl_mutex);
1007 br_ioctl_hook = hook;
1008 mutex_unlock(&br_ioctl_mutex);
1010 EXPORT_SYMBOL(brioctl_set);
1012 static DEFINE_MUTEX(vlan_ioctl_mutex);
1013 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1015 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1017 mutex_lock(&vlan_ioctl_mutex);
1018 vlan_ioctl_hook = hook;
1019 mutex_unlock(&vlan_ioctl_mutex);
1021 EXPORT_SYMBOL(vlan_ioctl_set);
1023 static DEFINE_MUTEX(dlci_ioctl_mutex);
1024 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1026 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1028 mutex_lock(&dlci_ioctl_mutex);
1029 dlci_ioctl_hook = hook;
1030 mutex_unlock(&dlci_ioctl_mutex);
1032 EXPORT_SYMBOL(dlci_ioctl_set);
1034 static long sock_do_ioctl(struct net *net, struct socket *sock,
1035 unsigned int cmd, unsigned long arg)
1038 void __user *argp = (void __user *)arg;
1040 err = sock->ops->ioctl(sock, cmd, arg);
1043 * If this ioctl is unknown try to hand it down
1044 * to the NIC driver.
1046 if (err != -ENOIOCTLCMD)
1049 if (cmd == SIOCGIFCONF) {
1051 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1054 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1056 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1061 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1063 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1064 if (!err && need_copyout)
1065 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1072 * With an ioctl, arg may well be a user mode pointer, but we don't know
1073 * what to do with it - that's up to the protocol still.
1077 * get_net_ns - increment the refcount of the network namespace
1078 * @ns: common namespace (net)
1080 * Returns the net's common namespace.
1083 struct ns_common *get_net_ns(struct ns_common *ns)
1085 return &get_net(container_of(ns, struct net, ns))->ns;
1087 EXPORT_SYMBOL_GPL(get_net_ns);
1089 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1091 struct socket *sock;
1093 void __user *argp = (void __user *)arg;
1097 sock = file->private_data;
1100 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1103 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1105 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1106 if (!err && need_copyout)
1107 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1110 #ifdef CONFIG_WEXT_CORE
1111 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1112 err = wext_handle_ioctl(net, cmd, argp);
1119 if (get_user(pid, (int __user *)argp))
1121 err = f_setown(sock->file, pid, 1);
1125 err = put_user(f_getown(sock->file),
1126 (int __user *)argp);
1134 request_module("bridge");
1136 mutex_lock(&br_ioctl_mutex);
1138 err = br_ioctl_hook(net, cmd, argp);
1139 mutex_unlock(&br_ioctl_mutex);
1144 if (!vlan_ioctl_hook)
1145 request_module("8021q");
1147 mutex_lock(&vlan_ioctl_mutex);
1148 if (vlan_ioctl_hook)
1149 err = vlan_ioctl_hook(net, argp);
1150 mutex_unlock(&vlan_ioctl_mutex);
1155 if (!dlci_ioctl_hook)
1156 request_module("dlci");
1158 mutex_lock(&dlci_ioctl_mutex);
1159 if (dlci_ioctl_hook)
1160 err = dlci_ioctl_hook(cmd, argp);
1161 mutex_unlock(&dlci_ioctl_mutex);
1165 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1168 err = open_related_ns(&net->ns, get_net_ns);
1170 case SIOCGSTAMP_OLD:
1171 case SIOCGSTAMPNS_OLD:
1172 if (!sock->ops->gettstamp) {
1176 err = sock->ops->gettstamp(sock, argp,
1177 cmd == SIOCGSTAMP_OLD,
1178 !IS_ENABLED(CONFIG_64BIT));
1180 case SIOCGSTAMP_NEW:
1181 case SIOCGSTAMPNS_NEW:
1182 if (!sock->ops->gettstamp) {
1186 err = sock->ops->gettstamp(sock, argp,
1187 cmd == SIOCGSTAMP_NEW,
1191 err = sock_do_ioctl(net, sock, cmd, arg);
1198 * sock_create_lite - creates a socket
1199 * @family: protocol family (AF_INET, ...)
1200 * @type: communication type (SOCK_STREAM, ...)
1201 * @protocol: protocol (0, ...)
1204 * Creates a new socket and assigns it to @res, passing through LSM.
1205 * The new socket initialization is not complete, see kernel_accept().
1206 * Returns 0 or an error. On failure @res is set to %NULL.
1207 * This function internally uses GFP_KERNEL.
1210 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1213 struct socket *sock = NULL;
1215 err = security_socket_create(family, type, protocol, 1);
1219 sock = sock_alloc();
1226 err = security_socket_post_create(sock, family, type, protocol, 1);
1238 EXPORT_SYMBOL(sock_create_lite);
1240 /* No kernel lock held - perfect */
1241 static __poll_t sock_poll(struct file *file, poll_table *wait)
1243 struct socket *sock = file->private_data;
1244 __poll_t events = poll_requested_events(wait), flag = 0;
1246 if (!sock->ops->poll)
1249 if (sk_can_busy_loop(sock->sk)) {
1250 /* poll once if requested by the syscall */
1251 if (events & POLL_BUSY_LOOP)
1252 sk_busy_loop(sock->sk, 1);
1254 /* if this socket can poll_ll, tell the system call */
1255 flag = POLL_BUSY_LOOP;
1258 return sock->ops->poll(file, sock, wait) | flag;
1261 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1263 struct socket *sock = file->private_data;
1265 return sock->ops->mmap(file, sock, vma);
1268 static int sock_close(struct inode *inode, struct file *filp)
1270 __sock_release(SOCKET_I(inode), inode);
1275 * Update the socket async list
1277 * Fasync_list locking strategy.
1279 * 1. fasync_list is modified only under process context socket lock
1280 * i.e. under semaphore.
1281 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1282 * or under socket lock
1285 static int sock_fasync(int fd, struct file *filp, int on)
1287 struct socket *sock = filp->private_data;
1288 struct sock *sk = sock->sk;
1289 struct socket_wq *wq = &sock->wq;
1295 fasync_helper(fd, filp, on, &wq->fasync_list);
1297 if (!wq->fasync_list)
1298 sock_reset_flag(sk, SOCK_FASYNC);
1300 sock_set_flag(sk, SOCK_FASYNC);
1306 /* This function may be called only under rcu_lock */
1308 int sock_wake_async(struct socket_wq *wq, int how, int band)
1310 if (!wq || !wq->fasync_list)
1314 case SOCK_WAKE_WAITD:
1315 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1318 case SOCK_WAKE_SPACE:
1319 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1324 kill_fasync(&wq->fasync_list, SIGIO, band);
1327 kill_fasync(&wq->fasync_list, SIGURG, band);
1332 EXPORT_SYMBOL(sock_wake_async);
1335 * __sock_create - creates a socket
1336 * @net: net namespace
1337 * @family: protocol family (AF_INET, ...)
