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 & FDPUT_FPUT;
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);
658 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
660 int err = security_socket_sendmsg(sock, msg,
663 return err ?: sock_sendmsg_nosec(sock, msg);
667 * sock_sendmsg - send a message through @sock
669 * @msg: message to send
671 * Sends @msg through @sock, passing through LSM.
672 * Returns the number of bytes sent, or an error code.
674 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
676 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
677 struct sockaddr_storage address;
678 int save_len = msg->msg_namelen;
682 memcpy(&address, msg->msg_name, msg->msg_namelen);
683 msg->msg_name = &address;
686 ret = __sock_sendmsg(sock, msg);
687 msg->msg_name = save_addr;
688 msg->msg_namelen = save_len;
692 EXPORT_SYMBOL(sock_sendmsg);
695 * kernel_sendmsg - send a message through @sock (kernel-space)
697 * @msg: message header
699 * @num: vec array length
700 * @size: total message data size
702 * Builds the message data with @vec and sends it through @sock.
703 * Returns the number of bytes sent, or an error code.
706 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
707 struct kvec *vec, size_t num, size_t size)
709 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
710 return sock_sendmsg(sock, msg);
712 EXPORT_SYMBOL(kernel_sendmsg);
715 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
717 * @msg: message header
718 * @vec: output s/g array
719 * @num: output s/g array length
720 * @size: total message data size
722 * Builds the message data with @vec and sends it through @sock.
723 * Returns the number of bytes sent, or an error code.
724 * Caller must hold @sk.
727 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
728 struct kvec *vec, size_t num, size_t size)
730 struct socket *sock = sk->sk_socket;
732 if (!sock->ops->sendmsg_locked)
733 return sock_no_sendmsg_locked(sk, msg, size);
735 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
737 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
739 EXPORT_SYMBOL(kernel_sendmsg_locked);
741 static bool skb_is_err_queue(const struct sk_buff *skb)
743 /* pkt_type of skbs enqueued on the error queue are set to
744 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
745 * in recvmsg, since skbs received on a local socket will never
746 * have a pkt_type of PACKET_OUTGOING.
748 return skb->pkt_type == PACKET_OUTGOING;
751 /* On transmit, software and hardware timestamps are returned independently.
752 * As the two skb clones share the hardware timestamp, which may be updated
753 * before the software timestamp is received, a hardware TX timestamp may be
754 * returned only if there is no software TX timestamp. Ignore false software
755 * timestamps, which may be made in the __sock_recv_timestamp() call when the
756 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
757 * hardware timestamp.
759 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
761 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
764 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
766 struct scm_ts_pktinfo ts_pktinfo;
767 struct net_device *orig_dev;
769 if (!skb_mac_header_was_set(skb))
772 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
775 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
777 ts_pktinfo.if_index = orig_dev->ifindex;
780 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
781 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
782 sizeof(ts_pktinfo), &ts_pktinfo);
786 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
788 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
791 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
792 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
793 struct scm_timestamping_internal tss;
795 int empty = 1, false_tstamp = 0;
796 struct skb_shared_hwtstamps *shhwtstamps =
799 /* Race occurred between timestamp enabling and packet
800 receiving. Fill in the current time for now. */
801 if (need_software_tstamp && skb->tstamp == 0) {
802 __net_timestamp(skb);
806 if (need_software_tstamp) {
807 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
809 struct __kernel_sock_timeval tv;
811 skb_get_new_timestamp(skb, &tv);
812 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
815 struct __kernel_old_timeval tv;
817 skb_get_timestamp(skb, &tv);
818 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
823 struct __kernel_timespec ts;
825 skb_get_new_timestampns(skb, &ts);
826 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
829 struct __kernel_old_timespec ts;
831 skb_get_timestampns(skb, &ts);
832 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
838 memset(&tss, 0, sizeof(tss));
839 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
840 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
843 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
844 !skb_is_swtx_tstamp(skb, false_tstamp) &&
845 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
847 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
848 !skb_is_err_queue(skb))
849 put_ts_pktinfo(msg, skb);
852 if (sock_flag(sk, SOCK_TSTAMP_NEW))
853 put_cmsg_scm_timestamping64(msg, &tss);
855 put_cmsg_scm_timestamping(msg, &tss);
857 if (skb_is_err_queue(skb) && skb->len &&
858 SKB_EXT_ERR(skb)->opt_stats)
859 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
860 skb->len, skb->data);
863 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
865 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
870 if (!sock_flag(sk, SOCK_WIFI_STATUS))
872 if (!skb->wifi_acked_valid)
875 ack = skb->wifi_acked;
877 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
879 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
881 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
884 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
885 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
886 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
889 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
892 sock_recv_timestamp(msg, sk, skb);
893 sock_recv_drops(msg, sk, skb);
895 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
897 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
899 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
901 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
904 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
905 inet_recvmsg, sock, msg, msg_data_left(msg),
910 * sock_recvmsg - receive a message from @sock
912 * @msg: message to receive
913 * @flags: message flags
915 * Receives @msg from @sock, passing through LSM. Returns the total number
916 * of bytes received, or an error.
918 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
920 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
922 return err ?: sock_recvmsg_nosec(sock, msg, flags);
924 EXPORT_SYMBOL(sock_recvmsg);
927 * kernel_recvmsg - Receive a message from a socket (kernel space)
928 * @sock: The socket to receive the message from
929 * @msg: Received message
930 * @vec: Input s/g array for message data
931 * @num: Size of input s/g array
932 * @size: Number of bytes to read
933 * @flags: Message flags (MSG_DONTWAIT, etc...)
935 * On return the msg structure contains the scatter/gather array passed in the
936 * vec argument. The array is modified so that it consists of the unfilled
937 * portion of the original array.
939 * The returned value is the total number of bytes received, or an error.
942 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
943 struct kvec *vec, size_t num, size_t size, int flags)
945 msg->msg_control_is_user = false;
946 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
947 return sock_recvmsg(sock, msg, flags);
949 EXPORT_SYMBOL(kernel_recvmsg);
951 static ssize_t sock_sendpage(struct file *file, struct page *page,
952 int offset, size_t size, loff_t *ppos, int more)
957 sock = file->private_data;
959 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
960 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
963 return kernel_sendpage(sock, page, offset, size, flags);
966 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
967 struct pipe_inode_info *pipe, size_t len,
970 struct socket *sock = file->private_data;
972 if (unlikely(!sock->ops->splice_read))
973 return generic_file_splice_read(file, ppos, pipe, len, flags);
975 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
978 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
980 struct file *file = iocb->ki_filp;
981 struct socket *sock = file->private_data;
982 struct msghdr msg = {.msg_iter = *to,
986 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
987 msg.msg_flags = MSG_DONTWAIT;
989 if (iocb->ki_pos != 0)
992 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
995 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1000 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1002 struct file *file = iocb->ki_filp;
1003 struct socket *sock = file->private_data;
1004 struct msghdr msg = {.msg_iter = *from,
1008 if (iocb->ki_pos != 0)
1011 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1012 msg.msg_flags = MSG_DONTWAIT;
1014 if (sock->type == SOCK_SEQPACKET)
1015 msg.msg_flags |= MSG_EOR;
1017 res = __sock_sendmsg(sock, &msg);
1018 *from = msg.msg_iter;
1023 * Atomic setting of ioctl hooks to avoid race
1024 * with module unload.
