1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/ethtool.h>
57 #include <linux/socket.h>
58 #include <linux/file.h>
59 #include <linux/net.h>
60 #include <linux/interrupt.h>
61 #include <linux/thread_info.h>
62 #include <linux/rcupdate.h>
63 #include <linux/netdevice.h>
64 #include <linux/proc_fs.h>
65 #include <linux/seq_file.h>
66 #include <linux/mutex.h>
67 #include <linux/if_bridge.h>
68 #include <linux/if_vlan.h>
69 #include <linux/ptp_classify.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74 #include <linux/highmem.h>
75 #include <linux/mount.h>
76 #include <linux/pseudo_fs.h>
77 #include <linux/security.h>
78 #include <linux/syscalls.h>
79 #include <linux/compat.h>
80 #include <linux/kmod.h>
81 #include <linux/audit.h>
82 #include <linux/wireless.h>
83 #include <linux/nsproxy.h>
84 #include <linux/magic.h>
85 #include <linux/slab.h>
86 #include <linux/xattr.h>
87 #include <linux/nospec.h>
88 #include <linux/indirect_call_wrapper.h>
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
95 #include <net/cls_cgroup.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/termios.h>
104 #include <linux/sockios.h>
105 #include <net/busy_poll.h>
106 #include <linux/errqueue.h>
108 #ifdef CONFIG_NET_RX_BUSY_POLL
109 unsigned int sysctl_net_busy_read __read_mostly;
110 unsigned int sysctl_net_busy_poll __read_mostly;
113 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
115 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
117 static int sock_close(struct inode *inode, struct file *file);
118 static __poll_t sock_poll(struct file *file,
119 struct poll_table_struct *wait);
120 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
122 static long compat_sock_ioctl(struct file *file,
123 unsigned int cmd, unsigned long arg);
125 static int sock_fasync(int fd, struct file *filp, int on);
126 static ssize_t sock_sendpage(struct file *file, struct page *page,
127 int offset, size_t size, loff_t *ppos, int more);
128 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
129 struct pipe_inode_info *pipe, size_t len,
132 #ifdef CONFIG_PROC_FS
133 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135 struct socket *sock = f->private_data;
137 if (sock->ops->show_fdinfo)
138 sock->ops->show_fdinfo(m, sock);
141 #define sock_show_fdinfo NULL
145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 * in the operation structures but are done directly via the socketcall() multiplexor.
149 static const struct file_operations socket_file_ops = {
150 .owner = THIS_MODULE,
152 .read_iter = sock_read_iter,
153 .write_iter = sock_write_iter,
155 .unlocked_ioctl = sock_ioctl,
157 .compat_ioctl = compat_sock_ioctl,
160 .release = sock_close,
161 .fasync = sock_fasync,
162 .sendpage = sock_sendpage,
163 .splice_write = generic_splice_sendpage,
164 .splice_read = sock_splice_read,
165 .show_fdinfo = sock_show_fdinfo,
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 user_namespace *mnt_userns,
338 struct dentry *dentry, struct inode *inode,
339 const char *suffix, const void *value,
340 size_t size, int flags)
342 /* Handled by LSM. */
346 static const struct xattr_handler sockfs_security_xattr_handler = {
347 .prefix = XATTR_SECURITY_PREFIX,
348 .set = sockfs_security_xattr_set,
351 static const struct xattr_handler *sockfs_xattr_handlers[] = {
352 &sockfs_xattr_handler,
353 &sockfs_security_xattr_handler,
357 static int sockfs_init_fs_context(struct fs_context *fc)
359 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
362 ctx->ops = &sockfs_ops;
363 ctx->dops = &sockfs_dentry_operations;
364 ctx->xattr = sockfs_xattr_handlers;
368 static struct vfsmount *sock_mnt __read_mostly;
370 static struct file_system_type sock_fs_type = {
372 .init_fs_context = sockfs_init_fs_context,
373 .kill_sb = kill_anon_super,
377 * Obtains the first available file descriptor and sets it up for use.
379 * These functions create file structures and maps them to fd space
380 * of the current process. On success it returns file descriptor
381 * and file struct implicitly stored in sock->file.
382 * Note that another thread may close file descriptor before we return
383 * from this function. We use the fact that now we do not refer
384 * to socket after mapping. If one day we will need it, this
385 * function will increment ref. count on file by 1.
387 * In any case returned fd MAY BE not valid!
388 * This race condition is unavoidable
389 * with shared fd spaces, we cannot solve it inside kernel,
390 * but we take care of internal coherence yet.
394 * sock_alloc_file - Bind a &socket to a &file
396 * @flags: file status flags
397 * @dname: protocol name
399 * Returns the &file bound with @sock, implicitly storing it
400 * in sock->file. If dname is %NULL, sets to "".
401 * On failure the return is a ERR pointer (see linux/err.h).
402 * This function uses GFP_KERNEL internally.
405 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
410 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
412 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
413 O_RDWR | (flags & O_NONBLOCK),
421 file->private_data = sock;
422 stream_open(SOCK_INODE(sock), file);
425 EXPORT_SYMBOL(sock_alloc_file);
427 static int sock_map_fd(struct socket *sock, int flags)
429 struct file *newfile;
430 int fd = get_unused_fd_flags(flags);
431 if (unlikely(fd < 0)) {
436 newfile = sock_alloc_file(sock, flags, NULL);
437 if (!IS_ERR(newfile)) {
438 fd_install(fd, newfile);
443 return PTR_ERR(newfile);
447 * sock_from_file - Return the &socket bounded to @file.
450 * On failure returns %NULL.
453 struct socket *sock_from_file(struct file *file)
455 if (file->f_op == &socket_file_ops)
456 return file->private_data; /* set in sock_map_fd */
460 EXPORT_SYMBOL(sock_from_file);
463 * sockfd_lookup - Go from a file number to its socket slot
465 * @err: pointer to an error code return
467 * The file handle passed in is locked and the socket it is bound
468 * to is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
472 * On a success the socket object pointer is returned.
475 struct socket *sockfd_lookup(int fd, int *err)
486 sock = sock_from_file(file);
493 EXPORT_SYMBOL(sockfd_lookup);
495 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
497 struct fd f = fdget(fd);
502 sock = sock_from_file(f.file);
504 *fput_needed = f.flags & FDPUT_FPUT;
513 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
519 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
529 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
534 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
541 static int sockfs_setattr(struct user_namespace *mnt_userns,
542 struct dentry *dentry, struct iattr *iattr)
544 int err = simple_setattr(&init_user_ns, dentry, iattr);
546 if (!err && (iattr->ia_valid & ATTR_UID)) {
547 struct socket *sock = SOCKET_I(d_inode(dentry));
550 sock->sk->sk_uid = iattr->ia_uid;
558 static const struct inode_operations sockfs_inode_ops = {
559 .listxattr = sockfs_listxattr,
560 .setattr = sockfs_setattr,
564 * sock_alloc - allocate a socket
566 * Allocate a new inode and socket object. The two are bound together
567 * and initialised. The socket is then returned. If we are out of inodes
568 * NULL is returned. This functions uses GFP_KERNEL internally.
571 struct socket *sock_alloc(void)
576 inode = new_inode_pseudo(sock_mnt->mnt_sb);
580 sock = SOCKET_I(inode);
582 inode->i_ino = get_next_ino();
583 inode->i_mode = S_IFSOCK | S_IRWXUGO;
584 inode->i_uid = current_fsuid();
585 inode->i_gid = current_fsgid();
586 inode->i_op = &sockfs_inode_ops;
590 EXPORT_SYMBOL(sock_alloc);
592 static void __sock_release(struct socket *sock, struct inode *inode)
595 struct module *owner = sock->ops->owner;
599 sock->ops->release(sock);
607 if (sock->wq.fasync_list)
608 pr_err("%s: fasync list not empty!\n", __func__);
611 iput(SOCK_INODE(sock));
618 * sock_release - close a socket
619 * @sock: socket to close
621 * The socket is released from the protocol stack if it has a release
622 * callback, and the inode is then released if the socket is bound to
623 * an inode not a file.
