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/sockios.h>
104 #include <net/busy_poll.h>
105 #include <linux/errqueue.h>
107 #ifdef CONFIG_NET_RX_BUSY_POLL
108 unsigned int sysctl_net_busy_read __read_mostly;
109 unsigned int sysctl_net_busy_poll __read_mostly;
112 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
113 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
114 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
116 static int sock_close(struct inode *inode, struct file *file);
117 static __poll_t sock_poll(struct file *file,
118 struct poll_table_struct *wait);
119 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
121 static long compat_sock_ioctl(struct file *file,
122 unsigned int cmd, unsigned long arg);
124 static int sock_fasync(int fd, struct file *filp, int on);
125 static ssize_t sock_sendpage(struct file *file, struct page *page,
126 int offset, size_t size, loff_t *ppos, int more);
127 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
128 struct pipe_inode_info *pipe, size_t len,
132 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
133 * in the operation structures but are done directly via the socketcall() multiplexor.
136 static const struct file_operations socket_file_ops = {
137 .owner = THIS_MODULE,
139 .read_iter = sock_read_iter,
140 .write_iter = sock_write_iter,
142 .unlocked_ioctl = sock_ioctl,
144 .compat_ioctl = compat_sock_ioctl,
147 .release = sock_close,
148 .fasync = sock_fasync,
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
151 .splice_read = sock_splice_read,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock);
159 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
163 * Move socket addresses back and forth across the kernel/user
164 * divide and look after the messy bits.
168 * move_addr_to_kernel - copy a socket address into kernel space
169 * @uaddr: Address in user space
170 * @kaddr: Address in kernel space
171 * @ulen: Length in user space
173 * The address is copied into kernel space. If the provided address is
174 * too long an error code of -EINVAL is returned. If the copy gives
175 * invalid addresses -EFAULT is returned. On a success 0 is returned.
178 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
180 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
184 if (copy_from_user(kaddr, uaddr, ulen))
186 return audit_sockaddr(ulen, kaddr);
190 * move_addr_to_user - copy an address to user space
191 * @kaddr: kernel space address
192 * @klen: length of address in kernel
193 * @uaddr: user space address
194 * @ulen: pointer to user length field
196 * The value pointed to by ulen on entry is the buffer length available.
197 * This is overwritten with the buffer space used. -EINVAL is returned
198 * if an overlong buffer is specified or a negative buffer size. -EFAULT
199 * is returned if either the buffer or the length field are not
201 * After copying the data up to the limit the user specifies, the true
202 * length of the data is written over the length limit the user
203 * specified. Zero is returned for a success.
206 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
207 void __user *uaddr, int __user *ulen)
212 BUG_ON(klen > sizeof(struct sockaddr_storage));
213 err = get_user(len, ulen);
221 if (audit_sockaddr(klen, kaddr))
223 if (copy_to_user(uaddr, kaddr, len))
227 * "fromlen shall refer to the value before truncation.."
230 return __put_user(klen, ulen);
233 static struct kmem_cache *sock_inode_cachep __ro_after_init;
235 static struct inode *sock_alloc_inode(struct super_block *sb)
237 struct socket_alloc *ei;
239 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
242 init_waitqueue_head(&ei->socket.wq.wait);
243 ei->socket.wq.fasync_list = NULL;
244 ei->socket.wq.flags = 0;
246 ei->socket.state = SS_UNCONNECTED;
247 ei->socket.flags = 0;
248 ei->socket.ops = NULL;
249 ei->socket.sk = NULL;
250 ei->socket.file = NULL;
252 return &ei->vfs_inode;
255 static void sock_free_inode(struct inode *inode)
257 struct socket_alloc *ei;
259 ei = container_of(inode, struct socket_alloc, vfs_inode);
260 kmem_cache_free(sock_inode_cachep, ei);
263 static void init_once(void *foo)
265 struct socket_alloc *ei = (struct socket_alloc *)foo;
267 inode_init_once(&ei->vfs_inode);
270 static void init_inodecache(void)
272 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
273 sizeof(struct socket_alloc),
275 (SLAB_HWCACHE_ALIGN |
276 SLAB_RECLAIM_ACCOUNT |
277 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
279 BUG_ON(sock_inode_cachep == NULL);
282 static const struct super_operations sockfs_ops = {
283 .alloc_inode = sock_alloc_inode,
284 .free_inode = sock_free_inode,
285 .statfs = simple_statfs,
289 * sockfs_dname() is called from d_path().
291 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
293 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
294 d_inode(dentry)->i_ino);
297 static const struct dentry_operations sockfs_dentry_operations = {
298 .d_dname = sockfs_dname,
301 static int sockfs_xattr_get(const struct xattr_handler *handler,
302 struct dentry *dentry, struct inode *inode,
303 const char *suffix, void *value, size_t size)
306 if (dentry->d_name.len + 1 > size)
308 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
310 return dentry->d_name.len + 1;
313 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
314 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
315 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
317 static const struct xattr_handler sockfs_xattr_handler = {
318 .name = XATTR_NAME_SOCKPROTONAME,
319 .get = sockfs_xattr_get,
322 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
323 struct dentry *dentry, struct inode *inode,
324 const char *suffix, const void *value,
325 size_t size, int flags)
327 /* Handled by LSM. */
331 static const struct xattr_handler sockfs_security_xattr_handler = {
332 .prefix = XATTR_SECURITY_PREFIX,
333 .set = sockfs_security_xattr_set,
336 static const struct xattr_handler *sockfs_xattr_handlers[] = {
337 &sockfs_xattr_handler,
338 &sockfs_security_xattr_handler,
342 static int sockfs_init_fs_context(struct fs_context *fc)
344 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
347 ctx->ops = &sockfs_ops;
348 ctx->dops = &sockfs_dentry_operations;
349 ctx->xattr = sockfs_xattr_handlers;
353 static struct vfsmount *sock_mnt __read_mostly;
355 static struct file_system_type sock_fs_type = {
357 .init_fs_context = sockfs_init_fs_context,
358 .kill_sb = kill_anon_super,
362 * Obtains the first available file descriptor and sets it up for use.