1027 static DEFINE_MUTEX(br_ioctl_mutex);
1028 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1030 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1032 mutex_lock(&br_ioctl_mutex);
1033 br_ioctl_hook = hook;
1034 mutex_unlock(&br_ioctl_mutex);
1036 EXPORT_SYMBOL(brioctl_set);
1038 static DEFINE_MUTEX(vlan_ioctl_mutex);
1039 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1041 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1043 mutex_lock(&vlan_ioctl_mutex);
1044 vlan_ioctl_hook = hook;
1045 mutex_unlock(&vlan_ioctl_mutex);
1047 EXPORT_SYMBOL(vlan_ioctl_set);
1049 static DEFINE_MUTEX(dlci_ioctl_mutex);
1050 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1052 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1054 mutex_lock(&dlci_ioctl_mutex);
1055 dlci_ioctl_hook = hook;
1056 mutex_unlock(&dlci_ioctl_mutex);
1058 EXPORT_SYMBOL(dlci_ioctl_set);
1060 static long sock_do_ioctl(struct net *net, struct socket *sock,
1061 unsigned int cmd, unsigned long arg)
1064 void __user *argp = (void __user *)arg;
1066 err = sock->ops->ioctl(sock, cmd, arg);
1069 * If this ioctl is unknown try to hand it down
1070 * to the NIC driver.
1072 if (err != -ENOIOCTLCMD)
1075 if (cmd == SIOCGIFCONF) {
1077 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1080 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1082 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1084 } else if (is_socket_ioctl_cmd(cmd)) {
1087 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1089 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1090 if (!err && need_copyout)
1091 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1100 * With an ioctl, arg may well be a user mode pointer, but we don't know
1101 * what to do with it - that's up to the protocol still.
1104 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1106 struct socket *sock;
1108 void __user *argp = (void __user *)arg;
1112 sock = file->private_data;
1115 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1118 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1120 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1121 if (!err && need_copyout)
1122 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1125 #ifdef CONFIG_WEXT_CORE
1126 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1127 err = wext_handle_ioctl(net, cmd, argp);
1134 if (get_user(pid, (int __user *)argp))
1136 err = f_setown(sock->file, pid, 1);
1140 err = put_user(f_getown(sock->file),
1141 (int __user *)argp);
1149 request_module("bridge");
1151 mutex_lock(&br_ioctl_mutex);
1153 err = br_ioctl_hook(net, cmd, argp);
1154 mutex_unlock(&br_ioctl_mutex);
1159 if (!vlan_ioctl_hook)
1160 request_module("8021q");
1162 mutex_lock(&vlan_ioctl_mutex);
1163 if (vlan_ioctl_hook)
1164 err = vlan_ioctl_hook(net, argp);
1165 mutex_unlock(&vlan_ioctl_mutex);
1170 if (!dlci_ioctl_hook)
1171 request_module("dlci");
1173 mutex_lock(&dlci_ioctl_mutex);
1174 if (dlci_ioctl_hook)
1175 err = dlci_ioctl_hook(cmd, argp);
1176 mutex_unlock(&dlci_ioctl_mutex);
1180 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1183 err = open_related_ns(&net->ns, get_net_ns);
1185 case SIOCGSTAMP_OLD:
1186 case SIOCGSTAMPNS_OLD:
1187 if (!sock->ops->gettstamp) {
1191 err = sock->ops->gettstamp(sock, argp,
1192 cmd == SIOCGSTAMP_OLD,
1193 !IS_ENABLED(CONFIG_64BIT));
1195 case SIOCGSTAMP_NEW:
1196 case SIOCGSTAMPNS_NEW:
1197 if (!sock->ops->gettstamp) {
1201 err = sock->ops->gettstamp(sock, argp,
1202 cmd == SIOCGSTAMP_NEW,
1206 err = sock_do_ioctl(net, sock, cmd, arg);
1213 * sock_create_lite - creates a socket
1214 * @family: protocol family (AF_INET, ...)
1215 * @type: communication type (SOCK_STREAM, ...)
1216 * @protocol: protocol (0, ...)
1219 * Creates a new socket and assigns it to @res, passing through LSM.
1220 * The new socket initialization is not complete, see kernel_accept().
1221 * Returns 0 or an error. On failure @res is set to %NULL.
1222 * This function internally uses GFP_KERNEL.
1225 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1228 struct socket *sock = NULL;
1230 err = security_socket_create(family, type, protocol, 1);
1234 sock = sock_alloc();
1241 err = security_socket_post_create(sock, family, type, protocol, 1);
1253 EXPORT_SYMBOL(sock_create_lite);
1255 /* No kernel lock held - perfect */
1256 static __poll_t sock_poll(struct file *file, poll_table *wait)
1258 struct socket *sock = file->private_data;
1259 __poll_t events = poll_requested_events(wait), flag = 0;
1261 if (!sock->ops->poll)
1264 if (sk_can_busy_loop(sock->sk)) {
1265 /* poll once if requested by the syscall */
1266 if (events & POLL_BUSY_LOOP)
1267 sk_busy_loop(sock->sk, 1);
1269 /* if this socket can poll_ll, tell the system call */
1270 flag = POLL_BUSY_LOOP;
1273 return sock->ops->poll(file, sock, wait) | flag;
1276 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1278 struct socket *sock = file->private_data;
1280 return sock->ops->mmap(file, sock, vma);
1283 static int sock_close(struct inode *inode, struct file *filp)
1285 __sock_release(SOCKET_I(inode), inode);
1290 * Update the socket async list
1292 * Fasync_list locking strategy.
1294 * 1. fasync_list is modified only under process context socket lock
1295 * i.e. under semaphore.
1296 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1297 * or under socket lock
1300 static int sock_fasync(int fd, struct file *filp, int on)
1302 struct socket *sock = filp->private_data;
1303 struct sock *sk = sock->sk;
1304 struct socket_wq *wq = &sock->wq;
1310 fasync_helper(fd, filp, on, &wq->fasync_list);
1312 if (!wq->fasync_list)
1313 sock_reset_flag(sk, SOCK_FASYNC);
1315 sock_set_flag(sk, SOCK_FASYNC);
1321 /* This function may be called only under rcu_lock */
1323 int sock_wake_async(struct socket_wq *wq, int how, int band)
1325 if (!wq || !wq->fasync_list)
1329 case SOCK_WAKE_WAITD:
1330 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1333 case SOCK_WAKE_SPACE:
1334 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1339 kill_fasync(&wq->fasync_list, SIGIO, band);
1342 kill_fasync(&wq->fasync_list, SIGURG, band);
1347 EXPORT_SYMBOL(sock_wake_async);
1350 * __sock_create - creates a socket
1351 * @net: net namespace
1352 * @family: protocol family (AF_INET, ...)
