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_vlan.h>
68 #include <linux/ptp_classify.h>
69 #include <linux/init.h>
70 #include <linux/poll.h>
71 #include <linux/cache.h>
72 #include <linux/module.h>
73 #include <linux/highmem.h>
74 #include <linux/mount.h>
75 #include <linux/pseudo_fs.h>
76 #include <linux/security.h>
77 #include <linux/syscalls.h>
78 #include <linux/compat.h>
79 #include <linux/kmod.h>
80 #include <linux/audit.h>
81 #include <linux/wireless.h>
82 #include <linux/nsproxy.h>
83 #include <linux/magic.h>
84 #include <linux/slab.h>
85 #include <linux/xattr.h>
86 #include <linux/nospec.h>
87 #include <linux/indirect_call_wrapper.h>
89 #include <linux/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
94 #include <net/cls_cgroup.h>
97 #include <linux/netfilter.h>
99 #include <linux/if_tun.h>
100 #include <linux/ipv6_route.h>
101 #include <linux/route.h>
102 #include <linux/termios.h>
103 #include <linux/sockios.h>
104 #include <net/busy_poll.h>
105 #include <linux/errqueue.h>
107 #ifdef CONFIG_NET_RX_BUSY_POLL
108 unsigned int sysctl_net_busy_read __read_mostly;
109 unsigned int sysctl_net_busy_poll __read_mostly;
112 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
113 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
114 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
116 static int sock_close(struct inode *inode, struct file *file);
117 static __poll_t sock_poll(struct file *file,
118 struct poll_table_struct *wait);
119 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
121 static long compat_sock_ioctl(struct file *file,
122 unsigned int cmd, unsigned long arg);
124 static int sock_fasync(int fd, struct file *filp, int on);
125 static ssize_t sock_sendpage(struct file *file, struct page *page,
126 int offset, size_t size, loff_t *ppos, int more);
127 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
128 struct pipe_inode_info *pipe, size_t len,
131 #ifdef CONFIG_PROC_FS
132 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
134 struct socket *sock = f->private_data;
136 if (sock->ops->show_fdinfo)
137 sock->ops->show_fdinfo(m, sock);
140 #define sock_show_fdinfo NULL
144 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
145 * in the operation structures but are done directly via the socketcall() multiplexor.
148 static const struct file_operations socket_file_ops = {
149 .owner = THIS_MODULE,
151 .read_iter = sock_read_iter,
152 .write_iter = sock_write_iter,
154 .unlocked_ioctl = sock_ioctl,
156 .compat_ioctl = compat_sock_ioctl,
159 .release = sock_close,
160 .fasync = sock_fasync,
161 .sendpage = sock_sendpage,
162 .splice_write = generic_splice_sendpage,
163 .splice_read = sock_splice_read,
164 .show_fdinfo = sock_show_fdinfo,
168 * The protocol list. Each protocol is registered in here.
171 static DEFINE_SPINLOCK(net_family_lock);
172 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
176 * Move socket addresses back and forth across the kernel/user
177 * divide and look after the messy bits.
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
191 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
193 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
197 if (copy_from_user(kaddr, uaddr, ulen))
199 return audit_sockaddr(ulen, kaddr);
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
219 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
220 void __user *uaddr, int __user *ulen)
225 BUG_ON(klen > sizeof(struct sockaddr_storage));
226 err = get_user(len, ulen);
234 if (audit_sockaddr(klen, kaddr))
236 if (copy_to_user(uaddr, kaddr, len))
240 * "fromlen shall refer to the value before truncation.."
243 return __put_user(klen, ulen);
246 static struct kmem_cache *sock_inode_cachep __ro_after_init;
248 static struct inode *sock_alloc_inode(struct super_block *sb)
250 struct socket_alloc *ei;
252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255 init_waitqueue_head(&ei->socket.wq.wait);
256 ei->socket.wq.fasync_list = NULL;
257 ei->socket.wq.flags = 0;
259 ei->socket.state = SS_UNCONNECTED;
260 ei->socket.flags = 0;
261 ei->socket.ops = NULL;
262 ei->socket.sk = NULL;
263 ei->socket.file = NULL;
265 return &ei->vfs_inode;
268 static void sock_free_inode(struct inode *inode)
270 struct socket_alloc *ei;
272 ei = container_of(inode, struct socket_alloc, vfs_inode);
273 kmem_cache_free(sock_inode_cachep, ei);
276 static void init_once(void *foo)
278 struct socket_alloc *ei = (struct socket_alloc *)foo;
280 inode_init_once(&ei->vfs_inode);
283 static void init_inodecache(void)
285 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
286 sizeof(struct socket_alloc),
288 (SLAB_HWCACHE_ALIGN |
289 SLAB_RECLAIM_ACCOUNT |
290 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
292 BUG_ON(sock_inode_cachep == NULL);
295 static const struct super_operations sockfs_ops = {
296 .alloc_inode = sock_alloc_inode,
297 .free_inode = sock_free_inode,
298 .statfs = simple_statfs,
302 * sockfs_dname() is called from d_path().
304 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
306 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
307 d_inode(dentry)->i_ino);
310 static const struct dentry_operations sockfs_dentry_operations = {
311 .d_dname = sockfs_dname,
314 static int sockfs_xattr_get(const struct xattr_handler *handler,
315 struct dentry *dentry, struct inode *inode,
316 const char *suffix, void *value, size_t size)
319 if (dentry->d_name.len + 1 > size)
321 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
323 return dentry->d_name.len + 1;
326 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
327 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
328 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
330 static const struct xattr_handler sockfs_xattr_handler = {
331 .name = XATTR_NAME_SOCKPROTONAME,
332 .get = sockfs_xattr_get,
335 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
336 struct dentry *dentry, struct inode *inode,
337 const char *suffix, const void *value,
338 size_t size, int flags)
340 /* Handled by LSM. */
344 static const struct xattr_handler sockfs_security_xattr_handler = {
345 .prefix = XATTR_SECURITY_PREFIX,
346 .set = sockfs_security_xattr_set,
349 static const struct xattr_handler *sockfs_xattr_handlers[] = {
350 &sockfs_xattr_handler,
351 &sockfs_security_xattr_handler,
355 static int sockfs_init_fs_context(struct fs_context *fc)
357 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
360 ctx->ops = &sockfs_ops;
361 ctx->dops = &sockfs_dentry_operations;
362 ctx->xattr = sockfs_xattr_handlers;
366 static struct vfsmount *sock_mnt __read_mostly;
368 static struct file_system_type sock_fs_type = {
370 .init_fs_context = sockfs_init_fs_context,
371 .kill_sb = kill_anon_super,
375 * Obtains the first available file descriptor and sets it up for use.
377 * These functions create file structures and maps them to fd space
378 * of the current process. On success it returns file descriptor
379 * and file struct implicitly stored in sock->file.
380 * Note that another thread may close file descriptor before we return
381 * from this function. We use the fact that now we do not refer
382 * to socket after mapping. If one day we will need it, this
383 * function will increment ref. count on file by 1.
385 * In any case returned fd MAY BE not valid!
386 * This race condition is unavoidable
387 * with shared fd spaces, we cannot solve it inside kernel,
388 * but we take care of internal coherence yet.
392 * sock_alloc_file - Bind a &socket to a &file
394 * @flags: file status flags
395 * @dname: protocol name
397 * Returns the &file bound with @sock, implicitly storing it
398 * in sock->file. If dname is %NULL, sets to "".
399 * On failure the return is a ERR pointer (see linux/err.h).
400 * This function uses GFP_KERNEL internally.
403 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
408 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
410 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
411 O_RDWR | (flags & O_NONBLOCK),
419 file->private_data = sock;
420 stream_open(SOCK_INODE(sock), file);
423 EXPORT_SYMBOL(sock_alloc_file);
425 static int sock_map_fd(struct socket *sock, int flags)
427 struct file *newfile;
428 int fd = get_unused_fd_flags(flags);
429 if (unlikely(fd < 0)) {
434 newfile = sock_alloc_file(sock, flags, NULL);
435 if (!IS_ERR(newfile)) {
436 fd_install(fd, newfile);
441 return PTR_ERR(newfile);
445 * sock_from_file - Return the &socket bounded to @file.