364 * These functions create file structures and maps them to fd space
365 * of the current process. On success it returns file descriptor
366 * and file struct implicitly stored in sock->file.
367 * Note that another thread may close file descriptor before we return
368 * from this function. We use the fact that now we do not refer
369 * to socket after mapping. If one day we will need it, this
370 * function will increment ref. count on file by 1.
372 * In any case returned fd MAY BE not valid!
373 * This race condition is unavoidable
374 * with shared fd spaces, we cannot solve it inside kernel,
375 * but we take care of internal coherence yet.
379 * sock_alloc_file - Bind a &socket to a &file
381 * @flags: file status flags
382 * @dname: protocol name
384 * Returns the &file bound with @sock, implicitly storing it
385 * in sock->file. If dname is %NULL, sets to "".
386 * On failure the return is a ERR pointer (see linux/err.h).
387 * This function uses GFP_KERNEL internally.
390 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
395 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
397 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
398 O_RDWR | (flags & O_NONBLOCK),
406 file->private_data = sock;
407 stream_open(SOCK_INODE(sock), file);
410 EXPORT_SYMBOL(sock_alloc_file);
412 static int sock_map_fd(struct socket *sock, int flags)
414 struct file *newfile;
415 int fd = get_unused_fd_flags(flags);
416 if (unlikely(fd < 0)) {
421 newfile = sock_alloc_file(sock, flags, NULL);
422 if (!IS_ERR(newfile)) {
423 fd_install(fd, newfile);
428 return PTR_ERR(newfile);
432 * sock_from_file - Return the &socket bounded to @file.
434 * @err: pointer to an error code return
436 * On failure returns %NULL and assigns -ENOTSOCK to @err.
439 struct socket *sock_from_file(struct file *file, int *err)
441 if (file->f_op == &socket_file_ops)
442 return file->private_data; /* set in sock_map_fd */
447 EXPORT_SYMBOL(sock_from_file);
450 * sockfd_lookup - Go from a file number to its socket slot
452 * @err: pointer to an error code return
454 * The file handle passed in is locked and the socket it is bound
455 * to is returned. If an error occurs the err pointer is overwritten
456 * with a negative errno code and NULL is returned. The function checks
457 * for both invalid handles and passing a handle which is not a socket.
459 * On a success the socket object pointer is returned.
462 struct socket *sockfd_lookup(int fd, int *err)
473 sock = sock_from_file(file, err);
478 EXPORT_SYMBOL(sockfd_lookup);
480 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
482 struct fd f = fdget(fd);
487 sock = sock_from_file(f.file, err);
489 *fput_needed = f.flags;
497 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
503 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
513 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
518 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
525 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
527 int err = simple_setattr(dentry, iattr);
529 if (!err && (iattr->ia_valid & ATTR_UID)) {
530 struct socket *sock = SOCKET_I(d_inode(dentry));
533 sock->sk->sk_uid = iattr->ia_uid;
541 static const struct inode_operations sockfs_inode_ops = {
542 .listxattr = sockfs_listxattr,
543 .setattr = sockfs_setattr,
547 * sock_alloc - allocate a socket
549 * Allocate a new inode and socket object. The two are bound together
550 * and initialised. The socket is then returned. If we are out of inodes
551 * NULL is returned. This functions uses GFP_KERNEL internally.
554 struct socket *sock_alloc(void)
559 inode = new_inode_pseudo(sock_mnt->mnt_sb);
563 sock = SOCKET_I(inode);
565 inode->i_ino = get_next_ino();
566 inode->i_mode = S_IFSOCK | S_IRWXUGO;
567 inode->i_uid = current_fsuid();
568 inode->i_gid = current_fsgid();
569 inode->i_op = &sockfs_inode_ops;
573 EXPORT_SYMBOL(sock_alloc);
576 * sock_release - close a socket
577 * @sock: socket to close
579 * The socket is released from the protocol stack if it has a release
580 * callback, and the inode is then released if the socket is bound to
581 * an inode not a file.
584 static void __sock_release(struct socket *sock, struct inode *inode)
587 struct module *owner = sock->ops->owner;
591 sock->ops->release(sock);
599 if (sock->wq.fasync_list)
600 pr_err("%s: fasync list not empty!\n", __func__);
603 iput(SOCK_INODE(sock));
609 void sock_release(struct socket *sock)
611 __sock_release(sock, NULL);
613 EXPORT_SYMBOL(sock_release);
615 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
617 u8 flags = *tx_flags;
619 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
620 flags |= SKBTX_HW_TSTAMP;
622 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
623 flags |= SKBTX_SW_TSTAMP;
625 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
626 flags |= SKBTX_SCHED_TSTAMP;
630 EXPORT_SYMBOL(__sock_tx_timestamp);
632 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
634 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
636 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
638 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
639 inet_sendmsg, sock, msg,
641 BUG_ON(ret == -EIOCBQUEUED);
646 * sock_sendmsg - send a message through @sock
648 * @msg: message to send
650 * Sends @msg through @sock, passing through LSM.
651 * Returns the number of bytes sent, or an error code.
653 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
655 int err = security_socket_sendmsg(sock, msg,
658 return err ?: sock_sendmsg_nosec(sock, msg);
660 EXPORT_SYMBOL(sock_sendmsg);
663 * kernel_sendmsg - send a message through @sock (kernel-space)
665 * @msg: message header
667 * @num: vec array length
668 * @size: total message data size
670 * Builds the message data with @vec and sends it through @sock.