1353 * @type: communication type (SOCK_STREAM, ...)
1354 * @protocol: protocol (0, ...)
1356 * @kern: boolean for kernel space sockets
1358 * Creates a new socket and assigns it to @res, passing through LSM.
1359 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1360 * be set to true if the socket resides in kernel space.
1361 * This function internally uses GFP_KERNEL.
1364 int __sock_create(struct net *net, int family, int type, int protocol,
1365 struct socket **res, int kern)
1368 struct socket *sock;
1369 const struct net_proto_family *pf;
1372 * Check protocol is in range
1374 if (family < 0 || family >= NPROTO)
1375 return -EAFNOSUPPORT;
1376 if (type < 0 || type >= SOCK_MAX)
1381 This uglymoron is moved from INET layer to here to avoid
1382 deadlock in module load.
1384 if (family == PF_INET && type == SOCK_PACKET) {
1385 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1390 err = security_socket_create(family, type, protocol, kern);
1395 * Allocate the socket and allow the family to set things up. if
1396 * the protocol is 0, the family is instructed to select an appropriate
1399 sock = sock_alloc();
1401 net_warn_ratelimited("socket: no more sockets\n");
1402 return -ENFILE; /* Not exactly a match, but its the
1403 closest posix thing */
1408 #ifdef CONFIG_MODULES
1409 /* Attempt to load a protocol module if the find failed.
1411 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1412 * requested real, full-featured networking support upon configuration.
1413 * Otherwise module support will break!
1415 if (rcu_access_pointer(net_families[family]) == NULL)
1416 request_module("net-pf-%d", family);
1420 pf = rcu_dereference(net_families[family]);
1421 err = -EAFNOSUPPORT;
1426 * We will call the ->create function, that possibly is in a loadable
1427 * module, so we have to bump that loadable module refcnt first.
1429 if (!try_module_get(pf->owner))
1432 /* Now protected by module ref count */
1435 err = pf->create(net, sock, protocol, kern);
1437 goto out_module_put;
1440 * Now to bump the refcnt of the [loadable] module that owns this
1441 * socket at sock_release time we decrement its refcnt.
1443 if (!try_module_get(sock->ops->owner))
1444 goto out_module_busy;
1447 * Now that we're done with the ->create function, the [loadable]
1448 * module can have its refcnt decremented
1450 module_put(pf->owner);
1451 err = security_socket_post_create(sock, family, type, protocol, kern);
1453 goto out_sock_release;
1459 err = -EAFNOSUPPORT;
1462 module_put(pf->owner);
1469 goto out_sock_release;
1471 EXPORT_SYMBOL(__sock_create);
1474 * sock_create - creates a socket
1475 * @family: protocol family (AF_INET, ...)
1476 * @type: communication type (SOCK_STREAM, ...)
1477 * @protocol: protocol (0, ...)
1480 * A wrapper around __sock_create().
1481 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1484 int sock_create(int family, int type, int protocol, struct socket **res)
1486 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1488 EXPORT_SYMBOL(sock_create);
1491 * sock_create_kern - creates a socket (kernel space)
1492 * @net: net namespace
1493 * @family: protocol family (AF_INET, ...)
1494 * @type: communication type (SOCK_STREAM, ...)
1495 * @protocol: protocol (0, ...)
1498 * A wrapper around __sock_create().
1499 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1502 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1504 return __sock_create(net, family, type, protocol, res, 1);
1506 EXPORT_SYMBOL(sock_create_kern);
1508 int __sys_socket(int family, int type, int protocol)
1511 struct socket *sock;
1514 /* Check the SOCK_* constants for consistency. */
1515 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1516 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1517 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1518 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1520 flags = type & ~SOCK_TYPE_MASK;
1521 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1523 type &= SOCK_TYPE_MASK;
1525 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1526 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1528 retval = sock_create(family, type, protocol, &sock);
1532 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1535 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1537 return __sys_socket(family, type, protocol);
1541 * Create a pair of connected sockets.
1544 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1546 struct socket *sock1, *sock2;
1548 struct file *newfile1, *newfile2;
1551 flags = type & ~SOCK_TYPE_MASK;
1552 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1554 type &= SOCK_TYPE_MASK;
1556 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1557 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1560 * reserve descriptors and make sure we won't fail
1561 * to return them to userland.
1563 fd1 = get_unused_fd_flags(flags);
1564 if (unlikely(fd1 < 0))
1567 fd2 = get_unused_fd_flags(flags);
1568 if (unlikely(fd2 < 0)) {
1573 err = put_user(fd1, &usockvec[0]);
1577 err = put_user(fd2, &usockvec[1]);
1582 * Obtain the first socket and check if the underlying protocol
1583 * supports the socketpair call.
1586 err = sock_create(family, type, protocol, &sock1);
1587 if (unlikely(err < 0))
1590 err = sock_create(family, type, protocol, &sock2);
1591 if (unlikely(err < 0)) {
1592 sock_release(sock1);
1596 err = security_socket_socketpair(sock1, sock2);
1597 if (unlikely(err)) {
1598 sock_release(sock2);
1599 sock_release(sock1);
1603 err = sock1->ops->socketpair(sock1, sock2);
1604 if (unlikely(err < 0)) {
1605 sock_release(sock2);
1606 sock_release(sock1);
1610 newfile1 = sock_alloc_file(sock1, flags, NULL);
1611 if (IS_ERR(newfile1)) {
1612 err = PTR_ERR(newfile1);
1613 sock_release(sock2);
1617 newfile2 = sock_alloc_file(sock2, flags, NULL);
1618 if (IS_ERR(newfile2)) {
1619 err = PTR_ERR(newfile2);
1624 audit_fd_pair(fd1, fd2);
1626 fd_install(fd1, newfile1);
1627 fd_install(fd2, newfile2);
1636 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1637 int __user *, usockvec)
1639 return __sys_socketpair(family, type, protocol, usockvec);
1643 * Bind a name to a socket. Nothing much to do here since it's
1644 * the protocol's responsibility to handle the local address.
1646 * We move the socket address to kernel space before we call
1647 * the protocol layer (having also checked the address is ok).
1650 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1652 struct socket *sock;
1653 struct sockaddr_storage address;
1654 int err, fput_needed;
1656 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1658 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1660 err = security_socket_bind(sock,
1661 (struct sockaddr *)&address,
1664 err = sock->ops->bind(sock,
1668 fput_light(sock->file, fput_needed);
1673 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1675 return __sys_bind(fd, umyaddr, addrlen);
1679 * Perform a listen. Basically, we allow the protocol to do anything
1680 * necessary for a listen, and if that works, we mark the socket as
1681 * ready for listening.
