1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/ethtool.h>
57 #include <linux/socket.h>
58 #include <linux/file.h>
59 #include <linux/net.h>
60 #include <linux/interrupt.h>
61 #include <linux/thread_info.h>
62 #include <linux/rcupdate.h>
63 #include <linux/netdevice.h>
64 #include <linux/proc_fs.h>
65 #include <linux/seq_file.h>
66 #include <linux/mutex.h>
67 #include <linux/if_bridge.h>
68 #include <linux/if_vlan.h>
69 #include <linux/ptp_classify.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74 #include <linux/highmem.h>
75 #include <linux/mount.h>
76 #include <linux/pseudo_fs.h>
77 #include <linux/security.h>
78 #include <linux/syscalls.h>
79 #include <linux/compat.h>
80 #include <linux/kmod.h>
81 #include <linux/audit.h>
82 #include <linux/wireless.h>
83 #include <linux/nsproxy.h>
84 #include <linux/magic.h>
85 #include <linux/slab.h>
86 #include <linux/xattr.h>
87 #include <linux/nospec.h>
88 #include <linux/indirect_call_wrapper.h>
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
95 #include <net/cls_cgroup.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/termios.h>
104 #include <linux/sockios.h>
105 #include <net/busy_poll.h>
106 #include <linux/errqueue.h>
108 #ifdef CONFIG_NET_RX_BUSY_POLL
109 unsigned int sysctl_net_busy_read __read_mostly;
110 unsigned int sysctl_net_busy_poll __read_mostly;
113 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
115 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
117 static int sock_close(struct inode *inode, struct file *file);
118 static __poll_t sock_poll(struct file *file,
119 struct poll_table_struct *wait);
120 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
122 static long compat_sock_ioctl(struct file *file,
123 unsigned int cmd, unsigned long arg);
125 static int sock_fasync(int fd, struct file *filp, int on);
126 static ssize_t sock_sendpage(struct file *file, struct page *page,
127 int offset, size_t size, loff_t *ppos, int more);
128 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
129 struct pipe_inode_info *pipe, size_t len,
132 #ifdef CONFIG_PROC_FS
133 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135 struct socket *sock = f->private_data;
137 if (sock->ops->show_fdinfo)
138 sock->ops->show_fdinfo(m, sock);
141 #define sock_show_fdinfo NULL
145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 * in the operation structures but are done directly via the socketcall() multiplexor.
149 static const struct file_operations socket_file_ops = {
150 .owner = THIS_MODULE,
152 .read_iter = sock_read_iter,
153 .write_iter = sock_write_iter,
155 .unlocked_ioctl = sock_ioctl,
157 .compat_ioctl = compat_sock_ioctl,
160 .release = sock_close,
161 .fasync = sock_fasync,
162 .sendpage = sock_sendpage,
163 .splice_write = generic_splice_sendpage,
164 .splice_read = sock_splice_read,
165 .show_fdinfo = sock_show_fdinfo,
168 static const char * const pf_family_names[] = {
169 [PF_UNSPEC] = "PF_UNSPEC",
170 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
171 [PF_INET] = "PF_INET",
172 [PF_AX25] = "PF_AX25",
174 [PF_APPLETALK] = "PF_APPLETALK",
175 [PF_NETROM] = "PF_NETROM",
176 [PF_BRIDGE] = "PF_BRIDGE",
177 [PF_ATMPVC] = "PF_ATMPVC",
179 [PF_INET6] = "PF_INET6",
180 [PF_ROSE] = "PF_ROSE",
181 [PF_DECnet] = "PF_DECnet",
182 [PF_NETBEUI] = "PF_NETBEUI",
183 [PF_SECURITY] = "PF_SECURITY",
185 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
186 [PF_PACKET] = "PF_PACKET",
188 [PF_ECONET] = "PF_ECONET",
189 [PF_ATMSVC] = "PF_ATMSVC",
192 [PF_IRDA] = "PF_IRDA",
193 [PF_PPPOX] = "PF_PPPOX",
194 [PF_WANPIPE] = "PF_WANPIPE",
197 [PF_MPLS] = "PF_MPLS",
199 [PF_TIPC] = "PF_TIPC",
200 [PF_BLUETOOTH] = "PF_BLUETOOTH",
201 [PF_IUCV] = "PF_IUCV",
202 [PF_RXRPC] = "PF_RXRPC",
203 [PF_ISDN] = "PF_ISDN",
204 [PF_PHONET] = "PF_PHONET",
205 [PF_IEEE802154] = "PF_IEEE802154",
206 [PF_CAIF] = "PF_CAIF",
209 [PF_VSOCK] = "PF_VSOCK",
211 [PF_QIPCRTR] = "PF_QIPCRTR",
217 * The protocol list. Each protocol is registered in here.
220 static DEFINE_SPINLOCK(net_family_lock);
221 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
225 * Move socket addresses back and forth across the kernel/user
226 * divide and look after the messy bits.
230 * move_addr_to_kernel - copy a socket address into kernel space
231 * @uaddr: Address in user space
232 * @kaddr: Address in kernel space
233 * @ulen: Length in user space
235 * The address is copied into kernel space. If the provided address is
236 * too long an error code of -EINVAL is returned. If the copy gives
237 * invalid addresses -EFAULT is returned. On a success 0 is returned.
240 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
242 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
246 if (copy_from_user(kaddr, uaddr, ulen))
248 return audit_sockaddr(ulen, kaddr);
252 * move_addr_to_user - copy an address to user space
253 * @kaddr: kernel space address
254 * @klen: length of address in kernel
255 * @uaddr: user space address
256 * @ulen: pointer to user length field
258 * The value pointed to by ulen on entry is the buffer length available.
259 * This is overwritten with the buffer space used. -EINVAL is returned
260 * if an overlong buffer is specified or a negative buffer size. -EFAULT
261 * is returned if either the buffer or the length field are not
263 * After copying the data up to the limit the user specifies, the true
264 * length of the data is written over the length limit the user
265 * specified. Zero is returned for a success.
268 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
269 void __user *uaddr, int __user *ulen)
274 BUG_ON(klen > sizeof(struct sockaddr_storage));
275 err = get_user(len, ulen);
283 if (audit_sockaddr(klen, kaddr))
285 if (copy_to_user(uaddr, kaddr, len))
289 * "fromlen shall refer to the value before truncation.."
292 return __put_user(klen, ulen);
295 static struct kmem_cache *sock_inode_cachep __ro_after_init;
297 static struct inode *sock_alloc_inode(struct super_block *sb)
299 struct socket_alloc *ei;
301 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
304 init_waitqueue_head(&ei->socket.wq.wait);
305 ei->socket.wq.fasync_list = NULL;
306 ei->socket.wq.flags = 0;
308 ei->socket.state = SS_UNCONNECTED;
309 ei->socket.flags = 0;
310 ei->socket.ops = NULL;
311 ei->socket.sk = NULL;
312 ei->socket.file = NULL;
314 return &ei->vfs_inode;
317 static void sock_free_inode(struct inode *inode)
319 struct socket_alloc *ei;
321 ei = container_of(inode, struct socket_alloc, vfs_inode);
322 kmem_cache_free(sock_inode_cachep, ei);
325 static void init_once(void *foo)
327 struct socket_alloc *ei = (struct socket_alloc *)foo;
329 inode_init_once(&ei->vfs_inode);
332 static void init_inodecache(void)
334 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
335 sizeof(struct socket_alloc),
337 (SLAB_HWCACHE_ALIGN |
338 SLAB_RECLAIM_ACCOUNT |
339 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
341 BUG_ON(sock_inode_cachep == NULL);
344 static const struct super_operations sockfs_ops = {
345 .alloc_inode = sock_alloc_inode,
346 .free_inode = sock_free_inode,
347 .statfs = simple_statfs,
351 * sockfs_dname() is called from d_path().
353 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
355 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
356 d_inode(dentry)->i_ino);
359 static const struct dentry_operations sockfs_dentry_operations = {
360 .d_dname = sockfs_dname,
363 static int sockfs_xattr_get(const struct xattr_handler *handler,
364 struct dentry *dentry, struct inode *inode,
365 const char *suffix, void *value, size_t size)
368 if (dentry->d_name.len + 1 > size)
370 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
372 return dentry->d_name.len + 1;
375 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
376 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
377 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
379 static const struct xattr_handler sockfs_xattr_handler = {
380 .name = XATTR_NAME_SOCKPROTONAME,
381 .get = sockfs_xattr_get,
384 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
385 struct user_namespace *mnt_userns,
386 struct dentry *dentry, struct inode *inode,
387 const char *suffix, const void *value,
388 size_t size, int flags)
390 /* Handled by LSM. */
394 static const struct xattr_handler sockfs_security_xattr_handler = {
395 .prefix = XATTR_SECURITY_PREFIX,
396 .set = sockfs_security_xattr_set,
399 static const struct xattr_handler *sockfs_xattr_handlers[] = {
400 &sockfs_xattr_handler,
401 &sockfs_security_xattr_handler,
405 static int sockfs_init_fs_context(struct fs_context *fc)
407 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
410 ctx->ops = &sockfs_ops;
411 ctx->dops = &sockfs_dentry_operations;
412 ctx->xattr = sockfs_xattr_handlers;
416 static struct vfsmount *sock_mnt __read_mostly;
418 static struct file_system_type sock_fs_type = {
420 .init_fs_context = sockfs_init_fs_context,
421 .kill_sb = kill_anon_super,
425 * Obtains the first available file descriptor and sets it up for use.