1338 * @type: communication type (SOCK_STREAM, ...)
1339 * @protocol: protocol (0, ...)
1341 * @kern: boolean for kernel space sockets
1343 * Creates a new socket and assigns it to @res, passing through LSM.
1344 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1345 * be set to true if the socket resides in kernel space.
1346 * This function internally uses GFP_KERNEL.
1349 int __sock_create(struct net *net, int family, int type, int protocol,
1350 struct socket **res, int kern)
1353 struct socket *sock;
1354 const struct net_proto_family *pf;
1357 * Check protocol is in range
1359 if (family < 0 || family >= NPROTO)
1360 return -EAFNOSUPPORT;
1361 if (type < 0 || type >= SOCK_MAX)
1366 This uglymoron is moved from INET layer to here to avoid
1367 deadlock in module load.
1369 if (family == PF_INET && type == SOCK_PACKET) {
1370 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1375 err = security_socket_create(family, type, protocol, kern);
1380 * Allocate the socket and allow the family to set things up. if
1381 * the protocol is 0, the family is instructed to select an appropriate
1384 sock = sock_alloc();
1386 net_warn_ratelimited("socket: no more sockets\n");
1387 return -ENFILE; /* Not exactly a match, but its the
1388 closest posix thing */
1393 #ifdef CONFIG_MODULES
1394 /* Attempt to load a protocol module if the find failed.
1396 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1397 * requested real, full-featured networking support upon configuration.
1398 * Otherwise module support will break!
1400 if (rcu_access_pointer(net_families[family]) == NULL)
1401 request_module("net-pf-%d", family);
1405 pf = rcu_dereference(net_families[family]);
1406 err = -EAFNOSUPPORT;
1411 * We will call the ->create function, that possibly is in a loadable
1412 * module, so we have to bump that loadable module refcnt first.
1414 if (!try_module_get(pf->owner))
1417 /* Now protected by module ref count */
1420 err = pf->create(net, sock, protocol, kern);
1422 goto out_module_put;
1425 * Now to bump the refcnt of the [loadable] module that owns this
1426 * socket at sock_release time we decrement its refcnt.
1428 if (!try_module_get(sock->ops->owner))
1429 goto out_module_busy;
1432 * Now that we're done with the ->create function, the [loadable]
1433 * module can have its refcnt decremented
1435 module_put(pf->owner);
1436 err = security_socket_post_create(sock, family, type, protocol, kern);
1438 goto out_sock_release;
1444 err = -EAFNOSUPPORT;
1447 module_put(pf->owner);
1454 goto out_sock_release;
1456 EXPORT_SYMBOL(__sock_create);
1459 * sock_create - creates a socket
1460 * @family: protocol family (AF_INET, ...)
1461 * @type: communication type (SOCK_STREAM, ...)
1462 * @protocol: protocol (0, ...)
1465 * A wrapper around __sock_create().
1466 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1469 int sock_create(int family, int type, int protocol, struct socket **res)
1471 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1473 EXPORT_SYMBOL(sock_create);
1476 * sock_create_kern - creates a socket (kernel space)
1477 * @net: net namespace
1478 * @family: protocol family (AF_INET, ...)
1479 * @type: communication type (SOCK_STREAM, ...)
1480 * @protocol: protocol (0, ...)
1483 * A wrapper around __sock_create().
1484 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1487 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1489 return __sock_create(net, family, type, protocol, res, 1);
1491 EXPORT_SYMBOL(sock_create_kern);
1493 int __sys_socket(int family, int type, int protocol)
1496 struct socket *sock;
1499 /* Check the SOCK_* constants for consistency. */
1500 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1501 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1502 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1503 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1505 flags = type & ~SOCK_TYPE_MASK;
1506 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1508 type &= SOCK_TYPE_MASK;
1510 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1511 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1513 retval = sock_create(family, type, protocol, &sock);
1517 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1520 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1522 return __sys_socket(family, type, protocol);
1526 * Create a pair of connected sockets.
1529 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1531 struct socket *sock1, *sock2;
1533 struct file *newfile1, *newfile2;
1536 flags = type & ~SOCK_TYPE_MASK;
1537 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1539 type &= SOCK_TYPE_MASK;
1541 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1542 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1545 * reserve descriptors and make sure we won't fail
1546 * to return them to userland.
1548 fd1 = get_unused_fd_flags(flags);
1549 if (unlikely(fd1 < 0))
1552 fd2 = get_unused_fd_flags(flags);
1553 if (unlikely(fd2 < 0)) {
1558 err = put_user(fd1, &usockvec[0]);
1562 err = put_user(fd2, &usockvec[1]);
1567 * Obtain the first socket and check if the underlying protocol
1568 * supports the socketpair call.
1571 err = sock_create(family, type, protocol, &sock1);
1572 if (unlikely(err < 0))
1575 err = sock_create(family, type, protocol, &sock2);
1576 if (unlikely(err < 0)) {
1577 sock_release(sock1);
1581 err = security_socket_socketpair(sock1, sock2);
1582 if (unlikely(err)) {
1583 sock_release(sock2);
1584 sock_release(sock1);
1588 err = sock1->ops->socketpair(sock1, sock2);
1589 if (unlikely(err < 0)) {
1590 sock_release(sock2);
1591 sock_release(sock1);
1595 newfile1 = sock_alloc_file(sock1, flags, NULL);
1596 if (IS_ERR(newfile1)) {
1597 err = PTR_ERR(newfile1);
1598 sock_release(sock2);
1602 newfile2 = sock_alloc_file(sock2, flags, NULL);
1603 if (IS_ERR(newfile2)) {
1604 err = PTR_ERR(newfile2);
1609 audit_fd_pair(fd1, fd2);
1611 fd_install(fd1, newfile1);
1612 fd_install(fd2, newfile2);
1621 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1622 int __user *, usockvec)
1624 return __sys_socketpair(family, type, protocol, usockvec);
1628 * Bind a name to a socket. Nothing much to do here since it's
1629 * the protocol's responsibility to handle the local address.
1631 * We move the socket address to kernel space before we call
1632 * the protocol layer (having also checked the address is ok).
1635 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1637 struct socket *sock;
1638 struct sockaddr_storage address;
1639 int err, fput_needed;
1641 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1643 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1645 err = security_socket_bind(sock,
1646 (struct sockaddr *)&address,
1649 err = sock->ops->bind(sock,
1653 fput_light(sock->file, fput_needed);
1658 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1660 return __sys_bind(fd, umyaddr, addrlen);
1664 * Perform a listen. Basically, we allow the protocol to do anything
1665 * necessary for a listen, and if that works, we mark the socket as
1666 * ready for listening.