625 void sock_release(struct socket *sock)
627 __sock_release(sock, NULL);
629 EXPORT_SYMBOL(sock_release);
631 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
633 u8 flags = *tx_flags;
635 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
636 flags |= SKBTX_HW_TSTAMP;
638 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
639 flags |= SKBTX_SW_TSTAMP;
641 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
642 flags |= SKBTX_SCHED_TSTAMP;
646 EXPORT_SYMBOL(__sock_tx_timestamp);
648 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
650 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
652 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
654 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
655 inet_sendmsg, sock, msg,
657 BUG_ON(ret == -EIOCBQUEUED);
662 * sock_sendmsg - send a message through @sock
664 * @msg: message to send
666 * Sends @msg through @sock, passing through LSM.
667 * Returns the number of bytes sent, or an error code.
669 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
671 int err = security_socket_sendmsg(sock, msg,
674 return err ?: sock_sendmsg_nosec(sock, msg);
676 EXPORT_SYMBOL(sock_sendmsg);
679 * kernel_sendmsg - send a message through @sock (kernel-space)
681 * @msg: message header
683 * @num: vec array length
684 * @size: total message data size
686 * Builds the message data with @vec and sends it through @sock.
687 * Returns the number of bytes sent, or an error code.
690 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
691 struct kvec *vec, size_t num, size_t size)
693 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
694 return sock_sendmsg(sock, msg);
696 EXPORT_SYMBOL(kernel_sendmsg);
699 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
701 * @msg: message header
702 * @vec: output s/g array
703 * @num: output s/g array length
704 * @size: total message data size
706 * Builds the message data with @vec and sends it through @sock.
707 * Returns the number of bytes sent, or an error code.
708 * Caller must hold @sk.
711 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
712 struct kvec *vec, size_t num, size_t size)
714 struct socket *sock = sk->sk_socket;
716 if (!sock->ops->sendmsg_locked)
717 return sock_no_sendmsg_locked(sk, msg, size);
719 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
721 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
723 EXPORT_SYMBOL(kernel_sendmsg_locked);
725 static bool skb_is_err_queue(const struct sk_buff *skb)
727 /* pkt_type of skbs enqueued on the error queue are set to
728 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
729 * in recvmsg, since skbs received on a local socket will never
730 * have a pkt_type of PACKET_OUTGOING.
732 return skb->pkt_type == PACKET_OUTGOING;
735 /* On transmit, software and hardware timestamps are returned independently.
736 * As the two skb clones share the hardware timestamp, which may be updated
737 * before the software timestamp is received, a hardware TX timestamp may be
738 * returned only if there is no software TX timestamp. Ignore false software
739 * timestamps, which may be made in the __sock_recv_timestamp() call when the
740 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
741 * hardware timestamp.
743 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
745 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
748 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
750 struct scm_ts_pktinfo ts_pktinfo;
751 struct net_device *orig_dev;
753 if (!skb_mac_header_was_set(skb))
756 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
759 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
761 ts_pktinfo.if_index = orig_dev->ifindex;
764 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
765 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
766 sizeof(ts_pktinfo), &ts_pktinfo);
770 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
772 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
775 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
776 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
777 struct scm_timestamping_internal tss;
779 int empty = 1, false_tstamp = 0;
780 struct skb_shared_hwtstamps *shhwtstamps =
783 /* Race occurred between timestamp enabling and packet
784 receiving. Fill in the current time for now. */
785 if (need_software_tstamp && skb->tstamp == 0) {
786 __net_timestamp(skb);
790 if (need_software_tstamp) {
791 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
793 struct __kernel_sock_timeval tv;
795 skb_get_new_timestamp(skb, &tv);
796 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
799 struct __kernel_old_timeval tv;
801 skb_get_timestamp(skb, &tv);
802 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
807 struct __kernel_timespec ts;
809 skb_get_new_timestampns(skb, &ts);
810 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
813 struct __kernel_old_timespec ts;
815 skb_get_timestampns(skb, &ts);
816 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
822 memset(&tss, 0, sizeof(tss));
823 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
824 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
827 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
828 !skb_is_swtx_tstamp(skb, false_tstamp) &&
829 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
831 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
832 !skb_is_err_queue(skb))
833 put_ts_pktinfo(msg, skb);
836 if (sock_flag(sk, SOCK_TSTAMP_NEW))
837 put_cmsg_scm_timestamping64(msg, &tss);
839 put_cmsg_scm_timestamping(msg, &tss);
841 if (skb_is_err_queue(skb) && skb->len &&
842 SKB_EXT_ERR(skb)->opt_stats)
843 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
844 skb->len, skb->data);
847 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
849 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
854 if (!sock_flag(sk, SOCK_WIFI_STATUS))
856 if (!skb->wifi_acked_valid)
859 ack = skb->wifi_acked;
861 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
863 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
865 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
868 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
869 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
870 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
873 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
876 sock_recv_timestamp(msg, sk, skb);
877 sock_recv_drops(msg, sk, skb);
879 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
881 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
883 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
885 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
888 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
889 inet_recvmsg, sock, msg, msg_data_left(msg),
894 * sock_recvmsg - receive a message from @sock
896 * @msg: message to receive
897 * @flags: message flags
899 * Receives @msg from @sock, passing through LSM. Returns the total number
900 * of bytes received, or an error.
902 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
904 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
906 return err ?: sock_recvmsg_nosec(sock, msg, flags);
908 EXPORT_SYMBOL(sock_recvmsg);
911 * kernel_recvmsg - Receive a message from a socket (kernel space)
912 * @sock: The socket to receive the message from
913 * @msg: Received message
914 * @vec: Input s/g array for message data
915 * @num: Size of input s/g array
916 * @size: Number of bytes to read
917 * @flags: Message flags (MSG_DONTWAIT, etc...)
919 * On return the msg structure contains the scatter/gather array passed in the
920 * vec argument. The array is modified so that it consists of the unfilled
921 * portion of the original array.
923 * The returned value is the total number of bytes received, or an error.
926 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
927 struct kvec *vec, size_t num, size_t size, int flags)
929 msg->msg_control_is_user = false;
930 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
931 return sock_recvmsg(sock, msg, flags);
933 EXPORT_SYMBOL(kernel_recvmsg);
935 static ssize_t sock_sendpage(struct file *file, struct page *page,
936 int offset, size_t size, loff_t *ppos, int more)
941 sock = file->private_data;
943 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
944 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
947 return kernel_sendpage(sock, page, offset, size, flags);
950 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
951 struct pipe_inode_info *pipe, size_t len,
954 struct socket *sock = file->private_data;
956 if (unlikely(!sock->ops->splice_read))
957 return generic_file_splice_read(file, ppos, pipe, len, flags);
959 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
962 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
964 struct file *file = iocb->ki_filp;
965 struct socket *sock = file->private_data;
966 struct msghdr msg = {.msg_iter = *to,
970 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
971 msg.msg_flags = MSG_DONTWAIT;
973 if (iocb->ki_pos != 0)
976 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
979 res = sock_recvmsg(sock, &msg, msg.msg_flags);
984 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
986 struct file *file = iocb->ki_filp;
987 struct socket *sock = file->private_data;
988 struct msghdr msg = {.msg_iter = *from,
992 if (iocb->ki_pos != 0)
995 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
996 msg.msg_flags = MSG_DONTWAIT;
998 if (sock->type == SOCK_SEQPACKET)
999 msg.msg_flags |= MSG_EOR;
1001 res = sock_sendmsg(sock, &msg);
1002 *from = msg.msg_iter;
1007 * Atomic setting of ioctl hooks to avoid race
1008 * with module unload.
1011 static DEFINE_MUTEX(br_ioctl_mutex);
1012 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1014 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1016 mutex_lock(&br_ioctl_mutex);
1017 br_ioctl_hook = hook;
1018 mutex_unlock(&br_ioctl_mutex);
1020 EXPORT_SYMBOL(brioctl_set);
1022 static DEFINE_MUTEX(vlan_ioctl_mutex);
1023 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1025 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1027 mutex_lock(&vlan_ioctl_mutex);
1028 vlan_ioctl_hook = hook;
1029 mutex_unlock(&vlan_ioctl_mutex);
1031 EXPORT_SYMBOL(vlan_ioctl_set);
1033 static long sock_do_ioctl(struct net *net, struct socket *sock,
1034 unsigned int cmd, unsigned long arg)
1037 void __user *argp = (void __user *)arg;
1039 err = sock->ops->ioctl(sock, cmd, arg);
1042 * If this ioctl is unknown try to hand it down
1043 * to the NIC driver.