427 * These functions create file structures and maps them to fd space
428 * of the current process. On success it returns file descriptor
429 * and file struct implicitly stored in sock->file.
430 * Note that another thread may close file descriptor before we return
431 * from this function. We use the fact that now we do not refer
432 * to socket after mapping. If one day we will need it, this
433 * function will increment ref. count on file by 1.
435 * In any case returned fd MAY BE not valid!
436 * This race condition is unavoidable
437 * with shared fd spaces, we cannot solve it inside kernel,
438 * but we take care of internal coherence yet.
442 * sock_alloc_file - Bind a &socket to a &file
444 * @flags: file status flags
445 * @dname: protocol name
447 * Returns the &file bound with @sock, implicitly storing it
448 * in sock->file. If dname is %NULL, sets to "".
449 * On failure the return is a ERR pointer (see linux/err.h).
450 * This function uses GFP_KERNEL internally.
453 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
458 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
460 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
461 O_RDWR | (flags & O_NONBLOCK),
469 file->private_data = sock;
470 stream_open(SOCK_INODE(sock), file);
473 EXPORT_SYMBOL(sock_alloc_file);
475 static int sock_map_fd(struct socket *sock, int flags)
477 struct file *newfile;
478 int fd = get_unused_fd_flags(flags);
479 if (unlikely(fd < 0)) {
484 newfile = sock_alloc_file(sock, flags, NULL);
485 if (!IS_ERR(newfile)) {
486 fd_install(fd, newfile);
491 return PTR_ERR(newfile);
495 * sock_from_file - Return the &socket bounded to @file.
498 * On failure returns %NULL.
501 struct socket *sock_from_file(struct file *file)
503 if (file->f_op == &socket_file_ops)
504 return file->private_data; /* set in sock_map_fd */
508 EXPORT_SYMBOL(sock_from_file);
511 * sockfd_lookup - Go from a file number to its socket slot
513 * @err: pointer to an error code return
515 * The file handle passed in is locked and the socket it is bound
516 * to is returned. If an error occurs the err pointer is overwritten
517 * with a negative errno code and NULL is returned. The function checks
518 * for both invalid handles and passing a handle which is not a socket.
520 * On a success the socket object pointer is returned.
523 struct socket *sockfd_lookup(int fd, int *err)
534 sock = sock_from_file(file);
541 EXPORT_SYMBOL(sockfd_lookup);
543 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
545 struct fd f = fdget(fd);
550 sock = sock_from_file(f.file);
552 *fput_needed = f.flags & FDPUT_FPUT;
561 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
567 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
577 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
582 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
589 static int sockfs_setattr(struct user_namespace *mnt_userns,
590 struct dentry *dentry, struct iattr *iattr)
592 int err = simple_setattr(&init_user_ns, dentry, iattr);
594 if (!err && (iattr->ia_valid & ATTR_UID)) {
595 struct socket *sock = SOCKET_I(d_inode(dentry));
598 sock->sk->sk_uid = iattr->ia_uid;
606 static const struct inode_operations sockfs_inode_ops = {
607 .listxattr = sockfs_listxattr,
608 .setattr = sockfs_setattr,
612 * sock_alloc - allocate a socket
614 * Allocate a new inode and socket object. The two are bound together
615 * and initialised. The socket is then returned. If we are out of inodes
616 * NULL is returned. This functions uses GFP_KERNEL internally.
619 struct socket *sock_alloc(void)
624 inode = new_inode_pseudo(sock_mnt->mnt_sb);
628 sock = SOCKET_I(inode);
630 inode->i_ino = get_next_ino();
631 inode->i_mode = S_IFSOCK | S_IRWXUGO;
632 inode->i_uid = current_fsuid();
633 inode->i_gid = current_fsgid();
634 inode->i_op = &sockfs_inode_ops;
638 EXPORT_SYMBOL(sock_alloc);
640 static void __sock_release(struct socket *sock, struct inode *inode)
643 struct module *owner = sock->ops->owner;
647 sock->ops->release(sock);
655 if (sock->wq.fasync_list)
656 pr_err("%s: fasync list not empty!\n", __func__);
659 iput(SOCK_INODE(sock));
666 * sock_release - close a socket
667 * @sock: socket to close
669 * The socket is released from the protocol stack if it has a release
670 * callback, and the inode is then released if the socket is bound to
671 * an inode not a file.
673 void sock_release(struct socket *sock)
675 __sock_release(sock, NULL);
677 EXPORT_SYMBOL(sock_release);
679 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
681 u8 flags = *tx_flags;
683 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
684 flags |= SKBTX_HW_TSTAMP;
686 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
687 flags |= SKBTX_SW_TSTAMP;
689 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
690 flags |= SKBTX_SCHED_TSTAMP;
694 EXPORT_SYMBOL(__sock_tx_timestamp);
696 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
698 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
700 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
702 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
703 inet_sendmsg, sock, msg,
705 BUG_ON(ret == -EIOCBQUEUED);
710 * sock_sendmsg - send a message through @sock
712 * @msg: message to send
714 * Sends @msg through @sock, passing through LSM.
715 * Returns the number of bytes sent, or an error code.
717 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
719 int err = security_socket_sendmsg(sock, msg,
722 return err ?: sock_sendmsg_nosec(sock, msg);
724 EXPORT_SYMBOL(sock_sendmsg);
727 * kernel_sendmsg - send a message through @sock (kernel-space)
729 * @msg: message header
731 * @num: vec array length
732 * @size: total message data size
734 * Builds the message data with @vec and sends it through @sock.
735 * Returns the number of bytes sent, or an error code.
738 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
739 struct kvec *vec, size_t num, size_t size)
741 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
742 return sock_sendmsg(sock, msg);
744 EXPORT_SYMBOL(kernel_sendmsg);
747 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
749 * @msg: message header
750 * @vec: output s/g array
751 * @num: output s/g array length
752 * @size: total message data size
754 * Builds the message data with @vec and sends it through @sock.
755 * Returns the number of bytes sent, or an error code.
756 * Caller must hold @sk.
759 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
760 struct kvec *vec, size_t num, size_t size)
762 struct socket *sock = sk->sk_socket;
764 if (!sock->ops->sendmsg_locked)
765 return sock_no_sendmsg_locked(sk, msg, size);
767 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
769 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
771 EXPORT_SYMBOL(kernel_sendmsg_locked);
773 static bool skb_is_err_queue(const struct sk_buff *skb)
775 /* pkt_type of skbs enqueued on the error queue are set to
776 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
777 * in recvmsg, since skbs received on a local socket will never
778 * have a pkt_type of PACKET_OUTGOING.
780 return skb->pkt_type == PACKET_OUTGOING;
783 /* On transmit, software and hardware timestamps are returned independently.
784 * As the two skb clones share the hardware timestamp, which may be updated
785 * before the software timestamp is received, a hardware TX timestamp may be
786 * returned only if there is no software TX timestamp. Ignore false software
787 * timestamps, which may be made in the __sock_recv_timestamp() call when the
788 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
789 * hardware timestamp.
791 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
793 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
796 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
798 struct scm_ts_pktinfo ts_pktinfo;
799 struct net_device *orig_dev;
801 if (!skb_mac_header_was_set(skb))
804 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
807 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
809 ts_pktinfo.if_index = orig_dev->ifindex;
812 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
813 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
814 sizeof(ts_pktinfo), &ts_pktinfo);
818 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
820 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
823 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
824 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
825 struct scm_timestamping_internal tss;
827 int empty = 1, false_tstamp = 0;
828 struct skb_shared_hwtstamps *shhwtstamps =
831 /* Race occurred between timestamp enabling and packet
832 receiving. Fill in the current time for now. */
833 if (need_software_tstamp && skb->tstamp == 0) {
834 __net_timestamp(skb);
838 if (need_software_tstamp) {
839 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
841 struct __kernel_sock_timeval tv;
843 skb_get_new_timestamp(skb, &tv);
844 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
847 struct __kernel_old_timeval tv;
849 skb_get_timestamp(skb, &tv);
850 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
855 struct __kernel_timespec ts;
857 skb_get_new_timestampns(skb, &ts);
858 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
861 struct __kernel_old_timespec ts;
863 skb_get_timestampns(skb, &ts);
864 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
870 memset(&tss, 0, sizeof(tss));
871 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
872 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
875 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
876 !skb_is_swtx_tstamp(skb, false_tstamp) &&
877 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
879 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
880 !skb_is_err_queue(skb))
881 put_ts_pktinfo(msg, skb);
884 if (sock_flag(sk, SOCK_TSTAMP_NEW))
885 put_cmsg_scm_timestamping64(msg, &tss);
887 put_cmsg_scm_timestamping(msg, &tss);
889 if (skb_is_err_queue(skb) && skb->len &&
890 SKB_EXT_ERR(skb)->opt_stats)
891 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
892 skb->len, skb->data);
895 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
897 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
902 if (!sock_flag(sk, SOCK_WIFI_STATUS))
904 if (!skb->wifi_acked_valid)
907 ack = skb->wifi_acked;
909 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
911 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
913 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
916 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
917 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
918 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
921 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
924 sock_recv_timestamp(msg, sk, skb);
925 sock_recv_drops(msg, sk, skb);
927 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
929 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
931 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
933 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
936 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
937 inet_recvmsg, sock, msg, msg_data_left(msg),
942 * sock_recvmsg - receive a message from @sock
944 * @msg: message to receive
945 * @flags: message flags
947 * Receives @msg from @sock, passing through LSM. Returns the total number
948 * of bytes received, or an error.