1684 int __sys_listen(int fd, int backlog)
1686 struct socket *sock;
1687 int err, fput_needed;
1690 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1692 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1693 if ((unsigned int)backlog > somaxconn)
1694 backlog = somaxconn;
1696 err = security_socket_listen(sock, backlog);
1698 err = sock->ops->listen(sock, backlog);
1700 fput_light(sock->file, fput_needed);
1705 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1707 return __sys_listen(fd, backlog);
1710 struct file *do_accept(struct file *file, unsigned file_flags,
1711 struct sockaddr __user *upeer_sockaddr,
1712 int __user *upeer_addrlen, int flags)
1714 struct socket *sock, *newsock;
1715 struct file *newfile;
1717 struct sockaddr_storage address;
1719 sock = sock_from_file(file, &err);
1721 return ERR_PTR(err);
1723 newsock = sock_alloc();
1725 return ERR_PTR(-ENFILE);
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 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1737 if (IS_ERR(newfile))
1740 err = security_socket_accept(sock, newsock);
1744 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1749 if (upeer_sockaddr) {
1750 len = newsock->ops->getname(newsock,
1751 (struct sockaddr *)&address, 2);
1753 err = -ECONNABORTED;
1756 err = move_addr_to_user(&address,
1757 len, upeer_sockaddr, upeer_addrlen);
1762 /* File flags are not inherited via accept() unlike another OSes. */
1766 return ERR_PTR(err);
1769 int __sys_accept4_file(struct file *file, unsigned file_flags,
1770 struct sockaddr __user *upeer_sockaddr,
1771 int __user *upeer_addrlen, int flags,
1772 unsigned long nofile)
1774 struct file *newfile;
1777 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1780 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1781 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1783 newfd = __get_unused_fd_flags(flags, nofile);
1784 if (unlikely(newfd < 0))
1787 newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen,
1789 if (IS_ERR(newfile)) {
1790 put_unused_fd(newfd);
1791 return PTR_ERR(newfile);
1793 fd_install(newfd, newfile);
1798 * For accept, we attempt to create a new socket, set up the link
1799 * with the client, wake up the client, then return the new
1800 * connected fd. We collect the address of the connector in kernel
1801 * space and move it to user at the very end. This is unclean because
1802 * we open the socket then return an error.
1804 * 1003.1g adds the ability to recvmsg() to query connection pending
1805 * status to recvmsg. We need to add that support in a way thats
1806 * clean when we restructure accept also.
1809 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1810 int __user *upeer_addrlen, int flags)
1817 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1818 upeer_addrlen, flags,
1819 rlimit(RLIMIT_NOFILE));
1826 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1827 int __user *, upeer_addrlen, int, flags)
1829 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1832 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1833 int __user *, upeer_addrlen)
1835 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1839 * Attempt to connect to a socket with the server address. The address
1840 * is in user space so we verify it is OK and move it to kernel space.
1842 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1845 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1846 * other SEQPACKET protocols that take time to connect() as it doesn't
1847 * include the -EINPROGRESS status for such sockets.
1850 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1851 int addrlen, int file_flags)
1853 struct socket *sock;
1856 sock = sock_from_file(file, &err);
1861 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1865 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1866 sock->file->f_flags | file_flags);
1871 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1878 struct sockaddr_storage address;
1880 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1882 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1889 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1892 return __sys_connect(fd, uservaddr, addrlen);
1896 * Get the local address ('name') of a socket object. Move the obtained
1897 * name to user space.
1900 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1901 int __user *usockaddr_len)
1903 struct socket *sock;
1904 struct sockaddr_storage address;
1905 int err, fput_needed;
1907 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1911 err = security_socket_getsockname(sock);
1915 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1918 /* "err" is actually length in this case */
1919 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1922 fput_light(sock->file, fput_needed);
1927 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1928 int __user *, usockaddr_len)
1930 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1934 * Get the remote address ('name') of a socket object. Move the obtained
1935 * name to user space.
1938 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1939 int __user *usockaddr_len)
1941 struct socket *sock;
1942 struct sockaddr_storage address;
1943 int err, fput_needed;
1945 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1947 err = security_socket_getpeername(sock);
1949 fput_light(sock->file, fput_needed);
1953 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1955 /* "err" is actually length in this case */
1956 err = move_addr_to_user(&address, err, usockaddr,
1958 fput_light(sock->file, fput_needed);
1963 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1964 int __user *, usockaddr_len)
1966 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1970 * Send a datagram to a given address. We move the address into kernel
1971 * space and check the user space data area is readable before invoking
1974 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1975 struct sockaddr __user *addr, int addr_len)
1977 struct socket *sock;
1978 struct sockaddr_storage address;
1984 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1987 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1991 msg.msg_name = NULL;
1992 msg.msg_control = NULL;
1993 msg.msg_controllen = 0;
1994 msg.msg_namelen = 0;
1996 err = move_addr_to_kernel(addr, addr_len, &address);
1999 msg.msg_name = (struct sockaddr *)&address;
2000 msg.msg_namelen = addr_len;
2002 if (sock->file->f_flags & O_NONBLOCK)
2003 flags |= MSG_DONTWAIT;
2004 msg.msg_flags = flags;
2005 err = __sock_sendmsg(sock, &msg);
2008 fput_light(sock->file, fput_needed);
2013 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2014 unsigned int, flags, struct sockaddr __user *, addr,
2017 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2021 * Send a datagram down a socket.
2024 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2025 unsigned int, flags)
2027 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2031 * Receive a frame from the socket and optionally record the address of the
2032 * sender. We verify the buffers are writable and if needed move the
2033 * sender address from kernel to user space.
2035 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2036 struct sockaddr __user *addr, int __user *addr_len)
2038 struct socket *sock;
2041 struct sockaddr_storage address;
2045 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2048 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2052 msg.msg_control = NULL;
2053 msg.msg_controllen = 0;
2054 /* Save some cycles and don't copy the address if not needed */
2055 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2056 /* We assume all kernel code knows the size of sockaddr_storage */
2057 msg.msg_namelen = 0;
2058 msg.msg_iocb = NULL;
2060 if (sock->file->f_flags & O_NONBLOCK)
2061 flags |= MSG_DONTWAIT;
2062 err = sock_recvmsg(sock, &msg, flags);
2064 if (err >= 0 && addr != NULL) {
2065 err2 = move_addr_to_user(&address,
2066 msg.msg_namelen, addr, addr_len);
2071 fput_light(sock->file, fput_needed);
2076 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2077 unsigned int, flags, struct sockaddr __user *, addr,
2078 int __user *, addr_len)
2080 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2084 * Receive a datagram from a socket.
2087 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2088 unsigned int, flags)
2090 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2093 static bool sock_use_custom_sol_socket(const struct socket *sock)
2095 const struct sock *sk = sock->sk;
2097 /* Use sock->ops->setsockopt() for MPTCP */
2098 return IS_ENABLED(CONFIG_MPTCP) &&
2099 sk->sk_protocol == IPPROTO_MPTCP &&
2100 sk->sk_type == SOCK_STREAM &&
2101 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2105 * Set a socket option. Because we don't know the option lengths we have
2106 * to pass the user mode parameter for the protocols to sort out.
