1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
56 #include <linux/socket.h>
57 #include <linux/file.h>
58 #include <linux/net.h>
59 #include <linux/interrupt.h>
60 #include <linux/thread_info.h>
61 #include <linux/rcupdate.h>
62 #include <linux/netdevice.h>
63 #include <linux/proc_fs.h>
64 #include <linux/seq_file.h>
65 #include <linux/mutex.h>
66 #include <linux/if_bridge.h>
67 #include <linux/if_frad.h>
68 #include <linux/if_vlan.h>
69 #include <linux/ptp_classify.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74 #include <linux/highmem.h>
75 #include <linux/mount.h>
76 #include <linux/pseudo_fs.h>
77 #include <linux/security.h>
78 #include <linux/syscalls.h>
79 #include <linux/compat.h>
80 #include <linux/kmod.h>
81 #include <linux/audit.h>
82 #include <linux/wireless.h>
83 #include <linux/nsproxy.h>
84 #include <linux/magic.h>
85 #include <linux/slab.h>
86 #include <linux/xattr.h>
87 #include <linux/nospec.h>
88 #include <linux/indirect_call_wrapper.h>
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
95 #include <net/cls_cgroup.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/termios.h>
104 #include <linux/sockios.h>
105 #include <net/busy_poll.h>
106 #include <linux/errqueue.h>
108 #ifdef CONFIG_NET_RX_BUSY_POLL
109 unsigned int sysctl_net_busy_read __read_mostly;
110 unsigned int sysctl_net_busy_poll __read_mostly;
113 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
115 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
117 static int sock_close(struct inode *inode, struct file *file);
118 static __poll_t sock_poll(struct file *file,
119 struct poll_table_struct *wait);
120 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
122 static long compat_sock_ioctl(struct file *file,
123 unsigned int cmd, unsigned long arg);
125 static int sock_fasync(int fd, struct file *filp, int on);
126 static ssize_t sock_sendpage(struct file *file, struct page *page,
127 int offset, size_t size, loff_t *ppos, int more);
128 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
129 struct pipe_inode_info *pipe, size_t len,
132 #ifdef CONFIG_PROC_FS
133 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135 struct socket *sock = f->private_data;
137 if (sock->ops->show_fdinfo)
138 sock->ops->show_fdinfo(m, sock);
141 #define sock_show_fdinfo NULL
145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 * in the operation structures but are done directly via the socketcall() multiplexor.
149 static const struct file_operations socket_file_ops = {
150 .owner = THIS_MODULE,
152 .read_iter = sock_read_iter,
153 .write_iter = sock_write_iter,
155 .unlocked_ioctl = sock_ioctl,
157 .compat_ioctl = compat_sock_ioctl,
160 .release = sock_close,
161 .fasync = sock_fasync,
162 .sendpage = sock_sendpage,
163 .splice_write = generic_splice_sendpage,
164 .splice_read = sock_splice_read,
165 .show_fdinfo = sock_show_fdinfo,
169 * The protocol list. Each protocol is registered in here.
172 static DEFINE_SPINLOCK(net_family_lock);
173 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
198 if (copy_from_user(kaddr, uaddr, ulen))
200 return audit_sockaddr(ulen, kaddr);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221 void __user *uaddr, int __user *ulen)
226 BUG_ON(klen > sizeof(struct sockaddr_storage));
227 err = get_user(len, ulen);
235 if (audit_sockaddr(klen, kaddr))
237 if (copy_to_user(uaddr, kaddr, len))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen, ulen);
247 static struct kmem_cache *sock_inode_cachep __ro_after_init;
249 static struct inode *sock_alloc_inode(struct super_block *sb)
251 struct socket_alloc *ei;
253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
256 init_waitqueue_head(&ei->socket.wq.wait);
257 ei->socket.wq.fasync_list = NULL;
258 ei->socket.wq.flags = 0;
260 ei->socket.state = SS_UNCONNECTED;
261 ei->socket.flags = 0;
262 ei->socket.ops = NULL;
263 ei->socket.sk = NULL;
264 ei->socket.file = NULL;
266 return &ei->vfs_inode;
269 static void sock_free_inode(struct inode *inode)
271 struct socket_alloc *ei;
273 ei = container_of(inode, struct socket_alloc, vfs_inode);
274 kmem_cache_free(sock_inode_cachep, ei);
277 static void init_once(void *foo)
279 struct socket_alloc *ei = (struct socket_alloc *)foo;
281 inode_init_once(&ei->vfs_inode);
284 static void init_inodecache(void)
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc),
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
291 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
293 BUG_ON(sock_inode_cachep == NULL);
296 static const struct super_operations sockfs_ops = {
297 .alloc_inode = sock_alloc_inode,
298 .free_inode = sock_free_inode,
299 .statfs = simple_statfs,
303 * sockfs_dname() is called from d_path().
305 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
307 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
308 d_inode(dentry)->i_ino);
311 static const struct dentry_operations sockfs_dentry_operations = {
312 .d_dname = sockfs_dname,
315 static int sockfs_xattr_get(const struct xattr_handler *handler,
316 struct dentry *dentry, struct inode *inode,
317 const char *suffix, void *value, size_t size)
320 if (dentry->d_name.len + 1 > size)
322 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
324 return dentry->d_name.len + 1;
327 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
331 static const struct xattr_handler sockfs_xattr_handler = {
332 .name = XATTR_NAME_SOCKPROTONAME,
333 .get = sockfs_xattr_get,
336 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337 struct dentry *dentry, struct inode *inode,
338 const char *suffix, const void *value,
339 size_t size, int flags)
341 /* Handled by LSM. */
345 static const struct xattr_handler sockfs_security_xattr_handler = {
346 .prefix = XATTR_SECURITY_PREFIX,
347 .set = sockfs_security_xattr_set,
350 static const struct xattr_handler *sockfs_xattr_handlers[] = {
351 &sockfs_xattr_handler,
352 &sockfs_security_xattr_handler,
356 static int sockfs_init_fs_context(struct fs_context *fc)
358 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
361 ctx->ops = &sockfs_ops;
362 ctx->dops = &sockfs_dentry_operations;
363 ctx->xattr = sockfs_xattr_handlers;
367 static struct vfsmount *sock_mnt __read_mostly;
369 static struct file_system_type sock_fs_type = {
371 .init_fs_context = sockfs_init_fs_context,
372 .kill_sb = kill_anon_super,
376 * Obtains the first available file descriptor and sets it up for use.
378 * These functions create file structures and maps them to fd space
379 * of the current process. On success it returns file descriptor
380 * and file struct implicitly stored in sock->file.
381 * Note that another thread may close file descriptor before we return
382 * from this function. We use the fact that now we do not refer
383 * to socket after mapping. If one day we will need it, this
384 * function will increment ref. count on file by 1.
386 * In any case returned fd MAY BE not valid!
387 * This race condition is unavoidable
388 * with shared fd spaces, we cannot solve it inside kernel,
389 * but we take care of internal coherence yet.
393 * sock_alloc_file - Bind a &socket to a &file
395 * @flags: file status flags
396 * @dname: protocol name
398 * Returns the &file bound with @sock, implicitly storing it
399 * in sock->file. If dname is %NULL, sets to "".
400 * On failure the return is a ERR pointer (see linux/err.h).
401 * This function uses GFP_KERNEL internally.
404 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
409 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
411 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
412 O_RDWR | (flags & O_NONBLOCK),
420 file->private_data = sock;
421 stream_open(SOCK_INODE(sock), file);
424 EXPORT_SYMBOL(sock_alloc_file);
426 static int sock_map_fd(struct socket *sock, int flags)
428 struct file *newfile;
429 int fd = get_unused_fd_flags(flags);
430 if (unlikely(fd < 0)) {
435 newfile = sock_alloc_file(sock, flags, NULL);
436 if (!IS_ERR(newfile)) {
437 fd_install(fd, newfile);
442 return PTR_ERR(newfile);
446 * sock_from_file - Return the &socket bounded to @file.
448 * @err: pointer to an error code return
450 * On failure returns %NULL and assigns -ENOTSOCK to @err.
453 struct socket *sock_from_file(struct file *file, int *err)
455 if (file->f_op == &socket_file_ops)
456 return file->private_data; /* set in sock_map_fd */
461 EXPORT_SYMBOL(sock_from_file);
464 * sockfd_lookup - Go from a file number to its socket slot
466 * @err: pointer to an error code return
468 * The file handle passed in is locked and the socket it is bound
469 * to is returned. If an error occurs the err pointer is overwritten
470 * with a negative errno code and NULL is returned. The function checks
471 * for both invalid handles and passing a handle which is not a socket.
473 * On a success the socket object pointer is returned.