950 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
952 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
954 return err ?: sock_recvmsg_nosec(sock, msg, flags);
956 EXPORT_SYMBOL(sock_recvmsg);
959 * kernel_recvmsg - Receive a message from a socket (kernel space)
960 * @sock: The socket to receive the message from
961 * @msg: Received message
962 * @vec: Input s/g array for message data
963 * @num: Size of input s/g array
964 * @size: Number of bytes to read
965 * @flags: Message flags (MSG_DONTWAIT, etc...)
967 * On return the msg structure contains the scatter/gather array passed in the
968 * vec argument. The array is modified so that it consists of the unfilled
969 * portion of the original array.
971 * The returned value is the total number of bytes received, or an error.
974 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
975 struct kvec *vec, size_t num, size_t size, int flags)
977 msg->msg_control_is_user = false;
978 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
979 return sock_recvmsg(sock, msg, flags);
981 EXPORT_SYMBOL(kernel_recvmsg);
983 static ssize_t sock_sendpage(struct file *file, struct page *page,
984 int offset, size_t size, loff_t *ppos, int more)
989 sock = file->private_data;
991 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
992 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
995 return kernel_sendpage(sock, page, offset, size, flags);
998 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
999 struct pipe_inode_info *pipe, size_t len,
1002 struct socket *sock = file->private_data;
1004 if (unlikely(!sock->ops->splice_read))
1005 return generic_file_splice_read(file, ppos, pipe, len, flags);
1007 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1010 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1012 struct file *file = iocb->ki_filp;
1013 struct socket *sock = file->private_data;
1014 struct msghdr msg = {.msg_iter = *to,
1018 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1019 msg.msg_flags = MSG_DONTWAIT;
1021 if (iocb->ki_pos != 0)
1024 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1027 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1032 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1034 struct file *file = iocb->ki_filp;
1035 struct socket *sock = file->private_data;
1036 struct msghdr msg = {.msg_iter = *from,
1040 if (iocb->ki_pos != 0)
1043 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1044 msg.msg_flags = MSG_DONTWAIT;
1046 if (sock->type == SOCK_SEQPACKET)
1047 msg.msg_flags |= MSG_EOR;
1049 res = sock_sendmsg(sock, &msg);
1050 *from = msg.msg_iter;
1055 * Atomic setting of ioctl hooks to avoid race
1056 * with module unload.
1059 static DEFINE_MUTEX(br_ioctl_mutex);
1060 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1062 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1064 mutex_lock(&br_ioctl_mutex);
1065 br_ioctl_hook = hook;
1066 mutex_unlock(&br_ioctl_mutex);
1068 EXPORT_SYMBOL(brioctl_set);
1070 static DEFINE_MUTEX(vlan_ioctl_mutex);
1071 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1073 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1075 mutex_lock(&vlan_ioctl_mutex);
1076 vlan_ioctl_hook = hook;
1077 mutex_unlock(&vlan_ioctl_mutex);
1079 EXPORT_SYMBOL(vlan_ioctl_set);
1081 static long sock_do_ioctl(struct net *net, struct socket *sock,
1082 unsigned int cmd, unsigned long arg)
1085 void __user *argp = (void __user *)arg;
1087 err = sock->ops->ioctl(sock, cmd, arg);
1090 * If this ioctl is unknown try to hand it down
1091 * to the NIC driver.
1093 if (err != -ENOIOCTLCMD)
1096 if (cmd == SIOCGIFCONF) {
1098 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1101 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1103 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1108 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1110 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1111 if (!err && need_copyout)
1112 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1119 * With an ioctl, arg may well be a user mode pointer, but we don't know
1120 * what to do with it - that's up to the protocol still.
1123 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1125 struct socket *sock;
1127 void __user *argp = (void __user *)arg;
1131 sock = file->private_data;
1134 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1137 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1139 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1140 if (!err && need_copyout)
1141 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1144 #ifdef CONFIG_WEXT_CORE
1145 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1146 err = wext_handle_ioctl(net, cmd, argp);
1153 if (get_user(pid, (int __user *)argp))
1155 err = f_setown(sock->file, pid, 1);
1159 err = put_user(f_getown(sock->file),
1160 (int __user *)argp);
1168 request_module("bridge");
1170 mutex_lock(&br_ioctl_mutex);
1172 err = br_ioctl_hook(net, cmd, argp);
1173 mutex_unlock(&br_ioctl_mutex);
1178 if (!vlan_ioctl_hook)
1179 request_module("8021q");
1181 mutex_lock(&vlan_ioctl_mutex);
1182 if (vlan_ioctl_hook)
1183 err = vlan_ioctl_hook(net, argp);
1184 mutex_unlock(&vlan_ioctl_mutex);
1188 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1191 err = open_related_ns(&net->ns, get_net_ns);
1193 case SIOCGSTAMP_OLD:
1194 case SIOCGSTAMPNS_OLD:
1195 if (!sock->ops->gettstamp) {
1199 err = sock->ops->gettstamp(sock, argp,
1200 cmd == SIOCGSTAMP_OLD,
1201 !IS_ENABLED(CONFIG_64BIT));
1203 case SIOCGSTAMP_NEW:
1204 case SIOCGSTAMPNS_NEW:
1205 if (!sock->ops->gettstamp) {
1209 err = sock->ops->gettstamp(sock, argp,
1210 cmd == SIOCGSTAMP_NEW,
1214 err = sock_do_ioctl(net, sock, cmd, arg);
1221 * sock_create_lite - creates a socket
1222 * @family: protocol family (AF_INET, ...)
1223 * @type: communication type (SOCK_STREAM, ...)
1224 * @protocol: protocol (0, ...)
1227 * Creates a new socket and assigns it to @res, passing through LSM.
1228 * The new socket initialization is not complete, see kernel_accept().
1229 * Returns 0 or an error. On failure @res is set to %NULL.
1230 * This function internally uses GFP_KERNEL.
1233 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1236 struct socket *sock = NULL;
1238 err = security_socket_create(family, type, protocol, 1);
1242 sock = sock_alloc();
1249 err = security_socket_post_create(sock, family, type, protocol, 1);
1261 EXPORT_SYMBOL(sock_create_lite);
1263 /* No kernel lock held - perfect */
1264 static __poll_t sock_poll(struct file *file, poll_table *wait)
1266 struct socket *sock = file->private_data;
1267 __poll_t events = poll_requested_events(wait), flag = 0;
1269 if (!sock->ops->poll)
1272 if (sk_can_busy_loop(sock->sk)) {
1273 /* poll once if requested by the syscall */
1274 if (events & POLL_BUSY_LOOP)
1275 sk_busy_loop(sock->sk, 1);
1277 /* if this socket can poll_ll, tell the system call */
1278 flag = POLL_BUSY_LOOP;
1281 return sock->ops->poll(file, sock, wait) | flag;
1284 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1286 struct socket *sock = file->private_data;
1288 return sock->ops->mmap(file, sock, vma);
1291 static int sock_close(struct inode *inode, struct file *filp)
1293 __sock_release(SOCKET_I(inode), inode);
1298 * Update the socket async list
1300 * Fasync_list locking strategy.
1302 * 1. fasync_list is modified only under process context socket lock
1303 * i.e. under semaphore.
1304 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1305 * or under socket lock
1308 static int sock_fasync(int fd, struct file *filp, int on)
1310 struct socket *sock = filp->private_data;
1311 struct sock *sk = sock->sk;
1312 struct socket_wq *wq = &sock->wq;
1318 fasync_helper(fd, filp, on, &wq->fasync_list);
1320 if (!wq->fasync_list)
1321 sock_reset_flag(sk, SOCK_FASYNC);
1323 sock_set_flag(sk, SOCK_FASYNC);
1329 /* This function may be called only under rcu_lock */
1331 int sock_wake_async(struct socket_wq *wq, int how, int band)
1333 if (!wq || !wq->fasync_list)
1337 case SOCK_WAKE_WAITD:
1338 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1341 case SOCK_WAKE_SPACE:
1342 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1347 kill_fasync(&wq->fasync_list, SIGIO, band);
1350 kill_fasync(&wq->fasync_list, SIGURG, band);
1355 EXPORT_SYMBOL(sock_wake_async);
1358 * __sock_create - creates a socket
1359 * @net: net namespace
1360 * @family: protocol family (AF_INET, ...)