671 * Returns the number of bytes sent, or an error code.
674 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
675 struct kvec *vec, size_t num, size_t size)
677 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
678 return sock_sendmsg(sock, msg);
680 EXPORT_SYMBOL(kernel_sendmsg);
683 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
685 * @msg: message header
686 * @vec: output s/g array
687 * @num: output s/g array length
688 * @size: total message data size
690 * Builds the message data with @vec and sends it through @sock.
691 * Returns the number of bytes sent, or an error code.
692 * Caller must hold @sk.
695 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
696 struct kvec *vec, size_t num, size_t size)
698 struct socket *sock = sk->sk_socket;
700 if (!sock->ops->sendmsg_locked)
701 return sock_no_sendmsg_locked(sk, msg, size);
703 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
705 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
707 EXPORT_SYMBOL(kernel_sendmsg_locked);
709 static bool skb_is_err_queue(const struct sk_buff *skb)
711 /* pkt_type of skbs enqueued on the error queue are set to
712 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
713 * in recvmsg, since skbs received on a local socket will never
714 * have a pkt_type of PACKET_OUTGOING.
716 return skb->pkt_type == PACKET_OUTGOING;
719 /* On transmit, software and hardware timestamps are returned independently.
720 * As the two skb clones share the hardware timestamp, which may be updated
721 * before the software timestamp is received, a hardware TX timestamp may be
722 * returned only if there is no software TX timestamp. Ignore false software
723 * timestamps, which may be made in the __sock_recv_timestamp() call when the
724 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
725 * hardware timestamp.
727 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
729 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
732 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
734 struct scm_ts_pktinfo ts_pktinfo;
735 struct net_device *orig_dev;
737 if (!skb_mac_header_was_set(skb))
740 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
743 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
745 ts_pktinfo.if_index = orig_dev->ifindex;
748 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
749 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
750 sizeof(ts_pktinfo), &ts_pktinfo);
754 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
756 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
759 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
760 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
761 struct scm_timestamping_internal tss;
763 int empty = 1, false_tstamp = 0;
764 struct skb_shared_hwtstamps *shhwtstamps =
767 /* Race occurred between timestamp enabling and packet
768 receiving. Fill in the current time for now. */
769 if (need_software_tstamp && skb->tstamp == 0) {
770 __net_timestamp(skb);
774 if (need_software_tstamp) {
775 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
777 struct __kernel_sock_timeval tv;
779 skb_get_new_timestamp(skb, &tv);
780 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
783 struct __kernel_old_timeval tv;
785 skb_get_timestamp(skb, &tv);
786 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
791 struct __kernel_timespec ts;
793 skb_get_new_timestampns(skb, &ts);
794 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
799 skb_get_timestampns(skb, &ts);
800 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
806 memset(&tss, 0, sizeof(tss));
807 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
808 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
811 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
812 !skb_is_swtx_tstamp(skb, false_tstamp) &&
813 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
815 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
816 !skb_is_err_queue(skb))
817 put_ts_pktinfo(msg, skb);
820 if (sock_flag(sk, SOCK_TSTAMP_NEW))
821 put_cmsg_scm_timestamping64(msg, &tss);
823 put_cmsg_scm_timestamping(msg, &tss);
825 if (skb_is_err_queue(skb) && skb->len &&
826 SKB_EXT_ERR(skb)->opt_stats)
827 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
828 skb->len, skb->data);
831 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
833 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
838 if (!sock_flag(sk, SOCK_WIFI_STATUS))
840 if (!skb->wifi_acked_valid)
843 ack = skb->wifi_acked;
845 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
847 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
849 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
852 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
853 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
854 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
857 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
860 sock_recv_timestamp(msg, sk, skb);
861 sock_recv_drops(msg, sk, skb);
863 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
865 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
867 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
869 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
872 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
873 inet_recvmsg, sock, msg, msg_data_left(msg),
878 * sock_recvmsg - receive a message from @sock
880 * @msg: message to receive
881 * @flags: message flags
883 * Receives @msg from @sock, passing through LSM. Returns the total number
884 * of bytes received, or an error.
886 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
888 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
890 return err ?: sock_recvmsg_nosec(sock, msg, flags);
892 EXPORT_SYMBOL(sock_recvmsg);
895 * kernel_recvmsg - Receive a message from a socket (kernel space)
896 * @sock: The socket to receive the message from
897 * @msg: Received message
898 * @vec: Input s/g array for message data
899 * @num: Size of input s/g array
900 * @size: Number of bytes to read
901 * @flags: Message flags (MSG_DONTWAIT, etc...)
903 * On return the msg structure contains the scatter/gather array passed in the
904 * vec argument. The array is modified so that it consists of the unfilled
905 * portion of the original array.
907 * The returned value is the total number of bytes received, or an error.