1669 int __sys_listen(int fd, int backlog)
1671 struct socket *sock;
1672 int err, fput_needed;
1675 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1677 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1678 if ((unsigned int)backlog > somaxconn)
1679 backlog = somaxconn;
1681 err = security_socket_listen(sock, backlog);
1683 err = sock->ops->listen(sock, backlog);
1685 fput_light(sock->file, fput_needed);
1690 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1692 return __sys_listen(fd, backlog);
1695 int __sys_accept4_file(struct file *file, unsigned file_flags,
1696 struct sockaddr __user *upeer_sockaddr,
1697 int __user *upeer_addrlen, int flags)
1699 struct socket *sock, *newsock;
1700 struct file *newfile;
1701 int err, len, newfd;
1702 struct sockaddr_storage address;
1704 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1707 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1708 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1710 sock = sock_from_file(file, &err);
1715 newsock = sock_alloc();
1719 newsock->type = sock->type;
1720 newsock->ops = sock->ops;
1723 * We don't need try_module_get here, as the listening socket (sock)
1724 * has the protocol module (sock->ops->owner) held.
1726 __module_get(newsock->ops->owner);
1728 newfd = get_unused_fd_flags(flags);
1729 if (unlikely(newfd < 0)) {
1731 sock_release(newsock);
1734 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1735 if (IS_ERR(newfile)) {
1736 err = PTR_ERR(newfile);
1737 put_unused_fd(newfd);
1741 err = security_socket_accept(sock, newsock);
1745 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1750 if (upeer_sockaddr) {
1751 len = newsock->ops->getname(newsock,
1752 (struct sockaddr *)&address, 2);
1754 err = -ECONNABORTED;
1757 err = move_addr_to_user(&address,
1758 len, upeer_sockaddr, upeer_addrlen);
1763 /* File flags are not inherited via accept() unlike another OSes. */
1765 fd_install(newfd, newfile);
1771 put_unused_fd(newfd);
1777 * For accept, we attempt to create a new socket, set up the link
1778 * with the client, wake up the client, then return the new
1779 * connected fd. We collect the address of the connector in kernel
1780 * space and move it to user at the very end. This is unclean because
1781 * we open the socket then return an error.
1783 * 1003.1g adds the ability to recvmsg() to query connection pending
1784 * status to recvmsg. We need to add that support in a way thats
1785 * clean when we restructure accept also.
1788 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1789 int __user *upeer_addrlen, int flags)
1796 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1797 upeer_addrlen, flags);
1805 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1806 int __user *, upeer_addrlen, int, flags)
1808 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1811 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1812 int __user *, upeer_addrlen)
1814 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1818 * Attempt to connect to a socket with the server address. The address
1819 * is in user space so we verify it is OK and move it to kernel space.
1821 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1824 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1825 * other SEQPACKET protocols that take time to connect() as it doesn't
1826 * include the -EINPROGRESS status for such sockets.
1829 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1830 int addrlen, int file_flags)
1832 struct socket *sock;
1835 sock = sock_from_file(file, &err);
1840 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1844 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1845 sock->file->f_flags | file_flags);
1850 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1857 struct sockaddr_storage address;
1859 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1861 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1869 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1872 return __sys_connect(fd, uservaddr, addrlen);
1876 * Get the local address ('name') of a socket object. Move the obtained
1877 * name to user space.
1880 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1881 int __user *usockaddr_len)
1883 struct socket *sock;
1884 struct sockaddr_storage address;
1885 int err, fput_needed;
1887 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1891 err = security_socket_getsockname(sock);
1895 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1898 /* "err" is actually length in this case */
1899 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1902 fput_light(sock->file, fput_needed);
1907 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1908 int __user *, usockaddr_len)
1910 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1914 * Get the remote address ('name') of a socket object. Move the obtained
1915 * name to user space.
1918 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1919 int __user *usockaddr_len)
1921 struct socket *sock;
1922 struct sockaddr_storage address;
1923 int err, fput_needed;
1925 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1927 err = security_socket_getpeername(sock);
1929 fput_light(sock->file, fput_needed);
1933 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1935 /* "err" is actually length in this case */
1936 err = move_addr_to_user(&address, err, usockaddr,
1938 fput_light(sock->file, fput_needed);
1943 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1944 int __user *, usockaddr_len)
1946 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1950 * Send a datagram to a given address. We move the address into kernel
1951 * space and check the user space data area is readable before invoking
1954 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1955 struct sockaddr __user *addr, int addr_len)
1957 struct socket *sock;
1958 struct sockaddr_storage address;
1964 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1967 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1971 msg.msg_name = NULL;
1972 msg.msg_control = NULL;
1973 msg.msg_controllen = 0;
1974 msg.msg_namelen = 0;
1976 err = move_addr_to_kernel(addr, addr_len, &address);
1979 msg.msg_name = (struct sockaddr *)&address;
1980 msg.msg_namelen = addr_len;
1982 if (sock->file->f_flags & O_NONBLOCK)
1983 flags |= MSG_DONTWAIT;
1984 msg.msg_flags = flags;
1985 err = sock_sendmsg(sock, &msg);
1988 fput_light(sock->file, fput_needed);
1993 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1994 unsigned int, flags, struct sockaddr __user *, addr,
1997 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2001 * Send a datagram down a socket.
2004 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2005 unsigned int, flags)
2007 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2011 * Receive a frame from the socket and optionally record the address of the
2012 * sender. We verify the buffers are writable and if needed move the
2013 * sender address from kernel to user space.
2015 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2016 struct sockaddr __user *addr, int __user *addr_len)
2018 struct socket *sock;
2021 struct sockaddr_storage address;
2025 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2028 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2032 msg.msg_control = NULL;
2033 msg.msg_controllen = 0;
2034 /* Save some cycles and don't copy the address if not needed */
2035 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2036 /* We assume all kernel code knows the size of sockaddr_storage */
2037 msg.msg_namelen = 0;
2038 msg.msg_iocb = NULL;
2040 if (sock->file->f_flags & O_NONBLOCK)
2041 flags |= MSG_DONTWAIT;
2042 err = sock_recvmsg(sock, &msg, flags);
2044 if (err >= 0 && addr != NULL) {
2045 err2 = move_addr_to_user(&address,
2046 msg.msg_namelen, addr, addr_len);
2051 fput_light(sock->file, fput_needed);
2056 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2057 unsigned int, flags, struct sockaddr __user *, addr,
2058 int __user *, addr_len)
2060 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2064 * Receive a datagram from a socket.