1361 * @type: communication type (SOCK_STREAM, ...)
1362 * @protocol: protocol (0, ...)
1364 * @kern: boolean for kernel space sockets
1366 * Creates a new socket and assigns it to @res, passing through LSM.
1367 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1368 * be set to true if the socket resides in kernel space.
1369 * This function internally uses GFP_KERNEL.
1372 int __sock_create(struct net *net, int family, int type, int protocol,
1373 struct socket **res, int kern)
1376 struct socket *sock;
1377 const struct net_proto_family *pf;
1380 * Check protocol is in range
1382 if (family < 0 || family >= NPROTO)
1383 return -EAFNOSUPPORT;
1384 if (type < 0 || type >= SOCK_MAX)
1389 This uglymoron is moved from INET layer to here to avoid
1390 deadlock in module load.
1392 if (family == PF_INET && type == SOCK_PACKET) {
1393 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1398 err = security_socket_create(family, type, protocol, kern);
1403 * Allocate the socket and allow the family to set things up. if
1404 * the protocol is 0, the family is instructed to select an appropriate
1407 sock = sock_alloc();
1409 net_warn_ratelimited("socket: no more sockets\n");
1410 return -ENFILE; /* Not exactly a match, but its the
1411 closest posix thing */
1416 #ifdef CONFIG_MODULES
1417 /* Attempt to load a protocol module if the find failed.
1419 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1420 * requested real, full-featured networking support upon configuration.
1421 * Otherwise module support will break!
1423 if (rcu_access_pointer(net_families[family]) == NULL)
1424 request_module("net-pf-%d", family);
1428 pf = rcu_dereference(net_families[family]);
1429 err = -EAFNOSUPPORT;
1434 * We will call the ->create function, that possibly is in a loadable
1435 * module, so we have to bump that loadable module refcnt first.
1437 if (!try_module_get(pf->owner))
1440 /* Now protected by module ref count */
1443 err = pf->create(net, sock, protocol, kern);
1445 goto out_module_put;
1448 * Now to bump the refcnt of the [loadable] module that owns this
1449 * socket at sock_release time we decrement its refcnt.
1451 if (!try_module_get(sock->ops->owner))
1452 goto out_module_busy;
1455 * Now that we're done with the ->create function, the [loadable]
1456 * module can have its refcnt decremented
1458 module_put(pf->owner);
1459 err = security_socket_post_create(sock, family, type, protocol, kern);
1461 goto out_sock_release;
1467 err = -EAFNOSUPPORT;
1470 module_put(pf->owner);
1477 goto out_sock_release;
1479 EXPORT_SYMBOL(__sock_create);
1482 * sock_create - creates a socket
1483 * @family: protocol family (AF_INET, ...)
1484 * @type: communication type (SOCK_STREAM, ...)
1485 * @protocol: protocol (0, ...)
1488 * A wrapper around __sock_create().
1489 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1492 int sock_create(int family, int type, int protocol, struct socket **res)
1494 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1496 EXPORT_SYMBOL(sock_create);
1499 * sock_create_kern - creates a socket (kernel space)
1500 * @net: net namespace
1501 * @family: protocol family (AF_INET, ...)
1502 * @type: communication type (SOCK_STREAM, ...)
1503 * @protocol: protocol (0, ...)
1506 * A wrapper around __sock_create().
1507 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1510 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1512 return __sock_create(net, family, type, protocol, res, 1);
1514 EXPORT_SYMBOL(sock_create_kern);
1516 int __sys_socket(int family, int type, int protocol)
1519 struct socket *sock;
1522 /* Check the SOCK_* constants for consistency. */
1523 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1524 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1525 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1526 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1528 flags = type & ~SOCK_TYPE_MASK;
1529 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1531 type &= SOCK_TYPE_MASK;
1533 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1534 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1536 retval = sock_create(family, type, protocol, &sock);
1540 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1543 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1545 return __sys_socket(family, type, protocol);
1549 * Create a pair of connected sockets.
1552 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1554 struct socket *sock1, *sock2;
1556 struct file *newfile1, *newfile2;
1559 flags = type & ~SOCK_TYPE_MASK;
1560 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1562 type &= SOCK_TYPE_MASK;
1564 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1565 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1568 * reserve descriptors and make sure we won't fail
1569 * to return them to userland.
1571 fd1 = get_unused_fd_flags(flags);
1572 if (unlikely(fd1 < 0))
1575 fd2 = get_unused_fd_flags(flags);
1576 if (unlikely(fd2 < 0)) {
1581 err = put_user(fd1, &usockvec[0]);
1585 err = put_user(fd2, &usockvec[1]);
1590 * Obtain the first socket and check if the underlying protocol
1591 * supports the socketpair call.
1594 err = sock_create(family, type, protocol, &sock1);
1595 if (unlikely(err < 0))
1598 err = sock_create(family, type, protocol, &sock2);
1599 if (unlikely(err < 0)) {
1600 sock_release(sock1);
1604 err = security_socket_socketpair(sock1, sock2);
1605 if (unlikely(err)) {
1606 sock_release(sock2);
1607 sock_release(sock1);
1611 err = sock1->ops->socketpair(sock1, sock2);
1612 if (unlikely(err < 0)) {
1613 sock_release(sock2);
1614 sock_release(sock1);
1618 newfile1 = sock_alloc_file(sock1, flags, NULL);
1619 if (IS_ERR(newfile1)) {
1620 err = PTR_ERR(newfile1);
1621 sock_release(sock2);
1625 newfile2 = sock_alloc_file(sock2, flags, NULL);
1626 if (IS_ERR(newfile2)) {
1627 err = PTR_ERR(newfile2);
1632 audit_fd_pair(fd1, fd2);
1634 fd_install(fd1, newfile1);
1635 fd_install(fd2, newfile2);
1644 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1645 int __user *, usockvec)
1647 return __sys_socketpair(family, type, protocol, usockvec);
1651 * Bind a name to a socket. Nothing much to do here since it's
1652 * the protocol's responsibility to handle the local address.
1654 * We move the socket address to kernel space before we call
1655 * the protocol layer (having also checked the address is ok).
1658 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1660 struct socket *sock;
1661 struct sockaddr_storage address;
1662 int err, fput_needed;
1664 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1666 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1668 err = security_socket_bind(sock,
1669 (struct sockaddr *)&address,
1672 err = sock->ops->bind(sock,
1676 fput_light(sock->file, fput_needed);
1681 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1683 return __sys_bind(fd, umyaddr, addrlen);
1687 * Perform a listen. Basically, we allow the protocol to do anything
1688 * necessary for a listen, and if that works, we mark the socket as
1689 * ready for listening.
1692 int __sys_listen(int fd, int backlog)
1694 struct socket *sock;
1695 int err, fput_needed;
1698 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1700 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1701 if ((unsigned int)backlog > somaxconn)
1702 backlog = somaxconn;
1704 err = security_socket_listen(sock, backlog);
1706 err = sock->ops->listen(sock, backlog);
1708 fput_light(sock->file, fput_needed);
1713 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1715 return __sys_listen(fd, backlog);
1718 int __sys_accept4_file(struct file *file, unsigned file_flags,
1719 struct sockaddr __user *upeer_sockaddr,
1720 int __user *upeer_addrlen, int flags,
1721 unsigned long nofile)
1723 struct socket *sock, *newsock;
1724 struct file *newfile;
1725 int err, len, newfd;
1726 struct sockaddr_storage address;
1728 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1731 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1732 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1734 sock = sock_from_file(file);
1741 newsock = sock_alloc();
1745 newsock->type = sock->type;
1746 newsock->ops = sock->ops;
1749 * We don't need try_module_get here, as the listening socket (sock)
1750 * has the protocol module (sock->ops->owner) held.
1752 __module_get(newsock->ops->owner);
1754 newfd = __get_unused_fd_flags(flags, nofile);
1755 if (unlikely(newfd < 0)) {
1757 sock_release(newsock);
1760 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1761 if (IS_ERR(newfile)) {
1762 err = PTR_ERR(newfile);
1763 put_unused_fd(newfd);
1767 err = security_socket_accept(sock, newsock);
1771 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1776 if (upeer_sockaddr) {
1777 len = newsock->ops->getname(newsock,
1778 (struct sockaddr *)&address, 2);
1780 err = -ECONNABORTED;
1783 err = move_addr_to_user(&address,
1784 len, upeer_sockaddr, upeer_addrlen);
1789 /* File flags are not inherited via accept() unlike another OSes. */
1791 fd_install(newfd, newfile);
1797 put_unused_fd(newfd);
1803 * For accept, we attempt to create a new socket, set up the link
1804 * with the client, wake up the client, then return the new
1805 * connected fd. We collect the address of the connector in kernel
1806 * space and move it to user at the very end. This is unclean because
1807 * we open the socket then return an error.