447 * @err: pointer to an error code return
449 * On failure returns %NULL and assigns -ENOTSOCK to @err.
452 struct socket *sock_from_file(struct file *file, int *err)
454 if (file->f_op == &socket_file_ops)
455 return file->private_data; /* set in sock_map_fd */
460 EXPORT_SYMBOL(sock_from_file);
463 * sockfd_lookup - Go from a file number to its socket slot
465 * @err: pointer to an error code return
467 * The file handle passed in is locked and the socket it is bound
468 * to is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
472 * On a success the socket object pointer is returned.
475 struct socket *sockfd_lookup(int fd, int *err)
486 sock = sock_from_file(file, err);
491 EXPORT_SYMBOL(sockfd_lookup);
493 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
495 struct fd f = fdget(fd);
500 sock = sock_from_file(f.file, err);
502 *fput_needed = f.flags & FDPUT_FPUT;
510 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
516 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
526 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
531 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
538 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
540 int err = simple_setattr(dentry, iattr);
542 if (!err && (iattr->ia_valid & ATTR_UID)) {
543 struct socket *sock = SOCKET_I(d_inode(dentry));
546 sock->sk->sk_uid = iattr->ia_uid;
554 static const struct inode_operations sockfs_inode_ops = {
555 .listxattr = sockfs_listxattr,
556 .setattr = sockfs_setattr,
560 * sock_alloc - allocate a socket
562 * Allocate a new inode and socket object. The two are bound together
563 * and initialised. The socket is then returned. If we are out of inodes
564 * NULL is returned. This functions uses GFP_KERNEL internally.
567 struct socket *sock_alloc(void)
572 inode = new_inode_pseudo(sock_mnt->mnt_sb);
576 sock = SOCKET_I(inode);
578 inode->i_ino = get_next_ino();
579 inode->i_mode = S_IFSOCK | S_IRWXUGO;
580 inode->i_uid = current_fsuid();
581 inode->i_gid = current_fsgid();
582 inode->i_op = &sockfs_inode_ops;
586 EXPORT_SYMBOL(sock_alloc);
588 static void __sock_release(struct socket *sock, struct inode *inode)
591 struct module *owner = sock->ops->owner;
595 sock->ops->release(sock);
603 if (sock->wq.fasync_list)
604 pr_err("%s: fasync list not empty!\n", __func__);
607 iput(SOCK_INODE(sock));
614 * sock_release - close a socket
615 * @sock: socket to close
617 * The socket is released from the protocol stack if it has a release
618 * callback, and the inode is then released if the socket is bound to
619 * an inode not a file.
621 void sock_release(struct socket *sock)
623 __sock_release(sock, NULL);
625 EXPORT_SYMBOL(sock_release);
627 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
629 u8 flags = *tx_flags;
631 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
632 flags |= SKBTX_HW_TSTAMP;
634 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
635 flags |= SKBTX_SW_TSTAMP;
637 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
638 flags |= SKBTX_SCHED_TSTAMP;
642 EXPORT_SYMBOL(__sock_tx_timestamp);
644 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
646 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
648 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
650 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
651 inet_sendmsg, sock, msg,
653 BUG_ON(ret == -EIOCBQUEUED);
658 * sock_sendmsg - send a message through @sock
660 * @msg: message to send
662 * Sends @msg through @sock, passing through LSM.
663 * Returns the number of bytes sent, or an error code.
665 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
667 int err = security_socket_sendmsg(sock, msg,
670 return err ?: sock_sendmsg_nosec(sock, msg);
672 EXPORT_SYMBOL(sock_sendmsg);
675 * kernel_sendmsg - send a message through @sock (kernel-space)
677 * @msg: message header
679 * @num: vec array length
680 * @size: total message data size
682 * Builds the message data with @vec and sends it through @sock.
683 * Returns the number of bytes sent, or an error code.
686 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
687 struct kvec *vec, size_t num, size_t size)
689 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
690 return sock_sendmsg(sock, msg);
692 EXPORT_SYMBOL(kernel_sendmsg);
695 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
697 * @msg: message header
698 * @vec: output s/g array
699 * @num: output s/g array length
700 * @size: total message data size
702 * Builds the message data with @vec and sends it through @sock.
703 * Returns the number of bytes sent, or an error code.
704 * Caller must hold @sk.
707 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
708 struct kvec *vec, size_t num, size_t size)
710 struct socket *sock = sk->sk_socket;
712 if (!sock->ops->sendmsg_locked)
713 return sock_no_sendmsg_locked(sk, msg, size);
715 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
717 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
719 EXPORT_SYMBOL(kernel_sendmsg_locked);
721 static bool skb_is_err_queue(const struct sk_buff *skb)
723 /* pkt_type of skbs enqueued on the error queue are set to
724 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
725 * in recvmsg, since skbs received on a local socket will never
726 * have a pkt_type of PACKET_OUTGOING.
728 return skb->pkt_type == PACKET_OUTGOING;
731 /* On transmit, software and hardware timestamps are returned independently.
732 * As the two skb clones share the hardware timestamp, which may be updated
733 * before the software timestamp is received, a hardware TX timestamp may be
734 * returned only if there is no software TX timestamp. Ignore false software
735 * timestamps, which may be made in the __sock_recv_timestamp() call when the
736 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
737 * hardware timestamp.
739 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
741 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
744 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
746 struct scm_ts_pktinfo ts_pktinfo;
747 struct net_device *orig_dev;
749 if (!skb_mac_header_was_set(skb))
752 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
755 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
757 ts_pktinfo.if_index = orig_dev->ifindex;
760 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
761 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
762 sizeof(ts_pktinfo), &ts_pktinfo);
766 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
768 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
771 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
772 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
773 struct scm_timestamping_internal tss;
775 int empty = 1, false_tstamp = 0;
776 struct skb_shared_hwtstamps *shhwtstamps =
779 /* Race occurred between timestamp enabling and packet
780 receiving. Fill in the current time for now. */
781 if (need_software_tstamp && skb->tstamp == 0) {
782 __net_timestamp(skb);
786 if (need_software_tstamp) {
787 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
789 struct __kernel_sock_timeval tv;
791 skb_get_new_timestamp(skb, &tv);
792 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
795 struct __kernel_old_timeval tv;
797 skb_get_timestamp(skb, &tv);
798 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
803 struct __kernel_timespec ts;
805 skb_get_new_timestampns(skb, &ts);
806 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
809 struct __kernel_old_timespec ts;
811 skb_get_timestampns(skb, &ts);
812 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
818 memset(&tss, 0, sizeof(tss));
819 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
820 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
823 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
824 !skb_is_swtx_tstamp(skb, false_tstamp) &&
825 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
827 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
828 !skb_is_err_queue(skb))
829 put_ts_pktinfo(msg, skb);
832 if (sock_flag(sk, SOCK_TSTAMP_NEW))
833 put_cmsg_scm_timestamping64(msg, &tss);
835 put_cmsg_scm_timestamping(msg, &tss);
837 if (skb_is_err_queue(skb) && skb->len &&
838 SKB_EXT_ERR(skb)->opt_stats)
839 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
840 skb->len, skb->data);
843 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
845 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
850 if (!sock_flag(sk, SOCK_WIFI_STATUS))
852 if (!skb->wifi_acked_valid)
855 ack = skb->wifi_acked;
857 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
859 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
861 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
864 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
865 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
866 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
869 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
872 sock_recv_timestamp(msg, sk, skb);
873 sock_recv_drops(msg, sk, skb);
875 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
877 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
879 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
881 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
884 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
885 inet_recvmsg, sock, msg, msg_data_left(msg),
890 * sock_recvmsg - receive a message from @sock
892 * @msg: message to receive
893 * @flags: message flags
895 * Receives @msg from @sock, passing through LSM. Returns the total number
896 * of bytes received, or an error.