2108 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2111 sockptr_t optval = USER_SOCKPTR(user_optval);
2112 char *kernel_optval = NULL;
2113 int err, fput_needed;
2114 struct socket *sock;
2119 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2123 err = security_socket_setsockopt(sock, level, optname);
2127 if (!in_compat_syscall())
2128 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2129 user_optval, &optlen,
2139 optval = KERNEL_SOCKPTR(kernel_optval);
2140 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2141 err = sock_setsockopt(sock, level, optname, optval, optlen);
2142 else if (unlikely(!sock->ops->setsockopt))
2145 err = sock->ops->setsockopt(sock, level, optname, optval,
2147 kfree(kernel_optval);
2149 fput_light(sock->file, fput_needed);
2153 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2154 char __user *, optval, int, optlen)
2156 return __sys_setsockopt(fd, level, optname, optval, optlen);
2159 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2163 * Get a socket option. Because we don't know the option lengths we have
2164 * to pass a user mode parameter for the protocols to sort out.
2166 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2169 int err, fput_needed;
2170 struct socket *sock;
2173 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2177 err = security_socket_getsockopt(sock, level, optname);
2181 if (!in_compat_syscall())
2182 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2184 if (level == SOL_SOCKET)
2185 err = sock_getsockopt(sock, level, optname, optval, optlen);
2186 else if (unlikely(!sock->ops->getsockopt))
2189 err = sock->ops->getsockopt(sock, level, optname, optval,
2192 if (!in_compat_syscall())
2193 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2194 optval, optlen, max_optlen,
2197 fput_light(sock->file, fput_needed);
2201 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2202 char __user *, optval, int __user *, optlen)
2204 return __sys_getsockopt(fd, level, optname, optval, optlen);
2208 * Shutdown a socket.
2211 int __sys_shutdown_sock(struct socket *sock, int how)
2215 err = security_socket_shutdown(sock, how);
2217 err = sock->ops->shutdown(sock, how);
2222 int __sys_shutdown(int fd, int how)
2224 int err, fput_needed;
2225 struct socket *sock;
2227 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2229 err = __sys_shutdown_sock(sock, how);
2230 fput_light(sock->file, fput_needed);
2235 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2237 return __sys_shutdown(fd, how);
2240 /* A couple of helpful macros for getting the address of the 32/64 bit
2241 * fields which are the same type (int / unsigned) on our platforms.
2243 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2244 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2245 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2247 struct used_address {
2248 struct sockaddr_storage name;
2249 unsigned int name_len;
2252 int __copy_msghdr_from_user(struct msghdr *kmsg,
2253 struct user_msghdr __user *umsg,
2254 struct sockaddr __user **save_addr,
2255 struct iovec __user **uiov, size_t *nsegs)
2257 struct user_msghdr msg;
2260 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2263 kmsg->msg_control_is_user = true;
2264 kmsg->msg_control_user = msg.msg_control;
2265 kmsg->msg_controllen = msg.msg_controllen;
2266 kmsg->msg_flags = msg.msg_flags;
2268 kmsg->msg_namelen = msg.msg_namelen;
2270 kmsg->msg_namelen = 0;
2272 if (kmsg->msg_namelen < 0)
2275 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2276 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2279 *save_addr = msg.msg_name;
2281 if (msg.msg_name && kmsg->msg_namelen) {
2283 err = move_addr_to_kernel(msg.msg_name,
2290 kmsg->msg_name = NULL;
2291 kmsg->msg_namelen = 0;
2294 if (msg.msg_iovlen > UIO_MAXIOV)
2297 kmsg->msg_iocb = NULL;
2298 *uiov = msg.msg_iov;
2299 *nsegs = msg.msg_iovlen;
2303 static int copy_msghdr_from_user(struct msghdr *kmsg,
2304 struct user_msghdr __user *umsg,
2305 struct sockaddr __user **save_addr,
2308 struct user_msghdr msg;
2311 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2316 err = import_iovec(save_addr ? READ : WRITE,
2317 msg.msg_iov, msg.msg_iovlen,
2318 UIO_FASTIOV, iov, &kmsg->msg_iter);
2319 return err < 0 ? err : 0;
2322 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2323 unsigned int flags, struct used_address *used_address,
2324 unsigned int allowed_msghdr_flags)
2326 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2327 __aligned(sizeof(__kernel_size_t));
2328 /* 20 is size of ipv6_pktinfo */
2329 unsigned char *ctl_buf = ctl;
2335 if (msg_sys->msg_controllen > INT_MAX)
2337 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2338 ctl_len = msg_sys->msg_controllen;
2339 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2341 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2345 ctl_buf = msg_sys->msg_control;
2346 ctl_len = msg_sys->msg_controllen;
2347 } else if (ctl_len) {
2348 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2349 CMSG_ALIGN(sizeof(struct cmsghdr)));
2350 if (ctl_len > sizeof(ctl)) {
2351 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2352 if (ctl_buf == NULL)
2356 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2358 msg_sys->msg_control = ctl_buf;
2359 msg_sys->msg_control_is_user = false;
2361 msg_sys->msg_flags = flags;
2363 if (sock->file->f_flags & O_NONBLOCK)
2364 msg_sys->msg_flags |= MSG_DONTWAIT;
2366 * If this is sendmmsg() and current destination address is same as
2367 * previously succeeded address, omit asking LSM's decision.
2368 * used_address->name_len is initialized to UINT_MAX so that the first
2369 * destination address never matches.
2371 if (used_address && msg_sys->msg_name &&
2372 used_address->name_len == msg_sys->msg_namelen &&
2373 !memcmp(&used_address->name, msg_sys->msg_name,
2374 used_address->name_len)) {
2375 err = sock_sendmsg_nosec(sock, msg_sys);
2378 err = __sock_sendmsg(sock, msg_sys);
2380 * If this is sendmmsg() and sending to current destination address was
2381 * successful, remember it.