2067 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2068 unsigned int, flags)
2070 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2074 * Set a socket option. Because we don't know the option lengths we have
2075 * to pass the user mode parameter for the protocols to sort out.
2078 static int __sys_setsockopt(int fd, int level, int optname,
2079 char __user *optval, int optlen)
2081 mm_segment_t oldfs = get_fs();
2082 char *kernel_optval = NULL;
2083 int err, fput_needed;
2084 struct socket *sock;
2089 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2091 err = security_socket_setsockopt(sock, level, optname);
2095 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2096 &optname, optval, &optlen,
2101 } else if (err > 0) {
2106 if (kernel_optval) {
2108 optval = (char __user __force *)kernel_optval;
2111 if (level == SOL_SOCKET)
2113 sock_setsockopt(sock, level, optname, optval,
2117 sock->ops->setsockopt(sock, level, optname, optval,
2120 if (kernel_optval) {
2122 kfree(kernel_optval);
2125 fput_light(sock->file, fput_needed);
2130 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2131 char __user *, optval, int, optlen)
2133 return __sys_setsockopt(fd, level, optname, optval, optlen);
2137 * Get a socket option. Because we don't know the option lengths we have
2138 * to pass a user mode parameter for the protocols to sort out.
2141 static int __sys_getsockopt(int fd, int level, int optname,
2142 char __user *optval, int __user *optlen)
2144 int err, fput_needed;
2145 struct socket *sock;
2148 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2150 err = security_socket_getsockopt(sock, level, optname);
2154 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2156 if (level == SOL_SOCKET)
2158 sock_getsockopt(sock, level, optname, optval,
2162 sock->ops->getsockopt(sock, level, optname, optval,
2165 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2169 fput_light(sock->file, fput_needed);
2174 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2175 char __user *, optval, int __user *, optlen)
2177 return __sys_getsockopt(fd, level, optname, optval, optlen);
2181 * Shutdown a socket.
2184 int __sys_shutdown(int fd, int how)
2186 int err, fput_needed;
2187 struct socket *sock;
2189 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2191 err = security_socket_shutdown(sock, how);
2193 err = sock->ops->shutdown(sock, how);
2194 fput_light(sock->file, fput_needed);
2199 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2201 return __sys_shutdown(fd, how);
2204 /* A couple of helpful macros for getting the address of the 32/64 bit
2205 * fields which are the same type (int / unsigned) on our platforms.
2207 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2208 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2209 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2211 struct used_address {
2212 struct sockaddr_storage name;
2213 unsigned int name_len;
2216 static int copy_msghdr_from_user(struct msghdr *kmsg,
2217 struct user_msghdr __user *umsg,
2218 struct sockaddr __user **save_addr,
2221 struct user_msghdr msg;
2224 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2227 kmsg->msg_control = (void __force *)msg.msg_control;
2228 kmsg->msg_controllen = msg.msg_controllen;
2229 kmsg->msg_flags = msg.msg_flags;
2231 kmsg->msg_namelen = msg.msg_namelen;
2233 kmsg->msg_namelen = 0;
2235 if (kmsg->msg_namelen < 0)
2238 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2239 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2242 *save_addr = msg.msg_name;
2244 if (msg.msg_name && kmsg->msg_namelen) {
2246 err = move_addr_to_kernel(msg.msg_name,
2253 kmsg->msg_name = NULL;
2254 kmsg->msg_namelen = 0;
2257 if (msg.msg_iovlen > UIO_MAXIOV)
2260 kmsg->msg_iocb = NULL;
2262 err = import_iovec(save_addr ? READ : WRITE,
2263 msg.msg_iov, msg.msg_iovlen,
2264 UIO_FASTIOV, iov, &kmsg->msg_iter);
2265 return err < 0 ? err : 0;
2268 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2269 unsigned int flags, struct used_address *used_address,
2270 unsigned int allowed_msghdr_flags)
2272 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2273 __aligned(sizeof(__kernel_size_t));
2274 /* 20 is size of ipv6_pktinfo */
2275 unsigned char *ctl_buf = ctl;
2281 if (msg_sys->msg_controllen > INT_MAX)
2283 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2284 ctl_len = msg_sys->msg_controllen;
2285 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2287 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2291 ctl_buf = msg_sys->msg_control;
2292 ctl_len = msg_sys->msg_controllen;
2293 } else if (ctl_len) {
2294 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2295 CMSG_ALIGN(sizeof(struct cmsghdr)));
2296 if (ctl_len > sizeof(ctl)) {
2297 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2298 if (ctl_buf == NULL)
2303 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2304 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2305 * checking falls down on this.
2307 if (copy_from_user(ctl_buf,
2308 (void __user __force *)msg_sys->msg_control,
2311 msg_sys->msg_control = ctl_buf;
2313 msg_sys->msg_flags = flags;
2315 if (sock->file->f_flags & O_NONBLOCK)
2316 msg_sys->msg_flags |= MSG_DONTWAIT;
2318 * If this is sendmmsg() and current destination address is same as
2319 * previously succeeded address, omit asking LSM's decision.
2320 * used_address->name_len is initialized to UINT_MAX so that the first
2321 * destination address never matches.
2323 if (used_address && msg_sys->msg_name &&
2324 used_address->name_len == msg_sys->msg_namelen &&
2325 !memcmp(&used_address->name, msg_sys->msg_name,
2326 used_address->name_len)) {
2327 err = sock_sendmsg_nosec(sock, msg_sys);
2330 err = sock_sendmsg(sock, msg_sys);
2332 * If this is sendmmsg() and sending to current destination address was
2333 * successful, remember it.