898 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
900 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
902 return err ?: sock_recvmsg_nosec(sock, msg, flags);
904 EXPORT_SYMBOL(sock_recvmsg);
907 * kernel_recvmsg - Receive a message from a socket (kernel space)
908 * @sock: The socket to receive the message from
909 * @msg: Received message
910 * @vec: Input s/g array for message data
911 * @num: Size of input s/g array
912 * @size: Number of bytes to read
913 * @flags: Message flags (MSG_DONTWAIT, etc...)
915 * On return the msg structure contains the scatter/gather array passed in the
916 * vec argument. The array is modified so that it consists of the unfilled
917 * portion of the original array.
919 * The returned value is the total number of bytes received, or an error.
922 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
923 struct kvec *vec, size_t num, size_t size, int flags)
925 msg->msg_control_is_user = false;
926 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
927 return sock_recvmsg(sock, msg, flags);
929 EXPORT_SYMBOL(kernel_recvmsg);
931 static ssize_t sock_sendpage(struct file *file, struct page *page,
932 int offset, size_t size, loff_t *ppos, int more)
937 sock = file->private_data;
939 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
940 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
943 return kernel_sendpage(sock, page, offset, size, flags);
946 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
947 struct pipe_inode_info *pipe, size_t len,
950 struct socket *sock = file->private_data;
952 if (unlikely(!sock->ops->splice_read))
953 return generic_file_splice_read(file, ppos, pipe, len, flags);
955 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
958 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
960 struct file *file = iocb->ki_filp;
961 struct socket *sock = file->private_data;
962 struct msghdr msg = {.msg_iter = *to,
966 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
967 msg.msg_flags = MSG_DONTWAIT;
969 if (iocb->ki_pos != 0)
972 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
975 res = sock_recvmsg(sock, &msg, msg.msg_flags);
980 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
982 struct file *file = iocb->ki_filp;
983 struct socket *sock = file->private_data;
984 struct msghdr msg = {.msg_iter = *from,
988 if (iocb->ki_pos != 0)
991 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
992 msg.msg_flags = MSG_DONTWAIT;
994 if (sock->type == SOCK_SEQPACKET)
995 msg.msg_flags |= MSG_EOR;
997 res = sock_sendmsg(sock, &msg);
998 *from = msg.msg_iter;
1003 * Atomic setting of ioctl hooks to avoid race
1004 * with module unload.
1007 static DEFINE_MUTEX(br_ioctl_mutex);
1008 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1010 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1012 mutex_lock(&br_ioctl_mutex);
1013 br_ioctl_hook = hook;
1014 mutex_unlock(&br_ioctl_mutex);
1016 EXPORT_SYMBOL(brioctl_set);
1018 static DEFINE_MUTEX(vlan_ioctl_mutex);
1019 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1021 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1023 mutex_lock(&vlan_ioctl_mutex);
1024 vlan_ioctl_hook = hook;
1025 mutex_unlock(&vlan_ioctl_mutex);
1027 EXPORT_SYMBOL(vlan_ioctl_set);
1029 static long sock_do_ioctl(struct net *net, struct socket *sock,
1030 unsigned int cmd, unsigned long arg)
1033 void __user *argp = (void __user *)arg;
1035 err = sock->ops->ioctl(sock, cmd, arg);
1038 * If this ioctl is unknown try to hand it down
1039 * to the NIC driver.
1041 if (err != -ENOIOCTLCMD)
1044 if (cmd == SIOCGIFCONF) {
1046 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1049 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1051 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1056 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1058 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1059 if (!err && need_copyout)
1060 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1067 * With an ioctl, arg may well be a user mode pointer, but we don't know
1068 * what to do with it - that's up to the protocol still.
1072 * get_net_ns - increment the refcount of the network namespace
1073 * @ns: common namespace (net)
1075 * Returns the net's common namespace.
1078 struct ns_common *get_net_ns(struct ns_common *ns)
1080 return &get_net(container_of(ns, struct net, ns))->ns;
1082 EXPORT_SYMBOL_GPL(get_net_ns);
1084 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1086 struct socket *sock;
1088 void __user *argp = (void __user *)arg;
1092 sock = file->private_data;
1095 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1098 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1100 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1101 if (!err && need_copyout)
1102 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1105 #ifdef CONFIG_WEXT_CORE
1106 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1107 err = wext_handle_ioctl(net, cmd, argp);
1114 if (get_user(pid, (int __user *)argp))
1116 err = f_setown(sock->file, pid, 1);
1120 err = put_user(f_getown(sock->file),
1121 (int __user *)argp);
1129 request_module("bridge");
1131 mutex_lock(&br_ioctl_mutex);
1133 err = br_ioctl_hook(net, cmd, argp);
1134 mutex_unlock(&br_ioctl_mutex);
1139 if (!vlan_ioctl_hook)
1140 request_module("8021q");
1142 mutex_lock(&vlan_ioctl_mutex);
1143 if (vlan_ioctl_hook)
1144 err = vlan_ioctl_hook(net, argp);
1145 mutex_unlock(&vlan_ioctl_mutex);
1149 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1152 err = open_related_ns(&net->ns, get_net_ns);
1154 case SIOCGSTAMP_OLD:
1155 case SIOCGSTAMPNS_OLD:
1156 if (!sock->ops->gettstamp) {
1160 err = sock->ops->gettstamp(sock, argp,
1161 cmd == SIOCGSTAMP_OLD,
1162 !IS_ENABLED(CONFIG_64BIT));
1164 case SIOCGSTAMP_NEW:
1165 case SIOCGSTAMPNS_NEW:
1166 if (!sock->ops->gettstamp) {
1170 err = sock->ops->gettstamp(sock, argp,
1171 cmd == SIOCGSTAMP_NEW,
1175 err = sock_do_ioctl(net, sock, cmd, arg);
1182 * sock_create_lite - creates a socket
1183 * @family: protocol family (AF_INET, ...)
1184 * @type: communication type (SOCK_STREAM, ...)
1185 * @protocol: protocol (0, ...)
1188 * Creates a new socket and assigns it to @res, passing through LSM.
1189 * The new socket initialization is not complete, see kernel_accept().
1190 * Returns 0 or an error. On failure @res is set to %NULL.
1191 * This function internally uses GFP_KERNEL.
1194 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1197 struct socket *sock = NULL;
1199 err = security_socket_create(family, type, protocol, 1);
1203 sock = sock_alloc();
1210 err = security_socket_post_create(sock, family, type, protocol, 1);
1222 EXPORT_SYMBOL(sock_create_lite);
1224 /* No kernel lock held - perfect */
1225 static __poll_t sock_poll(struct file *file, poll_table *wait)
1227 struct socket *sock = file->private_data;
1228 __poll_t events = poll_requested_events(wait), flag = 0;
1230 if (!sock->ops->poll)
1233 if (sk_can_busy_loop(sock->sk)) {
1234 /* poll once if requested by the syscall */
1235 if (events & POLL_BUSY_LOOP)
1236 sk_busy_loop(sock->sk, 1);
1238 /* if this socket can poll_ll, tell the system call */
1239 flag = POLL_BUSY_LOOP;
1242 return sock->ops->poll(file, sock, wait) | flag;
1245 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1247 struct socket *sock = file->private_data;
1249 return sock->ops->mmap(file, sock, vma);
1252 static int sock_close(struct inode *inode, struct file *filp)
1254 __sock_release(SOCKET_I(inode), inode);
1259 * Update the socket async list
1261 * Fasync_list locking strategy.
1263 * 1. fasync_list is modified only under process context socket lock
1264 * i.e. under semaphore.