910 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
911 struct kvec *vec, size_t num, size_t size, int flags)
913 mm_segment_t oldfs = get_fs();
916 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
918 result = sock_recvmsg(sock, msg, flags);
922 EXPORT_SYMBOL(kernel_recvmsg);
924 static ssize_t sock_sendpage(struct file *file, struct page *page,
925 int offset, size_t size, loff_t *ppos, int more)
930 sock = file->private_data;
932 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
933 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
936 return kernel_sendpage(sock, page, offset, size, flags);
939 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
940 struct pipe_inode_info *pipe, size_t len,
943 struct socket *sock = file->private_data;
945 if (unlikely(!sock->ops->splice_read))
946 return generic_file_splice_read(file, ppos, pipe, len, flags);
948 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
951 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
953 struct file *file = iocb->ki_filp;
954 struct socket *sock = file->private_data;
955 struct msghdr msg = {.msg_iter = *to,
959 if (file->f_flags & O_NONBLOCK)
960 msg.msg_flags = MSG_DONTWAIT;
962 if (iocb->ki_pos != 0)
965 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
968 res = sock_recvmsg(sock, &msg, msg.msg_flags);
973 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
975 struct file *file = iocb->ki_filp;
976 struct socket *sock = file->private_data;
977 struct msghdr msg = {.msg_iter = *from,
981 if (iocb->ki_pos != 0)
984 if (file->f_flags & O_NONBLOCK)
985 msg.msg_flags = MSG_DONTWAIT;
987 if (sock->type == SOCK_SEQPACKET)
988 msg.msg_flags |= MSG_EOR;
990 res = sock_sendmsg(sock, &msg);
991 *from = msg.msg_iter;
996 * Atomic setting of ioctl hooks to avoid race
997 * with module unload.
1000 static DEFINE_MUTEX(br_ioctl_mutex);
1001 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1003 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1005 mutex_lock(&br_ioctl_mutex);
1006 br_ioctl_hook = hook;
1007 mutex_unlock(&br_ioctl_mutex);
1009 EXPORT_SYMBOL(brioctl_set);
1011 static DEFINE_MUTEX(vlan_ioctl_mutex);
1012 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1014 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1016 mutex_lock(&vlan_ioctl_mutex);
1017 vlan_ioctl_hook = hook;
1018 mutex_unlock(&vlan_ioctl_mutex);
1020 EXPORT_SYMBOL(vlan_ioctl_set);
1022 static DEFINE_MUTEX(dlci_ioctl_mutex);
1023 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1025 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1027 mutex_lock(&dlci_ioctl_mutex);
1028 dlci_ioctl_hook = hook;
1029 mutex_unlock(&dlci_ioctl_mutex);
1031 EXPORT_SYMBOL(dlci_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.
1076 * get_net_ns - increment the refcount of the network namespace
1077 * @ns: common namespace (net)
1079 * Returns the net's common namespace.
1082 struct ns_common *get_net_ns(struct ns_common *ns)
1084 return &get_net(container_of(ns, struct net, ns))->ns;
1086 EXPORT_SYMBOL_GPL(get_net_ns);
1088 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1090 struct socket *sock;
1092 void __user *argp = (void __user *)arg;
1096 sock = file->private_data;
1099 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1102 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1104 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1105 if (!err && need_copyout)
1106 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1109 #ifdef CONFIG_WEXT_CORE
1110 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1111 err = wext_handle_ioctl(net, cmd, argp);
1118 if (get_user(pid, (int __user *)argp))
1120 err = f_setown(sock->file, pid, 1);
1124 err = put_user(f_getown(sock->file),
1125 (int __user *)argp);
1133 request_module("bridge");
1135 mutex_lock(&br_ioctl_mutex);
1137 err = br_ioctl_hook(net, cmd, argp);
1138 mutex_unlock(&br_ioctl_mutex);
1143 if (!vlan_ioctl_hook)
1144 request_module("8021q");
1146 mutex_lock(&vlan_ioctl_mutex);
1147 if (vlan_ioctl_hook)
1148 err = vlan_ioctl_hook(net, argp);
1149 mutex_unlock(&vlan_ioctl_mutex);
1154 if (!dlci_ioctl_hook)
1155 request_module("dlci");
1157 mutex_lock(&dlci_ioctl_mutex);
1158 if (dlci_ioctl_hook)
1159 err = dlci_ioctl_hook(cmd, argp);
1160 mutex_unlock(&dlci_ioctl_mutex);
1164 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1167 err = open_related_ns(&net->ns, get_net_ns);
1169 case SIOCGSTAMP_OLD:
1170 case SIOCGSTAMPNS_OLD:
1171 if (!sock->ops->gettstamp) {
1175 err = sock->ops->gettstamp(sock, argp,
1176 cmd == SIOCGSTAMP_OLD,
1177 !IS_ENABLED(CONFIG_64BIT));
1179 case SIOCGSTAMP_NEW:
1180 case SIOCGSTAMPNS_NEW:
1181 if (!sock->ops->gettstamp) {
1185 err = sock->ops->gettstamp(sock, argp,
1186 cmd == SIOCGSTAMP_NEW,
1190 err = sock_do_ioctl(net, sock, cmd, arg);
1197 * sock_create_lite - creates a socket
1198 * @family: protocol family (AF_INET, ...)
1199 * @type: communication type (SOCK_STREAM, ...)
1200 * @protocol: protocol (0, ...)
1203 * Creates a new socket and assigns it to @res, passing through LSM.
1204 * The new socket initialization is not complete, see kernel_accept().
1205 * Returns 0 or an error. On failure @res is set to %NULL.
1206 * This function internally uses GFP_KERNEL.
1209 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1212 struct socket *sock = NULL;
1214 err = security_socket_create(family, type, protocol, 1);
1218 sock = sock_alloc();
1225 err = security_socket_post_create(sock, family, type, protocol, 1);
1237 EXPORT_SYMBOL(sock_create_lite);
1239 /* No kernel lock held - perfect */
1240 static __poll_t sock_poll(struct file *file, poll_table *wait)
1242 struct socket *sock = file->private_data;
1243 __poll_t events = poll_requested_events(wait), flag = 0;
1245 if (!sock->ops->poll)
1248 if (sk_can_busy_loop(sock->sk)) {
1249 /* poll once if requested by the syscall */
1250 if (events & POLL_BUSY_LOOP)
1251 sk_busy_loop(sock->sk, 1);
1253 /* if this socket can poll_ll, tell the system call */
1254 flag = POLL_BUSY_LOOP;
1257 return sock->ops->poll(file, sock, wait) | flag;
1260 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1262 struct socket *sock = file->private_data;
1264 return sock->ops->mmap(file, sock, vma);
1267 static int sock_close(struct inode *inode, struct file *filp)
1269 __sock_release(SOCKET_I(inode), inode);
1274 * Update the socket async list
1276 * Fasync_list locking strategy.