476 struct socket *sockfd_lookup(int fd, int *err)
487 sock = sock_from_file(file, err);
492 EXPORT_SYMBOL(sockfd_lookup);
494 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
496 struct fd f = fdget(fd);
501 sock = sock_from_file(f.file, err);
503 *fput_needed = f.flags;
511 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
517 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
527 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
532 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
539 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
541 int err = simple_setattr(dentry, iattr);
543 if (!err && (iattr->ia_valid & ATTR_UID)) {
544 struct socket *sock = SOCKET_I(d_inode(dentry));
547 sock->sk->sk_uid = iattr->ia_uid;
555 static const struct inode_operations sockfs_inode_ops = {
556 .listxattr = sockfs_listxattr,
557 .setattr = sockfs_setattr,
561 * sock_alloc - allocate a socket
563 * Allocate a new inode and socket object. The two are bound together
564 * and initialised. The socket is then returned. If we are out of inodes
565 * NULL is returned. This functions uses GFP_KERNEL internally.
568 struct socket *sock_alloc(void)
573 inode = new_inode_pseudo(sock_mnt->mnt_sb);
577 sock = SOCKET_I(inode);
579 inode->i_ino = get_next_ino();
580 inode->i_mode = S_IFSOCK | S_IRWXUGO;
581 inode->i_uid = current_fsuid();
582 inode->i_gid = current_fsgid();
583 inode->i_op = &sockfs_inode_ops;
587 EXPORT_SYMBOL(sock_alloc);
589 static void __sock_release(struct socket *sock, struct inode *inode)
592 struct module *owner = sock->ops->owner;
596 sock->ops->release(sock);
604 if (sock->wq.fasync_list)
605 pr_err("%s: fasync list not empty!\n", __func__);
608 iput(SOCK_INODE(sock));
615 * sock_release - close a socket
616 * @sock: socket to close
618 * The socket is released from the protocol stack if it has a release
619 * callback, and the inode is then released if the socket is bound to
620 * an inode not a file.
622 void sock_release(struct socket *sock)
624 __sock_release(sock, NULL);
626 EXPORT_SYMBOL(sock_release);
628 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
630 u8 flags = *tx_flags;
632 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
633 flags |= SKBTX_HW_TSTAMP;
635 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
636 flags |= SKBTX_SW_TSTAMP;
638 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
639 flags |= SKBTX_SCHED_TSTAMP;
643 EXPORT_SYMBOL(__sock_tx_timestamp);
645 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
647 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
649 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
651 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
652 inet_sendmsg, sock, msg,
654 BUG_ON(ret == -EIOCBQUEUED);
659 * sock_sendmsg - send a message through @sock
661 * @msg: message to send
663 * Sends @msg through @sock, passing through LSM.
664 * Returns the number of bytes sent, or an error code.
666 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
668 int err = security_socket_sendmsg(sock, msg,
671 return err ?: sock_sendmsg_nosec(sock, msg);
673 EXPORT_SYMBOL(sock_sendmsg);
676 * kernel_sendmsg - send a message through @sock (kernel-space)
678 * @msg: message header
680 * @num: vec array length
681 * @size: total message data size
683 * Builds the message data with @vec and sends it through @sock.
684 * Returns the number of bytes sent, or an error code.
687 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
688 struct kvec *vec, size_t num, size_t size)
690 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
691 return sock_sendmsg(sock, msg);
693 EXPORT_SYMBOL(kernel_sendmsg);
696 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
698 * @msg: message header
699 * @vec: output s/g array
700 * @num: output s/g array length
701 * @size: total message data size
703 * Builds the message data with @vec and sends it through @sock.
704 * Returns the number of bytes sent, or an error code.
705 * Caller must hold @sk.
708 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
709 struct kvec *vec, size_t num, size_t size)
711 struct socket *sock = sk->sk_socket;
713 if (!sock->ops->sendmsg_locked)
714 return sock_no_sendmsg_locked(sk, msg, size);
716 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
718 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
720 EXPORT_SYMBOL(kernel_sendmsg_locked);
722 static bool skb_is_err_queue(const struct sk_buff *skb)
724 /* pkt_type of skbs enqueued on the error queue are set to
725 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
726 * in recvmsg, since skbs received on a local socket will never
727 * have a pkt_type of PACKET_OUTGOING.
729 return skb->pkt_type == PACKET_OUTGOING;
732 /* On transmit, software and hardware timestamps are returned independently.
733 * As the two skb clones share the hardware timestamp, which may be updated
734 * before the software timestamp is received, a hardware TX timestamp may be
735 * returned only if there is no software TX timestamp. Ignore false software
736 * timestamps, which may be made in the __sock_recv_timestamp() call when the
737 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
738 * hardware timestamp.
740 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
742 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
745 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
747 struct scm_ts_pktinfo ts_pktinfo;
748 struct net_device *orig_dev;
750 if (!skb_mac_header_was_set(skb))
753 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
756 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
758 ts_pktinfo.if_index = orig_dev->ifindex;
761 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
762 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
763 sizeof(ts_pktinfo), &ts_pktinfo);
767 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
769 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
772 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
773 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
774 struct scm_timestamping_internal tss;
776 int empty = 1, false_tstamp = 0;
777 struct skb_shared_hwtstamps *shhwtstamps =
780 /* Race occurred between timestamp enabling and packet
781 receiving. Fill in the current time for now. */
782 if (need_software_tstamp && skb->tstamp == 0) {
783 __net_timestamp(skb);
787 if (need_software_tstamp) {
788 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
790 struct __kernel_sock_timeval tv;
792 skb_get_new_timestamp(skb, &tv);
793 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
796 struct __kernel_old_timeval tv;
798 skb_get_timestamp(skb, &tv);
799 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
804 struct __kernel_timespec ts;
806 skb_get_new_timestampns(skb, &ts);
807 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
810 struct __kernel_old_timespec ts;
812 skb_get_timestampns(skb, &ts);
813 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
819 memset(&tss, 0, sizeof(tss));
820 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
821 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
824 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
825 !skb_is_swtx_tstamp(skb, false_tstamp) &&
826 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
828 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
829 !skb_is_err_queue(skb))
830 put_ts_pktinfo(msg, skb);
833 if (sock_flag(sk, SOCK_TSTAMP_NEW))
834 put_cmsg_scm_timestamping64(msg, &tss);
836 put_cmsg_scm_timestamping(msg, &tss);
838 if (skb_is_err_queue(skb) && skb->len &&
839 SKB_EXT_ERR(skb)->opt_stats)
840 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
841 skb->len, skb->data);
844 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
846 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
851 if (!sock_flag(sk, SOCK_WIFI_STATUS))
853 if (!skb->wifi_acked_valid)
856 ack = skb->wifi_acked;
858 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
860 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
862 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
865 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
866 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
867 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
870 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
873 sock_recv_timestamp(msg, sk, skb);
874 sock_recv_drops(msg, sk, skb);
876 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
878 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
880 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
882 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
885 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
886 inet_recvmsg, sock, msg, msg_data_left(msg),
891 * sock_recvmsg - receive a message from @sock
893 * @msg: message to receive
894 * @flags: message flags
896 * Receives @msg from @sock, passing through LSM. Returns the total number
897 * of bytes received, or an error.
899 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
901 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
903 return err ?: sock_recvmsg_nosec(sock, msg, flags);
905 EXPORT_SYMBOL(sock_recvmsg);
908 * kernel_recvmsg - Receive a message from a socket (kernel space)
909 * @sock: The socket to receive the message from
910 * @msg: Received message
911 * @vec: Input s/g array for message data
912 * @num: Size of input s/g array
913 * @size: Number of bytes to read
914 * @flags: Message flags (MSG_DONTWAIT, etc...)
916 * On return the msg structure contains the scatter/gather array passed in the
917 * vec argument. The array is modified so that it consists of the unfilled
918 * portion of the original array.
920 * The returned value is the total number of bytes received, or an error.