2335 if (used_address && err >= 0) {
2336 used_address->name_len = msg_sys->msg_namelen;
2337 if (msg_sys->msg_name)
2338 memcpy(&used_address->name, msg_sys->msg_name,
2339 used_address->name_len);
2344 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2349 int sendmsg_copy_msghdr(struct msghdr *msg,
2350 struct user_msghdr __user *umsg, unsigned flags,
2355 if (flags & MSG_CMSG_COMPAT) {
2356 struct compat_msghdr __user *msg_compat;
2358 msg_compat = (struct compat_msghdr __user *) umsg;
2359 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2361 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2369 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2370 struct msghdr *msg_sys, unsigned int flags,
2371 struct used_address *used_address,
2372 unsigned int allowed_msghdr_flags)
2374 struct sockaddr_storage address;
2375 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2378 msg_sys->msg_name = &address;
2380 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2384 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2385 allowed_msghdr_flags);
2391 * BSD sendmsg interface
2393 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2396 /* disallow ancillary data requests from this path */
2397 if (msg->msg_control || msg->msg_controllen)
2400 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2403 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2404 bool forbid_cmsg_compat)
2406 int fput_needed, err;
2407 struct msghdr msg_sys;
2408 struct socket *sock;
2410 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2413 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2417 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2419 fput_light(sock->file, fput_needed);
2424 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2426 return __sys_sendmsg(fd, msg, flags, true);
2430 * Linux sendmmsg interface
2433 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2434 unsigned int flags, bool forbid_cmsg_compat)
2436 int fput_needed, err, datagrams;
2437 struct socket *sock;
2438 struct mmsghdr __user *entry;
2439 struct compat_mmsghdr __user *compat_entry;
2440 struct msghdr msg_sys;
2441 struct used_address used_address;
2442 unsigned int oflags = flags;
2444 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2447 if (vlen > UIO_MAXIOV)
2452 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2456 used_address.name_len = UINT_MAX;
2458 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2462 while (datagrams < vlen) {
2463 if (datagrams == vlen - 1)
2466 if (MSG_CMSG_COMPAT & flags) {
2467 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2468 &msg_sys, flags, &used_address, MSG_EOR);
2471 err = __put_user(err, &compat_entry->msg_len);
2474 err = ___sys_sendmsg(sock,
2475 (struct user_msghdr __user *)entry,
2476 &msg_sys, flags, &used_address, MSG_EOR);
2479 err = put_user(err, &entry->msg_len);
2486 if (msg_data_left(&msg_sys))
2491 fput_light(sock->file, fput_needed);
2493 /* We only return an error if no datagrams were able to be sent */
2500 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2501 unsigned int, vlen, unsigned int, flags)
2503 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2506 int recvmsg_copy_msghdr(struct msghdr *msg,
2507 struct user_msghdr __user *umsg, unsigned flags,
2508 struct sockaddr __user **uaddr,
2513 if (MSG_CMSG_COMPAT & flags) {
2514 struct compat_msghdr __user *msg_compat;
2516 msg_compat = (struct compat_msghdr __user *) umsg;
2517 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2519 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2527 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2528 struct user_msghdr __user *msg,
2529 struct sockaddr __user *uaddr,
2530 unsigned int flags, int nosec)
2532 struct compat_msghdr __user *msg_compat =
2533 (struct compat_msghdr __user *) msg;
2534 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2535 struct sockaddr_storage addr;
2536 unsigned long cmsg_ptr;
2540 msg_sys->msg_name = &addr;
2541 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2542 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2544 /* We assume all kernel code knows the size of sockaddr_storage */
2545 msg_sys->msg_namelen = 0;
2547 if (sock->file->f_flags & O_NONBLOCK)
2548 flags |= MSG_DONTWAIT;
2550 if (unlikely(nosec))
2551 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2553 err = sock_recvmsg(sock, msg_sys, flags);
2559 if (uaddr != NULL) {
2560 err = move_addr_to_user(&addr,
2561 msg_sys->msg_namelen, uaddr,
2566 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2570 if (MSG_CMSG_COMPAT & flags)
2571 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2572 &msg_compat->msg_controllen);
2574 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2575 &msg->msg_controllen);
2583 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2584 struct msghdr *msg_sys, unsigned int flags, int nosec)
2586 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2587 /* user mode address pointers */
2588 struct sockaddr __user *uaddr;
2591 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2595 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2601 * BSD recvmsg interface
2604 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2605 struct user_msghdr __user *umsg,
2606 struct sockaddr __user *uaddr, unsigned int flags)
2608 /* disallow ancillary data requests from this path */
2609 if (msg->msg_control || msg->msg_controllen)
2612 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2615 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2616 bool forbid_cmsg_compat)
2618 int fput_needed, err;
2619 struct msghdr msg_sys;
2620 struct socket *sock;
2622 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2625 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2629 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2631 fput_light(sock->file, fput_needed);
2636 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2637 unsigned int, flags)
2639 return __sys_recvmsg(fd, msg, flags, true);
2643 * Linux recvmmsg interface
2646 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2647 unsigned int vlen, unsigned int flags,
2648 struct timespec64 *timeout)
2650 int fput_needed, err, datagrams;
2651 struct socket *sock;
2652 struct mmsghdr __user *entry;
2653 struct compat_mmsghdr __user *compat_entry;
2654 struct msghdr msg_sys;
2655 struct timespec64 end_time;
2656 struct timespec64 timeout64;
2659 poll_select_set_timeout(&end_time, timeout->tv_sec,
2665 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2669 if (likely(!(flags & MSG_ERRQUEUE))) {
2670 err = sock_error(sock->sk);
2678 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2680 while (datagrams < vlen) {
2682 * No need to ask LSM for more than the first datagram.
2684 if (MSG_CMSG_COMPAT & flags) {
2685 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2686 &msg_sys, flags & ~MSG_WAITFORONE,
2690 err = __put_user(err, &compat_entry->msg_len);
2693 err = ___sys_recvmsg(sock,
2694 (struct user_msghdr __user *)entry,
2695 &msg_sys, flags & ~MSG_WAITFORONE,
2699 err = put_user(err, &entry->msg_len);
2707 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2708 if (flags & MSG_WAITFORONE)
2709 flags |= MSG_DONTWAIT;
2712 ktime_get_ts64(&timeout64);
2713 *timeout = timespec64_sub(end_time, timeout64);
2714 if (timeout->tv_sec < 0) {
2715 timeout->tv_sec = timeout->tv_nsec = 0;
2719 /* Timeout, return less than vlen datagrams */
2720 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2724 /* Out of band data, return right away */
2725 if (msg_sys.msg_flags & MSG_OOB)
2733 if (datagrams == 0) {
2739 * We may return less entries than requested (vlen) if the
2740 * sock is non block and there aren't enough datagrams...
2742 if (err != -EAGAIN) {
2744 * ... or if recvmsg returns an error after we
2745 * received some datagrams, where we record the
2746 * error to return on the next call or if the
2747 * app asks about it using getsockopt(SO_ERROR).
2749 sock->sk->sk_err = -err;
2752 fput_light(sock->file, fput_needed);
2757 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2758 unsigned int vlen, unsigned int flags,
2759 struct __kernel_timespec __user *timeout,
2760 struct old_timespec32 __user *timeout32)
2763 struct timespec64 timeout_sys;
2765 if (timeout && get_timespec64(&timeout_sys, timeout))
2768 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2771 if (!timeout && !timeout32)
2772 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2774 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2779 if (timeout && put_timespec64(&timeout_sys, timeout))
2780 datagrams = -EFAULT;
2782 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2783 datagrams = -EFAULT;
2788 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2789 unsigned int, vlen, unsigned int, flags,
2790 struct __kernel_timespec __user *, timeout)
2792 if (flags & MSG_CMSG_COMPAT)
2795 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2798 #ifdef CONFIG_COMPAT_32BIT_TIME
2799 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2800 unsigned int, vlen, unsigned int, flags,
2801 struct old_timespec32 __user *, timeout)
2803 if (flags & MSG_CMSG_COMPAT)
2806 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2810 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2811 /* Argument list sizes for sys_socketcall */
2812 #define AL(x) ((x) * sizeof(unsigned long))
2813 static const unsigned char nargs[21] = {
2814 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2815 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2816 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2823 * System call vectors.