1278 * 1. fasync_list is modified only under process context socket lock
1279 * i.e. under semaphore.
1280 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1281 * or under socket lock
1284 static int sock_fasync(int fd, struct file *filp, int on)
1286 struct socket *sock = filp->private_data;
1287 struct sock *sk = sock->sk;
1288 struct socket_wq *wq = &sock->wq;
1294 fasync_helper(fd, filp, on, &wq->fasync_list);
1296 if (!wq->fasync_list)
1297 sock_reset_flag(sk, SOCK_FASYNC);
1299 sock_set_flag(sk, SOCK_FASYNC);
1305 /* This function may be called only under rcu_lock */
1307 int sock_wake_async(struct socket_wq *wq, int how, int band)
1309 if (!wq || !wq->fasync_list)
1313 case SOCK_WAKE_WAITD:
1314 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1317 case SOCK_WAKE_SPACE:
1318 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1323 kill_fasync(&wq->fasync_list, SIGIO, band);
1326 kill_fasync(&wq->fasync_list, SIGURG, band);
1331 EXPORT_SYMBOL(sock_wake_async);
1334 * __sock_create - creates a socket
1335 * @net: net namespace
1336 * @family: protocol family (AF_INET, ...)
1337 * @type: communication type (SOCK_STREAM, ...)
1338 * @protocol: protocol (0, ...)
1340 * @kern: boolean for kernel space sockets
1342 * Creates a new socket and assigns it to @res, passing through LSM.
1343 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1344 * be set to true if the socket resides in kernel space.
1345 * This function internally uses GFP_KERNEL.
1348 int __sock_create(struct net *net, int family, int type, int protocol,
1349 struct socket **res, int kern)
1352 struct socket *sock;
1353 const struct net_proto_family *pf;
1356 * Check protocol is in range
1358 if (family < 0 || family >= NPROTO)
1359 return -EAFNOSUPPORT;
1360 if (type < 0 || type >= SOCK_MAX)
1365 This uglymoron is moved from INET layer to here to avoid
1366 deadlock in module load.
1368 if (family == PF_INET && type == SOCK_PACKET) {
1369 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1374 err = security_socket_create(family, type, protocol, kern);
1379 * Allocate the socket and allow the family to set things up. if
1380 * the protocol is 0, the family is instructed to select an appropriate
1383 sock = sock_alloc();
1385 net_warn_ratelimited("socket: no more sockets\n");
1386 return -ENFILE; /* Not exactly a match, but its the
1387 closest posix thing */
1392 #ifdef CONFIG_MODULES
1393 /* Attempt to load a protocol module if the find failed.
1395 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1396 * requested real, full-featured networking support upon configuration.
1397 * Otherwise module support will break!
1399 if (rcu_access_pointer(net_families[family]) == NULL)
1400 request_module("net-pf-%d", family);
1404 pf = rcu_dereference(net_families[family]);
1405 err = -EAFNOSUPPORT;
1410 * We will call the ->create function, that possibly is in a loadable
1411 * module, so we have to bump that loadable module refcnt first.
1413 if (!try_module_get(pf->owner))
1416 /* Now protected by module ref count */
1419 err = pf->create(net, sock, protocol, kern);
1421 goto out_module_put;
1424 * Now to bump the refcnt of the [loadable] module that owns this
1425 * socket at sock_release time we decrement its refcnt.
1427 if (!try_module_get(sock->ops->owner))
1428 goto out_module_busy;
1431 * Now that we're done with the ->create function, the [loadable]
1432 * module can have its refcnt decremented
1434 module_put(pf->owner);
1435 err = security_socket_post_create(sock, family, type, protocol, kern);
1437 goto out_sock_release;
1443 err = -EAFNOSUPPORT;
1446 module_put(pf->owner);
1453 goto out_sock_release;
1455 EXPORT_SYMBOL(__sock_create);
1458 * sock_create - creates a socket
1459 * @family: protocol family (AF_INET, ...)
1460 * @type: communication type (SOCK_STREAM, ...)
1461 * @protocol: protocol (0, ...)
1464 * A wrapper around __sock_create().
1465 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1468 int sock_create(int family, int type, int protocol, struct socket **res)
1470 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1472 EXPORT_SYMBOL(sock_create);
1475 * sock_create_kern - creates a socket (kernel space)
1476 * @net: net namespace
1477 * @family: protocol family (AF_INET, ...)
1478 * @type: communication type (SOCK_STREAM, ...)
1479 * @protocol: protocol (0, ...)
1482 * A wrapper around __sock_create().
1483 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1486 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1488 return __sock_create(net, family, type, protocol, res, 1);
1490 EXPORT_SYMBOL(sock_create_kern);
1492 int __sys_socket(int family, int type, int protocol)
1495 struct socket *sock;
1498 /* Check the SOCK_* constants for consistency. */
1499 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1500 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1501 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1502 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1504 flags = type & ~SOCK_TYPE_MASK;
1505 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1507 type &= SOCK_TYPE_MASK;
1509 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1510 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1512 retval = sock_create(family, type, protocol, &sock);
1516 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1519 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1521 return __sys_socket(family, type, protocol);
1525 * Create a pair of connected sockets.
1528 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1530 struct socket *sock1, *sock2;
1532 struct file *newfile1, *newfile2;
1535 flags = type & ~SOCK_TYPE_MASK;
1536 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1538 type &= SOCK_TYPE_MASK;
1540 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1541 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1544 * reserve descriptors and make sure we won't fail
1545 * to return them to userland.