923 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
924 struct kvec *vec, size_t num, size_t size, int flags)
926 msg->msg_control_is_user = false;
927 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
928 return sock_recvmsg(sock, msg, flags);
930 EXPORT_SYMBOL(kernel_recvmsg);
932 static ssize_t sock_sendpage(struct file *file, struct page *page,
933 int offset, size_t size, loff_t *ppos, int more)
938 sock = file->private_data;
940 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
941 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
944 return kernel_sendpage(sock, page, offset, size, flags);
947 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
948 struct pipe_inode_info *pipe, size_t len,
951 struct socket *sock = file->private_data;
953 if (unlikely(!sock->ops->splice_read))
954 return generic_file_splice_read(file, ppos, pipe, len, flags);
956 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
959 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
961 struct file *file = iocb->ki_filp;
962 struct socket *sock = file->private_data;
963 struct msghdr msg = {.msg_iter = *to,
967 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
968 msg.msg_flags = MSG_DONTWAIT;
970 if (iocb->ki_pos != 0)
973 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
976 res = sock_recvmsg(sock, &msg, msg.msg_flags);
981 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
983 struct file *file = iocb->ki_filp;
984 struct socket *sock = file->private_data;
985 struct msghdr msg = {.msg_iter = *from,
989 if (iocb->ki_pos != 0)
992 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
993 msg.msg_flags = MSG_DONTWAIT;
995 if (sock->type == SOCK_SEQPACKET)
996 msg.msg_flags |= MSG_EOR;
998 res = sock_sendmsg(sock, &msg);
999 *from = msg.msg_iter;
1004 * Atomic setting of ioctl hooks to avoid race
1005 * with module unload.
1008 static DEFINE_MUTEX(br_ioctl_mutex);
1009 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1011 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1013 mutex_lock(&br_ioctl_mutex);
1014 br_ioctl_hook = hook;
1015 mutex_unlock(&br_ioctl_mutex);
1017 EXPORT_SYMBOL(brioctl_set);
1019 static DEFINE_MUTEX(vlan_ioctl_mutex);
1020 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1022 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1024 mutex_lock(&vlan_ioctl_mutex);
1025 vlan_ioctl_hook = hook;
1026 mutex_unlock(&vlan_ioctl_mutex);
1028 EXPORT_SYMBOL(vlan_ioctl_set);
1030 static DEFINE_MUTEX(dlci_ioctl_mutex);
1031 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1033 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1035 mutex_lock(&dlci_ioctl_mutex);
1036 dlci_ioctl_hook = hook;
1037 mutex_unlock(&dlci_ioctl_mutex);
1039 EXPORT_SYMBOL(dlci_ioctl_set);
1041 static long sock_do_ioctl(struct net *net, struct socket *sock,
1042 unsigned int cmd, unsigned long arg)
1045 void __user *argp = (void __user *)arg;
1047 err = sock->ops->ioctl(sock, cmd, arg);
1050 * If this ioctl is unknown try to hand it down
1051 * to the NIC driver.
1053 if (err != -ENOIOCTLCMD)
1056 if (cmd == SIOCGIFCONF) {
1058 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1061 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1063 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1068 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1070 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1071 if (!err && need_copyout)
1072 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1079 * With an ioctl, arg may well be a user mode pointer, but we don't know
1080 * what to do with it - that's up to the protocol still.
1084 * get_net_ns - increment the refcount of the network namespace
1085 * @ns: common namespace (net)
1087 * Returns the net's common namespace.
1090 struct ns_common *get_net_ns(struct ns_common *ns)
1092 return &get_net(container_of(ns, struct net, ns))->ns;
1094 EXPORT_SYMBOL_GPL(get_net_ns);
1096 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1098 struct socket *sock;
1100 void __user *argp = (void __user *)arg;
1104 sock = file->private_data;
1107 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1110 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1112 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1113 if (!err && need_copyout)
1114 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1117 #ifdef CONFIG_WEXT_CORE
1118 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1119 err = wext_handle_ioctl(net, cmd, argp);
1126 if (get_user(pid, (int __user *)argp))
1128 err = f_setown(sock->file, pid, 1);
1132 err = put_user(f_getown(sock->file),
1133 (int __user *)argp);
1141 request_module("bridge");
1143 mutex_lock(&br_ioctl_mutex);
1145 err = br_ioctl_hook(net, cmd, argp);
1146 mutex_unlock(&br_ioctl_mutex);
1151 if (!vlan_ioctl_hook)
1152 request_module("8021q");
1154 mutex_lock(&vlan_ioctl_mutex);
1155 if (vlan_ioctl_hook)
1156 err = vlan_ioctl_hook(net, argp);
1157 mutex_unlock(&vlan_ioctl_mutex);
1162 if (!dlci_ioctl_hook)
1163 request_module("dlci");
1165 mutex_lock(&dlci_ioctl_mutex);
1166 if (dlci_ioctl_hook)
1167 err = dlci_ioctl_hook(cmd, argp);
1168 mutex_unlock(&dlci_ioctl_mutex);
1172 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1175 err = open_related_ns(&net->ns, get_net_ns);
1177 case SIOCGSTAMP_OLD:
1178 case SIOCGSTAMPNS_OLD:
1179 if (!sock->ops->gettstamp) {
1183 err = sock->ops->gettstamp(sock, argp,
1184 cmd == SIOCGSTAMP_OLD,
1185 !IS_ENABLED(CONFIG_64BIT));
1187 case SIOCGSTAMP_NEW:
1188 case SIOCGSTAMPNS_NEW:
1189 if (!sock->ops->gettstamp) {
1193 err = sock->ops->gettstamp(sock, argp,
1194 cmd == SIOCGSTAMP_NEW,
1198 err = sock_do_ioctl(net, sock, cmd, arg);
1205 * sock_create_lite - creates a socket
1206 * @family: protocol family (AF_INET, ...)
1207 * @type: communication type (SOCK_STREAM, ...)
1208 * @protocol: protocol (0, ...)
1211 * Creates a new socket and assigns it to @res, passing through LSM.
1212 * The new socket initialization is not complete, see kernel_accept().
1213 * Returns 0 or an error. On failure @res is set to %NULL.
1214 * This function internally uses GFP_KERNEL.
1217 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1220 struct socket *sock = NULL;
1222 err = security_socket_create(family, type, protocol, 1);
1226 sock = sock_alloc();
1233 err = security_socket_post_create(sock, family, type, protocol, 1);
1245 EXPORT_SYMBOL(sock_create_lite);
1247 /* No kernel lock held - perfect */
1248 static __poll_t sock_poll(struct file *file, poll_table *wait)
1250 struct socket *sock = file->private_data;
1251 __poll_t events = poll_requested_events(wait), flag = 0;
1253 if (!sock->ops->poll)
1256 if (sk_can_busy_loop(sock->sk)) {
1257 /* poll once if requested by the syscall */
1258 if (events & POLL_BUSY_LOOP)
1259 sk_busy_loop(sock->sk, 1);
1261 /* if this socket can poll_ll, tell the system call */
1262 flag = POLL_BUSY_LOOP;
1265 return sock->ops->poll(file, sock, wait) | flag;
1268 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1270 struct socket *sock = file->private_data;
1272 return sock->ops->mmap(file, sock, vma);
1275 static int sock_close(struct inode *inode, struct file *filp)
1277 __sock_release(SOCKET_I(inode), inode);
1282 * Update the socket async list
1284 * Fasync_list locking strategy.
1286 * 1. fasync_list is modified only under process context socket lock
1287 * i.e. under semaphore.
1288 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1289 * or under socket lock
1292 static int sock_fasync(int fd, struct file *filp, int on)
1294 struct socket *sock = filp->private_data;
1295 struct sock *sk = sock->sk;
1296 struct socket_wq *wq = &sock->wq;
1302 fasync_helper(fd, filp, on, &wq->fasync_list);
1304 if (!wq->fasync_list)
1305 sock_reset_flag(sk, SOCK_FASYNC);
1307 sock_set_flag(sk, SOCK_FASYNC);
1313 /* This function may be called only under rcu_lock */
1315 int sock_wake_async(struct socket_wq *wq, int how, int band)
1317 if (!wq || !wq->fasync_list)
1321 case SOCK_WAKE_WAITD:
1322 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1325 case SOCK_WAKE_SPACE:
1326 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1331 kill_fasync(&wq->fasync_list, SIGIO, band);
1334 kill_fasync(&wq->fasync_list, SIGURG, band);
1339 EXPORT_SYMBOL(sock_wake_async);
1342 * __sock_create - creates a socket
1343 * @net: net namespace
1344 * @family: protocol family (AF_INET, ...)
1345 * @type: communication type (SOCK_STREAM, ...)
1346 * @protocol: protocol (0, ...)
1348 * @kern: boolean for kernel space sockets
1350 * Creates a new socket and assigns it to @res, passing through LSM.
1351 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1352 * be set to true if the socket resides in kernel space.
1353 * This function internally uses GFP_KERNEL.
1356 int __sock_create(struct net *net, int family, int type, int protocol,
1357 struct socket **res, int kern)
1360 struct socket *sock;
1361 const struct net_proto_family *pf;
1364 * Check protocol is in range
1366 if (family < 0 || family >= NPROTO)
1367 return -EAFNOSUPPORT;
1368 if (type < 0 || type >= SOCK_MAX)
1373 This uglymoron is moved from INET layer to here to avoid
1374 deadlock in module load.