1045 if (err != -ENOIOCTLCMD)
1048 if (cmd == SIOCGIFCONF) {
1050 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1053 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1055 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1060 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1062 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1063 if (!err && need_copyout)
1064 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1071 * With an ioctl, arg may well be a user mode pointer, but we don't know
1072 * what to do with it - that's up to the protocol still.
1075 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1077 struct socket *sock;
1079 void __user *argp = (void __user *)arg;
1083 sock = file->private_data;
1086 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1089 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1091 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1092 if (!err && need_copyout)
1093 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1096 #ifdef CONFIG_WEXT_CORE
1097 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1098 err = wext_handle_ioctl(net, cmd, argp);
1105 if (get_user(pid, (int __user *)argp))
1107 err = f_setown(sock->file, pid, 1);
1111 err = put_user(f_getown(sock->file),
1112 (int __user *)argp);
1120 request_module("bridge");
1122 mutex_lock(&br_ioctl_mutex);
1124 err = br_ioctl_hook(net, cmd, argp);
1125 mutex_unlock(&br_ioctl_mutex);
1130 if (!vlan_ioctl_hook)
1131 request_module("8021q");
1133 mutex_lock(&vlan_ioctl_mutex);
1134 if (vlan_ioctl_hook)
1135 err = vlan_ioctl_hook(net, argp);
1136 mutex_unlock(&vlan_ioctl_mutex);
1140 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1143 err = open_related_ns(&net->ns, get_net_ns);
1145 case SIOCGSTAMP_OLD:
1146 case SIOCGSTAMPNS_OLD:
1147 if (!sock->ops->gettstamp) {
1151 err = sock->ops->gettstamp(sock, argp,
1152 cmd == SIOCGSTAMP_OLD,
1153 !IS_ENABLED(CONFIG_64BIT));
1155 case SIOCGSTAMP_NEW:
1156 case SIOCGSTAMPNS_NEW:
1157 if (!sock->ops->gettstamp) {
1161 err = sock->ops->gettstamp(sock, argp,
1162 cmd == SIOCGSTAMP_NEW,
1166 err = sock_do_ioctl(net, sock, cmd, arg);
1173 * sock_create_lite - creates a socket
1174 * @family: protocol family (AF_INET, ...)
1175 * @type: communication type (SOCK_STREAM, ...)
1176 * @protocol: protocol (0, ...)
1179 * Creates a new socket and assigns it to @res, passing through LSM.
1180 * The new socket initialization is not complete, see kernel_accept().
1181 * Returns 0 or an error. On failure @res is set to %NULL.
1182 * This function internally uses GFP_KERNEL.
1185 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1188 struct socket *sock = NULL;
1190 err = security_socket_create(family, type, protocol, 1);
1194 sock = sock_alloc();
1201 err = security_socket_post_create(sock, family, type, protocol, 1);
1213 EXPORT_SYMBOL(sock_create_lite);
1215 /* No kernel lock held - perfect */
1216 static __poll_t sock_poll(struct file *file, poll_table *wait)
1218 struct socket *sock = file->private_data;
1219 __poll_t events = poll_requested_events(wait), flag = 0;
1221 if (!sock->ops->poll)
1224 if (sk_can_busy_loop(sock->sk)) {
1225 /* poll once if requested by the syscall */
1226 if (events & POLL_BUSY_LOOP)
1227 sk_busy_loop(sock->sk, 1);
1229 /* if this socket can poll_ll, tell the system call */
1230 flag = POLL_BUSY_LOOP;
1233 return sock->ops->poll(file, sock, wait) | flag;
1236 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1238 struct socket *sock = file->private_data;
1240 return sock->ops->mmap(file, sock, vma);
1243 static int sock_close(struct inode *inode, struct file *filp)
1245 __sock_release(SOCKET_I(inode), inode);
1250 * Update the socket async list
1252 * Fasync_list locking strategy.
1254 * 1. fasync_list is modified only under process context socket lock
1255 * i.e. under semaphore.
1256 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1257 * or under socket lock
1260 static int sock_fasync(int fd, struct file *filp, int on)
1262 struct socket *sock = filp->private_data;
1263 struct sock *sk = sock->sk;
1264 struct socket_wq *wq = &sock->wq;
1270 fasync_helper(fd, filp, on, &wq->fasync_list);
1272 if (!wq->fasync_list)
1273 sock_reset_flag(sk, SOCK_FASYNC);
1275 sock_set_flag(sk, SOCK_FASYNC);
1281 /* This function may be called only under rcu_lock */
1283 int sock_wake_async(struct socket_wq *wq, int how, int band)
1285 if (!wq || !wq->fasync_list)
1289 case SOCK_WAKE_WAITD:
1290 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1293 case SOCK_WAKE_SPACE:
1294 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1299 kill_fasync(&wq->fasync_list, SIGIO, band);
1302 kill_fasync(&wq->fasync_list, SIGURG, band);
1307 EXPORT_SYMBOL(sock_wake_async);
1310 * __sock_create - creates a socket
1311 * @net: net namespace
1312 * @family: protocol family (AF_INET, ...)
1313 * @type: communication type (SOCK_STREAM, ...)
1314 * @protocol: protocol (0, ...)
1316 * @kern: boolean for kernel space sockets
1318 * Creates a new socket and assigns it to @res, passing through LSM.
1319 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1320 * be set to true if the socket resides in kernel space.
1321 * This function internally uses GFP_KERNEL.
1324 int __sock_create(struct net *net, int family, int type, int protocol,
1325 struct socket **res, int kern)
1328 struct socket *sock;
1329 const struct net_proto_family *pf;
1332 * Check protocol is in range
1334 if (family < 0 || family >= NPROTO)
1335 return -EAFNOSUPPORT;
1336 if (type < 0 || type >= SOCK_MAX)
1341 This uglymoron is moved from INET layer to here to avoid
1342 deadlock in module load.
1344 if (family == PF_INET && type == SOCK_PACKET) {
1345 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1350 err = security_socket_create(family, type, protocol, kern);
1355 * Allocate the socket and allow the family to set things up. if
1356 * the protocol is 0, the family is instructed to select an appropriate
1359 sock = sock_alloc();
1361 net_warn_ratelimited("socket: no more sockets\n");
1362 return -ENFILE; /* Not exactly a match, but its the
1363 closest posix thing */
1368 #ifdef CONFIG_MODULES
1369 /* Attempt to load a protocol module if the find failed.
1371 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1372 * requested real, full-featured networking support upon configuration.
1373 * Otherwise module support will break!
1375 if (rcu_access_pointer(net_families[family]) == NULL)
1376 request_module("net-pf-%d", family);
1380 pf = rcu_dereference(net_families[family]);
1381 err = -EAFNOSUPPORT;
1386 * We will call the ->create function, that possibly is in a loadable
1387 * module, so we have to bump that loadable module refcnt first.
1389 if (!try_module_get(pf->owner))
1392 /* Now protected by module ref count */
1395 err = pf->create(net, sock, protocol, kern);
1397 goto out_module_put;
1400 * Now to bump the refcnt of the [loadable] module that owns this
1401 * socket at sock_release time we decrement its refcnt.
1403 if (!try_module_get(sock->ops->owner))
1404 goto out_module_busy;
1407 * Now that we're done with the ->create function, the [loadable]
1408 * module can have its refcnt decremented
1410 module_put(pf->owner);
1411 err = security_socket_post_create(sock, family, type, protocol, kern);
1413 goto out_sock_release;
1419 err = -EAFNOSUPPORT;
1422 module_put(pf->owner);
1429 goto out_sock_release;
1431 EXPORT_SYMBOL(__sock_create);
1434 * sock_create - creates a socket
1435 * @family: protocol family (AF_INET, ...)
1436 * @type: communication type (SOCK_STREAM, ...)
1437 * @protocol: protocol (0, ...)
1440 * A wrapper around __sock_create().
1441 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1444 int sock_create(int family, int type, int protocol, struct socket **res)
1446 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1448 EXPORT_SYMBOL(sock_create);
1451 * sock_create_kern - creates a socket (kernel space)
1452 * @net: net namespace
1453 * @family: protocol family (AF_INET, ...)