2383 if (used_address && err >= 0) {
2384 used_address->name_len = msg_sys->msg_namelen;
2385 if (msg_sys->msg_name)
2386 memcpy(&used_address->name, msg_sys->msg_name,
2387 used_address->name_len);
2392 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2397 int sendmsg_copy_msghdr(struct msghdr *msg,
2398 struct user_msghdr __user *umsg, unsigned flags,
2403 if (flags & MSG_CMSG_COMPAT) {
2404 struct compat_msghdr __user *msg_compat;
2406 msg_compat = (struct compat_msghdr __user *) umsg;
2407 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2409 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2417 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2418 struct msghdr *msg_sys, unsigned int flags,
2419 struct used_address *used_address,
2420 unsigned int allowed_msghdr_flags)
2422 struct sockaddr_storage address;
2423 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2426 msg_sys->msg_name = &address;
2428 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2432 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2433 allowed_msghdr_flags);
2439 * BSD sendmsg interface
2441 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2444 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2447 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2448 bool forbid_cmsg_compat)
2450 int fput_needed, err;
2451 struct msghdr msg_sys;
2452 struct socket *sock;
2454 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2457 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2461 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2463 fput_light(sock->file, fput_needed);
2468 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2470 return __sys_sendmsg(fd, msg, flags, true);
2474 * Linux sendmmsg interface
2477 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2478 unsigned int flags, bool forbid_cmsg_compat)
2480 int fput_needed, err, datagrams;
2481 struct socket *sock;
2482 struct mmsghdr __user *entry;
2483 struct compat_mmsghdr __user *compat_entry;
2484 struct msghdr msg_sys;
2485 struct used_address used_address;
2486 unsigned int oflags = flags;
2488 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2491 if (vlen > UIO_MAXIOV)
2496 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2500 used_address.name_len = UINT_MAX;
2502 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2506 while (datagrams < vlen) {
2507 if (datagrams == vlen - 1)
2510 if (MSG_CMSG_COMPAT & flags) {
2511 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2512 &msg_sys, flags, &used_address, MSG_EOR);
2515 err = __put_user(err, &compat_entry->msg_len);
2518 err = ___sys_sendmsg(sock,
2519 (struct user_msghdr __user *)entry,
2520 &msg_sys, flags, &used_address, MSG_EOR);
2523 err = put_user(err, &entry->msg_len);
2530 if (msg_data_left(&msg_sys))
2535 fput_light(sock->file, fput_needed);
2537 /* We only return an error if no datagrams were able to be sent */
2544 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2545 unsigned int, vlen, unsigned int, flags)
2547 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2550 int recvmsg_copy_msghdr(struct msghdr *msg,
2551 struct user_msghdr __user *umsg, unsigned flags,
2552 struct sockaddr __user **uaddr,
2557 if (MSG_CMSG_COMPAT & flags) {
2558 struct compat_msghdr __user *msg_compat;
2560 msg_compat = (struct compat_msghdr __user *) umsg;
2561 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2563 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2571 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2572 struct user_msghdr __user *msg,
2573 struct sockaddr __user *uaddr,
2574 unsigned int flags, int nosec)
2576 struct compat_msghdr __user *msg_compat =
2577 (struct compat_msghdr __user *) msg;
2578 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2579 struct sockaddr_storage addr;
2580 unsigned long cmsg_ptr;
2584 msg_sys->msg_name = &addr;
2585 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2586 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2588 /* We assume all kernel code knows the size of sockaddr_storage */
2589 msg_sys->msg_namelen = 0;
2591 if (sock->file->f_flags & O_NONBLOCK)
2592 flags |= MSG_DONTWAIT;
2594 if (unlikely(nosec))
2595 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2597 err = sock_recvmsg(sock, msg_sys, flags);
2603 if (uaddr != NULL) {
2604 err = move_addr_to_user(&addr,
2605 msg_sys->msg_namelen, uaddr,
2610 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2614 if (MSG_CMSG_COMPAT & flags)
2615 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2616 &msg_compat->msg_controllen);
2618 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2619 &msg->msg_controllen);
2627 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2628 struct msghdr *msg_sys, unsigned int flags, int nosec)
2630 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2631 /* user mode address pointers */
2632 struct sockaddr __user *uaddr;
2635 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2639 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2645 * BSD recvmsg interface
2648 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2649 struct user_msghdr __user *umsg,
2650 struct sockaddr __user *uaddr, unsigned int flags)
2652 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2655 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2656 bool forbid_cmsg_compat)
2658 int fput_needed, err;
2659 struct msghdr msg_sys;
2660 struct socket *sock;
2662 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2665 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2669 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2671 fput_light(sock->file, fput_needed);
2676 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2677 unsigned int, flags)
2679 return __sys_recvmsg(fd, msg, flags, true);
2683 * Linux recvmmsg interface
2686 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2687 unsigned int vlen, unsigned int flags,
2688 struct timespec64 *timeout)
2690 int fput_needed, err, datagrams;
2691 struct socket *sock;
2692 struct mmsghdr __user *entry;
2693 struct compat_mmsghdr __user *compat_entry;
2694 struct msghdr msg_sys;
2695 struct timespec64 end_time;
2696 struct timespec64 timeout64;
2699 poll_select_set_timeout(&end_time, timeout->tv_sec,
2705 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2709 if (likely(!(flags & MSG_ERRQUEUE))) {
2710 err = sock_error(sock->sk);
2718 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2720 while (datagrams < vlen) {
2722 * No need to ask LSM for more than the first datagram.
2724 if (MSG_CMSG_COMPAT & flags) {
2725 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2726 &msg_sys, flags & ~MSG_WAITFORONE,
2730 err = __put_user(err, &compat_entry->msg_len);
2733 err = ___sys_recvmsg(sock,
2734 (struct user_msghdr __user *)entry,
2735 &msg_sys, flags & ~MSG_WAITFORONE,
2739 err = put_user(err, &entry->msg_len);
2747 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2748 if (flags & MSG_WAITFORONE)
2749 flags |= MSG_DONTWAIT;
2752 ktime_get_ts64(&timeout64);
2753 *timeout = timespec64_sub(end_time, timeout64);
2754 if (timeout->tv_sec < 0) {
2755 timeout->tv_sec = timeout->tv_nsec = 0;
2759 /* Timeout, return less than vlen datagrams */
2760 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2764 /* Out of band data, return right away */
2765 if (msg_sys.msg_flags & MSG_OOB)
2773 if (datagrams == 0) {
2779 * We may return less entries than requested (vlen) if the
2780 * sock is non block and there aren't enough datagrams...
2782 if (err != -EAGAIN) {
2784 * ... or if recvmsg returns an error after we
2785 * received some datagrams, where we record the
2786 * error to return on the next call or if the
2787 * app asks about it using getsockopt(SO_ERROR).
2789 WRITE_ONCE(sock->sk->sk_err, -err);
2792 fput_light(sock->file, fput_needed);
2797 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2798 unsigned int vlen, unsigned int flags,
2799 struct __kernel_timespec __user *timeout,
2800 struct old_timespec32 __user *timeout32)
2803 struct timespec64 timeout_sys;
2805 if (timeout && get_timespec64(&timeout_sys, timeout))
2808 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2811 if (!timeout && !timeout32)
2812 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2814 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2819 if (timeout && put_timespec64(&timeout_sys, timeout))
2820 datagrams = -EFAULT;
2822 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2823 datagrams = -EFAULT;
2828 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2829 unsigned int, vlen, unsigned int, flags,
2830 struct __kernel_timespec __user *, timeout)
2832 if (flags & MSG_CMSG_COMPAT)
2835 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2838 #ifdef CONFIG_COMPAT_32BIT_TIME
2839 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2840 unsigned int, vlen, unsigned int, flags,
2841 struct old_timespec32 __user *, timeout)
2843 if (flags & MSG_CMSG_COMPAT)
2846 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2850 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2851 /* Argument list sizes for sys_socketcall */
2852 #define AL(x) ((x) * sizeof(unsigned long))
2853 static const unsigned char nargs[21] = {
2854 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2855 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2856 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2863 * System call vectors.