1265 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1266 * or under socket lock
1269 static int sock_fasync(int fd, struct file *filp, int on)
1271 struct socket *sock = filp->private_data;
1272 struct sock *sk = sock->sk;
1273 struct socket_wq *wq = &sock->wq;
1279 fasync_helper(fd, filp, on, &wq->fasync_list);
1281 if (!wq->fasync_list)
1282 sock_reset_flag(sk, SOCK_FASYNC);
1284 sock_set_flag(sk, SOCK_FASYNC);
1290 /* This function may be called only under rcu_lock */
1292 int sock_wake_async(struct socket_wq *wq, int how, int band)
1294 if (!wq || !wq->fasync_list)
1298 case SOCK_WAKE_WAITD:
1299 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1302 case SOCK_WAKE_SPACE:
1303 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1308 kill_fasync(&wq->fasync_list, SIGIO, band);
1311 kill_fasync(&wq->fasync_list, SIGURG, band);
1316 EXPORT_SYMBOL(sock_wake_async);
1319 * __sock_create - creates a socket
1320 * @net: net namespace
1321 * @family: protocol family (AF_INET, ...)
1322 * @type: communication type (SOCK_STREAM, ...)
1323 * @protocol: protocol (0, ...)
1325 * @kern: boolean for kernel space sockets
1327 * Creates a new socket and assigns it to @res, passing through LSM.
1328 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1329 * be set to true if the socket resides in kernel space.
1330 * This function internally uses GFP_KERNEL.
1333 int __sock_create(struct net *net, int family, int type, int protocol,
1334 struct socket **res, int kern)
1337 struct socket *sock;
1338 const struct net_proto_family *pf;
1341 * Check protocol is in range
1343 if (family < 0 || family >= NPROTO)
1344 return -EAFNOSUPPORT;
1345 if (type < 0 || type >= SOCK_MAX)
1350 This uglymoron is moved from INET layer to here to avoid
1351 deadlock in module load.
1353 if (family == PF_INET && type == SOCK_PACKET) {
1354 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1359 err = security_socket_create(family, type, protocol, kern);
1364 * Allocate the socket and allow the family to set things up. if
1365 * the protocol is 0, the family is instructed to select an appropriate
1368 sock = sock_alloc();
1370 net_warn_ratelimited("socket: no more sockets\n");
1371 return -ENFILE; /* Not exactly a match, but its the
1372 closest posix thing */
1377 #ifdef CONFIG_MODULES
1378 /* Attempt to load a protocol module if the find failed.
1380 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1381 * requested real, full-featured networking support upon configuration.
1382 * Otherwise module support will break!
1384 if (rcu_access_pointer(net_families[family]) == NULL)
1385 request_module("net-pf-%d", family);
1389 pf = rcu_dereference(net_families[family]);
1390 err = -EAFNOSUPPORT;
1395 * We will call the ->create function, that possibly is in a loadable
1396 * module, so we have to bump that loadable module refcnt first.
1398 if (!try_module_get(pf->owner))
1401 /* Now protected by module ref count */
1404 err = pf->create(net, sock, protocol, kern);
1406 goto out_module_put;
1409 * Now to bump the refcnt of the [loadable] module that owns this
1410 * socket at sock_release time we decrement its refcnt.
1412 if (!try_module_get(sock->ops->owner))
1413 goto out_module_busy;
1416 * Now that we're done with the ->create function, the [loadable]
1417 * module can have its refcnt decremented
1419 module_put(pf->owner);
1420 err = security_socket_post_create(sock, family, type, protocol, kern);
1422 goto out_sock_release;
1428 err = -EAFNOSUPPORT;
1431 module_put(pf->owner);
1438 goto out_sock_release;
1440 EXPORT_SYMBOL(__sock_create);
1443 * sock_create - creates a socket
1444 * @family: protocol family (AF_INET, ...)
1445 * @type: communication type (SOCK_STREAM, ...)
1446 * @protocol: protocol (0, ...)
1449 * A wrapper around __sock_create().
1450 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1453 int sock_create(int family, int type, int protocol, struct socket **res)
1455 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1457 EXPORT_SYMBOL(sock_create);
1460 * sock_create_kern - creates a socket (kernel space)
1461 * @net: net namespace
1462 * @family: protocol family (AF_INET, ...)
1463 * @type: communication type (SOCK_STREAM, ...)
1464 * @protocol: protocol (0, ...)
1467 * A wrapper around __sock_create().
1468 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1471 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1473 return __sock_create(net, family, type, protocol, res, 1);
1475 EXPORT_SYMBOL(sock_create_kern);
1477 int __sys_socket(int family, int type, int protocol)
1480 struct socket *sock;
1483 /* Check the SOCK_* constants for consistency. */
1484 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1485 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1486 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1487 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1489 flags = type & ~SOCK_TYPE_MASK;
1490 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1492 type &= SOCK_TYPE_MASK;
1494 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1495 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1497 retval = sock_create(family, type, protocol, &sock);
1501 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1504 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1506 return __sys_socket(family, type, protocol);
1510 * Create a pair of connected sockets.
1513 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1515 struct socket *sock1, *sock2;
1517 struct file *newfile1, *newfile2;
1520 flags = type & ~SOCK_TYPE_MASK;
1521 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1523 type &= SOCK_TYPE_MASK;
1525 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1526 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1529 * reserve descriptors and make sure we won't fail
1530 * to return them to userland.
1532 fd1 = get_unused_fd_flags(flags);
1533 if (unlikely(fd1 < 0))
1536 fd2 = get_unused_fd_flags(flags);
1537 if (unlikely(fd2 < 0)) {
1542 err = put_user(fd1, &usockvec[0]);
1546 err = put_user(fd2, &usockvec[1]);
1551 * Obtain the first socket and check if the underlying protocol
1552 * supports the socketpair call.
1555 err = sock_create(family, type, protocol, &sock1);
1556 if (unlikely(err < 0))
1559 err = sock_create(family, type, protocol, &sock2);
1560 if (unlikely(err < 0)) {
1561 sock_release(sock1);
1565 err = security_socket_socketpair(sock1, sock2);
1566 if (unlikely(err)) {
1567 sock_release(sock2);
1568 sock_release(sock1);
1572 err = sock1->ops->socketpair(sock1, sock2);
1573 if (unlikely(err < 0)) {
1574 sock_release(sock2);
1575 sock_release(sock1);
1579 newfile1 = sock_alloc_file(sock1, flags, NULL);
1580 if (IS_ERR(newfile1)) {
1581 err = PTR_ERR(newfile1);
1582 sock_release(sock2);
1586 newfile2 = sock_alloc_file(sock2, flags, NULL);
1587 if (IS_ERR(newfile2)) {
1588 err = PTR_ERR(newfile2);
1593 audit_fd_pair(fd1, fd2);
1595 fd_install(fd1, newfile1);
1596 fd_install(fd2, newfile2);
1605 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1606 int __user *, usockvec)
1608 return __sys_socketpair(family, type, protocol, usockvec);
1612 * Bind a name to a socket. Nothing much to do here since it's
1613 * the protocol's responsibility to handle the local address.
1615 * We move the socket address to kernel space before we call
1616 * the protocol layer (having also checked the address is ok).
1619 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1621 struct socket *sock;
1622 struct sockaddr_storage address;
1623 int err, fput_needed;
1625 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1627 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1629 err = security_socket_bind(sock,
1630 (struct sockaddr *)&address,
1633 err = sock->ops->bind(sock,
1637 fput_light(sock->file, fput_needed);
1642 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1644 return __sys_bind(fd, umyaddr, addrlen);
1648 * Perform a listen. Basically, we allow the protocol to do anything
1649 * necessary for a listen, and if that works, we mark the socket as
1650 * ready for listening.
1653 int __sys_listen(int fd, int backlog)
1655 struct socket *sock;
1656 int err, fput_needed;
1659 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1661 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1662 if ((unsigned int)backlog > somaxconn)
1663 backlog = somaxconn;
1665 err = security_socket_listen(sock, backlog);
1667 err = sock->ops->listen(sock, backlog);
1669 fput_light(sock->file, fput_needed);
1674 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1676 return __sys_listen(fd, backlog);
1679 int __sys_accept4_file(struct file *file, unsigned file_flags,
1680 struct sockaddr __user *upeer_sockaddr,
1681 int __user *upeer_addrlen, int flags,
1682 unsigned long nofile)
1684 struct socket *sock, *newsock;
1685 struct file *newfile;
1686 int err, len, newfd;
1687 struct sockaddr_storage address;
1689 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1692 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1693 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1695 sock = sock_from_file(file, &err);
1700 newsock = sock_alloc();
1704 newsock->type = sock->type;
1705 newsock->ops = sock->ops;
1708 * We don't need try_module_get here, as the listening socket (sock)
1709 * has the protocol module (sock->ops->owner) held.