1376 if (family == PF_INET && type == SOCK_PACKET) {
1377 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1382 err = security_socket_create(family, type, protocol, kern);
1387 * Allocate the socket and allow the family to set things up. if
1388 * the protocol is 0, the family is instructed to select an appropriate
1391 sock = sock_alloc();
1393 net_warn_ratelimited("socket: no more sockets\n");
1394 return -ENFILE; /* Not exactly a match, but its the
1395 closest posix thing */
1400 #ifdef CONFIG_MODULES
1401 /* Attempt to load a protocol module if the find failed.
1403 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1404 * requested real, full-featured networking support upon configuration.
1405 * Otherwise module support will break!
1407 if (rcu_access_pointer(net_families[family]) == NULL)
1408 request_module("net-pf-%d", family);
1412 pf = rcu_dereference(net_families[family]);
1413 err = -EAFNOSUPPORT;
1418 * We will call the ->create function, that possibly is in a loadable
1419 * module, so we have to bump that loadable module refcnt first.
1421 if (!try_module_get(pf->owner))
1424 /* Now protected by module ref count */
1427 err = pf->create(net, sock, protocol, kern);
1429 goto out_module_put;
1432 * Now to bump the refcnt of the [loadable] module that owns this
1433 * socket at sock_release time we decrement its refcnt.
1435 if (!try_module_get(sock->ops->owner))
1436 goto out_module_busy;
1439 * Now that we're done with the ->create function, the [loadable]
1440 * module can have its refcnt decremented
1442 module_put(pf->owner);
1443 err = security_socket_post_create(sock, family, type, protocol, kern);
1445 goto out_sock_release;
1451 err = -EAFNOSUPPORT;
1454 module_put(pf->owner);
1461 goto out_sock_release;
1463 EXPORT_SYMBOL(__sock_create);
1466 * sock_create - creates a socket
1467 * @family: protocol family (AF_INET, ...)
1468 * @type: communication type (SOCK_STREAM, ...)
1469 * @protocol: protocol (0, ...)
1472 * A wrapper around __sock_create().
1473 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1476 int sock_create(int family, int type, int protocol, struct socket **res)
1478 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1480 EXPORT_SYMBOL(sock_create);
1483 * sock_create_kern - creates a socket (kernel space)
1484 * @net: net namespace
1485 * @family: protocol family (AF_INET, ...)
1486 * @type: communication type (SOCK_STREAM, ...)
1487 * @protocol: protocol (0, ...)
1490 * A wrapper around __sock_create().
1491 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1494 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1496 return __sock_create(net, family, type, protocol, res, 1);
1498 EXPORT_SYMBOL(sock_create_kern);
1500 int __sys_socket(int family, int type, int protocol)
1503 struct socket *sock;
1506 /* Check the SOCK_* constants for consistency. */
1507 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1508 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1509 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1510 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1512 flags = type & ~SOCK_TYPE_MASK;
1513 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1515 type &= SOCK_TYPE_MASK;
1517 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1518 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1520 retval = sock_create(family, type, protocol, &sock);
1524 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1527 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1529 return __sys_socket(family, type, protocol);
1533 * Create a pair of connected sockets.
1536 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1538 struct socket *sock1, *sock2;
1540 struct file *newfile1, *newfile2;
1543 flags = type & ~SOCK_TYPE_MASK;
1544 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1546 type &= SOCK_TYPE_MASK;
1548 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1549 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1552 * reserve descriptors and make sure we won't fail
1553 * to return them to userland.
1555 fd1 = get_unused_fd_flags(flags);
1556 if (unlikely(fd1 < 0))
1559 fd2 = get_unused_fd_flags(flags);
1560 if (unlikely(fd2 < 0)) {
1565 err = put_user(fd1, &usockvec[0]);
1569 err = put_user(fd2, &usockvec[1]);
1574 * Obtain the first socket and check if the underlying protocol
1575 * supports the socketpair call.
1578 err = sock_create(family, type, protocol, &sock1);
1579 if (unlikely(err < 0))
1582 err = sock_create(family, type, protocol, &sock2);
1583 if (unlikely(err < 0)) {
1584 sock_release(sock1);
1588 err = security_socket_socketpair(sock1, sock2);
1589 if (unlikely(err)) {
1590 sock_release(sock2);
1591 sock_release(sock1);
1595 err = sock1->ops->socketpair(sock1, sock2);
1596 if (unlikely(err < 0)) {
1597 sock_release(sock2);
1598 sock_release(sock1);
1602 newfile1 = sock_alloc_file(sock1, flags, NULL);
1603 if (IS_ERR(newfile1)) {
1604 err = PTR_ERR(newfile1);
1605 sock_release(sock2);
1609 newfile2 = sock_alloc_file(sock2, flags, NULL);
1610 if (IS_ERR(newfile2)) {
1611 err = PTR_ERR(newfile2);
1616 audit_fd_pair(fd1, fd2);
1618 fd_install(fd1, newfile1);
1619 fd_install(fd2, newfile2);
1628 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1629 int __user *, usockvec)
1631 return __sys_socketpair(family, type, protocol, usockvec);
1635 * Bind a name to a socket. Nothing much to do here since it's
1636 * the protocol's responsibility to handle the local address.
1638 * We move the socket address to kernel space before we call
1639 * the protocol layer (having also checked the address is ok).
1642 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1644 struct socket *sock;
1645 struct sockaddr_storage address;
1646 int err, fput_needed;
1648 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1650 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1652 err = security_socket_bind(sock,
1653 (struct sockaddr *)&address,
1656 err = sock->ops->bind(sock,
1660 fput_light(sock->file, fput_needed);
1665 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1667 return __sys_bind(fd, umyaddr, addrlen);
1671 * Perform a listen. Basically, we allow the protocol to do anything
1672 * necessary for a listen, and if that works, we mark the socket as
1673 * ready for listening.
1676 int __sys_listen(int fd, int backlog)
1678 struct socket *sock;
1679 int err, fput_needed;
1682 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1684 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1685 if ((unsigned int)backlog > somaxconn)
1686 backlog = somaxconn;
1688 err = security_socket_listen(sock, backlog);
1690 err = sock->ops->listen(sock, backlog);
1692 fput_light(sock->file, fput_needed);
1697 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1699 return __sys_listen(fd, backlog);
1702 int __sys_accept4_file(struct file *file, unsigned file_flags,
1703 struct sockaddr __user *upeer_sockaddr,
1704 int __user *upeer_addrlen, int flags,
1705 unsigned long nofile)
1707 struct socket *sock, *newsock;
1708 struct file *newfile;
1709 int err, len, newfd;
1710 struct sockaddr_storage address;
1712 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1715 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1716 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1718 sock = sock_from_file(file, &err);
1723 newsock = sock_alloc();
1727 newsock->type = sock->type;
1728 newsock->ops = sock->ops;
1731 * We don't need try_module_get here, as the listening socket (sock)
1732 * has the protocol module (sock->ops->owner) held.
1734 __module_get(newsock->ops->owner);
1736 newfd = __get_unused_fd_flags(flags, nofile);
1737 if (unlikely(newfd < 0)) {
1739 sock_release(newsock);
1742 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1743 if (IS_ERR(newfile)) {
1744 err = PTR_ERR(newfile);
1745 put_unused_fd(newfd);
1749 err = security_socket_accept(sock, newsock);
1753 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1758 if (upeer_sockaddr) {
1759 len = newsock->ops->getname(newsock,
1760 (struct sockaddr *)&address, 2);
1762 err = -ECONNABORTED;
1765 err = move_addr_to_user(&address,
1766 len, upeer_sockaddr, upeer_addrlen);
1771 /* File flags are not inherited via accept() unlike another OSes. */
1773 fd_install(newfd, newfile);
1779 put_unused_fd(newfd);
1785 * For accept, we attempt to create a new socket, set up the link
1786 * with the client, wake up the client, then return the new
1787 * connected fd. We collect the address of the connector in kernel
1788 * space and move it to user at the very end. This is unclean because
1789 * we open the socket then return an error.
1791 * 1003.1g adds the ability to recvmsg() to query connection pending
1792 * status to recvmsg. We need to add that support in a way thats
1793 * clean when we restructure accept also.
1796 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1797 int __user *upeer_addrlen, int flags)
1804 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1805 upeer_addrlen, flags,
1806 rlimit(RLIMIT_NOFILE));
1814 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1815 int __user *, upeer_addrlen, int, flags)
1817 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1820 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1821 int __user *, upeer_addrlen)
1823 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1827 * Attempt to connect to a socket with the server address. The address
1828 * is in user space so we verify it is OK and move it to kernel space.
1830 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1833 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1834 * other SEQPACKET protocols that take time to connect() as it doesn't
1835 * include the -EINPROGRESS status for such sockets.