1547 fd1 = get_unused_fd_flags(flags);
1548 if (unlikely(fd1 < 0))
1551 fd2 = get_unused_fd_flags(flags);
1552 if (unlikely(fd2 < 0)) {
1557 err = put_user(fd1, &usockvec[0]);
1561 err = put_user(fd2, &usockvec[1]);
1566 * Obtain the first socket and check if the underlying protocol
1567 * supports the socketpair call.
1570 err = sock_create(family, type, protocol, &sock1);
1571 if (unlikely(err < 0))
1574 err = sock_create(family, type, protocol, &sock2);
1575 if (unlikely(err < 0)) {
1576 sock_release(sock1);
1580 err = security_socket_socketpair(sock1, sock2);
1581 if (unlikely(err)) {
1582 sock_release(sock2);
1583 sock_release(sock1);
1587 err = sock1->ops->socketpair(sock1, sock2);
1588 if (unlikely(err < 0)) {
1589 sock_release(sock2);
1590 sock_release(sock1);
1594 newfile1 = sock_alloc_file(sock1, flags, NULL);
1595 if (IS_ERR(newfile1)) {
1596 err = PTR_ERR(newfile1);
1597 sock_release(sock2);
1601 newfile2 = sock_alloc_file(sock2, flags, NULL);
1602 if (IS_ERR(newfile2)) {
1603 err = PTR_ERR(newfile2);
1608 audit_fd_pair(fd1, fd2);
1610 fd_install(fd1, newfile1);
1611 fd_install(fd2, newfile2);
1620 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1621 int __user *, usockvec)
1623 return __sys_socketpair(family, type, protocol, usockvec);
1627 * Bind a name to a socket. Nothing much to do here since it's
1628 * the protocol's responsibility to handle the local address.
1630 * We move the socket address to kernel space before we call
1631 * the protocol layer (having also checked the address is ok).
1634 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1636 struct socket *sock;
1637 struct sockaddr_storage address;
1638 int err, fput_needed;
1640 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1642 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1644 err = security_socket_bind(sock,
1645 (struct sockaddr *)&address,
1648 err = sock->ops->bind(sock,
1652 fput_light(sock->file, fput_needed);
1657 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1659 return __sys_bind(fd, umyaddr, addrlen);
1663 * Perform a listen. Basically, we allow the protocol to do anything
1664 * necessary for a listen, and if that works, we mark the socket as
1665 * ready for listening.
1668 int __sys_listen(int fd, int backlog)
1670 struct socket *sock;
1671 int err, fput_needed;
1674 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1676 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1677 if ((unsigned int)backlog > somaxconn)
1678 backlog = somaxconn;
1680 err = security_socket_listen(sock, backlog);
1682 err = sock->ops->listen(sock, backlog);
1684 fput_light(sock->file, fput_needed);
1689 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1691 return __sys_listen(fd, backlog);
1695 * For accept, we attempt to create a new socket, set up the link
1696 * with the client, wake up the client, then return the new
1697 * connected fd. We collect the address of the connector in kernel
1698 * space and move it to user at the very end. This is unclean because
1699 * we open the socket then return an error.
1701 * 1003.1g adds the ability to recvmsg() to query connection pending
1702 * status to recvmsg. We need to add that support in a way thats
1703 * clean when we restructure accept also.
1706 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1707 int __user *upeer_addrlen, int flags)
1709 struct socket *sock, *newsock;
1710 struct file *newfile;
1711 int err, len, newfd, fput_needed;
1712 struct sockaddr_storage address;
1714 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1717 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1718 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1720 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1725 newsock = sock_alloc();
1729 newsock->type = sock->type;
1730 newsock->ops = sock->ops;
1733 * We don't need try_module_get here, as the listening socket (sock)
1734 * has the protocol module (sock->ops->owner) held.
1736 __module_get(newsock->ops->owner);
1738 newfd = get_unused_fd_flags(flags);
1739 if (unlikely(newfd < 0)) {
1741 sock_release(newsock);
1744 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1745 if (IS_ERR(newfile)) {
1746 err = PTR_ERR(newfile);
1747 put_unused_fd(newfd);
1751 err = security_socket_accept(sock, newsock);
1755 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1759 if (upeer_sockaddr) {
1760 len = newsock->ops->getname(newsock,
1761 (struct sockaddr *)&address, 2);
1763 err = -ECONNABORTED;
1766 err = move_addr_to_user(&address,
1767 len, upeer_sockaddr, upeer_addrlen);
1772 /* File flags are not inherited via accept() unlike another OSes. */
1774 fd_install(newfd, newfile);
1778 fput_light(sock->file, fput_needed);
1783 put_unused_fd(newfd);
1787 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1788 int __user *, upeer_addrlen, int, flags)
1790 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1793 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1794 int __user *, upeer_addrlen)
1796 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1800 * Attempt to connect to a socket with the server address. The address
1801 * is in user space so we verify it is OK and move it to kernel space.
1803 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1806 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1807 * other SEQPACKET protocols that take time to connect() as it doesn't
1808 * include the -EINPROGRESS status for such sockets.
1811 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1813 struct socket *sock;
1814 struct sockaddr_storage address;
1815 int err, fput_needed;
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1820 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1825 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1829 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1830 sock->file->f_flags);
1832 fput_light(sock->file, fput_needed);
1837 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1840 return __sys_connect(fd, uservaddr, addrlen);
1844 * Get the local address ('name') of a socket object. Move the obtained
1845 * name to user space.