2865 * Argument checking cleaned up. Saved 20% in size.
2866 * This function doesn't need to set the kernel lock because
2867 * it is set by the callees.
2870 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2872 unsigned long a[AUDITSC_ARGS];
2873 unsigned long a0, a1;
2877 if (call < 1 || call > SYS_SENDMMSG)
2879 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2882 if (len > sizeof(a))
2885 /* copy_from_user should be SMP safe. */
2886 if (copy_from_user(a, args, len))
2889 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2898 err = __sys_socket(a0, a1, a[2]);
2901 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2904 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2907 err = __sys_listen(a0, a1);
2910 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2911 (int __user *)a[2], 0);
2913 case SYS_GETSOCKNAME:
2915 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2916 (int __user *)a[2]);
2918 case SYS_GETPEERNAME:
2920 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2921 (int __user *)a[2]);
2923 case SYS_SOCKETPAIR:
2924 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2927 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2931 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2932 (struct sockaddr __user *)a[4], a[5]);
2935 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2939 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2940 (struct sockaddr __user *)a[4],
2941 (int __user *)a[5]);
2944 err = __sys_shutdown(a0, a1);
2946 case SYS_SETSOCKOPT:
2947 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2950 case SYS_GETSOCKOPT:
2952 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2953 (int __user *)a[4]);
2956 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2960 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2964 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2968 if (IS_ENABLED(CONFIG_64BIT))
2969 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2971 (struct __kernel_timespec __user *)a[4],
2974 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2976 (struct old_timespec32 __user *)a[4]);
2979 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2980 (int __user *)a[2], a[3]);
2989 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2992 * sock_register - add a socket protocol handler
2993 * @ops: description of protocol
2995 * This function is called by a protocol handler that wants to
2996 * advertise its address family, and have it linked into the
2997 * socket interface. The value ops->family corresponds to the
2998 * socket system call protocol family.
3000 int sock_register(const struct net_proto_family *ops)
3004 if (ops->family >= NPROTO) {
3005 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3009 spin_lock(&net_family_lock);
3010 if (rcu_dereference_protected(net_families[ops->family],
3011 lockdep_is_held(&net_family_lock)))
3014 rcu_assign_pointer(net_families[ops->family], ops);
3017 spin_unlock(&net_family_lock);
3019 pr_info("NET: Registered protocol family %d\n", ops->family);
3022 EXPORT_SYMBOL(sock_register);
3025 * sock_unregister - remove a protocol handler
3026 * @family: protocol family to remove
3028 * This function is called by a protocol handler that wants to
3029 * remove its address family, and have it unlinked from the
3030 * new socket creation.
3032 * If protocol handler is a module, then it can use module reference
3033 * counts to protect against new references. If protocol handler is not
3034 * a module then it needs to provide its own protection in
3035 * the ops->create routine.
3037 void sock_unregister(int family)
3039 BUG_ON(family < 0 || family >= NPROTO);
3041 spin_lock(&net_family_lock);
3042 RCU_INIT_POINTER(net_families[family], NULL);
3043 spin_unlock(&net_family_lock);
3047 pr_info("NET: Unregistered protocol family %d\n", family);
3049 EXPORT_SYMBOL(sock_unregister);
3051 bool sock_is_registered(int family)
3053 return family < NPROTO && rcu_access_pointer(net_families[family]);
3056 static int __init sock_init(void)
3060 * Initialize the network sysctl infrastructure.
3062 err = net_sysctl_init();
3067 * Initialize skbuff SLAB cache
3072 * Initialize the protocols module.
3077 err = register_filesystem(&sock_fs_type);
3080 sock_mnt = kern_mount(&sock_fs_type);
3081 if (IS_ERR(sock_mnt)) {
3082 err = PTR_ERR(sock_mnt);
3086 /* The real protocol initialization is performed in later initcalls.
3089 #ifdef CONFIG_NETFILTER
3090 err = netfilter_init();
3095 ptp_classifier_init();
3101 unregister_filesystem(&sock_fs_type);
3105 core_initcall(sock_init); /* early initcall */
3107 #ifdef CONFIG_PROC_FS
3108 void socket_seq_show(struct seq_file *seq)
3110 seq_printf(seq, "sockets: used %d\n",
3111 sock_inuse_get(seq->private));
3113 #endif /* CONFIG_PROC_FS */
3115 #ifdef CONFIG_COMPAT
3116 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3118 struct compat_ifconf ifc32;
3122 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3125 ifc.ifc_len = ifc32.ifc_len;
3126 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3129 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3134 ifc32.ifc_len = ifc.ifc_len;
3135 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3141 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3143 compat_uptr_t uptr32;
3148 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3151 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3154 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3155 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3157 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3159 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3160 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3166 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3167 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3168 struct compat_ifreq __user *u_ifreq32)
3173 if (!is_socket_ioctl_cmd(cmd))
3175 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3177 if (get_user(data32, &u_ifreq32->ifr_data))
3179 ifreq.ifr_data = compat_ptr(data32);
3181 return dev_ioctl(net, cmd, &ifreq, NULL);
3184 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3186 struct compat_ifreq __user *uifr32)
3188 struct ifreq __user *uifr;
3191 /* Handle the fact that while struct ifreq has the same *layout* on
3192 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3193 * which are handled elsewhere, it still has different *size* due to
3194 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3195 * resulting in struct ifreq being 32 and 40 bytes respectively).
3196 * As a result, if the struct happens to be at the end of a page and
3197 * the next page isn't readable/writable, we get a fault. To prevent
3198 * that, copy back and forth to the full size.