1838 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1839 int addrlen, int file_flags)
1841 struct socket *sock;
1844 sock = sock_from_file(file, &err);
1849 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1853 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1854 sock->file->f_flags | file_flags);
1859 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1866 struct sockaddr_storage address;
1868 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1870 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1878 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1881 return __sys_connect(fd, uservaddr, addrlen);
1885 * Get the local address ('name') of a socket object. Move the obtained
1886 * name to user space.
1889 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1890 int __user *usockaddr_len)
1892 struct socket *sock;
1893 struct sockaddr_storage address;
1894 int err, fput_needed;
1896 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1900 err = security_socket_getsockname(sock);
1904 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1907 /* "err" is actually length in this case */
1908 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1911 fput_light(sock->file, fput_needed);
1916 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1917 int __user *, usockaddr_len)
1919 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1923 * Get the remote address ('name') of a socket object. Move the obtained
1924 * name to user space.
1927 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1928 int __user *usockaddr_len)
1930 struct socket *sock;
1931 struct sockaddr_storage address;
1932 int err, fput_needed;
1934 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1936 err = security_socket_getpeername(sock);
1938 fput_light(sock->file, fput_needed);
1942 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1944 /* "err" is actually length in this case */
1945 err = move_addr_to_user(&address, err, usockaddr,
1947 fput_light(sock->file, fput_needed);
1952 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1953 int __user *, usockaddr_len)
1955 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1959 * Send a datagram to a given address. We move the address into kernel
1960 * space and check the user space data area is readable before invoking
1963 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1964 struct sockaddr __user *addr, int addr_len)
1966 struct socket *sock;
1967 struct sockaddr_storage address;
1973 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1976 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1980 msg.msg_name = NULL;
1981 msg.msg_control = NULL;
1982 msg.msg_controllen = 0;
1983 msg.msg_namelen = 0;
1985 err = move_addr_to_kernel(addr, addr_len, &address);
1988 msg.msg_name = (struct sockaddr *)&address;
1989 msg.msg_namelen = addr_len;
1991 if (sock->file->f_flags & O_NONBLOCK)
1992 flags |= MSG_DONTWAIT;
1993 msg.msg_flags = flags;
1994 err = sock_sendmsg(sock, &msg);
1997 fput_light(sock->file, fput_needed);
2002 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2003 unsigned int, flags, struct sockaddr __user *, addr,
2006 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2010 * Send a datagram down a socket.
2013 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2014 unsigned int, flags)
2016 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2020 * Receive a frame from the socket and optionally record the address of the
2021 * sender. We verify the buffers are writable and if needed move the
2022 * sender address from kernel to user space.
2024 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2025 struct sockaddr __user *addr, int __user *addr_len)
2027 struct socket *sock;
2030 struct sockaddr_storage address;
2034 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2037 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2041 msg.msg_control = NULL;
2042 msg.msg_controllen = 0;
2043 /* Save some cycles and don't copy the address if not needed */
2044 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2045 /* We assume all kernel code knows the size of sockaddr_storage */
2046 msg.msg_namelen = 0;
2047 msg.msg_iocb = NULL;
2049 if (sock->file->f_flags & O_NONBLOCK)
2050 flags |= MSG_DONTWAIT;
2051 err = sock_recvmsg(sock, &msg, flags);
2053 if (err >= 0 && addr != NULL) {
2054 err2 = move_addr_to_user(&address,
2055 msg.msg_namelen, addr, addr_len);
2060 fput_light(sock->file, fput_needed);
2065 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2066 unsigned int, flags, struct sockaddr __user *, addr,
2067 int __user *, addr_len)
2069 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2073 * Receive a datagram from a socket.
2076 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2077 unsigned int, flags)
2079 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2082 static bool sock_use_custom_sol_socket(const struct socket *sock)
2084 const struct sock *sk = sock->sk;
2086 /* Use sock->ops->setsockopt() for MPTCP */
2087 return IS_ENABLED(CONFIG_MPTCP) &&
2088 sk->sk_protocol == IPPROTO_MPTCP &&
2089 sk->sk_type == SOCK_STREAM &&
2090 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2094 * Set a socket option. Because we don't know the option lengths we have
2095 * to pass the user mode parameter for the protocols to sort out.
2097 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2101 char *kernel_optval = NULL;
2102 int err, fput_needed;
2103 struct socket *sock;
2108 err = init_user_sockptr(&optval, user_optval, optlen);
2112 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2116 err = security_socket_setsockopt(sock, level, optname);
2120 if (!in_compat_syscall())
2121 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2122 user_optval, &optlen,
2132 optval = KERNEL_SOCKPTR(kernel_optval);
2133 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2134 err = sock_setsockopt(sock, level, optname, optval, optlen);
2135 else if (unlikely(!sock->ops->setsockopt))
2138 err = sock->ops->setsockopt(sock, level, optname, optval,
2140 kfree(kernel_optval);
2142 fput_light(sock->file, fput_needed);
2146 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2147 char __user *, optval, int, optlen)
2149 return __sys_setsockopt(fd, level, optname, optval, optlen);
2153 * Get a socket option. Because we don't know the option lengths we have
2154 * to pass a user mode parameter for the protocols to sort out.
2156 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2159 int err, fput_needed;
2160 struct socket *sock;
2163 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2167 err = security_socket_getsockopt(sock, level, optname);
2171 if (!in_compat_syscall())
2172 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2174 if (level == SOL_SOCKET)
2175 err = sock_getsockopt(sock, level, optname, optval, optlen);
2176 else if (unlikely(!sock->ops->getsockopt))
2179 err = sock->ops->getsockopt(sock, level, optname, optval,
2182 if (!in_compat_syscall())
2183 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2184 optval, optlen, max_optlen,
2187 fput_light(sock->file, fput_needed);
2191 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2192 char __user *, optval, int __user *, optlen)
2194 return __sys_getsockopt(fd, level, optname, optval, optlen);
2198 * Shutdown a socket.
2201 int __sys_shutdown(int fd, int how)
2203 int err, fput_needed;
2204 struct socket *sock;
2206 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2208 err = security_socket_shutdown(sock, how);
2210 err = sock->ops->shutdown(sock, how);
2211 fput_light(sock->file, fput_needed);
2216 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2218 return __sys_shutdown(fd, how);
2221 /* A couple of helpful macros for getting the address of the 32/64 bit
2222 * fields which are the same type (int / unsigned) on our platforms.
2224 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2225 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2226 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2228 struct used_address {
2229 struct sockaddr_storage name;
2230 unsigned int name_len;
2233 int __copy_msghdr_from_user(struct msghdr *kmsg,
2234 struct user_msghdr __user *umsg,
2235 struct sockaddr __user **save_addr,
2236 struct iovec __user **uiov, size_t *nsegs)
2238 struct user_msghdr msg;
2241 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2244 kmsg->msg_control_is_user = true;
2245 kmsg->msg_control_user = msg.msg_control;
2246 kmsg->msg_controllen = msg.msg_controllen;
2247 kmsg->msg_flags = msg.msg_flags;
2249 kmsg->msg_namelen = msg.msg_namelen;
2251 kmsg->msg_namelen = 0;
2253 if (kmsg->msg_namelen < 0)
2256 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2257 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2260 *save_addr = msg.msg_name;
2262 if (msg.msg_name && kmsg->msg_namelen) {
2264 err = move_addr_to_kernel(msg.msg_name,
2271 kmsg->msg_name = NULL;
2272 kmsg->msg_namelen = 0;
2275 if (msg.msg_iovlen > UIO_MAXIOV)
2278 kmsg->msg_iocb = NULL;
2279 *uiov = msg.msg_iov;
2280 *nsegs = msg.msg_iovlen;
2284 static int copy_msghdr_from_user(struct msghdr *kmsg,
2285 struct user_msghdr __user *umsg,
2286 struct sockaddr __user **save_addr,
2289 struct user_msghdr msg;
2292 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2297 err = import_iovec(save_addr ? READ : WRITE,
2298 msg.msg_iov, msg.msg_iovlen,
2299 UIO_FASTIOV, iov, &kmsg->msg_iter);
2300 return err < 0 ? err : 0;
2303 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2304 unsigned int flags, struct used_address *used_address,
2305 unsigned int allowed_msghdr_flags)
2307 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2308 __aligned(sizeof(__kernel_size_t));
2309 /* 20 is size of ipv6_pktinfo */
2310 unsigned char *ctl_buf = ctl;
2316 if (msg_sys->msg_controllen > INT_MAX)
2318 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2319 ctl_len = msg_sys->msg_controllen;
2320 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2322 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2326 ctl_buf = msg_sys->msg_control;
2327 ctl_len = msg_sys->msg_controllen;
2328 } else if (ctl_len) {
2329 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2330 CMSG_ALIGN(sizeof(struct cmsghdr)));
2331 if (ctl_len > sizeof(ctl)) {
2332 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2333 if (ctl_buf == NULL)
2337 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2339 msg_sys->msg_control = ctl_buf;
2340 msg_sys->msg_control_is_user = false;
2342 msg_sys->msg_flags = flags;
2344 if (sock->file->f_flags & O_NONBLOCK)
2345 msg_sys->msg_flags |= MSG_DONTWAIT;
2347 * If this is sendmmsg() and current destination address is same as
2348 * previously succeeded address, omit asking LSM's decision.