2825 * Argument checking cleaned up. Saved 20% in size.
2826 * This function doesn't need to set the kernel lock because
2827 * it is set by the callees.
2830 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2832 unsigned long a[AUDITSC_ARGS];
2833 unsigned long a0, a1;
2837 if (call < 1 || call > SYS_SENDMMSG)
2839 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2842 if (len > sizeof(a))
2845 /* copy_from_user should be SMP safe. */
2846 if (copy_from_user(a, args, len))
2849 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2858 err = __sys_socket(a0, a1, a[2]);
2861 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2864 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2867 err = __sys_listen(a0, a1);
2870 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2871 (int __user *)a[2], 0);
2873 case SYS_GETSOCKNAME:
2875 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2876 (int __user *)a[2]);
2878 case SYS_GETPEERNAME:
2880 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2881 (int __user *)a[2]);
2883 case SYS_SOCKETPAIR:
2884 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2887 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2891 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2892 (struct sockaddr __user *)a[4], a[5]);
2895 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2899 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2900 (struct sockaddr __user *)a[4],
2901 (int __user *)a[5]);
2904 err = __sys_shutdown(a0, a1);
2906 case SYS_SETSOCKOPT:
2907 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2910 case SYS_GETSOCKOPT:
2912 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2913 (int __user *)a[4]);
2916 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2920 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2924 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2928 if (IS_ENABLED(CONFIG_64BIT))
2929 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2931 (struct __kernel_timespec __user *)a[4],
2934 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2936 (struct old_timespec32 __user *)a[4]);
2939 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2940 (int __user *)a[2], a[3]);
2949 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2952 * sock_register - add a socket protocol handler
2953 * @ops: description of protocol
2955 * This function is called by a protocol handler that wants to
2956 * advertise its address family, and have it linked into the
2957 * socket interface. The value ops->family corresponds to the
2958 * socket system call protocol family.
2960 int sock_register(const struct net_proto_family *ops)
2964 if (ops->family >= NPROTO) {
2965 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2969 spin_lock(&net_family_lock);
2970 if (rcu_dereference_protected(net_families[ops->family],
2971 lockdep_is_held(&net_family_lock)))
2974 rcu_assign_pointer(net_families[ops->family], ops);
2977 spin_unlock(&net_family_lock);
2979 pr_info("NET: Registered protocol family %d\n", ops->family);
2982 EXPORT_SYMBOL(sock_register);
2985 * sock_unregister - remove a protocol handler
2986 * @family: protocol family to remove
2988 * This function is called by a protocol handler that wants to
2989 * remove its address family, and have it unlinked from the
2990 * new socket creation.
2992 * If protocol handler is a module, then it can use module reference
2993 * counts to protect against new references. If protocol handler is not
2994 * a module then it needs to provide its own protection in
2995 * the ops->create routine.
2997 void sock_unregister(int family)
2999 BUG_ON(family < 0 || family >= NPROTO);
3001 spin_lock(&net_family_lock);
3002 RCU_INIT_POINTER(net_families[family], NULL);
3003 spin_unlock(&net_family_lock);
3007 pr_info("NET: Unregistered protocol family %d\n", family);
3009 EXPORT_SYMBOL(sock_unregister);
3011 bool sock_is_registered(int family)
3013 return family < NPROTO && rcu_access_pointer(net_families[family]);
3016 static int __init sock_init(void)
3020 * Initialize the network sysctl infrastructure.
3022 err = net_sysctl_init();
3027 * Initialize skbuff SLAB cache
3032 * Initialize the protocols module.
3037 err = register_filesystem(&sock_fs_type);
3040 sock_mnt = kern_mount(&sock_fs_type);
3041 if (IS_ERR(sock_mnt)) {
3042 err = PTR_ERR(sock_mnt);
3046 /* The real protocol initialization is performed in later initcalls.
3049 #ifdef CONFIG_NETFILTER
3050 err = netfilter_init();
3055 ptp_classifier_init();
3061 unregister_filesystem(&sock_fs_type);
3066 core_initcall(sock_init); /* early initcall */
3068 #ifdef CONFIG_PROC_FS
3069 void socket_seq_show(struct seq_file *seq)
3071 seq_printf(seq, "sockets: used %d\n",
3072 sock_inuse_get(seq->private));
3074 #endif /* CONFIG_PROC_FS */
3076 #ifdef CONFIG_COMPAT
3077 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3079 struct compat_ifconf ifc32;
3083 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3086 ifc.ifc_len = ifc32.ifc_len;
3087 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3090 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3095 ifc32.ifc_len = ifc.ifc_len;
3096 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3102 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3104 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3105 bool convert_in = false, convert_out = false;
3106 size_t buf_size = 0;
3107 struct ethtool_rxnfc __user *rxnfc = NULL;
3109 u32 rule_cnt = 0, actual_rule_cnt;
3114 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3117 compat_rxnfc = compat_ptr(data);
3119 if (get_user(ethcmd, &compat_rxnfc->cmd))
3122 /* Most ethtool structures are defined without padding.
3123 * Unfortunately struct ethtool_rxnfc is an exception.
3128 case ETHTOOL_GRXCLSRLALL:
3129 /* Buffer size is variable */
3130 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3132 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3134 buf_size += rule_cnt * sizeof(u32);
3136 case ETHTOOL_GRXRINGS:
3137 case ETHTOOL_GRXCLSRLCNT:
3138 case ETHTOOL_GRXCLSRULE:
3139 case ETHTOOL_SRXCLSRLINS:
3142 case ETHTOOL_SRXCLSRLDEL:
3143 buf_size += sizeof(struct ethtool_rxnfc);
3145 rxnfc = compat_alloc_user_space(buf_size);
3149 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3152 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3155 /* We expect there to be holes between fs.m_ext and
3156 * fs.ring_cookie and at the end of fs, but nowhere else.