1454 * @type: communication type (SOCK_STREAM, ...)
1455 * @protocol: protocol (0, ...)
1458 * A wrapper around __sock_create().
1459 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1462 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1464 return __sock_create(net, family, type, protocol, res, 1);
1466 EXPORT_SYMBOL(sock_create_kern);
1468 int __sys_socket(int family, int type, int protocol)
1471 struct socket *sock;
1474 /* Check the SOCK_* constants for consistency. */
1475 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1476 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1477 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1478 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1480 flags = type & ~SOCK_TYPE_MASK;
1481 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1483 type &= SOCK_TYPE_MASK;
1485 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1486 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1488 retval = sock_create(family, type, protocol, &sock);
1492 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1495 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1497 return __sys_socket(family, type, protocol);
1501 * Create a pair of connected sockets.
1504 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1506 struct socket *sock1, *sock2;
1508 struct file *newfile1, *newfile2;
1511 flags = type & ~SOCK_TYPE_MASK;
1512 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1514 type &= SOCK_TYPE_MASK;
1516 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1517 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1520 * reserve descriptors and make sure we won't fail
1521 * to return them to userland.
1523 fd1 = get_unused_fd_flags(flags);
1524 if (unlikely(fd1 < 0))
1527 fd2 = get_unused_fd_flags(flags);
1528 if (unlikely(fd2 < 0)) {
1533 err = put_user(fd1, &usockvec[0]);
1537 err = put_user(fd2, &usockvec[1]);
1542 * Obtain the first socket and check if the underlying protocol
1543 * supports the socketpair call.
1546 err = sock_create(family, type, protocol, &sock1);
1547 if (unlikely(err < 0))
1550 err = sock_create(family, type, protocol, &sock2);
1551 if (unlikely(err < 0)) {
1552 sock_release(sock1);
1556 err = security_socket_socketpair(sock1, sock2);
1557 if (unlikely(err)) {
1558 sock_release(sock2);
1559 sock_release(sock1);
1563 err = sock1->ops->socketpair(sock1, sock2);
1564 if (unlikely(err < 0)) {
1565 sock_release(sock2);
1566 sock_release(sock1);
1570 newfile1 = sock_alloc_file(sock1, flags, NULL);
1571 if (IS_ERR(newfile1)) {
1572 err = PTR_ERR(newfile1);
1573 sock_release(sock2);
1577 newfile2 = sock_alloc_file(sock2, flags, NULL);
1578 if (IS_ERR(newfile2)) {
1579 err = PTR_ERR(newfile2);
1584 audit_fd_pair(fd1, fd2);
1586 fd_install(fd1, newfile1);
1587 fd_install(fd2, newfile2);
1596 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1597 int __user *, usockvec)
1599 return __sys_socketpair(family, type, protocol, usockvec);
1603 * Bind a name to a socket. Nothing much to do here since it's
1604 * the protocol's responsibility to handle the local address.
1606 * We move the socket address to kernel space before we call
1607 * the protocol layer (having also checked the address is ok).
1610 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1612 struct socket *sock;
1613 struct sockaddr_storage address;
1614 int err, fput_needed;
1616 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1618 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1620 err = security_socket_bind(sock,
1621 (struct sockaddr *)&address,
1624 err = sock->ops->bind(sock,
1628 fput_light(sock->file, fput_needed);
1633 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1635 return __sys_bind(fd, umyaddr, addrlen);
1639 * Perform a listen. Basically, we allow the protocol to do anything
1640 * necessary for a listen, and if that works, we mark the socket as
1641 * ready for listening.
1644 int __sys_listen(int fd, int backlog)
1646 struct socket *sock;
1647 int err, fput_needed;
1650 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1652 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1653 if ((unsigned int)backlog > somaxconn)
1654 backlog = somaxconn;
1656 err = security_socket_listen(sock, backlog);
1658 err = sock->ops->listen(sock, backlog);
1660 fput_light(sock->file, fput_needed);
1665 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1667 return __sys_listen(fd, backlog);
1670 int __sys_accept4_file(struct file *file, unsigned file_flags,
1671 struct sockaddr __user *upeer_sockaddr,
1672 int __user *upeer_addrlen, int flags,
1673 unsigned long nofile)
1675 struct socket *sock, *newsock;
1676 struct file *newfile;
1677 int err, len, newfd;
1678 struct sockaddr_storage address;
1680 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1683 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1684 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1686 sock = sock_from_file(file);
1693 newsock = sock_alloc();
1697 newsock->type = sock->type;
1698 newsock->ops = sock->ops;
1701 * We don't need try_module_get here, as the listening socket (sock)
1702 * has the protocol module (sock->ops->owner) held.
1704 __module_get(newsock->ops->owner);
1706 newfd = __get_unused_fd_flags(flags, nofile);
1707 if (unlikely(newfd < 0)) {
1709 sock_release(newsock);
1712 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1713 if (IS_ERR(newfile)) {
1714 err = PTR_ERR(newfile);
1715 put_unused_fd(newfd);
1719 err = security_socket_accept(sock, newsock);
1723 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1728 if (upeer_sockaddr) {
1729 len = newsock->ops->getname(newsock,
1730 (struct sockaddr *)&address, 2);
1732 err = -ECONNABORTED;
1735 err = move_addr_to_user(&address,
1736 len, upeer_sockaddr, upeer_addrlen);
1741 /* File flags are not inherited via accept() unlike another OSes. */
1743 fd_install(newfd, newfile);
1749 put_unused_fd(newfd);
1755 * For accept, we attempt to create a new socket, set up the link
1756 * with the client, wake up the client, then return the new
1757 * connected fd. We collect the address of the connector in kernel
1758 * space and move it to user at the very end. This is unclean because
1759 * we open the socket then return an error.
1761 * 1003.1g adds the ability to recvmsg() to query connection pending
1762 * status to recvmsg. We need to add that support in a way thats
1763 * clean when we restructure accept also.
1766 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1767 int __user *upeer_addrlen, int flags)
1774 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1775 upeer_addrlen, flags,
1776 rlimit(RLIMIT_NOFILE));
1783 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1784 int __user *, upeer_addrlen, int, flags)
1786 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1789 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1790 int __user *, upeer_addrlen)
1792 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1796 * Attempt to connect to a socket with the server address. The address
1797 * is in user space so we verify it is OK and move it to kernel space.
1799 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1802 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1803 * other SEQPACKET protocols that take time to connect() as it doesn't
1804 * include the -EINPROGRESS status for such sockets.
1807 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1808 int addrlen, int file_flags)
1810 struct socket *sock;
1813 sock = sock_from_file(file);
1820 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1824 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1825 sock->file->f_flags | file_flags);
1830 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1837 struct sockaddr_storage address;
1839 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1841 ret = __sys_connect_file(f.file, &address, addrlen, 0);
1848 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1851 return __sys_connect(fd, uservaddr, addrlen);
1855 * Get the local address ('name') of a socket object. Move the obtained
1856 * name to user space.
1859 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1860 int __user *usockaddr_len)
1862 struct socket *sock;
1863 struct sockaddr_storage address;
1864 int err, fput_needed;
1866 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1870 err = security_socket_getsockname(sock);
1874 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1877 /* "err" is actually length in this case */
1878 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1881 fput_light(sock->file, fput_needed);
1886 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1887 int __user *, usockaddr_len)
1889 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1893 * Get the remote address ('name') of a socket object. Move the obtained
1894 * name to user space.
1897 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1898 int __user *usockaddr_len)
1900 struct socket *sock;
1901 struct sockaddr_storage address;
1902 int err, fput_needed;
1904 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1906 err = security_socket_getpeername(sock);
1908 fput_light(sock->file, fput_needed);
1912 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1914 /* "err" is actually length in this case */
1915 err = move_addr_to_user(&address, err, usockaddr,
1917 fput_light(sock->file, fput_needed);
1922 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1923 int __user *, usockaddr_len)
1925 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1929 * Send a datagram to a given address. We move the address into kernel
1930 * space and check the user space data area is readable before invoking
1933 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1934 struct sockaddr __user *addr, int addr_len)
1936 struct socket *sock;
1937 struct sockaddr_storage address;
1943 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1946 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1950 msg.msg_name = NULL;
1951 msg.msg_control = NULL;
1952 msg.msg_controllen = 0;
1953 msg.msg_namelen = 0;
1955 err = move_addr_to_kernel(addr, addr_len, &address);
1958 msg.msg_name = (struct sockaddr *)&address;
1959 msg.msg_namelen = addr_len;
1961 if (sock->file->f_flags & O_NONBLOCK)
1962 flags |= MSG_DONTWAIT;
1963 msg.msg_flags = flags;
1964 err = sock_sendmsg(sock, &msg);
1967 fput_light(sock->file, fput_needed);
1972 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1973 unsigned int, flags, struct sockaddr __user *, addr,
1976 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1980 * Send a datagram down a socket.