2349 * used_address->name_len is initialized to UINT_MAX so that the first
2350 * destination address never matches.
2352 if (used_address && msg_sys->msg_name &&
2353 used_address->name_len == msg_sys->msg_namelen &&
2354 !memcmp(&used_address->name, msg_sys->msg_name,
2355 used_address->name_len)) {
2356 err = sock_sendmsg_nosec(sock, msg_sys);
2359 err = sock_sendmsg(sock, msg_sys);
2361 * If this is sendmmsg() and sending to current destination address was
2362 * successful, remember it.
2364 if (used_address && err >= 0) {
2365 used_address->name_len = msg_sys->msg_namelen;
2366 if (msg_sys->msg_name)
2367 memcpy(&used_address->name, msg_sys->msg_name,
2368 used_address->name_len);
2373 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2378 int sendmsg_copy_msghdr(struct msghdr *msg,
2379 struct user_msghdr __user *umsg, unsigned flags,
2384 if (flags & MSG_CMSG_COMPAT) {
2385 struct compat_msghdr __user *msg_compat;
2387 msg_compat = (struct compat_msghdr __user *) umsg;
2388 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2390 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2398 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2399 struct msghdr *msg_sys, unsigned int flags,
2400 struct used_address *used_address,
2401 unsigned int allowed_msghdr_flags)
2403 struct sockaddr_storage address;
2404 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2407 msg_sys->msg_name = &address;
2409 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2413 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2414 allowed_msghdr_flags);
2420 * BSD sendmsg interface
2422 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2425 /* disallow ancillary data requests from this path */
2426 if (msg->msg_control || msg->msg_controllen)
2429 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2432 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2433 bool forbid_cmsg_compat)
2435 int fput_needed, err;
2436 struct msghdr msg_sys;
2437 struct socket *sock;
2439 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2442 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2446 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2448 fput_light(sock->file, fput_needed);
2453 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2455 return __sys_sendmsg(fd, msg, flags, true);
2459 * Linux sendmmsg interface
2462 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2463 unsigned int flags, bool forbid_cmsg_compat)
2465 int fput_needed, err, datagrams;
2466 struct socket *sock;
2467 struct mmsghdr __user *entry;
2468 struct compat_mmsghdr __user *compat_entry;
2469 struct msghdr msg_sys;
2470 struct used_address used_address;
2471 unsigned int oflags = flags;
2473 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2476 if (vlen > UIO_MAXIOV)
2481 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2485 used_address.name_len = UINT_MAX;
2487 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2491 while (datagrams < vlen) {
2492 if (datagrams == vlen - 1)
2495 if (MSG_CMSG_COMPAT & flags) {
2496 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2497 &msg_sys, flags, &used_address, MSG_EOR);
2500 err = __put_user(err, &compat_entry->msg_len);
2503 err = ___sys_sendmsg(sock,
2504 (struct user_msghdr __user *)entry,
2505 &msg_sys, flags, &used_address, MSG_EOR);
2508 err = put_user(err, &entry->msg_len);
2515 if (msg_data_left(&msg_sys))
2520 fput_light(sock->file, fput_needed);
2522 /* We only return an error if no datagrams were able to be sent */
2529 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2530 unsigned int, vlen, unsigned int, flags)
2532 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2535 int recvmsg_copy_msghdr(struct msghdr *msg,
2536 struct user_msghdr __user *umsg, unsigned flags,
2537 struct sockaddr __user **uaddr,
2542 if (MSG_CMSG_COMPAT & flags) {
2543 struct compat_msghdr __user *msg_compat;
2545 msg_compat = (struct compat_msghdr __user *) umsg;
2546 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2548 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2556 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2557 struct user_msghdr __user *msg,
2558 struct sockaddr __user *uaddr,
2559 unsigned int flags, int nosec)
2561 struct compat_msghdr __user *msg_compat =
2562 (struct compat_msghdr __user *) msg;
2563 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2564 struct sockaddr_storage addr;
2565 unsigned long cmsg_ptr;
2569 msg_sys->msg_name = &addr;
2570 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2571 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2573 /* We assume all kernel code knows the size of sockaddr_storage */
2574 msg_sys->msg_namelen = 0;
2576 if (sock->file->f_flags & O_NONBLOCK)
2577 flags |= MSG_DONTWAIT;
2579 if (unlikely(nosec))
2580 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2582 err = sock_recvmsg(sock, msg_sys, flags);
2588 if (uaddr != NULL) {
2589 err = move_addr_to_user(&addr,
2590 msg_sys->msg_namelen, uaddr,
2595 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2599 if (MSG_CMSG_COMPAT & flags)
2600 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2601 &msg_compat->msg_controllen);
2603 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2604 &msg->msg_controllen);
2612 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2613 struct msghdr *msg_sys, unsigned int flags, int nosec)
2615 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2616 /* user mode address pointers */
2617 struct sockaddr __user *uaddr;
2620 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2624 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2630 * BSD recvmsg interface
2633 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2634 struct user_msghdr __user *umsg,
2635 struct sockaddr __user *uaddr, unsigned int flags)
2637 /* disallow ancillary data requests from this path */
2638 if (msg->msg_control || msg->msg_controllen)
2641 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2644 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2645 bool forbid_cmsg_compat)
2647 int fput_needed, err;
2648 struct msghdr msg_sys;
2649 struct socket *sock;
2651 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2654 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2658 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2660 fput_light(sock->file, fput_needed);
2665 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2666 unsigned int, flags)
2668 return __sys_recvmsg(fd, msg, flags, true);
2672 * Linux recvmmsg interface
2675 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2676 unsigned int vlen, unsigned int flags,
2677 struct timespec64 *timeout)
2679 int fput_needed, err, datagrams;
2680 struct socket *sock;
2681 struct mmsghdr __user *entry;
2682 struct compat_mmsghdr __user *compat_entry;
2683 struct msghdr msg_sys;
2684 struct timespec64 end_time;
2685 struct timespec64 timeout64;
2688 poll_select_set_timeout(&end_time, timeout->tv_sec,
2694 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2698 if (likely(!(flags & MSG_ERRQUEUE))) {
2699 err = sock_error(sock->sk);
2707 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2709 while (datagrams < vlen) {
2711 * No need to ask LSM for more than the first datagram.
2713 if (MSG_CMSG_COMPAT & flags) {
2714 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2715 &msg_sys, flags & ~MSG_WAITFORONE,
2719 err = __put_user(err, &compat_entry->msg_len);
2722 err = ___sys_recvmsg(sock,
2723 (struct user_msghdr __user *)entry,
2724 &msg_sys, flags & ~MSG_WAITFORONE,
2728 err = put_user(err, &entry->msg_len);
2736 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2737 if (flags & MSG_WAITFORONE)
2738 flags |= MSG_DONTWAIT;
2741 ktime_get_ts64(&timeout64);
2742 *timeout = timespec64_sub(end_time, timeout64);
2743 if (timeout->tv_sec < 0) {
2744 timeout->tv_sec = timeout->tv_nsec = 0;
2748 /* Timeout, return less than vlen datagrams */
2749 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2753 /* Out of band data, return right away */
2754 if (msg_sys.msg_flags & MSG_OOB)
2762 if (datagrams == 0) {
2768 * We may return less entries than requested (vlen) if the
2769 * sock is non block and there aren't enough datagrams...
2771 if (err != -EAGAIN) {
2773 * ... or if recvmsg returns an error after we
2774 * received some datagrams, where we record the
2775 * error to return on the next call or if the
2776 * app asks about it using getsockopt(SO_ERROR).