3201 uifr = compat_alloc_user_space(sizeof(*uifr));
3202 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3205 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3216 case SIOCGIFBRDADDR:
3217 case SIOCGIFDSTADDR:
3218 case SIOCGIFNETMASK:
3224 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3232 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3233 struct compat_ifreq __user *uifr32)
3236 struct compat_ifmap __user *uifmap32;
3239 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3240 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3241 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3242 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3243 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3244 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3245 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3246 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3250 err = dev_ioctl(net, cmd, &ifr, NULL);
3252 if (cmd == SIOCGIFMAP && !err) {
3253 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3254 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3255 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3256 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3257 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3258 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3259 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3266 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3267 * for some operations; this forces use of the newer bridge-utils that
3268 * use compatible ioctls
3270 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3274 if (get_user(tmp, argp))
3276 if (tmp == BRCTL_GET_VERSION)
3277 return BRCTL_VERSION + 1;
3281 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3282 unsigned int cmd, unsigned long arg)
3284 void __user *argp = compat_ptr(arg);
3285 struct sock *sk = sock->sk;
3286 struct net *net = sock_net(sk);
3288 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3289 return compat_ifr_data_ioctl(net, cmd, argp);
3294 return old_bridge_ioctl(argp);
3296 return compat_dev_ifconf(net, argp);
3298 return compat_siocwandev(net, argp);
3301 return compat_sioc_ifmap(net, cmd, argp);
3302 case SIOCGSTAMP_OLD:
3303 case SIOCGSTAMPNS_OLD:
3304 if (!sock->ops->gettstamp)
3305 return -ENOIOCTLCMD;
3306 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3307 !COMPAT_USE_64BIT_TIME);
3310 case SIOCBONDSLAVEINFOQUERY:
3311 case SIOCBONDINFOQUERY:
3314 return compat_ifr_data_ioctl(net, cmd, argp);
3327 case SIOCGSTAMP_NEW:
3328 case SIOCGSTAMPNS_NEW:
3329 return sock_ioctl(file, cmd, arg);
3346 case SIOCSIFHWBROADCAST:
3348 case SIOCGIFBRDADDR:
3349 case SIOCSIFBRDADDR:
3350 case SIOCGIFDSTADDR:
3351 case SIOCSIFDSTADDR:
3352 case SIOCGIFNETMASK:
3353 case SIOCSIFNETMASK:
3365 case SIOCBONDENSLAVE:
3366 case SIOCBONDRELEASE:
3367 case SIOCBONDSETHWADDR:
3368 case SIOCBONDCHANGEACTIVE:
3369 return compat_ifreq_ioctl(net, sock, cmd, argp);
3377 return sock_do_ioctl(net, sock, cmd, arg);
3380 return -ENOIOCTLCMD;
3383 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3386 struct socket *sock = file->private_data;
3387 int ret = -ENOIOCTLCMD;
3394 if (sock->ops->compat_ioctl)
3395 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3397 if (ret == -ENOIOCTLCMD &&
3398 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3399 ret = compat_wext_handle_ioctl(net, cmd, arg);
3401 if (ret == -ENOIOCTLCMD)
3402 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3409 * kernel_bind - bind an address to a socket (kernel space)
3412 * @addrlen: length of address
3414 * Returns 0 or an error.
3417 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3419 struct sockaddr_storage address;
3421 memcpy(&address, addr, addrlen);
3423 return sock->ops->bind(sock, (struct sockaddr *)&address, addrlen);
3425 EXPORT_SYMBOL(kernel_bind);
3428 * kernel_listen - move socket to listening state (kernel space)
3430 * @backlog: pending connections queue size
3432 * Returns 0 or an error.
3435 int kernel_listen(struct socket *sock, int backlog)
3437 return sock->ops->listen(sock, backlog);
3439 EXPORT_SYMBOL(kernel_listen);
3442 * kernel_accept - accept a connection (kernel space)
3443 * @sock: listening socket
3444 * @newsock: new connected socket
3447 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3448 * If it fails, @newsock is guaranteed to be %NULL.
3449 * Returns 0 or an error.
3452 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3454 struct sock *sk = sock->sk;
3457 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3462 err = sock->ops->accept(sock, *newsock, flags, true);
3464 sock_release(*newsock);
3469 (*newsock)->ops = sock->ops;
3470 __module_get((*newsock)->ops->owner);
3475 EXPORT_SYMBOL(kernel_accept);
3478 * kernel_connect - connect a socket (kernel space)
3481 * @addrlen: address length
3482 * @flags: flags (O_NONBLOCK, ...)
3484 * For datagram sockets, @addr is the addres to which datagrams are sent
3485 * by default, and the only address from which datagrams are received.
3486 * For stream sockets, attempts to connect to @addr.
3487 * Returns 0 or an error code.
3490 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3493 struct sockaddr_storage address;
3495 memcpy(&address, addr, addrlen);
3497 return sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, flags);
3499 EXPORT_SYMBOL(kernel_connect);
3502 * kernel_getsockname - get the address which the socket is bound (kernel space)
3504 * @addr: address holder
3506 * Fills the @addr pointer with the address which the socket is bound.
3507 * Returns 0 or an error code.
3510 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3512 return sock->ops->getname(sock, addr, 0);
3514 EXPORT_SYMBOL(kernel_getsockname);
3517 * kernel_getpeername - get the address which the socket is connected (kernel space)
3519 * @addr: address holder
3521 * Fills the @addr pointer with the address which the socket is connected.
3522 * Returns 0 or an error code.
3525 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3527 return sock->ops->getname(sock, addr, 1);
3529 EXPORT_SYMBOL(kernel_getpeername);
3532 * kernel_sendpage - send a &page through a socket (kernel space)
3535 * @offset: page offset
3536 * @size: total size in bytes
3537 * @flags: flags (MSG_DONTWAIT, ...)
3539 * Returns the total amount sent in bytes or an error.
3542 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3543 size_t size, int flags)
3545 if (sock->ops->sendpage) {
3546 /* Warn in case the improper page to zero-copy send */
3547 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3548 return sock->ops->sendpage(sock, page, offset, size, flags);
3550 return sock_no_sendpage(sock, page, offset, size, flags);
3552 EXPORT_SYMBOL(kernel_sendpage);
3555 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3558 * @offset: page offset
3559 * @size: total size in bytes
3560 * @flags: flags (MSG_DONTWAIT, ...)
3562 * Returns the total amount sent in bytes or an error.
3563 * Caller must hold @sk.
3566 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3567 size_t size, int flags)
3569 struct socket *sock = sk->sk_socket;
3571 if (sock->ops->sendpage_locked)
3572 return sock->ops->sendpage_locked(sk, page, offset, size,
3575 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3577 EXPORT_SYMBOL(kernel_sendpage_locked);
3580 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3582 * @how: connection part
3584 * Returns 0 or an error.
3587 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3589 return sock->ops->shutdown(sock, how);
3591 EXPORT_SYMBOL(kernel_sock_shutdown);
3594 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3597 * This routine returns the IP overhead imposed by a socket i.e.
3598 * the length of the underlying IP header, depending on whether
3599 * this is an IPv4 or IPv6 socket and the length from IP options turned
3600 * on at the socket. Assumes that the caller has a lock on the socket.
3603 u32 kernel_sock_ip_overhead(struct sock *sk)
3605 struct inet_sock *inet;
3606 struct ip_options_rcu *opt;
3608 #if IS_ENABLED(CONFIG_IPV6)
3609 struct ipv6_pinfo *np;
3610 struct ipv6_txoptions *optv6 = NULL;
3611 #endif /* IS_ENABLED(CONFIG_IPV6) */
3616 switch (sk->sk_family) {
3619 overhead += sizeof(struct iphdr);
3620 opt = rcu_dereference_protected(inet->inet_opt,
3621 sock_owned_by_user(sk));
3623 overhead += opt->opt.optlen;
3625 #if IS_ENABLED(CONFIG_IPV6)
3628 overhead += sizeof(struct ipv6hdr);
3630 optv6 = rcu_dereference_protected(np->opt,
3631 sock_owned_by_user(sk));
3633 overhead += (optv6->opt_flen + optv6->opt_nflen);
3635 #endif /* IS_ENABLED(CONFIG_IPV6) */
3636 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3640 EXPORT_SYMBOL(kernel_sock_ip_overhead);