1809 * 1003.1g adds the ability to recvmsg() to query connection pending
1810 * status to recvmsg. We need to add that support in a way thats
1811 * clean when we restructure accept also.
1814 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1815 int __user *upeer_addrlen, int flags)
1822 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1823 upeer_addrlen, flags,
1824 rlimit(RLIMIT_NOFILE));
1831 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1832 int __user *, upeer_addrlen, int, flags)
1834 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1837 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1838 int __user *, upeer_addrlen)
1840 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1844 * Attempt to connect to a socket with the server address. The address
1845 * is in user space so we verify it is OK and move it to kernel space.
1847 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1850 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1851 * other SEQPACKET protocols that take time to connect() as it doesn't
1852 * include the -EINPROGRESS status for such sockets.
1855 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1856 int addrlen, int file_flags)
1858 struct socket *sock;
1861 sock = sock_from_file(file);
1868 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1872 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1873 sock->file->f_flags | file_flags);
1878 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1885 struct sockaddr_storage address;
1887 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1889 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1896 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1899 return __sys_connect(fd, uservaddr, addrlen);
1903 * Get the local address ('name') of a socket object. Move the obtained
1904 * name to user space.
1907 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1908 int __user *usockaddr_len)
1910 struct socket *sock;
1911 struct sockaddr_storage address;
1912 int err, fput_needed;
1914 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1918 err = security_socket_getsockname(sock);
1922 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1925 /* "err" is actually length in this case */
1926 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1929 fput_light(sock->file, fput_needed);
1934 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1935 int __user *, usockaddr_len)
1937 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1941 * Get the remote address ('name') of a socket object. Move the obtained
1942 * name to user space.
1945 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1946 int __user *usockaddr_len)
1948 struct socket *sock;
1949 struct sockaddr_storage address;
1950 int err, fput_needed;
1952 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1954 err = security_socket_getpeername(sock);
1956 fput_light(sock->file, fput_needed);
1960 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1962 /* "err" is actually length in this case */
1963 err = move_addr_to_user(&address, err, usockaddr,
1965 fput_light(sock->file, fput_needed);
1970 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1971 int __user *, usockaddr_len)
1973 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1977 * Send a datagram to a given address. We move the address into kernel
1978 * space and check the user space data area is readable before invoking
1981 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1982 struct sockaddr __user *addr, int addr_len)
1984 struct socket *sock;
1985 struct sockaddr_storage address;
1991 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1994 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1998 msg.msg_name = NULL;
1999 msg.msg_control = NULL;
2000 msg.msg_controllen = 0;
2001 msg.msg_namelen = 0;
2003 err = move_addr_to_kernel(addr, addr_len, &address);
2006 msg.msg_name = (struct sockaddr *)&address;
2007 msg.msg_namelen = addr_len;
2009 if (sock->file->f_flags & O_NONBLOCK)
2010 flags |= MSG_DONTWAIT;
2011 msg.msg_flags = flags;
2012 err = sock_sendmsg(sock, &msg);
2015 fput_light(sock->file, fput_needed);
2020 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2021 unsigned int, flags, struct sockaddr __user *, addr,
2024 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2028 * Send a datagram down a socket.
2031 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2032 unsigned int, flags)
2034 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2038 * Receive a frame from the socket and optionally record the address of the
2039 * sender. We verify the buffers are writable and if needed move the
2040 * sender address from kernel to user space.
2042 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2043 struct sockaddr __user *addr, int __user *addr_len)
2045 struct socket *sock;
2048 struct sockaddr_storage address;
2052 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2055 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2059 msg.msg_control = NULL;
2060 msg.msg_controllen = 0;
2061 /* Save some cycles and don't copy the address if not needed */
2062 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2063 /* We assume all kernel code knows the size of sockaddr_storage */
2064 msg.msg_namelen = 0;
2065 msg.msg_iocb = NULL;
2067 if (sock->file->f_flags & O_NONBLOCK)
2068 flags |= MSG_DONTWAIT;
2069 err = sock_recvmsg(sock, &msg, flags);
2071 if (err >= 0 && addr != NULL) {
2072 err2 = move_addr_to_user(&address,
2073 msg.msg_namelen, addr, addr_len);
2078 fput_light(sock->file, fput_needed);
2083 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2084 unsigned int, flags, struct sockaddr __user *, addr,
2085 int __user *, addr_len)
2087 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2091 * Receive a datagram from a socket.
2094 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2095 unsigned int, flags)
2097 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2100 static bool sock_use_custom_sol_socket(const struct socket *sock)
2102 const struct sock *sk = sock->sk;
2104 /* Use sock->ops->setsockopt() for MPTCP */
2105 return IS_ENABLED(CONFIG_MPTCP) &&
2106 sk->sk_protocol == IPPROTO_MPTCP &&
2107 sk->sk_type == SOCK_STREAM &&
2108 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2112 * Set a socket option. Because we don't know the option lengths we have
2113 * to pass the user mode parameter for the protocols to sort out.
2115 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2118 sockptr_t optval = USER_SOCKPTR(user_optval);
2119 char *kernel_optval = NULL;
2120 int err, fput_needed;
2121 struct socket *sock;
2126 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2130 err = security_socket_setsockopt(sock, level, optname);
2134 if (!in_compat_syscall())
2135 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2136 user_optval, &optlen,
2146 optval = KERNEL_SOCKPTR(kernel_optval);
2147 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2148 err = sock_setsockopt(sock, level, optname, optval, optlen);
2149 else if (unlikely(!sock->ops->setsockopt))
2152 err = sock->ops->setsockopt(sock, level, optname, optval,
2154 kfree(kernel_optval);
2156 fput_light(sock->file, fput_needed);
2160 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2161 char __user *, optval, int, optlen)
2163 return __sys_setsockopt(fd, level, optname, optval, optlen);
2166 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2170 * Get a socket option. Because we don't know the option lengths we have
2171 * to pass a user mode parameter for the protocols to sort out.
2173 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2176 int err, fput_needed;
2177 struct socket *sock;
2180 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2184 err = security_socket_getsockopt(sock, level, optname);
2188 if (!in_compat_syscall())
2189 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2191 if (level == SOL_SOCKET)
2192 err = sock_getsockopt(sock, level, optname, optval, optlen);
2193 else if (unlikely(!sock->ops->getsockopt))
2196 err = sock->ops->getsockopt(sock, level, optname, optval,
2199 if (!in_compat_syscall())
2200 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2201 optval, optlen, max_optlen,
2204 fput_light(sock->file, fput_needed);
2208 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2209 char __user *, optval, int __user *, optlen)
2211 return __sys_getsockopt(fd, level, optname, optval, optlen);
2215 * Shutdown a socket.
2218 int __sys_shutdown_sock(struct socket *sock, int how)
2222 err = security_socket_shutdown(sock, how);
2224 err = sock->ops->shutdown(sock, how);
2229 int __sys_shutdown(int fd, int how)
2231 int err, fput_needed;
2232 struct socket *sock;
2234 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2236 err = __sys_shutdown_sock(sock, how);
2237 fput_light(sock->file, fput_needed);
2242 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2244 return __sys_shutdown(fd, how);
2247 /* A couple of helpful macros for getting the address of the 32/64 bit
2248 * fields which are the same type (int / unsigned) on our platforms.