1711 __module_get(newsock->ops->owner);
1713 newfd = __get_unused_fd_flags(flags, nofile);
1714 if (unlikely(newfd < 0)) {
1716 sock_release(newsock);
1719 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1720 if (IS_ERR(newfile)) {
1721 err = PTR_ERR(newfile);
1722 put_unused_fd(newfd);
1726 err = security_socket_accept(sock, newsock);
1730 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1735 if (upeer_sockaddr) {
1736 len = newsock->ops->getname(newsock,
1737 (struct sockaddr *)&address, 2);
1739 err = -ECONNABORTED;
1742 err = move_addr_to_user(&address,
1743 len, upeer_sockaddr, upeer_addrlen);
1748 /* File flags are not inherited via accept() unlike another OSes. */
1750 fd_install(newfd, newfile);
1756 put_unused_fd(newfd);
1762 * For accept, we attempt to create a new socket, set up the link
1763 * with the client, wake up the client, then return the new
1764 * connected fd. We collect the address of the connector in kernel
1765 * space and move it to user at the very end. This is unclean because
1766 * we open the socket then return an error.
1768 * 1003.1g adds the ability to recvmsg() to query connection pending
1769 * status to recvmsg. We need to add that support in a way thats
1770 * clean when we restructure accept also.
1773 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1774 int __user *upeer_addrlen, int flags)
1781 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1782 upeer_addrlen, flags,
1783 rlimit(RLIMIT_NOFILE));
1790 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1791 int __user *, upeer_addrlen, int, flags)
1793 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1796 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1797 int __user *, upeer_addrlen)
1799 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1803 * Attempt to connect to a socket with the server address. The address
1804 * is in user space so we verify it is OK and move it to kernel space.
1806 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1809 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1810 * other SEQPACKET protocols that take time to connect() as it doesn't
1811 * include the -EINPROGRESS status for such sockets.
1814 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1815 int addrlen, int file_flags)
1817 struct socket *sock;
1820 sock = sock_from_file(file, &err);
1825 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1829 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1830 sock->file->f_flags | file_flags);
1835 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1842 struct sockaddr_storage address;
1844 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1846 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1853 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1856 return __sys_connect(fd, uservaddr, addrlen);
1860 * Get the local address ('name') of a socket object. Move the obtained
1861 * name to user space.
1864 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1865 int __user *usockaddr_len)
1867 struct socket *sock;
1868 struct sockaddr_storage address;
1869 int err, fput_needed;
1871 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1875 err = security_socket_getsockname(sock);
1879 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1882 /* "err" is actually length in this case */
1883 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1886 fput_light(sock->file, fput_needed);
1891 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1892 int __user *, usockaddr_len)
1894 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1898 * Get the remote address ('name') of a socket object. Move the obtained
1899 * name to user space.
1902 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1903 int __user *usockaddr_len)
1905 struct socket *sock;
1906 struct sockaddr_storage address;
1907 int err, fput_needed;
1909 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1911 err = security_socket_getpeername(sock);
1913 fput_light(sock->file, fput_needed);
1917 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1919 /* "err" is actually length in this case */
1920 err = move_addr_to_user(&address, err, usockaddr,
1922 fput_light(sock->file, fput_needed);
1927 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1928 int __user *, usockaddr_len)
1930 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1934 * Send a datagram to a given address. We move the address into kernel
1935 * space and check the user space data area is readable before invoking
1938 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1939 struct sockaddr __user *addr, int addr_len)
1941 struct socket *sock;
1942 struct sockaddr_storage address;
1948 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1951 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1955 msg.msg_name = NULL;
1956 msg.msg_control = NULL;
1957 msg.msg_controllen = 0;
1958 msg.msg_namelen = 0;
1960 err = move_addr_to_kernel(addr, addr_len, &address);
1963 msg.msg_name = (struct sockaddr *)&address;
1964 msg.msg_namelen = addr_len;
1966 if (sock->file->f_flags & O_NONBLOCK)
1967 flags |= MSG_DONTWAIT;
1968 msg.msg_flags = flags;
1969 err = sock_sendmsg(sock, &msg);
1972 fput_light(sock->file, fput_needed);
1977 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1978 unsigned int, flags, struct sockaddr __user *, addr,
1981 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1985 * Send a datagram down a socket.
1988 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1989 unsigned int, flags)
1991 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1995 * Receive a frame from the socket and optionally record the address of the
1996 * sender. We verify the buffers are writable and if needed move the
1997 * sender address from kernel to user space.
1999 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2000 struct sockaddr __user *addr, int __user *addr_len)
2002 struct socket *sock;
2005 struct sockaddr_storage address;
2009 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2012 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2016 msg.msg_control = NULL;
2017 msg.msg_controllen = 0;
2018 /* Save some cycles and don't copy the address if not needed */
2019 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2020 /* We assume all kernel code knows the size of sockaddr_storage */
2021 msg.msg_namelen = 0;
2022 msg.msg_iocb = NULL;
2024 if (sock->file->f_flags & O_NONBLOCK)
2025 flags |= MSG_DONTWAIT;
2026 err = sock_recvmsg(sock, &msg, flags);
2028 if (err >= 0 && addr != NULL) {
2029 err2 = move_addr_to_user(&address,
2030 msg.msg_namelen, addr, addr_len);
2035 fput_light(sock->file, fput_needed);
2040 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2041 unsigned int, flags, struct sockaddr __user *, addr,
2042 int __user *, addr_len)
2044 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2048 * Receive a datagram from a socket.
2051 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2052 unsigned int, flags)
2054 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2057 static bool sock_use_custom_sol_socket(const struct socket *sock)
2059 const struct sock *sk = sock->sk;
2061 /* Use sock->ops->setsockopt() for MPTCP */
2062 return IS_ENABLED(CONFIG_MPTCP) &&
2063 sk->sk_protocol == IPPROTO_MPTCP &&
2064 sk->sk_type == SOCK_STREAM &&
2065 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2069 * Set a socket option. Because we don't know the option lengths we have
2070 * to pass the user mode parameter for the protocols to sort out.
2072 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2075 sockptr_t optval = USER_SOCKPTR(user_optval);
2076 char *kernel_optval = NULL;
2077 int err, fput_needed;
2078 struct socket *sock;
2083 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2087 err = security_socket_setsockopt(sock, level, optname);
2091 if (!in_compat_syscall())
2092 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2093 user_optval, &optlen,
2103 optval = KERNEL_SOCKPTR(kernel_optval);
2104 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2105 err = sock_setsockopt(sock, level, optname, optval, optlen);
2106 else if (unlikely(!sock->ops->setsockopt))
2109 err = sock->ops->setsockopt(sock, level, optname, optval,
2111 kfree(kernel_optval);
2113 fput_light(sock->file, fput_needed);
2117 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2118 char __user *, optval, int, optlen)
2120 return __sys_setsockopt(fd, level, optname, optval, optlen);
2124 * Get a socket option. Because we don't know the option lengths we have
2125 * to pass a user mode parameter for the protocols to sort out.
2127 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2130 int err, fput_needed;
2131 struct socket *sock;
2134 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2138 err = security_socket_getsockopt(sock, level, optname);
2142 if (!in_compat_syscall())
2143 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2145 if (level == SOL_SOCKET)
2146 err = sock_getsockopt(sock, level, optname, optval, optlen);
2147 else if (unlikely(!sock->ops->getsockopt))
2150 err = sock->ops->getsockopt(sock, level, optname, optval,
2153 if (!in_compat_syscall())
2154 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2155 optval, optlen, max_optlen,
2158 fput_light(sock->file, fput_needed);
2162 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2163 char __user *, optval, int __user *, optlen)
2165 return __sys_getsockopt(fd, level, optname, optval, optlen);
2169 * Shutdown a socket.