1983 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1984 unsigned int, flags)
1986 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1990 * Receive a frame from the socket and optionally record the address of the
1991 * sender. We verify the buffers are writable and if needed move the
1992 * sender address from kernel to user space.
1994 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1995 struct sockaddr __user *addr, int __user *addr_len)
1997 struct socket *sock;
2000 struct sockaddr_storage address;
2004 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2007 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2011 msg.msg_control = NULL;
2012 msg.msg_controllen = 0;
2013 /* Save some cycles and don't copy the address if not needed */
2014 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2015 /* We assume all kernel code knows the size of sockaddr_storage */
2016 msg.msg_namelen = 0;
2017 msg.msg_iocb = NULL;
2019 if (sock->file->f_flags & O_NONBLOCK)
2020 flags |= MSG_DONTWAIT;
2021 err = sock_recvmsg(sock, &msg, flags);
2023 if (err >= 0 && addr != NULL) {
2024 err2 = move_addr_to_user(&address,
2025 msg.msg_namelen, addr, addr_len);
2030 fput_light(sock->file, fput_needed);
2035 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2036 unsigned int, flags, struct sockaddr __user *, addr,
2037 int __user *, addr_len)
2039 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2043 * Receive a datagram from a socket.
2046 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2047 unsigned int, flags)
2049 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2052 static bool sock_use_custom_sol_socket(const struct socket *sock)
2054 const struct sock *sk = sock->sk;
2056 /* Use sock->ops->setsockopt() for MPTCP */
2057 return IS_ENABLED(CONFIG_MPTCP) &&
2058 sk->sk_protocol == IPPROTO_MPTCP &&
2059 sk->sk_type == SOCK_STREAM &&
2060 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2064 * Set a socket option. Because we don't know the option lengths we have
2065 * to pass the user mode parameter for the protocols to sort out.
2067 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2070 sockptr_t optval = USER_SOCKPTR(user_optval);
2071 char *kernel_optval = NULL;
2072 int err, fput_needed;
2073 struct socket *sock;
2078 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2082 err = security_socket_setsockopt(sock, level, optname);
2086 if (!in_compat_syscall())
2087 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2088 user_optval, &optlen,
2098 optval = KERNEL_SOCKPTR(kernel_optval);
2099 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2100 err = sock_setsockopt(sock, level, optname, optval, optlen);
2101 else if (unlikely(!sock->ops->setsockopt))
2104 err = sock->ops->setsockopt(sock, level, optname, optval,
2106 kfree(kernel_optval);
2108 fput_light(sock->file, fput_needed);
2112 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2113 char __user *, optval, int, optlen)
2115 return __sys_setsockopt(fd, level, optname, optval, optlen);
2118 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2122 * Get a socket option. Because we don't know the option lengths we have
2123 * to pass a user mode parameter for the protocols to sort out.
2125 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2128 int err, fput_needed;
2129 struct socket *sock;
2132 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2136 err = security_socket_getsockopt(sock, level, optname);
2140 if (!in_compat_syscall())
2141 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2143 if (level == SOL_SOCKET)
2144 err = sock_getsockopt(sock, level, optname, optval, optlen);
2145 else if (unlikely(!sock->ops->getsockopt))
2148 err = sock->ops->getsockopt(sock, level, optname, optval,
2151 if (!in_compat_syscall())
2152 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2153 optval, optlen, max_optlen,
2156 fput_light(sock->file, fput_needed);
2160 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2161 char __user *, optval, int __user *, optlen)
2163 return __sys_getsockopt(fd, level, optname, optval, optlen);
2167 * Shutdown a socket.
2170 int __sys_shutdown_sock(struct socket *sock, int how)
2174 err = security_socket_shutdown(sock, how);
2176 err = sock->ops->shutdown(sock, how);
2181 int __sys_shutdown(int fd, int how)
2183 int err, fput_needed;
2184 struct socket *sock;
2186 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2188 err = __sys_shutdown_sock(sock, how);
2189 fput_light(sock->file, fput_needed);
2194 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2196 return __sys_shutdown(fd, how);
2199 /* A couple of helpful macros for getting the address of the 32/64 bit
2200 * fields which are the same type (int / unsigned) on our platforms.
2202 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2203 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2204 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2206 struct used_address {
2207 struct sockaddr_storage name;
2208 unsigned int name_len;
2211 int __copy_msghdr_from_user(struct msghdr *kmsg,
2212 struct user_msghdr __user *umsg,
2213 struct sockaddr __user **save_addr,
2214 struct iovec __user **uiov, size_t *nsegs)
2216 struct user_msghdr msg;
2219 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2222 kmsg->msg_control_is_user = true;
2223 kmsg->msg_control_user = msg.msg_control;
2224 kmsg->msg_controllen = msg.msg_controllen;
2225 kmsg->msg_flags = msg.msg_flags;
2227 kmsg->msg_namelen = msg.msg_namelen;
2229 kmsg->msg_namelen = 0;
2231 if (kmsg->msg_namelen < 0)
2234 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2235 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2238 *save_addr = msg.msg_name;
2240 if (msg.msg_name && kmsg->msg_namelen) {
2242 err = move_addr_to_kernel(msg.msg_name,
2249 kmsg->msg_name = NULL;
2250 kmsg->msg_namelen = 0;
2253 if (msg.msg_iovlen > UIO_MAXIOV)
2256 kmsg->msg_iocb = NULL;
2257 *uiov = msg.msg_iov;
2258 *nsegs = msg.msg_iovlen;
2262 static int copy_msghdr_from_user(struct msghdr *kmsg,
2263 struct user_msghdr __user *umsg,
2264 struct sockaddr __user **save_addr,
2267 struct user_msghdr msg;
2270 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2275 err = import_iovec(save_addr ? READ : WRITE,
2276 msg.msg_iov, msg.msg_iovlen,
2277 UIO_FASTIOV, iov, &kmsg->msg_iter);
2278 return err < 0 ? err : 0;
2281 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2282 unsigned int flags, struct used_address *used_address,
2283 unsigned int allowed_msghdr_flags)
2285 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2286 __aligned(sizeof(__kernel_size_t));
2287 /* 20 is size of ipv6_pktinfo */
2288 unsigned char *ctl_buf = ctl;
2294 if (msg_sys->msg_controllen > INT_MAX)
2296 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2297 ctl_len = msg_sys->msg_controllen;
2298 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2300 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2304 ctl_buf = msg_sys->msg_control;
2305 ctl_len = msg_sys->msg_controllen;
2306 } else if (ctl_len) {
2307 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2308 CMSG_ALIGN(sizeof(struct cmsghdr)));
2309 if (ctl_len > sizeof(ctl)) {
2310 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2311 if (ctl_buf == NULL)
2315 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2317 msg_sys->msg_control = ctl_buf;
2318 msg_sys->msg_control_is_user = false;
2320 msg_sys->msg_flags = flags;
2322 if (sock->file->f_flags & O_NONBLOCK)
2323 msg_sys->msg_flags |= MSG_DONTWAIT;
2325 * If this is sendmmsg() and current destination address is same as
2326 * previously succeeded address, omit asking LSM's decision.
2327 * used_address->name_len is initialized to UINT_MAX so that the first
2328 * destination address never matches.
2330 if (used_address && msg_sys->msg_name &&
2331 used_address->name_len == msg_sys->msg_namelen &&
2332 !memcmp(&used_address->name, msg_sys->msg_name,
2333 used_address->name_len)) {
2334 err = sock_sendmsg_nosec(sock, msg_sys);
2337 err = sock_sendmsg(sock, msg_sys);
2339 * If this is sendmmsg() and sending to current destination address was
2340 * successful, remember it.