3158 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3159 sizeof(compat_rxnfc->fs.m_ext) !=
3160 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3161 sizeof(rxnfc->fs.m_ext));
3163 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3164 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3165 offsetof(struct ethtool_rxnfc, fs.location) -
3166 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3168 if (copy_in_user(rxnfc, compat_rxnfc,
3169 (void __user *)(&rxnfc->fs.m_ext + 1) -
3170 (void __user *)rxnfc) ||
3171 copy_in_user(&rxnfc->fs.ring_cookie,
3172 &compat_rxnfc->fs.ring_cookie,
3173 (void __user *)(&rxnfc->fs.location + 1) -
3174 (void __user *)&rxnfc->fs.ring_cookie))
3176 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3177 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3179 } else if (copy_in_user(&rxnfc->rule_cnt,
3180 &compat_rxnfc->rule_cnt,
3181 sizeof(rxnfc->rule_cnt)))
3185 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3190 if (copy_in_user(compat_rxnfc, rxnfc,
3191 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3192 (const void __user *)rxnfc) ||
3193 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3194 &rxnfc->fs.ring_cookie,
3195 (const void __user *)(&rxnfc->fs.location + 1) -
3196 (const void __user *)&rxnfc->fs.ring_cookie) ||
3197 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3198 sizeof(rxnfc->rule_cnt)))
3201 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3202 /* As an optimisation, we only copy the actual
3203 * number of rules that the underlying
3204 * function returned. Since Mallory might
3205 * change the rule count in user memory, we
3206 * check that it is less than the rule count
3207 * originally given (as the user buffer size),
3208 * which has been range-checked.
3210 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3212 if (actual_rule_cnt < rule_cnt)
3213 rule_cnt = actual_rule_cnt;
3214 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3215 &rxnfc->rule_locs[0],
3216 rule_cnt * sizeof(u32)))
3224 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3226 compat_uptr_t uptr32;
3231 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3234 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3237 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3238 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3240 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3242 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3243 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3249 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3250 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3251 struct compat_ifreq __user *u_ifreq32)
3256 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3258 if (get_user(data32, &u_ifreq32->ifr_data))
3260 ifreq.ifr_data = compat_ptr(data32);
3262 return dev_ioctl(net, cmd, &ifreq, NULL);
3265 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3267 struct compat_ifreq __user *uifr32)
3269 struct ifreq __user *uifr;
3272 /* Handle the fact that while struct ifreq has the same *layout* on
3273 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3274 * which are handled elsewhere, it still has different *size* due to
3275 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3276 * resulting in struct ifreq being 32 and 40 bytes respectively).
3277 * As a result, if the struct happens to be at the end of a page and
3278 * the next page isn't readable/writable, we get a fault. To prevent
3279 * that, copy back and forth to the full size.
3282 uifr = compat_alloc_user_space(sizeof(*uifr));
3283 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3286 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3297 case SIOCGIFBRDADDR:
3298 case SIOCGIFDSTADDR:
3299 case SIOCGIFNETMASK:
3305 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3313 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3314 struct compat_ifreq __user *uifr32)
3317 struct compat_ifmap __user *uifmap32;
3320 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3321 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3322 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3323 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3324 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3325 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3326 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3327 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3331 err = dev_ioctl(net, cmd, &ifr, NULL);
3333 if (cmd == SIOCGIFMAP && !err) {
3334 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3335 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3336 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3337 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3338 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3339 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3340 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3349 struct sockaddr rt_dst; /* target address */
3350 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3351 struct sockaddr rt_genmask; /* target network mask (IP) */
3352 unsigned short rt_flags;
3355 unsigned char rt_tos;
3356 unsigned char rt_class;
3358 short rt_metric; /* +1 for binary compatibility! */
3359 /* char * */ u32 rt_dev; /* forcing the device at add */
3360 u32 rt_mtu; /* per route MTU/Window */
3361 u32 rt_window; /* Window clamping */
3362 unsigned short rt_irtt; /* Initial RTT */
3365 struct in6_rtmsg32 {
3366 struct in6_addr rtmsg_dst;
3367 struct in6_addr rtmsg_src;
3368 struct in6_addr rtmsg_gateway;
3378 static int routing_ioctl(struct net *net, struct socket *sock,
3379 unsigned int cmd, void __user *argp)
3383 struct in6_rtmsg r6;
3387 mm_segment_t old_fs = get_fs();
3389 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3390 struct in6_rtmsg32 __user *ur6 = argp;
3391 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3392 3 * sizeof(struct in6_addr));
3393 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3394 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3395 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3396 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3397 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3398 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3399 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3403 struct rtentry32 __user *ur4 = argp;
3404 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3405 3 * sizeof(struct sockaddr));
3406 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3407 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3408 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3409 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3410 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3411 ret |= get_user(rtdev, &(ur4->rt_dev));
3413 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3414 r4.rt_dev = (char __user __force *)devname;
3428 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3435 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3436 * for some operations; this forces use of the newer bridge-utils that
3437 * use compatible ioctls
3439 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3443 if (get_user(tmp, argp))
3445 if (tmp == BRCTL_GET_VERSION)
3446 return BRCTL_VERSION + 1;
3450 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3451 unsigned int cmd, unsigned long arg)
3453 void __user *argp = compat_ptr(arg);
3454 struct sock *sk = sock->sk;
3455 struct net *net = sock_net(sk);
3457 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3458 return compat_ifr_data_ioctl(net, cmd, argp);
3463 return old_bridge_ioctl(argp);
3465 return compat_dev_ifconf(net, argp);
3467 return ethtool_ioctl(net, argp);
3469 return compat_siocwandev(net, argp);
3472 return compat_sioc_ifmap(net, cmd, argp);
3475 return routing_ioctl(net, sock, cmd, argp);
3476 case SIOCGSTAMP_OLD:
3477 case SIOCGSTAMPNS_OLD:
3478 if (!sock->ops->gettstamp)
3479 return -ENOIOCTLCMD;
3480 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3481 !COMPAT_USE_64BIT_TIME);
3483 case SIOCBONDSLAVEINFOQUERY:
3484 case SIOCBONDINFOQUERY:
3487 return compat_ifr_data_ioctl(net, cmd, argp);
3500 case SIOCGSTAMP_NEW:
3501 case SIOCGSTAMPNS_NEW:
3502 return sock_ioctl(file, cmd, arg);
3519 case SIOCSIFHWBROADCAST:
3521 case SIOCGIFBRDADDR:
3522 case SIOCSIFBRDADDR:
3523 case SIOCGIFDSTADDR:
3524 case SIOCSIFDSTADDR:
3525 case SIOCGIFNETMASK:
3526 case SIOCSIFNETMASK:
3538 case SIOCBONDENSLAVE:
3539 case SIOCBONDRELEASE:
3540 case SIOCBONDSETHWADDR:
3541 case SIOCBONDCHANGEACTIVE:
3542 return compat_ifreq_ioctl(net, sock, cmd, argp);
3550 return sock_do_ioctl(net, sock, cmd, arg);
3553 return -ENOIOCTLCMD;
3556 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3559 struct socket *sock = file->private_data;
3560 int ret = -ENOIOCTLCMD;
3567 if (sock->ops->compat_ioctl)
3568 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3570 if (ret == -ENOIOCTLCMD &&
3571 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3572 ret = compat_wext_handle_ioctl(net, cmd, arg);
3574 if (ret == -ENOIOCTLCMD)
3575 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3582 * kernel_bind - bind an address to a socket (kernel space)
3585 * @addrlen: length of address
3587 * Returns 0 or an error.