2172 int __sys_shutdown(int fd, int how)
2174 int err, fput_needed;
2175 struct socket *sock;
2177 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2179 err = security_socket_shutdown(sock, how);
2181 err = sock->ops->shutdown(sock, how);
2182 fput_light(sock->file, fput_needed);
2187 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2189 return __sys_shutdown(fd, how);
2192 /* A couple of helpful macros for getting the address of the 32/64 bit
2193 * fields which are the same type (int / unsigned) on our platforms.
2195 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2196 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2197 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2199 struct used_address {
2200 struct sockaddr_storage name;
2201 unsigned int name_len;
2204 int __copy_msghdr_from_user(struct msghdr *kmsg,
2205 struct user_msghdr __user *umsg,
2206 struct sockaddr __user **save_addr,
2207 struct iovec __user **uiov, size_t *nsegs)
2209 struct user_msghdr msg;
2212 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2215 kmsg->msg_control_is_user = true;
2216 kmsg->msg_control_user = msg.msg_control;
2217 kmsg->msg_controllen = msg.msg_controllen;
2218 kmsg->msg_flags = msg.msg_flags;
2220 kmsg->msg_namelen = msg.msg_namelen;
2222 kmsg->msg_namelen = 0;
2224 if (kmsg->msg_namelen < 0)
2227 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2228 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2231 *save_addr = msg.msg_name;
2233 if (msg.msg_name && kmsg->msg_namelen) {
2235 err = move_addr_to_kernel(msg.msg_name,
2242 kmsg->msg_name = NULL;
2243 kmsg->msg_namelen = 0;
2246 if (msg.msg_iovlen > UIO_MAXIOV)
2249 kmsg->msg_iocb = NULL;
2250 *uiov = msg.msg_iov;
2251 *nsegs = msg.msg_iovlen;
2255 static int copy_msghdr_from_user(struct msghdr *kmsg,
2256 struct user_msghdr __user *umsg,
2257 struct sockaddr __user **save_addr,
2260 struct user_msghdr msg;
2263 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2268 err = import_iovec(save_addr ? READ : WRITE,
2269 msg.msg_iov, msg.msg_iovlen,
2270 UIO_FASTIOV, iov, &kmsg->msg_iter);
2271 return err < 0 ? err : 0;
2274 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2275 unsigned int flags, struct used_address *used_address,
2276 unsigned int allowed_msghdr_flags)
2278 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2279 __aligned(sizeof(__kernel_size_t));
2280 /* 20 is size of ipv6_pktinfo */
2281 unsigned char *ctl_buf = ctl;
2287 if (msg_sys->msg_controllen > INT_MAX)
2289 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2290 ctl_len = msg_sys->msg_controllen;
2291 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2293 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2297 ctl_buf = msg_sys->msg_control;
2298 ctl_len = msg_sys->msg_controllen;
2299 } else if (ctl_len) {
2300 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2301 CMSG_ALIGN(sizeof(struct cmsghdr)));
2302 if (ctl_len > sizeof(ctl)) {
2303 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2304 if (ctl_buf == NULL)
2308 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2310 msg_sys->msg_control = ctl_buf;
2311 msg_sys->msg_control_is_user = false;
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 if (msg->msg_control || msg->msg_controllen) {
2609 /* disallow ancillary data reqs unless cmsg is plain data */
2610 if (!(sock->ops->flags & PROTO_CMSG_DATA_ONLY))
2614 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2617 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2618 bool forbid_cmsg_compat)
2620 int fput_needed, err;
2621 struct msghdr msg_sys;
2622 struct socket *sock;
2624 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2627 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2631 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2633 fput_light(sock->file, fput_needed);
2638 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2639 unsigned int, flags)
2641 return __sys_recvmsg(fd, msg, flags, true);
2645 * Linux recvmmsg interface
2648 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2649 unsigned int vlen, unsigned int flags,
2650 struct timespec64 *timeout)
2652 int fput_needed, err, datagrams;
2653 struct socket *sock;
2654 struct mmsghdr __user *entry;
2655 struct compat_mmsghdr __user *compat_entry;
2656 struct msghdr msg_sys;
2657 struct timespec64 end_time;
2658 struct timespec64 timeout64;
2661 poll_select_set_timeout(&end_time, timeout->tv_sec,
2667 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2671 if (likely(!(flags & MSG_ERRQUEUE))) {
2672 err = sock_error(sock->sk);
2680 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2682 while (datagrams < vlen) {
2684 * No need to ask LSM for more than the first datagram.
2686 if (MSG_CMSG_COMPAT & flags) {
2687 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2688 &msg_sys, flags & ~MSG_WAITFORONE,
2692 err = __put_user(err, &compat_entry->msg_len);
2695 err = ___sys_recvmsg(sock,
2696 (struct user_msghdr __user *)entry,
2697 &msg_sys, flags & ~MSG_WAITFORONE,
2701 err = put_user(err, &entry->msg_len);
2709 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2710 if (flags & MSG_WAITFORONE)
2711 flags |= MSG_DONTWAIT;
2714 ktime_get_ts64(&timeout64);
2715 *timeout = timespec64_sub(end_time, timeout64);
2716 if (timeout->tv_sec < 0) {
2717 timeout->tv_sec = timeout->tv_nsec = 0;
2721 /* Timeout, return less than vlen datagrams */
2722 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2726 /* Out of band data, return right away */
2727 if (msg_sys.msg_flags & MSG_OOB)
2735 if (datagrams == 0) {
2741 * We may return less entries than requested (vlen) if the
2742 * sock is non block and there aren't enough datagrams...
2744 if (err != -EAGAIN) {
2746 * ... or if recvmsg returns an error after we
2747 * received some datagrams, where we record the
2748 * error to return on the next call or if the
2749 * app asks about it using getsockopt(SO_ERROR).
2751 sock->sk->sk_err = -err;
2754 fput_light(sock->file, fput_needed);
2759 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2760 unsigned int vlen, unsigned int flags,
2761 struct __kernel_timespec __user *timeout,
2762 struct old_timespec32 __user *timeout32)
2765 struct timespec64 timeout_sys;
2767 if (timeout && get_timespec64(&timeout_sys, timeout))
2770 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2773 if (!timeout && !timeout32)
2774 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2776 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2781 if (timeout && put_timespec64(&timeout_sys, timeout))
2782 datagrams = -EFAULT;
2784 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2785 datagrams = -EFAULT;
2790 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2791 unsigned int, vlen, unsigned int, flags,
2792 struct __kernel_timespec __user *, timeout)
2794 if (flags & MSG_CMSG_COMPAT)
2797 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2800 #ifdef CONFIG_COMPAT_32BIT_TIME
2801 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2802 unsigned int, vlen, unsigned int, flags,
2803 struct old_timespec32 __user *, timeout)
2805 if (flags & MSG_CMSG_COMPAT)
2808 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2812 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2813 /* Argument list sizes for sys_socketcall */
2814 #define AL(x) ((x) * sizeof(unsigned long))
2815 static const unsigned char nargs[21] = {
2816 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2817 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2818 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2825 * System call vectors.
2827 * Argument checking cleaned up. Saved 20% in size.
2828 * This function doesn't need to set the kernel lock because
2829 * it is set by the callees.