2342 if (used_address && err >= 0) {
2343 used_address->name_len = msg_sys->msg_namelen;
2344 if (msg_sys->msg_name)
2345 memcpy(&used_address->name, msg_sys->msg_name,
2346 used_address->name_len);
2351 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2356 int sendmsg_copy_msghdr(struct msghdr *msg,
2357 struct user_msghdr __user *umsg, unsigned flags,
2362 if (flags & MSG_CMSG_COMPAT) {
2363 struct compat_msghdr __user *msg_compat;
2365 msg_compat = (struct compat_msghdr __user *) umsg;
2366 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2368 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2376 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2377 struct msghdr *msg_sys, unsigned int flags,
2378 struct used_address *used_address,
2379 unsigned int allowed_msghdr_flags)
2381 struct sockaddr_storage address;
2382 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2385 msg_sys->msg_name = &address;
2387 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2391 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2392 allowed_msghdr_flags);
2398 * BSD sendmsg interface
2400 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2403 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2406 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2407 bool forbid_cmsg_compat)
2409 int fput_needed, err;
2410 struct msghdr msg_sys;
2411 struct socket *sock;
2413 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2416 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2420 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2422 fput_light(sock->file, fput_needed);
2427 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2429 return __sys_sendmsg(fd, msg, flags, true);
2433 * Linux sendmmsg interface
2436 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2437 unsigned int flags, bool forbid_cmsg_compat)
2439 int fput_needed, err, datagrams;
2440 struct socket *sock;
2441 struct mmsghdr __user *entry;
2442 struct compat_mmsghdr __user *compat_entry;
2443 struct msghdr msg_sys;
2444 struct used_address used_address;
2445 unsigned int oflags = flags;
2447 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2450 if (vlen > UIO_MAXIOV)
2455 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2459 used_address.name_len = UINT_MAX;
2461 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2465 while (datagrams < vlen) {
2466 if (datagrams == vlen - 1)
2469 if (MSG_CMSG_COMPAT & flags) {
2470 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2471 &msg_sys, flags, &used_address, MSG_EOR);
2474 err = __put_user(err, &compat_entry->msg_len);
2477 err = ___sys_sendmsg(sock,
2478 (struct user_msghdr __user *)entry,
2479 &msg_sys, flags, &used_address, MSG_EOR);
2482 err = put_user(err, &entry->msg_len);
2489 if (msg_data_left(&msg_sys))
2494 fput_light(sock->file, fput_needed);
2496 /* We only return an error if no datagrams were able to be sent */
2503 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2504 unsigned int, vlen, unsigned int, flags)
2506 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2509 int recvmsg_copy_msghdr(struct msghdr *msg,
2510 struct user_msghdr __user *umsg, unsigned flags,
2511 struct sockaddr __user **uaddr,
2516 if (MSG_CMSG_COMPAT & flags) {
2517 struct compat_msghdr __user *msg_compat;
2519 msg_compat = (struct compat_msghdr __user *) umsg;
2520 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2522 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2530 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2531 struct user_msghdr __user *msg,
2532 struct sockaddr __user *uaddr,
2533 unsigned int flags, int nosec)
2535 struct compat_msghdr __user *msg_compat =
2536 (struct compat_msghdr __user *) msg;
2537 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2538 struct sockaddr_storage addr;
2539 unsigned long cmsg_ptr;
2543 msg_sys->msg_name = &addr;
2544 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2545 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2547 /* We assume all kernel code knows the size of sockaddr_storage */
2548 msg_sys->msg_namelen = 0;
2550 if (sock->file->f_flags & O_NONBLOCK)
2551 flags |= MSG_DONTWAIT;
2553 if (unlikely(nosec))
2554 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2556 err = sock_recvmsg(sock, msg_sys, flags);
2562 if (uaddr != NULL) {
2563 err = move_addr_to_user(&addr,
2564 msg_sys->msg_namelen, uaddr,
2569 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2573 if (MSG_CMSG_COMPAT & flags)
2574 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2575 &msg_compat->msg_controllen);
2577 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2578 &msg->msg_controllen);
2586 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2587 struct msghdr *msg_sys, unsigned int flags, int nosec)
2589 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2590 /* user mode address pointers */
2591 struct sockaddr __user *uaddr;
2594 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2598 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2604 * BSD recvmsg interface
2607 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2608 struct user_msghdr __user *umsg,
2609 struct sockaddr __user *uaddr, unsigned int flags)
2611 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2614 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2615 bool forbid_cmsg_compat)
2617 int fput_needed, err;
2618 struct msghdr msg_sys;
2619 struct socket *sock;
2621 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2624 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2628 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2630 fput_light(sock->file, fput_needed);
2635 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2636 unsigned int, flags)
2638 return __sys_recvmsg(fd, msg, flags, true);
2642 * Linux recvmmsg interface
2645 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2646 unsigned int vlen, unsigned int flags,
2647 struct timespec64 *timeout)
2649 int fput_needed, err, datagrams;
2650 struct socket *sock;
2651 struct mmsghdr __user *entry;
2652 struct compat_mmsghdr __user *compat_entry;
2653 struct msghdr msg_sys;
2654 struct timespec64 end_time;
2655 struct timespec64 timeout64;
2658 poll_select_set_timeout(&end_time, timeout->tv_sec,
2664 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2668 if (likely(!(flags & MSG_ERRQUEUE))) {
2669 err = sock_error(sock->sk);
2677 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2679 while (datagrams < vlen) {
2681 * No need to ask LSM for more than the first datagram.
2683 if (MSG_CMSG_COMPAT & flags) {
2684 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2685 &msg_sys, flags & ~MSG_WAITFORONE,
2689 err = __put_user(err, &compat_entry->msg_len);
2692 err = ___sys_recvmsg(sock,
2693 (struct user_msghdr __user *)entry,
2694 &msg_sys, flags & ~MSG_WAITFORONE,
2698 err = put_user(err, &entry->msg_len);
2706 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2707 if (flags & MSG_WAITFORONE)
2708 flags |= MSG_DONTWAIT;
2711 ktime_get_ts64(&timeout64);
2712 *timeout = timespec64_sub(end_time, timeout64);
2713 if (timeout->tv_sec < 0) {
2714 timeout->tv_sec = timeout->tv_nsec = 0;
2718 /* Timeout, return less than vlen datagrams */
2719 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2723 /* Out of band data, return right away */
2724 if (msg_sys.msg_flags & MSG_OOB)
2732 if (datagrams == 0) {
2738 * We may return less entries than requested (vlen) if the
2739 * sock is non block and there aren't enough datagrams...
2741 if (err != -EAGAIN) {
2743 * ... or if recvmsg returns an error after we
2744 * received some datagrams, where we record the
2745 * error to return on the next call or if the
2746 * app asks about it using getsockopt(SO_ERROR).
2748 sock->sk->sk_err = -err;
2751 fput_light(sock->file, fput_needed);
2756 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2757 unsigned int vlen, unsigned int flags,
2758 struct __kernel_timespec __user *timeout,
2759 struct old_timespec32 __user *timeout32)
2762 struct timespec64 timeout_sys;
2764 if (timeout && get_timespec64(&timeout_sys, timeout))
2767 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2770 if (!timeout && !timeout32)
2771 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2773 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2778 if (timeout && put_timespec64(&timeout_sys, timeout))
2779 datagrams = -EFAULT;
2781 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2782 datagrams = -EFAULT;
2787 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2788 unsigned int, vlen, unsigned int, flags,
2789 struct __kernel_timespec __user *, timeout)
2791 if (flags & MSG_CMSG_COMPAT)
2794 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2797 #ifdef CONFIG_COMPAT_32BIT_TIME
2798 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2799 unsigned int, vlen, unsigned int, flags,
2800 struct old_timespec32 __user *, timeout)
2802 if (flags & MSG_CMSG_COMPAT)
2805 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2809 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2810 /* Argument list sizes for sys_socketcall */
2811 #define AL(x) ((x) * sizeof(unsigned long))
2812 static const unsigned char nargs[21] = {
2813 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2814 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2815 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2822 * System call vectors.
2824 * Argument checking cleaned up. Saved 20% in size.
2825 * This function doesn't need to set the kernel lock because
2826 * it is set by the callees.