2778 sock->sk->sk_err = -err;
2781 fput_light(sock->file, fput_needed);
2786 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2787 unsigned int vlen, unsigned int flags,
2788 struct __kernel_timespec __user *timeout,
2789 struct old_timespec32 __user *timeout32)
2792 struct timespec64 timeout_sys;
2794 if (timeout && get_timespec64(&timeout_sys, timeout))
2797 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2800 if (!timeout && !timeout32)
2801 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2803 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2808 if (timeout && put_timespec64(&timeout_sys, timeout))
2809 datagrams = -EFAULT;
2811 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2812 datagrams = -EFAULT;
2817 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2818 unsigned int, vlen, unsigned int, flags,
2819 struct __kernel_timespec __user *, timeout)
2821 if (flags & MSG_CMSG_COMPAT)
2824 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2827 #ifdef CONFIG_COMPAT_32BIT_TIME
2828 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2829 unsigned int, vlen, unsigned int, flags,
2830 struct old_timespec32 __user *, timeout)
2832 if (flags & MSG_CMSG_COMPAT)
2835 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2839 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2840 /* Argument list sizes for sys_socketcall */
2841 #define AL(x) ((x) * sizeof(unsigned long))
2842 static const unsigned char nargs[21] = {
2843 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2844 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2845 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2852 * System call vectors.
2854 * Argument checking cleaned up. Saved 20% in size.
2855 * This function doesn't need to set the kernel lock because
2856 * it is set by the callees.
2859 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2861 unsigned long a[AUDITSC_ARGS];
2862 unsigned long a0, a1;
2866 if (call < 1 || call > SYS_SENDMMSG)
2868 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2871 if (len > sizeof(a))
2874 /* copy_from_user should be SMP safe. */
2875 if (copy_from_user(a, args, len))
2878 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2887 err = __sys_socket(a0, a1, a[2]);
2890 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2893 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2896 err = __sys_listen(a0, a1);
2899 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2900 (int __user *)a[2], 0);
2902 case SYS_GETSOCKNAME:
2904 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2905 (int __user *)a[2]);
2907 case SYS_GETPEERNAME:
2909 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2910 (int __user *)a[2]);
2912 case SYS_SOCKETPAIR:
2913 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2916 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2920 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2921 (struct sockaddr __user *)a[4], a[5]);
2924 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2928 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2929 (struct sockaddr __user *)a[4],
2930 (int __user *)a[5]);
2933 err = __sys_shutdown(a0, a1);
2935 case SYS_SETSOCKOPT:
2936 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2939 case SYS_GETSOCKOPT:
2941 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2942 (int __user *)a[4]);
2945 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2949 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2953 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2957 if (IS_ENABLED(CONFIG_64BIT))
2958 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2960 (struct __kernel_timespec __user *)a[4],
2963 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2965 (struct old_timespec32 __user *)a[4]);
2968 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2969 (int __user *)a[2], a[3]);
2978 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2981 * sock_register - add a socket protocol handler
2982 * @ops: description of protocol
2984 * This function is called by a protocol handler that wants to
2985 * advertise its address family, and have it linked into the
2986 * socket interface. The value ops->family corresponds to the
2987 * socket system call protocol family.
2989 int sock_register(const struct net_proto_family *ops)
2993 if (ops->family >= NPROTO) {
2994 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2998 spin_lock(&net_family_lock);
2999 if (rcu_dereference_protected(net_families[ops->family],
3000 lockdep_is_held(&net_family_lock)))
3003 rcu_assign_pointer(net_families[ops->family], ops);
3006 spin_unlock(&net_family_lock);
3008 pr_info("NET: Registered protocol family %d\n", ops->family);
3011 EXPORT_SYMBOL(sock_register);
3014 * sock_unregister - remove a protocol handler
3015 * @family: protocol family to remove
3017 * This function is called by a protocol handler that wants to
3018 * remove its address family, and have it unlinked from the
3019 * new socket creation.
3021 * If protocol handler is a module, then it can use module reference
3022 * counts to protect against new references. If protocol handler is not
3023 * a module then it needs to provide its own protection in
3024 * the ops->create routine.
3026 void sock_unregister(int family)
3028 BUG_ON(family < 0 || family >= NPROTO);
3030 spin_lock(&net_family_lock);
3031 RCU_INIT_POINTER(net_families[family], NULL);
3032 spin_unlock(&net_family_lock);
3036 pr_info("NET: Unregistered protocol family %d\n", family);
3038 EXPORT_SYMBOL(sock_unregister);
3040 bool sock_is_registered(int family)
3042 return family < NPROTO && rcu_access_pointer(net_families[family]);
3045 static int __init sock_init(void)
3049 * Initialize the network sysctl infrastructure.
3051 err = net_sysctl_init();
3056 * Initialize skbuff SLAB cache
3061 * Initialize the protocols module.
3066 err = register_filesystem(&sock_fs_type);
3069 sock_mnt = kern_mount(&sock_fs_type);
3070 if (IS_ERR(sock_mnt)) {
3071 err = PTR_ERR(sock_mnt);
3075 /* The real protocol initialization is performed in later initcalls.
3078 #ifdef CONFIG_NETFILTER
3079 err = netfilter_init();
3084 ptp_classifier_init();
3090 unregister_filesystem(&sock_fs_type);
3095 core_initcall(sock_init); /* early initcall */
3097 #ifdef CONFIG_PROC_FS
3098 void socket_seq_show(struct seq_file *seq)
3100 seq_printf(seq, "sockets: used %d\n",
3101 sock_inuse_get(seq->private));
3103 #endif /* CONFIG_PROC_FS */
3105 #ifdef CONFIG_COMPAT
3106 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3108 struct compat_ifconf ifc32;
3112 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3115 ifc.ifc_len = ifc32.ifc_len;
3116 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3119 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3124 ifc32.ifc_len = ifc.ifc_len;
3125 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3131 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3133 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3134 bool convert_in = false, convert_out = false;
3135 size_t buf_size = 0;
3136 struct ethtool_rxnfc __user *rxnfc = NULL;
3138 u32 rule_cnt = 0, actual_rule_cnt;
3143 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3146 compat_rxnfc = compat_ptr(data);
3148 if (get_user(ethcmd, &compat_rxnfc->cmd))
3151 /* Most ethtool structures are defined without padding.
3152 * Unfortunately struct ethtool_rxnfc is an exception.
3157 case ETHTOOL_GRXCLSRLALL:
3158 /* Buffer size is variable */
3159 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3161 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3163 buf_size += rule_cnt * sizeof(u32);
3165 case ETHTOOL_GRXRINGS:
3166 case ETHTOOL_GRXCLSRLCNT:
3167 case ETHTOOL_GRXCLSRULE:
3168 case ETHTOOL_SRXCLSRLINS:
3171 case ETHTOOL_SRXCLSRLDEL:
3172 buf_size += sizeof(struct ethtool_rxnfc);
3174 rxnfc = compat_alloc_user_space(buf_size);
3178 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3181 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3184 /* We expect there to be holes between fs.m_ext and
3185 * fs.ring_cookie and at the end of fs, but nowhere else.
3187 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3188 sizeof(compat_rxnfc->fs.m_ext) !=
3189 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3190 sizeof(rxnfc->fs.m_ext));
3192 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3193 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3194 offsetof(struct ethtool_rxnfc, fs.location) -
3195 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3197 if (copy_in_user(rxnfc, compat_rxnfc,
3198 (void __user *)(&rxnfc->fs.m_ext + 1) -
3199 (void __user *)rxnfc) ||
3200 copy_in_user(&rxnfc->fs.ring_cookie,
3201 &compat_rxnfc->fs.ring_cookie,
3202 (void __user *)(&rxnfc->fs.location + 1) -
3203 (void __user *)&rxnfc->fs.ring_cookie))
3205 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3206 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3208 } else if (copy_in_user(&rxnfc->rule_cnt,
3209 &compat_rxnfc->rule_cnt,
3210 sizeof(rxnfc->rule_cnt)))
3214 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3219 if (copy_in_user(compat_rxnfc, rxnfc,
3220 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3221 (const void __user *)rxnfc) ||
3222 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3223 &rxnfc->fs.ring_cookie,
3224 (const void __user *)(&rxnfc->fs.location + 1) -
3225 (const void __user *)&rxnfc->fs.ring_cookie) ||
3226 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3227 sizeof(rxnfc->rule_cnt)))
3230 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3231 /* As an optimisation, we only copy the actual
3232 * number of rules that the underlying
3233 * function returned. Since Mallory might
3234 * change the rule count in user memory, we
3235 * check that it is less than the rule count
3236 * originally given (as the user buffer size),
3237 * which has been range-checked.
3239 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3241 if (actual_rule_cnt < rule_cnt)
3242 rule_cnt = actual_rule_cnt;
3243 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3244 &rxnfc->rule_locs[0],
3245 rule_cnt * sizeof(u32)))
3253 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3255 compat_uptr_t uptr32;
3260 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3263 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3266 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3267 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3269 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3271 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3272 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3278 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3279 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3280 struct compat_ifreq __user *u_ifreq32)
3285 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3287 if (get_user(data32, &u_ifreq32->ifr_data))
3289 ifreq.ifr_data = compat_ptr(data32);
3291 return dev_ioctl(net, cmd, &ifreq, NULL);
3294 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3296 struct compat_ifreq __user *uifr32)
3298 struct ifreq __user *uifr;
3301 /* Handle the fact that while struct ifreq has the same *layout* on
3302 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3303 * which are handled elsewhere, it still has different *size* due to
3304 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3305 * resulting in struct ifreq being 32 and 40 bytes respectively).