2832 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2834 unsigned long a[AUDITSC_ARGS];
2835 unsigned long a0, a1;
2839 if (call < 1 || call > SYS_SENDMMSG)
2841 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2844 if (len > sizeof(a))
2847 /* copy_from_user should be SMP safe. */
2848 if (copy_from_user(a, args, len))
2851 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2860 err = __sys_socket(a0, a1, a[2]);
2863 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2866 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2869 err = __sys_listen(a0, a1);
2872 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2873 (int __user *)a[2], 0);
2875 case SYS_GETSOCKNAME:
2877 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2878 (int __user *)a[2]);
2880 case SYS_GETPEERNAME:
2882 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2883 (int __user *)a[2]);
2885 case SYS_SOCKETPAIR:
2886 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2889 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2893 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2894 (struct sockaddr __user *)a[4], a[5]);
2897 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2901 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2902 (struct sockaddr __user *)a[4],
2903 (int __user *)a[5]);
2906 err = __sys_shutdown(a0, a1);
2908 case SYS_SETSOCKOPT:
2909 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2912 case SYS_GETSOCKOPT:
2914 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2915 (int __user *)a[4]);
2918 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2922 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2926 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2930 if (IS_ENABLED(CONFIG_64BIT))
2931 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2933 (struct __kernel_timespec __user *)a[4],
2936 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2938 (struct old_timespec32 __user *)a[4]);
2941 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2942 (int __user *)a[2], a[3]);
2951 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2954 * sock_register - add a socket protocol handler
2955 * @ops: description of protocol
2957 * This function is called by a protocol handler that wants to
2958 * advertise its address family, and have it linked into the
2959 * socket interface. The value ops->family corresponds to the
2960 * socket system call protocol family.
2962 int sock_register(const struct net_proto_family *ops)
2966 if (ops->family >= NPROTO) {
2967 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2971 spin_lock(&net_family_lock);
2972 if (rcu_dereference_protected(net_families[ops->family],
2973 lockdep_is_held(&net_family_lock)))
2976 rcu_assign_pointer(net_families[ops->family], ops);
2979 spin_unlock(&net_family_lock);
2981 pr_info("NET: Registered protocol family %d\n", ops->family);
2984 EXPORT_SYMBOL(sock_register);
2987 * sock_unregister - remove a protocol handler
2988 * @family: protocol family to remove
2990 * This function is called by a protocol handler that wants to
2991 * remove its address family, and have it unlinked from the
2992 * new socket creation.
2994 * If protocol handler is a module, then it can use module reference
2995 * counts to protect against new references. If protocol handler is not
2996 * a module then it needs to provide its own protection in
2997 * the ops->create routine.
2999 void sock_unregister(int family)
3001 BUG_ON(family < 0 || family >= NPROTO);
3003 spin_lock(&net_family_lock);
3004 RCU_INIT_POINTER(net_families[family], NULL);
3005 spin_unlock(&net_family_lock);
3009 pr_info("NET: Unregistered protocol family %d\n", family);
3011 EXPORT_SYMBOL(sock_unregister);
3013 bool sock_is_registered(int family)
3015 return family < NPROTO && rcu_access_pointer(net_families[family]);
3018 static int __init sock_init(void)
3022 * Initialize the network sysctl infrastructure.
3024 err = net_sysctl_init();
3029 * Initialize skbuff SLAB cache
3034 * Initialize the protocols module.
3039 err = register_filesystem(&sock_fs_type);
3042 sock_mnt = kern_mount(&sock_fs_type);
3043 if (IS_ERR(sock_mnt)) {
3044 err = PTR_ERR(sock_mnt);
3048 /* The real protocol initialization is performed in later initcalls.
3051 #ifdef CONFIG_NETFILTER
3052 err = netfilter_init();
3057 ptp_classifier_init();
3063 unregister_filesystem(&sock_fs_type);
3067 core_initcall(sock_init); /* early initcall */
3069 #ifdef CONFIG_PROC_FS
3070 void socket_seq_show(struct seq_file *seq)
3072 seq_printf(seq, "sockets: used %d\n",
3073 sock_inuse_get(seq->private));
3075 #endif /* CONFIG_PROC_FS */
3077 #ifdef CONFIG_COMPAT
3078 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3080 struct compat_ifconf ifc32;
3084 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3087 ifc.ifc_len = ifc32.ifc_len;
3088 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3091 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3096 ifc32.ifc_len = ifc.ifc_len;
3097 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3103 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3105 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3106 bool convert_in = false, convert_out = false;
3107 size_t buf_size = 0;
3108 struct ethtool_rxnfc __user *rxnfc = NULL;
3110 u32 rule_cnt = 0, actual_rule_cnt;
3115 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3118 compat_rxnfc = compat_ptr(data);
3120 if (get_user(ethcmd, &compat_rxnfc->cmd))
3123 /* Most ethtool structures are defined without padding.
3124 * Unfortunately struct ethtool_rxnfc is an exception.
3129 case ETHTOOL_GRXCLSRLALL:
3130 /* Buffer size is variable */
3131 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3133 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3135 buf_size += rule_cnt * sizeof(u32);
3137 case ETHTOOL_GRXRINGS:
3138 case ETHTOOL_GRXCLSRLCNT:
3139 case ETHTOOL_GRXCLSRULE:
3140 case ETHTOOL_SRXCLSRLINS:
3143 case ETHTOOL_SRXCLSRLDEL:
3144 buf_size += sizeof(struct ethtool_rxnfc);
3146 rxnfc = compat_alloc_user_space(buf_size);
3150 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3153 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3156 /* We expect there to be holes between fs.m_ext and
3157 * fs.ring_cookie and at the end of fs, but nowhere else.
3159 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3160 sizeof(compat_rxnfc->fs.m_ext) !=
3161 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3162 sizeof(rxnfc->fs.m_ext));
3164 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3165 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3166 offsetof(struct ethtool_rxnfc, fs.location) -
3167 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3169 if (copy_in_user(rxnfc, compat_rxnfc,
3170 (void __user *)(&rxnfc->fs.m_ext + 1) -
3171 (void __user *)rxnfc) ||
3172 copy_in_user(&rxnfc->fs.ring_cookie,
3173 &compat_rxnfc->fs.ring_cookie,
3174 (void __user *)(&rxnfc->fs.location + 1) -
3175 (void __user *)&rxnfc->fs.ring_cookie))
3177 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3178 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3180 } else if (copy_in_user(&rxnfc->rule_cnt,
3181 &compat_rxnfc->rule_cnt,
3182 sizeof(rxnfc->rule_cnt)))
3186 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3191 if (copy_in_user(compat_rxnfc, rxnfc,
3192 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3193 (const void __user *)rxnfc) ||
3194 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3195 &rxnfc->fs.ring_cookie,
3196 (const void __user *)(&rxnfc->fs.location + 1) -
3197 (const void __user *)&rxnfc->fs.ring_cookie) ||
3198 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3199 sizeof(rxnfc->rule_cnt)))
3202 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3203 /* As an optimisation, we only copy the actual
3204 * number of rules that the underlying
3205 * function returned. Since Mallory might
3206 * change the rule count in user memory, we
3207 * check that it is less than the rule count
3208 * originally given (as the user buffer size),
3209 * which has been range-checked.
3211 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3213 if (actual_rule_cnt < rule_cnt)
3214 rule_cnt = actual_rule_cnt;
3215 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3216 &rxnfc->rule_locs[0],
3217 rule_cnt * sizeof(u32)))
3225 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3227 compat_uptr_t uptr32;
3232 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3235 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3238 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3239 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3241 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3243 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3244 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3250 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3251 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3252 struct compat_ifreq __user *u_ifreq32)
3257 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3259 if (get_user(data32, &u_ifreq32->ifr_data))
3261 ifreq.ifr_data = compat_ptr(data32);
3263 return dev_ioctl(net, cmd, &ifreq, NULL);
3266 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3268 struct compat_ifreq __user *uifr32)
3270 struct ifreq __user *uifr;
3273 /* Handle the fact that while struct ifreq has the same *layout* on
3274 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3275 * which are handled elsewhere, it still has different *size* due to
3276 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3277 * resulting in struct ifreq being 32 and 40 bytes respectively).
3278 * As a result, if the struct happens to be at the end of a page and
3279 * the next page isn't readable/writable, we get a fault. To prevent
3280 * that, copy back and forth to the full size.