2829 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2831 unsigned long a[AUDITSC_ARGS];
2832 unsigned long a0, a1;
2836 if (call < 1 || call > SYS_SENDMMSG)
2838 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2841 if (len > sizeof(a))
2844 /* copy_from_user should be SMP safe. */
2845 if (copy_from_user(a, args, len))
2848 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2857 err = __sys_socket(a0, a1, a[2]);
2860 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2863 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2866 err = __sys_listen(a0, a1);
2869 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2870 (int __user *)a[2], 0);
2872 case SYS_GETSOCKNAME:
2874 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2875 (int __user *)a[2]);
2877 case SYS_GETPEERNAME:
2879 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2880 (int __user *)a[2]);
2882 case SYS_SOCKETPAIR:
2883 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2886 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2890 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2891 (struct sockaddr __user *)a[4], a[5]);
2894 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2898 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2899 (struct sockaddr __user *)a[4],
2900 (int __user *)a[5]);
2903 err = __sys_shutdown(a0, a1);
2905 case SYS_SETSOCKOPT:
2906 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2909 case SYS_GETSOCKOPT:
2911 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2912 (int __user *)a[4]);
2915 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2919 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2923 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2927 if (IS_ENABLED(CONFIG_64BIT))
2928 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2930 (struct __kernel_timespec __user *)a[4],
2933 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2935 (struct old_timespec32 __user *)a[4]);
2938 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2939 (int __user *)a[2], a[3]);
2948 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2951 * sock_register - add a socket protocol handler
2952 * @ops: description of protocol
2954 * This function is called by a protocol handler that wants to
2955 * advertise its address family, and have it linked into the
2956 * socket interface. The value ops->family corresponds to the
2957 * socket system call protocol family.
2959 int sock_register(const struct net_proto_family *ops)
2963 if (ops->family >= NPROTO) {
2964 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2968 spin_lock(&net_family_lock);
2969 if (rcu_dereference_protected(net_families[ops->family],
2970 lockdep_is_held(&net_family_lock)))
2973 rcu_assign_pointer(net_families[ops->family], ops);
2976 spin_unlock(&net_family_lock);
2978 pr_info("NET: Registered protocol family %d\n", ops->family);
2981 EXPORT_SYMBOL(sock_register);
2984 * sock_unregister - remove a protocol handler
2985 * @family: protocol family to remove
2987 * This function is called by a protocol handler that wants to
2988 * remove its address family, and have it unlinked from the
2989 * new socket creation.
2991 * If protocol handler is a module, then it can use module reference
2992 * counts to protect against new references. If protocol handler is not
2993 * a module then it needs to provide its own protection in
2994 * the ops->create routine.
2996 void sock_unregister(int family)
2998 BUG_ON(family < 0 || family >= NPROTO);
3000 spin_lock(&net_family_lock);
3001 RCU_INIT_POINTER(net_families[family], NULL);
3002 spin_unlock(&net_family_lock);
3006 pr_info("NET: Unregistered protocol family %d\n", family);
3008 EXPORT_SYMBOL(sock_unregister);
3010 bool sock_is_registered(int family)
3012 return family < NPROTO && rcu_access_pointer(net_families[family]);
3015 static int __init sock_init(void)
3019 * Initialize the network sysctl infrastructure.
3021 err = net_sysctl_init();
3026 * Initialize skbuff SLAB cache
3031 * Initialize the protocols module.
3036 err = register_filesystem(&sock_fs_type);
3039 sock_mnt = kern_mount(&sock_fs_type);
3040 if (IS_ERR(sock_mnt)) {
3041 err = PTR_ERR(sock_mnt);
3045 /* The real protocol initialization is performed in later initcalls.
3048 #ifdef CONFIG_NETFILTER
3049 err = netfilter_init();
3054 ptp_classifier_init();
3060 unregister_filesystem(&sock_fs_type);
3064 core_initcall(sock_init); /* early initcall */
3066 #ifdef CONFIG_PROC_FS
3067 void socket_seq_show(struct seq_file *seq)
3069 seq_printf(seq, "sockets: used %d\n",
3070 sock_inuse_get(seq->private));
3072 #endif /* CONFIG_PROC_FS */
3074 #ifdef CONFIG_COMPAT
3075 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3077 struct compat_ifconf ifc32;
3081 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3084 ifc.ifc_len = ifc32.ifc_len;
3085 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3088 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3093 ifc32.ifc_len = ifc.ifc_len;
3094 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3100 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3102 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3103 bool convert_in = false, convert_out = false;
3104 size_t buf_size = 0;
3105 struct ethtool_rxnfc __user *rxnfc = NULL;
3107 u32 rule_cnt = 0, actual_rule_cnt;
3112 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3115 compat_rxnfc = compat_ptr(data);
3117 if (get_user(ethcmd, &compat_rxnfc->cmd))
3120 /* Most ethtool structures are defined without padding.
3121 * Unfortunately struct ethtool_rxnfc is an exception.
3126 case ETHTOOL_GRXCLSRLALL:
3127 /* Buffer size is variable */
3128 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3130 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3132 buf_size += rule_cnt * sizeof(u32);
3134 case ETHTOOL_GRXRINGS:
3135 case ETHTOOL_GRXCLSRLCNT:
3136 case ETHTOOL_GRXCLSRULE:
3137 case ETHTOOL_SRXCLSRLINS:
3140 case ETHTOOL_SRXCLSRLDEL:
3141 buf_size += sizeof(struct ethtool_rxnfc);
3143 rxnfc = compat_alloc_user_space(buf_size);
3147 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3150 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3153 /* We expect there to be holes between fs.m_ext and
3154 * fs.ring_cookie and at the end of fs, but nowhere else.
3156 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3157 sizeof(compat_rxnfc->fs.m_ext) !=
3158 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3159 sizeof(rxnfc->fs.m_ext));
3161 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3162 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3163 offsetof(struct ethtool_rxnfc, fs.location) -
3164 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3166 if (copy_in_user(rxnfc, compat_rxnfc,
3167 (void __user *)(&rxnfc->fs.m_ext + 1) -
3168 (void __user *)rxnfc) ||
3169 copy_in_user(&rxnfc->fs.ring_cookie,
3170 &compat_rxnfc->fs.ring_cookie,
3171 (void __user *)(&rxnfc->fs.location + 1) -
3172 (void __user *)&rxnfc->fs.ring_cookie))
3174 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3175 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3177 } else if (copy_in_user(&rxnfc->rule_cnt,
3178 &compat_rxnfc->rule_cnt,
3179 sizeof(rxnfc->rule_cnt)))
3183 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3188 if (copy_in_user(compat_rxnfc, rxnfc,
3189 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3190 (const void __user *)rxnfc) ||
3191 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3192 &rxnfc->fs.ring_cookie,
3193 (const void __user *)(&rxnfc->fs.location + 1) -
3194 (const void __user *)&rxnfc->fs.ring_cookie) ||
3195 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3196 sizeof(rxnfc->rule_cnt)))
3199 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3200 /* As an optimisation, we only copy the actual
3201 * number of rules that the underlying
3202 * function returned. Since Mallory might
3203 * change the rule count in user memory, we
3204 * check that it is less than the rule count
3205 * originally given (as the user buffer size),
3206 * which has been range-checked.
3208 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3210 if (actual_rule_cnt < rule_cnt)
3211 rule_cnt = actual_rule_cnt;
3212 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3213 &rxnfc->rule_locs[0],
3214 rule_cnt * sizeof(u32)))
3222 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3224 compat_uptr_t uptr32;
3229 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3232 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3235 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3236 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3238 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3240 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3241 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3247 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3248 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3249 struct compat_ifreq __user *u_ifreq32)
3254 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3256 if (get_user(data32, &u_ifreq32->ifr_data))
3258 ifreq.ifr_data = compat_ptr(data32);
3260 return dev_ioctl(net, cmd, &ifreq, NULL);
3263 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3265 struct compat_ifreq __user *uifr32)
3267 struct ifreq __user *uifr;
3270 /* Handle the fact that while struct ifreq has the same *layout* on
3271 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3272 * which are handled elsewhere, it still has different *size* due to
3273 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3274 * resulting in struct ifreq being 32 and 40 bytes respectively).
3275 * As a result, if the struct happens to be at the end of a page and
3276 * the next page isn't readable/writable, we get a fault. To prevent
3277 * that, copy back and forth to the full size.