3306 * As a result, if the struct happens to be at the end of a page and
3307 * the next page isn't readable/writable, we get a fault. To prevent
3308 * that, copy back and forth to the full size.
3311 uifr = compat_alloc_user_space(sizeof(*uifr));
3312 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3315 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3326 case SIOCGIFBRDADDR:
3327 case SIOCGIFDSTADDR:
3328 case SIOCGIFNETMASK:
3334 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3342 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3343 struct compat_ifreq __user *uifr32)
3346 struct compat_ifmap __user *uifmap32;
3349 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3350 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3351 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3352 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3353 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3354 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3355 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3356 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3360 err = dev_ioctl(net, cmd, &ifr, NULL);
3362 if (cmd == SIOCGIFMAP && !err) {
3363 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3364 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3365 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3366 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3367 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3368 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3369 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3376 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3377 * for some operations; this forces use of the newer bridge-utils that
3378 * use compatible ioctls
3380 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3384 if (get_user(tmp, argp))
3386 if (tmp == BRCTL_GET_VERSION)
3387 return BRCTL_VERSION + 1;
3391 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3392 unsigned int cmd, unsigned long arg)
3394 void __user *argp = compat_ptr(arg);
3395 struct sock *sk = sock->sk;
3396 struct net *net = sock_net(sk);
3398 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3399 return compat_ifr_data_ioctl(net, cmd, argp);
3404 return old_bridge_ioctl(argp);
3406 return compat_dev_ifconf(net, argp);
3408 return ethtool_ioctl(net, argp);
3410 return compat_siocwandev(net, argp);
3413 return compat_sioc_ifmap(net, cmd, argp);
3414 case SIOCGSTAMP_OLD:
3415 case SIOCGSTAMPNS_OLD:
3416 if (!sock->ops->gettstamp)
3417 return -ENOIOCTLCMD;
3418 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3419 !COMPAT_USE_64BIT_TIME);
3421 case SIOCBONDSLAVEINFOQUERY:
3422 case SIOCBONDINFOQUERY:
3425 return compat_ifr_data_ioctl(net, cmd, argp);
3438 case SIOCGSTAMP_NEW:
3439 case SIOCGSTAMPNS_NEW:
3440 return sock_ioctl(file, cmd, arg);
3457 case SIOCSIFHWBROADCAST:
3459 case SIOCGIFBRDADDR:
3460 case SIOCSIFBRDADDR:
3461 case SIOCGIFDSTADDR:
3462 case SIOCSIFDSTADDR:
3463 case SIOCGIFNETMASK:
3464 case SIOCSIFNETMASK:
3476 case SIOCBONDENSLAVE:
3477 case SIOCBONDRELEASE:
3478 case SIOCBONDSETHWADDR:
3479 case SIOCBONDCHANGEACTIVE:
3480 return compat_ifreq_ioctl(net, sock, cmd, argp);
3488 return sock_do_ioctl(net, sock, cmd, arg);
3491 return -ENOIOCTLCMD;
3494 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3497 struct socket *sock = file->private_data;
3498 int ret = -ENOIOCTLCMD;
3505 if (sock->ops->compat_ioctl)
3506 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3508 if (ret == -ENOIOCTLCMD &&
3509 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3510 ret = compat_wext_handle_ioctl(net, cmd, arg);
3512 if (ret == -ENOIOCTLCMD)
3513 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3520 * kernel_bind - bind an address to a socket (kernel space)
3523 * @addrlen: length of address
3525 * Returns 0 or an error.
3528 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3530 return sock->ops->bind(sock, addr, addrlen);
3532 EXPORT_SYMBOL(kernel_bind);
3535 * kernel_listen - move socket to listening state (kernel space)
3537 * @backlog: pending connections queue size
3539 * Returns 0 or an error.
3542 int kernel_listen(struct socket *sock, int backlog)
3544 return sock->ops->listen(sock, backlog);
3546 EXPORT_SYMBOL(kernel_listen);
3549 * kernel_accept - accept a connection (kernel space)
3550 * @sock: listening socket
3551 * @newsock: new connected socket
3554 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3555 * If it fails, @newsock is guaranteed to be %NULL.
3556 * Returns 0 or an error.
3559 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3561 struct sock *sk = sock->sk;
3564 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3569 err = sock->ops->accept(sock, *newsock, flags, true);
3571 sock_release(*newsock);
3576 (*newsock)->ops = sock->ops;
3577 __module_get((*newsock)->ops->owner);
3582 EXPORT_SYMBOL(kernel_accept);
3585 * kernel_connect - connect a socket (kernel space)
3588 * @addrlen: address length
3589 * @flags: flags (O_NONBLOCK, ...)
3591 * For datagram sockets, @addr is the addres to which datagrams are sent
3592 * by default, and the only address from which datagrams are received.
3593 * For stream sockets, attempts to connect to @addr.
3594 * Returns 0 or an error code.
3597 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3600 return sock->ops->connect(sock, addr, addrlen, flags);
3602 EXPORT_SYMBOL(kernel_connect);
3605 * kernel_getsockname - get the address which the socket is bound (kernel space)
3607 * @addr: address holder
3609 * Fills the @addr pointer with the address which the socket is bound.
3610 * Returns 0 or an error code.
3613 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3615 return sock->ops->getname(sock, addr, 0);
3617 EXPORT_SYMBOL(kernel_getsockname);
3620 * kernel_peername - get the address which the socket is connected (kernel space)
3622 * @addr: address holder
3624 * Fills the @addr pointer with the address which the socket is connected.
3625 * Returns 0 or an error code.
3628 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3630 return sock->ops->getname(sock, addr, 1);
3632 EXPORT_SYMBOL(kernel_getpeername);
3635 * kernel_sendpage - send a &page through a socket (kernel space)
3638 * @offset: page offset
3639 * @size: total size in bytes
3640 * @flags: flags (MSG_DONTWAIT, ...)
3642 * Returns the total amount sent in bytes or an error.
3645 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3646 size_t size, int flags)
3648 if (sock->ops->sendpage)
3649 return sock->ops->sendpage(sock, page, offset, size, flags);
3651 return sock_no_sendpage(sock, page, offset, size, flags);
3653 EXPORT_SYMBOL(kernel_sendpage);
3656 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3659 * @offset: page offset
3660 * @size: total size in bytes
3661 * @flags: flags (MSG_DONTWAIT, ...)
3663 * Returns the total amount sent in bytes or an error.
3664 * Caller must hold @sk.
3667 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3668 size_t size, int flags)
3670 struct socket *sock = sk->sk_socket;
3672 if (sock->ops->sendpage_locked)
3673 return sock->ops->sendpage_locked(sk, page, offset, size,
3676 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3678 EXPORT_SYMBOL(kernel_sendpage_locked);
3681 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3683 * @how: connection part
3685 * Returns 0 or an error.
3688 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3690 return sock->ops->shutdown(sock, how);
3692 EXPORT_SYMBOL(kernel_sock_shutdown);
3695 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3698 * This routine returns the IP overhead imposed by a socket i.e.
3699 * the length of the underlying IP header, depending on whether
3700 * this is an IPv4 or IPv6 socket and the length from IP options turned
3701 * on at the socket. Assumes that the caller has a lock on the socket.
3704 u32 kernel_sock_ip_overhead(struct sock *sk)
3706 struct inet_sock *inet;
3707 struct ip_options_rcu *opt;
3709 #if IS_ENABLED(CONFIG_IPV6)
3710 struct ipv6_pinfo *np;
3711 struct ipv6_txoptions *optv6 = NULL;
3712 #endif /* IS_ENABLED(CONFIG_IPV6) */
3717 switch (sk->sk_family) {
3720 overhead += sizeof(struct iphdr);
3721 opt = rcu_dereference_protected(inet->inet_opt,
3722 sock_owned_by_user(sk));
3724 overhead += opt->opt.optlen;
3726 #if IS_ENABLED(CONFIG_IPV6)
3729 overhead += sizeof(struct ipv6hdr);
3731 optv6 = rcu_dereference_protected(np->opt,
3732 sock_owned_by_user(sk));
3734 overhead += (optv6->opt_flen + optv6->opt_nflen);
3736 #endif /* IS_ENABLED(CONFIG_IPV6) */
3737 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3741 EXPORT_SYMBOL(kernel_sock_ip_overhead);