1848 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1849 int __user *usockaddr_len)
1851 struct socket *sock;
1852 struct sockaddr_storage address;
1853 int err, fput_needed;
1855 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1859 err = security_socket_getsockname(sock);
1863 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1866 /* "err" is actually length in this case */
1867 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1870 fput_light(sock->file, fput_needed);
1875 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1876 int __user *, usockaddr_len)
1878 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1882 * Get the remote address ('name') of a socket object. Move the obtained
1883 * name to user space.
1886 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1887 int __user *usockaddr_len)
1889 struct socket *sock;
1890 struct sockaddr_storage address;
1891 int err, fput_needed;
1893 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1895 err = security_socket_getpeername(sock);
1897 fput_light(sock->file, fput_needed);
1901 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1903 /* "err" is actually length in this case */
1904 err = move_addr_to_user(&address, err, usockaddr,
1906 fput_light(sock->file, fput_needed);
1911 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1912 int __user *, usockaddr_len)
1914 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1918 * Send a datagram to a given address. We move the address into kernel
1919 * space and check the user space data area is readable before invoking
1922 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1923 struct sockaddr __user *addr, int addr_len)
1925 struct socket *sock;
1926 struct sockaddr_storage address;
1932 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1935 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1939 msg.msg_name = NULL;
1940 msg.msg_control = NULL;
1941 msg.msg_controllen = 0;
1942 msg.msg_namelen = 0;
1944 err = move_addr_to_kernel(addr, addr_len, &address);
1947 msg.msg_name = (struct sockaddr *)&address;
1948 msg.msg_namelen = addr_len;
1950 if (sock->file->f_flags & O_NONBLOCK)
1951 flags |= MSG_DONTWAIT;
1952 msg.msg_flags = flags;
1953 err = sock_sendmsg(sock, &msg);
1956 fput_light(sock->file, fput_needed);
1961 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1962 unsigned int, flags, struct sockaddr __user *, addr,
1965 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1969 * Send a datagram down a socket.
1972 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1973 unsigned int, flags)
1975 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1979 * Receive a frame from the socket and optionally record the address of the
1980 * sender. We verify the buffers are writable and if needed move the
1981 * sender address from kernel to user space.
1983 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1984 struct sockaddr __user *addr, int __user *addr_len)
1986 struct socket *sock;
1989 struct sockaddr_storage address;
1993 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1996 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2000 msg.msg_control = NULL;
2001 msg.msg_controllen = 0;
2002 /* Save some cycles and don't copy the address if not needed */
2003 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2004 /* We assume all kernel code knows the size of sockaddr_storage */
2005 msg.msg_namelen = 0;
2006 msg.msg_iocb = NULL;
2008 if (sock->file->f_flags & O_NONBLOCK)
2009 flags |= MSG_DONTWAIT;
2010 err = sock_recvmsg(sock, &msg, flags);
2012 if (err >= 0 && addr != NULL) {
2013 err2 = move_addr_to_user(&address,
2014 msg.msg_namelen, addr, addr_len);
2019 fput_light(sock->file, fput_needed);
2024 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2025 unsigned int, flags, struct sockaddr __user *, addr,
2026 int __user *, addr_len)
2028 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2032 * Receive a datagram from a socket.
2035 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2036 unsigned int, flags)
2038 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2042 * Set a socket option. Because we don't know the option lengths we have
2043 * to pass the user mode parameter for the protocols to sort out.
2046 static int __sys_setsockopt(int fd, int level, int optname,
2047 char __user *optval, int optlen)
2049 mm_segment_t oldfs = get_fs();
2050 char *kernel_optval = NULL;
2051 int err, fput_needed;
2052 struct socket *sock;
2057 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2059 err = security_socket_setsockopt(sock, level, optname);
2063 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2064 &optname, optval, &optlen,
2069 } else if (err > 0) {
2074 if (kernel_optval) {
2076 optval = (char __user __force *)kernel_optval;
2079 if (level == SOL_SOCKET)
2081 sock_setsockopt(sock, level, optname, optval,
2085 sock->ops->setsockopt(sock, level, optname, optval,
2088 if (kernel_optval) {
2090 kfree(kernel_optval);
2093 fput_light(sock->file, fput_needed);
2098 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2099 char __user *, optval, int, optlen)
2101 return __sys_setsockopt(fd, level, optname, optval, optlen);
2105 * Get a socket option. Because we don't know the option lengths we have
2106 * to pass a user mode parameter for the protocols to sort out.
2109 static int __sys_getsockopt(int fd, int level, int optname,
2110 char __user *optval, int __user *optlen)
2112 int err, fput_needed;
2113 struct socket *sock;
2116 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2118 err = security_socket_getsockopt(sock, level, optname);
2122 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2124 if (level == SOL_SOCKET)
2126 sock_getsockopt(sock, level, optname, optval,
2130 sock->ops->getsockopt(sock, level, optname, optval,
2133 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2137 fput_light(sock->file, fput_needed);
2142 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2143 char __user *, optval, int __user *, optlen)
2145 return __sys_getsockopt(fd, level, optname, optval, optlen);
2149 * Shutdown a socket.
2152 int __sys_shutdown(int fd, int how)
2154 int err, fput_needed;
2155 struct socket *sock;
2157 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159 err = security_socket_shutdown(sock, how);
2161 err = sock->ops->shutdown(sock, how);
2162 fput_light(sock->file, fput_needed);
2167 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2169 return __sys_shutdown(fd, how);
2172 /* A couple of helpful macros for getting the address of the 32/64 bit
2173 * fields which are the same type (int / unsigned) on our platforms.