2250 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2251 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2252 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2254 struct used_address {
2255 struct sockaddr_storage name;
2256 unsigned int name_len;
2259 int __copy_msghdr_from_user(struct msghdr *kmsg,
2260 struct user_msghdr __user *umsg,
2261 struct sockaddr __user **save_addr,
2262 struct iovec __user **uiov, size_t *nsegs)
2264 struct user_msghdr msg;
2267 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2270 kmsg->msg_control_is_user = true;
2271 kmsg->msg_control_user = msg.msg_control;
2272 kmsg->msg_controllen = msg.msg_controllen;
2273 kmsg->msg_flags = msg.msg_flags;
2275 kmsg->msg_namelen = msg.msg_namelen;
2277 kmsg->msg_namelen = 0;
2279 if (kmsg->msg_namelen < 0)
2282 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2283 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2286 *save_addr = msg.msg_name;
2288 if (msg.msg_name && kmsg->msg_namelen) {
2290 err = move_addr_to_kernel(msg.msg_name,
2297 kmsg->msg_name = NULL;
2298 kmsg->msg_namelen = 0;
2301 if (msg.msg_iovlen > UIO_MAXIOV)
2304 kmsg->msg_iocb = NULL;
2305 *uiov = msg.msg_iov;
2306 *nsegs = msg.msg_iovlen;
2310 static int copy_msghdr_from_user(struct msghdr *kmsg,
2311 struct user_msghdr __user *umsg,
2312 struct sockaddr __user **save_addr,
2315 struct user_msghdr msg;
2318 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2323 err = import_iovec(save_addr ? READ : WRITE,
2324 msg.msg_iov, msg.msg_iovlen,
2325 UIO_FASTIOV, iov, &kmsg->msg_iter);
2326 return err < 0 ? err : 0;
2329 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2330 unsigned int flags, struct used_address *used_address,
2331 unsigned int allowed_msghdr_flags)
2333 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2334 __aligned(sizeof(__kernel_size_t));
2335 /* 20 is size of ipv6_pktinfo */
2336 unsigned char *ctl_buf = ctl;
2342 if (msg_sys->msg_controllen > INT_MAX)
2344 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2345 ctl_len = msg_sys->msg_controllen;
2346 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2348 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2352 ctl_buf = msg_sys->msg_control;
2353 ctl_len = msg_sys->msg_controllen;
2354 } else if (ctl_len) {
2355 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2356 CMSG_ALIGN(sizeof(struct cmsghdr)));
2357 if (ctl_len > sizeof(ctl)) {
2358 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2359 if (ctl_buf == NULL)
2363 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2365 msg_sys->msg_control = ctl_buf;
2366 msg_sys->msg_control_is_user = false;
2368 msg_sys->msg_flags = flags;
2370 if (sock->file->f_flags & O_NONBLOCK)
2371 msg_sys->msg_flags |= MSG_DONTWAIT;
2373 * If this is sendmmsg() and current destination address is same as
2374 * previously succeeded address, omit asking LSM's decision.
2375 * used_address->name_len is initialized to UINT_MAX so that the first
2376 * destination address never matches.
2378 if (used_address && msg_sys->msg_name &&
2379 used_address->name_len == msg_sys->msg_namelen &&
2380 !memcmp(&used_address->name, msg_sys->msg_name,
2381 used_address->name_len)) {
2382 err = sock_sendmsg_nosec(sock, msg_sys);
2385 err = sock_sendmsg(sock, msg_sys);
2387 * If this is sendmmsg() and sending to current destination address was
2388 * successful, remember it.
2390 if (used_address && err >= 0) {
2391 used_address->name_len = msg_sys->msg_namelen;
2392 if (msg_sys->msg_name)
2393 memcpy(&used_address->name, msg_sys->msg_name,
2394 used_address->name_len);
2399 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2404 int sendmsg_copy_msghdr(struct msghdr *msg,
2405 struct user_msghdr __user *umsg, unsigned flags,
2410 if (flags & MSG_CMSG_COMPAT) {
2411 struct compat_msghdr __user *msg_compat;
2413 msg_compat = (struct compat_msghdr __user *) umsg;
2414 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2416 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2424 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2425 struct msghdr *msg_sys, unsigned int flags,
2426 struct used_address *used_address,
2427 unsigned int allowed_msghdr_flags)
2429 struct sockaddr_storage address;
2430 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2433 msg_sys->msg_name = &address;
2435 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2439 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2440 allowed_msghdr_flags);
2446 * BSD sendmsg interface
2448 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2451 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2454 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2455 bool forbid_cmsg_compat)
2457 int fput_needed, err;
2458 struct msghdr msg_sys;
2459 struct socket *sock;
2461 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2464 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2468 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2470 fput_light(sock->file, fput_needed);
2475 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2477 return __sys_sendmsg(fd, msg, flags, true);
2481 * Linux sendmmsg interface
2484 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2485 unsigned int flags, bool forbid_cmsg_compat)
2487 int fput_needed, err, datagrams;
2488 struct socket *sock;
2489 struct mmsghdr __user *entry;
2490 struct compat_mmsghdr __user *compat_entry;
2491 struct msghdr msg_sys;
2492 struct used_address used_address;
2493 unsigned int oflags = flags;
2495 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2498 if (vlen > UIO_MAXIOV)
2503 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2507 used_address.name_len = UINT_MAX;
2509 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2513 while (datagrams < vlen) {
2514 if (datagrams == vlen - 1)
2517 if (MSG_CMSG_COMPAT & flags) {
2518 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2519 &msg_sys, flags, &used_address, MSG_EOR);
2522 err = __put_user(err, &compat_entry->msg_len);
2525 err = ___sys_sendmsg(sock,
2526 (struct user_msghdr __user *)entry,
2527 &msg_sys, flags, &used_address, MSG_EOR);
2530 err = put_user(err, &entry->msg_len);
2537 if (msg_data_left(&msg_sys))
2542 fput_light(sock->file, fput_needed);
2544 /* We only return an error if no datagrams were able to be sent */
2551 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2552 unsigned int, vlen, unsigned int, flags)
2554 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2557 int recvmsg_copy_msghdr(struct msghdr *msg,
2558 struct user_msghdr __user *umsg, unsigned flags,
2559 struct sockaddr __user **uaddr,
2564 if (MSG_CMSG_COMPAT & flags) {
2565 struct compat_msghdr __user *msg_compat;
2567 msg_compat = (struct compat_msghdr __user *) umsg;
2568 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2570 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2578 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2579 struct user_msghdr __user *msg,
2580 struct sockaddr __user *uaddr,
2581 unsigned int flags, int nosec)
2583 struct compat_msghdr __user *msg_compat =
2584 (struct compat_msghdr __user *) msg;
2585 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2586 struct sockaddr_storage addr;
2587 unsigned long cmsg_ptr;
2591 msg_sys->msg_name = &addr;
2592 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2593 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2595 /* We assume all kernel code knows the size of sockaddr_storage */
2596 msg_sys->msg_namelen = 0;
2598 if (sock->file->f_flags & O_NONBLOCK)
2599 flags |= MSG_DONTWAIT;
2601 if (unlikely(nosec))
2602 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2604 err = sock_recvmsg(sock, msg_sys, flags);
2610 if (uaddr != NULL) {
2611 err = move_addr_to_user(&addr,
2612 msg_sys->msg_namelen, uaddr,
2617 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2621 if (MSG_CMSG_COMPAT & flags)
2622 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2623 &msg_compat->msg_controllen);
2625 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2626 &msg->msg_controllen);
2634 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2635 struct msghdr *msg_sys, unsigned int flags, int nosec)
2637 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2638 /* user mode address pointers */
2639 struct sockaddr __user *uaddr;
2642 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2646 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2652 * BSD recvmsg interface
2655 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2656 struct user_msghdr __user *umsg,
2657 struct sockaddr __user *uaddr, unsigned int flags)
2659 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2662 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2663 bool forbid_cmsg_compat)
2665 int fput_needed, err;
2666 struct msghdr msg_sys;
2667 struct socket *sock;
2669 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2672 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2676 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2678 fput_light(sock->file, fput_needed);
2683 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2684 unsigned int, flags)
2686 return __sys_recvmsg(fd, msg, flags, true);
2690 * Linux recvmmsg interface
2693 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2694 unsigned int vlen, unsigned int flags,
2695 struct timespec64 *timeout)
2697 int fput_needed, err, datagrams;
2698 struct socket *sock;
2699 struct mmsghdr __user *entry;
2700 struct compat_mmsghdr __user *compat_entry;
2701 struct msghdr msg_sys;
2702 struct timespec64 end_time;
2703 struct timespec64 timeout64;
2706 poll_select_set_timeout(&end_time, timeout->tv_sec,
2712 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2716 if (likely(!(flags & MSG_ERRQUEUE))) {
2717 err = sock_error(sock->sk);
2725 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2727 while (datagrams < vlen) {
2729 * No need to ask LSM for more than the first datagram.
2731 if (MSG_CMSG_COMPAT & flags) {
2732 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2733 &msg_sys, flags & ~MSG_WAITFORONE,
2737 err = __put_user(err, &compat_entry->msg_len);
2740 err = ___sys_recvmsg(sock,
2741 (struct user_msghdr __user *)entry,
2742 &msg_sys, flags & ~MSG_WAITFORONE,
2746 err = put_user(err, &entry->msg_len);
2754 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2755 if (flags & MSG_WAITFORONE)
2756 flags |= MSG_DONTWAIT;
2759 ktime_get_ts64(&timeout64);
2760 *timeout = timespec64_sub(end_time, timeout64);
2761 if (timeout->tv_sec < 0) {
2762 timeout->tv_sec = timeout->tv_nsec = 0;
2766 /* Timeout, return less than vlen datagrams */
2767 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2771 /* Out of band data, return right away */
2772 if (msg_sys.msg_flags & MSG_OOB)
2780 if (datagrams == 0) {
2786 * We may return less entries than requested (vlen) if the
2787 * sock is non block and there aren't enough datagrams...
2789 if (err != -EAGAIN) {
2791 * ... or if recvmsg returns an error after we
2792 * received some datagrams, where we record the
2793 * error to return on the next call or if the
2794 * app asks about it using getsockopt(SO_ERROR).