3280 uifr = compat_alloc_user_space(sizeof(*uifr));
3281 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3284 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3295 case SIOCGIFBRDADDR:
3296 case SIOCGIFDSTADDR:
3297 case SIOCGIFNETMASK:
3303 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3311 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3312 struct compat_ifreq __user *uifr32)
3315 struct compat_ifmap __user *uifmap32;
3318 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3319 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3320 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3321 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3322 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3323 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3324 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3325 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3329 err = dev_ioctl(net, cmd, &ifr, NULL);
3331 if (cmd == SIOCGIFMAP && !err) {
3332 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3333 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3334 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3335 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3336 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3337 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3338 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3345 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3346 * for some operations; this forces use of the newer bridge-utils that
3347 * use compatible ioctls
3349 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3353 if (get_user(tmp, argp))
3355 if (tmp == BRCTL_GET_VERSION)
3356 return BRCTL_VERSION + 1;
3360 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3361 unsigned int cmd, unsigned long arg)
3363 void __user *argp = compat_ptr(arg);
3364 struct sock *sk = sock->sk;
3365 struct net *net = sock_net(sk);
3367 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3368 return compat_ifr_data_ioctl(net, cmd, argp);
3373 return old_bridge_ioctl(argp);
3375 return compat_dev_ifconf(net, argp);
3377 return ethtool_ioctl(net, argp);
3379 return compat_siocwandev(net, argp);
3382 return compat_sioc_ifmap(net, cmd, argp);
3383 case SIOCGSTAMP_OLD:
3384 case SIOCGSTAMPNS_OLD:
3385 if (!sock->ops->gettstamp)
3386 return -ENOIOCTLCMD;
3387 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3388 !COMPAT_USE_64BIT_TIME);
3390 case SIOCBONDSLAVEINFOQUERY:
3391 case SIOCBONDINFOQUERY:
3394 return compat_ifr_data_ioctl(net, cmd, argp);
3405 case SIOCGSTAMP_NEW:
3406 case SIOCGSTAMPNS_NEW:
3407 return sock_ioctl(file, cmd, arg);
3424 case SIOCSIFHWBROADCAST:
3426 case SIOCGIFBRDADDR:
3427 case SIOCSIFBRDADDR:
3428 case SIOCGIFDSTADDR:
3429 case SIOCSIFDSTADDR:
3430 case SIOCGIFNETMASK:
3431 case SIOCSIFNETMASK:
3443 case SIOCBONDENSLAVE:
3444 case SIOCBONDRELEASE:
3445 case SIOCBONDSETHWADDR:
3446 case SIOCBONDCHANGEACTIVE:
3447 return compat_ifreq_ioctl(net, sock, cmd, argp);
3455 return sock_do_ioctl(net, sock, cmd, arg);
3458 return -ENOIOCTLCMD;
3461 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3464 struct socket *sock = file->private_data;
3465 int ret = -ENOIOCTLCMD;
3472 if (sock->ops->compat_ioctl)
3473 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3475 if (ret == -ENOIOCTLCMD &&
3476 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3477 ret = compat_wext_handle_ioctl(net, cmd, arg);
3479 if (ret == -ENOIOCTLCMD)
3480 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3487 * kernel_bind - bind an address to a socket (kernel space)
3490 * @addrlen: length of address
3492 * Returns 0 or an error.
3495 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3497 return sock->ops->bind(sock, addr, addrlen);
3499 EXPORT_SYMBOL(kernel_bind);
3502 * kernel_listen - move socket to listening state (kernel space)
3504 * @backlog: pending connections queue size
3506 * Returns 0 or an error.
3509 int kernel_listen(struct socket *sock, int backlog)
3511 return sock->ops->listen(sock, backlog);
3513 EXPORT_SYMBOL(kernel_listen);
3516 * kernel_accept - accept a connection (kernel space)
3517 * @sock: listening socket
3518 * @newsock: new connected socket
3521 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3522 * If it fails, @newsock is guaranteed to be %NULL.
3523 * Returns 0 or an error.
3526 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3528 struct sock *sk = sock->sk;
3531 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3536 err = sock->ops->accept(sock, *newsock, flags, true);
3538 sock_release(*newsock);
3543 (*newsock)->ops = sock->ops;
3544 __module_get((*newsock)->ops->owner);
3549 EXPORT_SYMBOL(kernel_accept);
3552 * kernel_connect - connect a socket (kernel space)
3555 * @addrlen: address length
3556 * @flags: flags (O_NONBLOCK, ...)
3558 * For datagram sockets, @addr is the address to which datagrams are sent
3559 * by default, and the only address from which datagrams are received.
3560 * For stream sockets, attempts to connect to @addr.
3561 * Returns 0 or an error code.
3564 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3567 return sock->ops->connect(sock, addr, addrlen, flags);
3569 EXPORT_SYMBOL(kernel_connect);
3572 * kernel_getsockname - get the address which the socket is bound (kernel space)
3574 * @addr: address holder
3576 * Fills the @addr pointer with the address which the socket is bound.
3577 * Returns 0 or an error code.
3580 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3582 return sock->ops->getname(sock, addr, 0);
3584 EXPORT_SYMBOL(kernel_getsockname);
3587 * kernel_getpeername - get the address which the socket is connected (kernel space)
3589 * @addr: address holder
3591 * Fills the @addr pointer with the address which the socket is connected.
3592 * Returns 0 or an error code.
3595 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3597 return sock->ops->getname(sock, addr, 1);
3599 EXPORT_SYMBOL(kernel_getpeername);
3602 * kernel_sendpage - send a &page through a socket (kernel space)
3605 * @offset: page offset
3606 * @size: total size in bytes
3607 * @flags: flags (MSG_DONTWAIT, ...)
3609 * Returns the total amount sent in bytes or an error.
3612 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3613 size_t size, int flags)
3615 if (sock->ops->sendpage) {
3616 /* Warn in case the improper page to zero-copy send */
3617 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3618 return sock->ops->sendpage(sock, page, offset, size, flags);
3620 return sock_no_sendpage(sock, page, offset, size, flags);
3622 EXPORT_SYMBOL(kernel_sendpage);
3625 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3628 * @offset: page offset
3629 * @size: total size in bytes
3630 * @flags: flags (MSG_DONTWAIT, ...)
3632 * Returns the total amount sent in bytes or an error.
3633 * Caller must hold @sk.
3636 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3637 size_t size, int flags)
3639 struct socket *sock = sk->sk_socket;
3641 if (sock->ops->sendpage_locked)
3642 return sock->ops->sendpage_locked(sk, page, offset, size,
3645 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3647 EXPORT_SYMBOL(kernel_sendpage_locked);
3650 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3652 * @how: connection part
3654 * Returns 0 or an error.
3657 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3659 return sock->ops->shutdown(sock, how);
3661 EXPORT_SYMBOL(kernel_sock_shutdown);
3664 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3667 * This routine returns the IP overhead imposed by a socket i.e.
3668 * the length of the underlying IP header, depending on whether
3669 * this is an IPv4 or IPv6 socket and the length from IP options turned
3670 * on at the socket. Assumes that the caller has a lock on the socket.
3673 u32 kernel_sock_ip_overhead(struct sock *sk)
3675 struct inet_sock *inet;
3676 struct ip_options_rcu *opt;
3678 #if IS_ENABLED(CONFIG_IPV6)
3679 struct ipv6_pinfo *np;
3680 struct ipv6_txoptions *optv6 = NULL;
3681 #endif /* IS_ENABLED(CONFIG_IPV6) */
3686 switch (sk->sk_family) {
3689 overhead += sizeof(struct iphdr);
3690 opt = rcu_dereference_protected(inet->inet_opt,
3691 sock_owned_by_user(sk));
3693 overhead += opt->opt.optlen;
3695 #if IS_ENABLED(CONFIG_IPV6)
3698 overhead += sizeof(struct ipv6hdr);
3700 optv6 = rcu_dereference_protected(np->opt,
3701 sock_owned_by_user(sk));
3703 overhead += (optv6->opt_flen + optv6->opt_nflen);
3705 #endif /* IS_ENABLED(CONFIG_IPV6) */
3706 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3710 EXPORT_SYMBOL(kernel_sock_ip_overhead);