3590 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3592 return sock->ops->bind(sock, addr, addrlen);
3594 EXPORT_SYMBOL(kernel_bind);
3597 * kernel_listen - move socket to listening state (kernel space)
3599 * @backlog: pending connections queue size
3601 * Returns 0 or an error.
3604 int kernel_listen(struct socket *sock, int backlog)
3606 return sock->ops->listen(sock, backlog);
3608 EXPORT_SYMBOL(kernel_listen);
3611 * kernel_accept - accept a connection (kernel space)
3612 * @sock: listening socket
3613 * @newsock: new connected socket
3616 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3617 * If it fails, @newsock is guaranteed to be %NULL.
3618 * Returns 0 or an error.
3621 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3623 struct sock *sk = sock->sk;
3626 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3631 err = sock->ops->accept(sock, *newsock, flags, true);
3633 sock_release(*newsock);
3638 (*newsock)->ops = sock->ops;
3639 __module_get((*newsock)->ops->owner);
3644 EXPORT_SYMBOL(kernel_accept);
3647 * kernel_connect - connect a socket (kernel space)
3650 * @addrlen: address length
3651 * @flags: flags (O_NONBLOCK, ...)
3653 * For datagram sockets, @addr is the addres to which datagrams are sent
3654 * by default, and the only address from which datagrams are received.
3655 * For stream sockets, attempts to connect to @addr.
3656 * Returns 0 or an error code.
3659 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3662 return sock->ops->connect(sock, addr, addrlen, flags);
3664 EXPORT_SYMBOL(kernel_connect);
3667 * kernel_getsockname - get the address which the socket is bound (kernel space)
3669 * @addr: address holder
3671 * Fills the @addr pointer with the address which the socket is bound.
3672 * Returns 0 or an error code.
3675 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3677 return sock->ops->getname(sock, addr, 0);
3679 EXPORT_SYMBOL(kernel_getsockname);
3682 * kernel_peername - get the address which the socket is connected (kernel space)
3684 * @addr: address holder
3686 * Fills the @addr pointer with the address which the socket is connected.
3687 * Returns 0 or an error code.
3690 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3692 return sock->ops->getname(sock, addr, 1);
3694 EXPORT_SYMBOL(kernel_getpeername);
3697 * kernel_getsockopt - get a socket option (kernel space)
3699 * @level: API level (SOL_SOCKET, ...)
3700 * @optname: option tag
3701 * @optval: option value
3702 * @optlen: option length
3704 * Assigns the option length to @optlen.
3705 * Returns 0 or an error.
3708 int kernel_getsockopt(struct socket *sock, int level, int optname,
3709 char *optval, int *optlen)
3711 mm_segment_t oldfs = get_fs();
3712 char __user *uoptval;
3713 int __user *uoptlen;
3716 uoptval = (char __user __force *) optval;
3717 uoptlen = (int __user __force *) optlen;
3720 if (level == SOL_SOCKET)
3721 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3723 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3728 EXPORT_SYMBOL(kernel_getsockopt);
3731 * kernel_setsockopt - set a socket option (kernel space)
3733 * @level: API level (SOL_SOCKET, ...)
3734 * @optname: option tag
3735 * @optval: option value
3736 * @optlen: option length
3738 * Returns 0 or an error.
3741 int kernel_setsockopt(struct socket *sock, int level, int optname,
3742 char *optval, unsigned int optlen)
3744 mm_segment_t oldfs = get_fs();
3745 char __user *uoptval;
3748 uoptval = (char __user __force *) optval;
3751 if (level == SOL_SOCKET)
3752 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3754 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3759 EXPORT_SYMBOL(kernel_setsockopt);
3762 * kernel_sendpage - send a &page through a socket (kernel space)
3765 * @offset: page offset
3766 * @size: total size in bytes
3767 * @flags: flags (MSG_DONTWAIT, ...)
3769 * Returns the total amount sent in bytes or an error.
3772 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3773 size_t size, int flags)
3775 if (sock->ops->sendpage)
3776 return sock->ops->sendpage(sock, page, offset, size, flags);
3778 return sock_no_sendpage(sock, page, offset, size, flags);
3780 EXPORT_SYMBOL(kernel_sendpage);
3783 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3786 * @offset: page offset
3787 * @size: total size in bytes
3788 * @flags: flags (MSG_DONTWAIT, ...)
3790 * Returns the total amount sent in bytes or an error.
3791 * Caller must hold @sk.
3794 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3795 size_t size, int flags)
3797 struct socket *sock = sk->sk_socket;
3799 if (sock->ops->sendpage_locked)
3800 return sock->ops->sendpage_locked(sk, page, offset, size,
3803 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3805 EXPORT_SYMBOL(kernel_sendpage_locked);
3808 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3810 * @how: connection part
3812 * Returns 0 or an error.
3815 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3817 return sock->ops->shutdown(sock, how);
3819 EXPORT_SYMBOL(kernel_sock_shutdown);
3822 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3825 * This routine returns the IP overhead imposed by a socket i.e.
3826 * the length of the underlying IP header, depending on whether
3827 * this is an IPv4 or IPv6 socket and the length from IP options turned
3828 * on at the socket. Assumes that the caller has a lock on the socket.
3831 u32 kernel_sock_ip_overhead(struct sock *sk)
3833 struct inet_sock *inet;
3834 struct ip_options_rcu *opt;
3836 #if IS_ENABLED(CONFIG_IPV6)
3837 struct ipv6_pinfo *np;
3838 struct ipv6_txoptions *optv6 = NULL;
3839 #endif /* IS_ENABLED(CONFIG_IPV6) */
3844 switch (sk->sk_family) {
3847 overhead += sizeof(struct iphdr);
3848 opt = rcu_dereference_protected(inet->inet_opt,
3849 sock_owned_by_user(sk));
3851 overhead += opt->opt.optlen;
3853 #if IS_ENABLED(CONFIG_IPV6)
3856 overhead += sizeof(struct ipv6hdr);
3858 optv6 = rcu_dereference_protected(np->opt,
3859 sock_owned_by_user(sk));
3861 overhead += (optv6->opt_flen + optv6->opt_nflen);
3863 #endif /* IS_ENABLED(CONFIG_IPV6) */
3864 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3868 EXPORT_SYMBOL(kernel_sock_ip_overhead);