2796 sock->sk->sk_err = -err;
2799 fput_light(sock->file, fput_needed);
2804 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2805 unsigned int vlen, unsigned int flags,
2806 struct __kernel_timespec __user *timeout,
2807 struct old_timespec32 __user *timeout32)
2810 struct timespec64 timeout_sys;
2812 if (timeout && get_timespec64(&timeout_sys, timeout))
2815 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2818 if (!timeout && !timeout32)
2819 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2821 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2826 if (timeout && put_timespec64(&timeout_sys, timeout))
2827 datagrams = -EFAULT;
2829 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2830 datagrams = -EFAULT;
2835 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2836 unsigned int, vlen, unsigned int, flags,
2837 struct __kernel_timespec __user *, timeout)
2839 if (flags & MSG_CMSG_COMPAT)
2842 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2845 #ifdef CONFIG_COMPAT_32BIT_TIME
2846 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2847 unsigned int, vlen, unsigned int, flags,
2848 struct old_timespec32 __user *, timeout)
2850 if (flags & MSG_CMSG_COMPAT)
2853 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2857 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2858 /* Argument list sizes for sys_socketcall */
2859 #define AL(x) ((x) * sizeof(unsigned long))
2860 static const unsigned char nargs[21] = {
2861 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2862 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2863 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2870 * System call vectors.
2872 * Argument checking cleaned up. Saved 20% in size.
2873 * This function doesn't need to set the kernel lock because
2874 * it is set by the callees.
2877 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2879 unsigned long a[AUDITSC_ARGS];
2880 unsigned long a0, a1;
2884 if (call < 1 || call > SYS_SENDMMSG)
2886 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2889 if (len > sizeof(a))
2892 /* copy_from_user should be SMP safe. */
2893 if (copy_from_user(a, args, len))
2896 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2905 err = __sys_socket(a0, a1, a[2]);
2908 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2911 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2914 err = __sys_listen(a0, a1);
2917 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2918 (int __user *)a[2], 0);
2920 case SYS_GETSOCKNAME:
2922 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2923 (int __user *)a[2]);
2925 case SYS_GETPEERNAME:
2927 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2928 (int __user *)a[2]);
2930 case SYS_SOCKETPAIR:
2931 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2934 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2938 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2939 (struct sockaddr __user *)a[4], a[5]);
2942 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2946 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2947 (struct sockaddr __user *)a[4],
2948 (int __user *)a[5]);
2951 err = __sys_shutdown(a0, a1);
2953 case SYS_SETSOCKOPT:
2954 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2957 case SYS_GETSOCKOPT:
2959 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2960 (int __user *)a[4]);
2963 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2967 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2971 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2975 if (IS_ENABLED(CONFIG_64BIT))
2976 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2978 (struct __kernel_timespec __user *)a[4],
2981 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2983 (struct old_timespec32 __user *)a[4]);
2986 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2987 (int __user *)a[2], a[3]);
2996 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2999 * sock_register - add a socket protocol handler
3000 * @ops: description of protocol
3002 * This function is called by a protocol handler that wants to
3003 * advertise its address family, and have it linked into the
3004 * socket interface. The value ops->family corresponds to the
3005 * socket system call protocol family.
3007 int sock_register(const struct net_proto_family *ops)
3011 if (ops->family >= NPROTO) {
3012 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3016 spin_lock(&net_family_lock);
3017 if (rcu_dereference_protected(net_families[ops->family],
3018 lockdep_is_held(&net_family_lock)))
3021 rcu_assign_pointer(net_families[ops->family], ops);
3024 spin_unlock(&net_family_lock);
3026 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3029 EXPORT_SYMBOL(sock_register);
3032 * sock_unregister - remove a protocol handler
3033 * @family: protocol family to remove
3035 * This function is called by a protocol handler that wants to
3036 * remove its address family, and have it unlinked from the
3037 * new socket creation.
3039 * If protocol handler is a module, then it can use module reference
3040 * counts to protect against new references. If protocol handler is not
3041 * a module then it needs to provide its own protection in
3042 * the ops->create routine.
3044 void sock_unregister(int family)
3046 BUG_ON(family < 0 || family >= NPROTO);
3048 spin_lock(&net_family_lock);
3049 RCU_INIT_POINTER(net_families[family], NULL);
3050 spin_unlock(&net_family_lock);
3054 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3056 EXPORT_SYMBOL(sock_unregister);
3058 bool sock_is_registered(int family)
3060 return family < NPROTO && rcu_access_pointer(net_families[family]);
3063 static int __init sock_init(void)
3067 * Initialize the network sysctl infrastructure.
3069 err = net_sysctl_init();
3074 * Initialize skbuff SLAB cache
3079 * Initialize the protocols module.
3084 err = register_filesystem(&sock_fs_type);
3087 sock_mnt = kern_mount(&sock_fs_type);
3088 if (IS_ERR(sock_mnt)) {
3089 err = PTR_ERR(sock_mnt);
3093 /* The real protocol initialization is performed in later initcalls.
3096 #ifdef CONFIG_NETFILTER
3097 err = netfilter_init();
3102 ptp_classifier_init();
3108 unregister_filesystem(&sock_fs_type);
3112 core_initcall(sock_init); /* early initcall */
3114 #ifdef CONFIG_PROC_FS
3115 void socket_seq_show(struct seq_file *seq)
3117 seq_printf(seq, "sockets: used %d\n",
3118 sock_inuse_get(seq->private));
3120 #endif /* CONFIG_PROC_FS */
3122 #ifdef CONFIG_COMPAT
3123 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3125 struct compat_ifconf ifc32;
3129 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3132 ifc.ifc_len = ifc32.ifc_len;
3133 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3136 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3141 ifc32.ifc_len = ifc.ifc_len;
3142 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3148 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3150 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3151 bool convert_in = false, convert_out = false;
3152 size_t buf_size = 0;
3153 struct ethtool_rxnfc __user *rxnfc = NULL;
3155 u32 rule_cnt = 0, actual_rule_cnt;
3160 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3163 compat_rxnfc = compat_ptr(data);
3165 if (get_user(ethcmd, &compat_rxnfc->cmd))
3168 /* Most ethtool structures are defined without padding.
3169 * Unfortunately struct ethtool_rxnfc is an exception.
3174 case ETHTOOL_GRXCLSRLALL:
3175 /* Buffer size is variable */
3176 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3178 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3180 buf_size += rule_cnt * sizeof(u32);
3182 case ETHTOOL_GRXRINGS:
3183 case ETHTOOL_GRXCLSRLCNT:
3184 case ETHTOOL_GRXCLSRULE:
3185 case ETHTOOL_SRXCLSRLINS:
3188 case ETHTOOL_SRXCLSRLDEL:
3189 buf_size += sizeof(struct ethtool_rxnfc);
3191 rxnfc = compat_alloc_user_space(buf_size);
3195 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3198 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3201 /* We expect there to be holes between fs.m_ext and
3202 * fs.ring_cookie and at the end of fs, but nowhere else.
3204 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3205 sizeof(compat_rxnfc->fs.m_ext) !=
3206 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3207 sizeof(rxnfc->fs.m_ext));
3209 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3210 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3211 offsetof(struct ethtool_rxnfc, fs.location) -
3212 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3214 if (copy_in_user(rxnfc, compat_rxnfc,
3215 (void __user *)(&rxnfc->fs.m_ext + 1) -
3216 (void __user *)rxnfc) ||
3217 copy_in_user(&rxnfc->fs.ring_cookie,
3218 &compat_rxnfc->fs.ring_cookie,
3219 (void __user *)(&rxnfc->fs.location + 1) -
3220 (void __user *)&rxnfc->fs.ring_cookie))
3222 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3223 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3225 } else if (copy_in_user(&rxnfc->rule_cnt,
3226 &compat_rxnfc->rule_cnt,
3227 sizeof(rxnfc->rule_cnt)))
3231 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3236 if (copy_in_user(compat_rxnfc, rxnfc,
3237 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3238 (const void __user *)rxnfc) ||
3239 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3240 &rxnfc->fs.ring_cookie,
3241 (const void __user *)(&rxnfc->fs.location + 1) -
3242 (const void __user *)&rxnfc->fs.ring_cookie) ||
3243 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3244 sizeof(rxnfc->rule_cnt)))
3247 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3248 /* As an optimisation, we only copy the actual
3249 * number of rules that the underlying
3250 * function returned. Since Mallory might
3251 * change the rule count in user memory, we
3252 * check that it is less than the rule count
3253 * originally given (as the user buffer size),
3254 * which has been range-checked.