3283 uifr = compat_alloc_user_space(sizeof(*uifr));
3284 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3287 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3298 case SIOCGIFBRDADDR:
3299 case SIOCGIFDSTADDR:
3300 case SIOCGIFNETMASK:
3306 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3314 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3315 struct compat_ifreq __user *uifr32)
3318 struct compat_ifmap __user *uifmap32;
3321 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3322 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3323 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3324 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3325 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3326 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3327 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3328 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3332 err = dev_ioctl(net, cmd, &ifr, NULL);
3334 if (cmd == SIOCGIFMAP && !err) {
3335 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3336 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3337 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3338 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3339 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3340 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3341 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3348 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3349 * for some operations; this forces use of the newer bridge-utils that
3350 * use compatible ioctls
3352 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3356 if (get_user(tmp, argp))
3358 if (tmp == BRCTL_GET_VERSION)
3359 return BRCTL_VERSION + 1;
3363 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3364 unsigned int cmd, unsigned long arg)
3366 void __user *argp = compat_ptr(arg);
3367 struct sock *sk = sock->sk;
3368 struct net *net = sock_net(sk);
3370 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3371 return compat_ifr_data_ioctl(net, cmd, argp);
3376 return old_bridge_ioctl(argp);
3378 return compat_dev_ifconf(net, argp);
3380 return ethtool_ioctl(net, argp);
3382 return compat_siocwandev(net, argp);
3385 return compat_sioc_ifmap(net, cmd, argp);
3386 case SIOCGSTAMP_OLD:
3387 case SIOCGSTAMPNS_OLD:
3388 if (!sock->ops->gettstamp)
3389 return -ENOIOCTLCMD;
3390 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3391 !COMPAT_USE_64BIT_TIME);
3393 case SIOCBONDSLAVEINFOQUERY:
3394 case SIOCBONDINFOQUERY:
3397 return compat_ifr_data_ioctl(net, cmd, argp);
3408 case SIOCGSTAMP_NEW:
3409 case SIOCGSTAMPNS_NEW:
3410 return sock_ioctl(file, cmd, arg);
3427 case SIOCSIFHWBROADCAST:
3429 case SIOCGIFBRDADDR:
3430 case SIOCSIFBRDADDR:
3431 case SIOCGIFDSTADDR:
3432 case SIOCSIFDSTADDR:
3433 case SIOCGIFNETMASK:
3434 case SIOCSIFNETMASK:
3446 case SIOCBONDENSLAVE:
3447 case SIOCBONDRELEASE:
3448 case SIOCBONDSETHWADDR:
3449 case SIOCBONDCHANGEACTIVE:
3450 return compat_ifreq_ioctl(net, sock, cmd, argp);
3458 return sock_do_ioctl(net, sock, cmd, arg);
3461 return -ENOIOCTLCMD;
3464 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3467 struct socket *sock = file->private_data;
3468 int ret = -ENOIOCTLCMD;
3475 if (sock->ops->compat_ioctl)
3476 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3478 if (ret == -ENOIOCTLCMD &&
3479 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3480 ret = compat_wext_handle_ioctl(net, cmd, arg);
3482 if (ret == -ENOIOCTLCMD)
3483 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3490 * kernel_bind - bind an address to a socket (kernel space)
3493 * @addrlen: length of address
3495 * Returns 0 or an error.
3498 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3500 return sock->ops->bind(sock, addr, addrlen);
3502 EXPORT_SYMBOL(kernel_bind);
3505 * kernel_listen - move socket to listening state (kernel space)
3507 * @backlog: pending connections queue size
3509 * Returns 0 or an error.
3512 int kernel_listen(struct socket *sock, int backlog)
3514 return sock->ops->listen(sock, backlog);
3516 EXPORT_SYMBOL(kernel_listen);
3519 * kernel_accept - accept a connection (kernel space)
3520 * @sock: listening socket
3521 * @newsock: new connected socket
3524 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3525 * If it fails, @newsock is guaranteed to be %NULL.
3526 * Returns 0 or an error.
3529 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3531 struct sock *sk = sock->sk;
3534 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3539 err = sock->ops->accept(sock, *newsock, flags, true);
3541 sock_release(*newsock);
3546 (*newsock)->ops = sock->ops;
3547 __module_get((*newsock)->ops->owner);
3552 EXPORT_SYMBOL(kernel_accept);
3555 * kernel_connect - connect a socket (kernel space)
3558 * @addrlen: address length
3559 * @flags: flags (O_NONBLOCK, ...)
3561 * For datagram sockets, @addr is the addres to which datagrams are sent
3562 * by default, and the only address from which datagrams are received.
3563 * For stream sockets, attempts to connect to @addr.
3564 * Returns 0 or an error code.
3567 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3570 return sock->ops->connect(sock, addr, addrlen, flags);
3572 EXPORT_SYMBOL(kernel_connect);
3575 * kernel_getsockname - get the address which the socket is bound (kernel space)
3577 * @addr: address holder
3579 * Fills the @addr pointer with the address which the socket is bound.
3580 * Returns 0 or an error code.
3583 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3585 return sock->ops->getname(sock, addr, 0);
3587 EXPORT_SYMBOL(kernel_getsockname);
3590 * kernel_getpeername - get the address which the socket is connected (kernel space)
3592 * @addr: address holder
3594 * Fills the @addr pointer with the address which the socket is connected.
3595 * Returns 0 or an error code.
3598 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3600 return sock->ops->getname(sock, addr, 1);
3602 EXPORT_SYMBOL(kernel_getpeername);
3605 * kernel_sendpage - send a &page through a socket (kernel space)
3608 * @offset: page offset
3609 * @size: total size in bytes
3610 * @flags: flags (MSG_DONTWAIT, ...)
3612 * Returns the total amount sent in bytes or an error.
3615 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3616 size_t size, int flags)
3618 if (sock->ops->sendpage) {
3619 /* Warn in case the improper page to zero-copy send */
3620 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3621 return sock->ops->sendpage(sock, page, offset, size, flags);
3623 return sock_no_sendpage(sock, page, offset, size, flags);
3625 EXPORT_SYMBOL(kernel_sendpage);
3628 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3631 * @offset: page offset
3632 * @size: total size in bytes
3633 * @flags: flags (MSG_DONTWAIT, ...)
3635 * Returns the total amount sent in bytes or an error.
3636 * Caller must hold @sk.
3639 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3640 size_t size, int flags)
3642 struct socket *sock = sk->sk_socket;
3644 if (sock->ops->sendpage_locked)
3645 return sock->ops->sendpage_locked(sk, page, offset, size,
3648 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3650 EXPORT_SYMBOL(kernel_sendpage_locked);
3653 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3655 * @how: connection part
3657 * Returns 0 or an error.
3660 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3662 return sock->ops->shutdown(sock, how);
3664 EXPORT_SYMBOL(kernel_sock_shutdown);
3667 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3670 * This routine returns the IP overhead imposed by a socket i.e.
3671 * the length of the underlying IP header, depending on whether
3672 * this is an IPv4 or IPv6 socket and the length from IP options turned
3673 * on at the socket. Assumes that the caller has a lock on the socket.
3676 u32 kernel_sock_ip_overhead(struct sock *sk)
3678 struct inet_sock *inet;
3679 struct ip_options_rcu *opt;
3681 #if IS_ENABLED(CONFIG_IPV6)
3682 struct ipv6_pinfo *np;
3683 struct ipv6_txoptions *optv6 = NULL;
3684 #endif /* IS_ENABLED(CONFIG_IPV6) */
3689 switch (sk->sk_family) {
3692 overhead += sizeof(struct iphdr);
3693 opt = rcu_dereference_protected(inet->inet_opt,
3694 sock_owned_by_user(sk));
3696 overhead += opt->opt.optlen;
3698 #if IS_ENABLED(CONFIG_IPV6)
3701 overhead += sizeof(struct ipv6hdr);
3703 optv6 = rcu_dereference_protected(np->opt,
3704 sock_owned_by_user(sk));
3706 overhead += (optv6->opt_flen + optv6->opt_nflen);
3708 #endif /* IS_ENABLED(CONFIG_IPV6) */
3709 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3713 EXPORT_SYMBOL(kernel_sock_ip_overhead);