3256 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3258 if (actual_rule_cnt < rule_cnt)
3259 rule_cnt = actual_rule_cnt;
3260 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3261 &rxnfc->rule_locs[0],
3262 rule_cnt * sizeof(u32)))
3270 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3272 compat_uptr_t uptr32;
3277 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3280 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3283 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3284 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3286 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3288 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3289 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3295 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3296 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3297 struct compat_ifreq __user *u_ifreq32)
3302 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3304 if (get_user(data32, &u_ifreq32->ifr_data))
3306 ifreq.ifr_data = compat_ptr(data32);
3308 return dev_ioctl(net, cmd, &ifreq, NULL);
3311 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3313 struct compat_ifreq __user *uifr32)
3315 struct ifreq __user *uifr;
3318 /* Handle the fact that while struct ifreq has the same *layout* on
3319 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3320 * which are handled elsewhere, it still has different *size* due to
3321 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3322 * resulting in struct ifreq being 32 and 40 bytes respectively).
3323 * As a result, if the struct happens to be at the end of a page and
3324 * the next page isn't readable/writable, we get a fault. To prevent
3325 * that, copy back and forth to the full size.
3328 uifr = compat_alloc_user_space(sizeof(*uifr));
3329 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3332 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3343 case SIOCGIFBRDADDR:
3344 case SIOCGIFDSTADDR:
3345 case SIOCGIFNETMASK:
3351 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3359 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3360 struct compat_ifreq __user *uifr32)
3363 struct compat_ifmap __user *uifmap32;
3366 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3367 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3368 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3369 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3370 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3371 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3372 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3373 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3377 err = dev_ioctl(net, cmd, &ifr, NULL);
3379 if (cmd == SIOCGIFMAP && !err) {
3380 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3381 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3382 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3383 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3384 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3385 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3386 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3393 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3394 * for some operations; this forces use of the newer bridge-utils that
3395 * use compatible ioctls
3397 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3401 if (get_user(tmp, argp))
3403 if (tmp == BRCTL_GET_VERSION)
3404 return BRCTL_VERSION + 1;
3408 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3409 unsigned int cmd, unsigned long arg)
3411 void __user *argp = compat_ptr(arg);
3412 struct sock *sk = sock->sk;
3413 struct net *net = sock_net(sk);
3415 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3416 return compat_ifr_data_ioctl(net, cmd, argp);
3421 return old_bridge_ioctl(argp);
3423 return compat_dev_ifconf(net, argp);
3425 return ethtool_ioctl(net, argp);
3427 return compat_siocwandev(net, argp);
3430 return compat_sioc_ifmap(net, cmd, argp);
3431 case SIOCGSTAMP_OLD:
3432 case SIOCGSTAMPNS_OLD:
3433 if (!sock->ops->gettstamp)
3434 return -ENOIOCTLCMD;
3435 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3436 !COMPAT_USE_64BIT_TIME);
3438 case SIOCBONDSLAVEINFOQUERY:
3439 case SIOCBONDINFOQUERY:
3442 return compat_ifr_data_ioctl(net, cmd, argp);
3453 case SIOCGSTAMP_NEW:
3454 case SIOCGSTAMPNS_NEW:
3455 return sock_ioctl(file, cmd, arg);
3472 case SIOCSIFHWBROADCAST:
3474 case SIOCGIFBRDADDR:
3475 case SIOCSIFBRDADDR:
3476 case SIOCGIFDSTADDR:
3477 case SIOCSIFDSTADDR:
3478 case SIOCGIFNETMASK:
3479 case SIOCSIFNETMASK:
3491 case SIOCBONDENSLAVE:
3492 case SIOCBONDRELEASE:
3493 case SIOCBONDSETHWADDR:
3494 case SIOCBONDCHANGEACTIVE:
3495 return compat_ifreq_ioctl(net, sock, cmd, argp);
3503 return sock_do_ioctl(net, sock, cmd, arg);
3506 return -ENOIOCTLCMD;
3509 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3512 struct socket *sock = file->private_data;
3513 int ret = -ENOIOCTLCMD;
3520 if (sock->ops->compat_ioctl)
3521 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3523 if (ret == -ENOIOCTLCMD &&
3524 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3525 ret = compat_wext_handle_ioctl(net, cmd, arg);
3527 if (ret == -ENOIOCTLCMD)
3528 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3535 * kernel_bind - bind an address to a socket (kernel space)
3538 * @addrlen: length of address
3540 * Returns 0 or an error.
3543 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3545 return sock->ops->bind(sock, addr, addrlen);
3547 EXPORT_SYMBOL(kernel_bind);
3550 * kernel_listen - move socket to listening state (kernel space)
3552 * @backlog: pending connections queue size
3554 * Returns 0 or an error.
3557 int kernel_listen(struct socket *sock, int backlog)
3559 return sock->ops->listen(sock, backlog);
3561 EXPORT_SYMBOL(kernel_listen);
3564 * kernel_accept - accept a connection (kernel space)
3565 * @sock: listening socket
3566 * @newsock: new connected socket
3569 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3570 * If it fails, @newsock is guaranteed to be %NULL.
3571 * Returns 0 or an error.
3574 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3576 struct sock *sk = sock->sk;
3579 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3584 err = sock->ops->accept(sock, *newsock, flags, true);
3586 sock_release(*newsock);
3591 (*newsock)->ops = sock->ops;
3592 __module_get((*newsock)->ops->owner);
3597 EXPORT_SYMBOL(kernel_accept);
3600 * kernel_connect - connect a socket (kernel space)
3603 * @addrlen: address length
3604 * @flags: flags (O_NONBLOCK, ...)
3606 * For datagram sockets, @addr is the address to which datagrams are sent
3607 * by default, and the only address from which datagrams are received.
3608 * For stream sockets, attempts to connect to @addr.
3609 * Returns 0 or an error code.
3612 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3615 return sock->ops->connect(sock, addr, addrlen, flags);
3617 EXPORT_SYMBOL(kernel_connect);
3620 * kernel_getsockname - get the address which the socket is bound (kernel space)
3622 * @addr: address holder
3624 * Fills the @addr pointer with the address which the socket is bound.
3625 * Returns 0 or an error code.
3628 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3630 return sock->ops->getname(sock, addr, 0);
3632 EXPORT_SYMBOL(kernel_getsockname);
3635 * kernel_getpeername - get the address which the socket is connected (kernel space)
3637 * @addr: address holder
3639 * Fills the @addr pointer with the address which the socket is connected.
3640 * Returns 0 or an error code.
3643 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3645 return sock->ops->getname(sock, addr, 1);
3647 EXPORT_SYMBOL(kernel_getpeername);
3650 * kernel_sendpage - send a &page through a socket (kernel space)
3653 * @offset: page offset
3654 * @size: total size in bytes
3655 * @flags: flags (MSG_DONTWAIT, ...)
3657 * Returns the total amount sent in bytes or an error.
3660 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3661 size_t size, int flags)
3663 if (sock->ops->sendpage) {
3664 /* Warn in case the improper page to zero-copy send */
3665 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3666 return sock->ops->sendpage(sock, page, offset, size, flags);
3668 return sock_no_sendpage(sock, page, offset, size, flags);
3670 EXPORT_SYMBOL(kernel_sendpage);
3673 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3676 * @offset: page offset
3677 * @size: total size in bytes
3678 * @flags: flags (MSG_DONTWAIT, ...)
3680 * Returns the total amount sent in bytes or an error.
3681 * Caller must hold @sk.
3684 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3685 size_t size, int flags)
3687 struct socket *sock = sk->sk_socket;
3689 if (sock->ops->sendpage_locked)
3690 return sock->ops->sendpage_locked(sk, page, offset, size,
3693 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3695 EXPORT_SYMBOL(kernel_sendpage_locked);
3698 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3700 * @how: connection part
3702 * Returns 0 or an error.
3705 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3707 return sock->ops->shutdown(sock, how);
3709 EXPORT_SYMBOL(kernel_sock_shutdown);
3712 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3715 * This routine returns the IP overhead imposed by a socket i.e.
3716 * the length of the underlying IP header, depending on whether
3717 * this is an IPv4 or IPv6 socket and the length from IP options turned
3718 * on at the socket. Assumes that the caller has a lock on the socket.
3721 u32 kernel_sock_ip_overhead(struct sock *sk)
3723 struct inet_sock *inet;
3724 struct ip_options_rcu *opt;
3726 #if IS_ENABLED(CONFIG_IPV6)
3727 struct ipv6_pinfo *np;
3728 struct ipv6_txoptions *optv6 = NULL;
3729 #endif /* IS_ENABLED(CONFIG_IPV6) */
3734 switch (sk->sk_family) {
3737 overhead += sizeof(struct iphdr);
3738 opt = rcu_dereference_protected(inet->inet_opt,
3739 sock_owned_by_user(sk));
3741 overhead += opt->opt.optlen;
3743 #if IS_ENABLED(CONFIG_IPV6)
3746 overhead += sizeof(struct ipv6hdr);
3748 optv6 = rcu_dereference_protected(np->opt,
3749 sock_owned_by_user(sk));
3751 overhead += (optv6->opt_flen + optv6->opt_nflen);
3753 #endif /* IS_ENABLED(CONFIG_IPV6) */
3754 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3758 EXPORT_SYMBOL(kernel_sock_ip_overhead);