2175 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2176 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2177 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2179 struct used_address {
2180 struct sockaddr_storage name;
2181 unsigned int name_len;
2184 static int copy_msghdr_from_user(struct msghdr *kmsg,
2185 struct user_msghdr __user *umsg,
2186 struct sockaddr __user **save_addr,
2189 struct user_msghdr msg;
2192 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2195 kmsg->msg_control = (void __force *)msg.msg_control;
2196 kmsg->msg_controllen = msg.msg_controllen;
2197 kmsg->msg_flags = msg.msg_flags;
2199 kmsg->msg_namelen = msg.msg_namelen;
2201 kmsg->msg_namelen = 0;
2203 if (kmsg->msg_namelen < 0)
2206 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2207 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2210 *save_addr = msg.msg_name;
2212 if (msg.msg_name && kmsg->msg_namelen) {
2214 err = move_addr_to_kernel(msg.msg_name,
2221 kmsg->msg_name = NULL;
2222 kmsg->msg_namelen = 0;
2225 if (msg.msg_iovlen > UIO_MAXIOV)
2228 kmsg->msg_iocb = NULL;
2230 err = import_iovec(save_addr ? READ : WRITE,
2231 msg.msg_iov, msg.msg_iovlen,
2232 UIO_FASTIOV, iov, &kmsg->msg_iter);
2233 return err < 0 ? err : 0;
2236 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2237 struct msghdr *msg_sys, unsigned int flags,
2238 struct used_address *used_address,
2239 unsigned int allowed_msghdr_flags)
2241 struct compat_msghdr __user *msg_compat =
2242 (struct compat_msghdr __user *)msg;
2243 struct sockaddr_storage address;
2244 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2245 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2246 __aligned(sizeof(__kernel_size_t));
2247 /* 20 is size of ipv6_pktinfo */
2248 unsigned char *ctl_buf = ctl;
2252 msg_sys->msg_name = &address;
2254 if (MSG_CMSG_COMPAT & flags)
2255 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2257 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2263 if (msg_sys->msg_controllen > INT_MAX)
2265 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2266 ctl_len = msg_sys->msg_controllen;
2267 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2269 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2273 ctl_buf = msg_sys->msg_control;
2274 ctl_len = msg_sys->msg_controllen;
2275 } else if (ctl_len) {
2276 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2277 CMSG_ALIGN(sizeof(struct cmsghdr)));
2278 if (ctl_len > sizeof(ctl)) {
2279 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2280 if (ctl_buf == NULL)
2285 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2286 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2287 * checking falls down on this.
2289 if (copy_from_user(ctl_buf,
2290 (void __user __force *)msg_sys->msg_control,
2293 msg_sys->msg_control = ctl_buf;
2295 msg_sys->msg_flags = flags;
2297 if (sock->file->f_flags & O_NONBLOCK)
2298 msg_sys->msg_flags |= MSG_DONTWAIT;
2300 * If this is sendmmsg() and current destination address is same as
2301 * previously succeeded address, omit asking LSM's decision.
2302 * used_address->name_len is initialized to UINT_MAX so that the first
2303 * destination address never matches.
2305 if (used_address && msg_sys->msg_name &&
2306 used_address->name_len == msg_sys->msg_namelen &&
2307 !memcmp(&used_address->name, msg_sys->msg_name,
2308 used_address->name_len)) {
2309 err = sock_sendmsg_nosec(sock, msg_sys);
2312 err = sock_sendmsg(sock, msg_sys);
2314 * If this is sendmmsg() and sending to current destination address was
2315 * successful, remember it.
2317 if (used_address && err >= 0) {
2318 used_address->name_len = msg_sys->msg_namelen;
2319 if (msg_sys->msg_name)
2320 memcpy(&used_address->name, msg_sys->msg_name,
2321 used_address->name_len);
2326 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2333 * BSD sendmsg interface
2335 long __sys_sendmsg_sock(struct socket *sock, struct user_msghdr __user *msg,
2338 struct msghdr msg_sys;
2340 return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2343 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2344 bool forbid_cmsg_compat)
2346 int fput_needed, err;
2347 struct msghdr msg_sys;
2348 struct socket *sock;
2350 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2353 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2357 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2359 fput_light(sock->file, fput_needed);
2364 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2366 return __sys_sendmsg(fd, msg, flags, true);
2370 * Linux sendmmsg interface
2373 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2374 unsigned int flags, bool forbid_cmsg_compat)
2376 int fput_needed, err, datagrams;
2377 struct socket *sock;
2378 struct mmsghdr __user *entry;
2379 struct compat_mmsghdr __user *compat_entry;
2380 struct msghdr msg_sys;
2381 struct used_address used_address;
2382 unsigned int oflags = flags;
2384 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2387 if (vlen > UIO_MAXIOV)
2392 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2396 used_address.name_len = UINT_MAX;
2398 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2402 while (datagrams < vlen) {
2403 if (datagrams == vlen - 1)
2406 if (MSG_CMSG_COMPAT & flags) {
2407 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2408 &msg_sys, flags, &used_address, MSG_EOR);
2411 err = __put_user(err, &compat_entry->msg_len);
2414 err = ___sys_sendmsg(sock,
2415 (struct user_msghdr __user *)entry,
2416 &msg_sys, flags, &used_address, MSG_EOR);
2419 err = put_user(err, &entry->msg_len);
2426 if (msg_data_left(&msg_sys))
2431 fput_light(sock->file, fput_needed);
2433 /* We only return an error if no datagrams were able to be sent */
2440 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2441 unsigned int, vlen, unsigned int, flags)
2443 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2446 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2447 struct msghdr *msg_sys, unsigned int flags, int nosec)
2449 struct compat_msghdr __user *msg_compat =
2450 (struct compat_msghdr __user *)msg;
2451 struct iovec iovstack[UIO_FASTIOV];
2452 struct iovec *iov = iovstack;
2453 unsigned long cmsg_ptr;
2457 /* kernel mode address */
2458 struct sockaddr_storage addr;
2460 /* user mode address pointers */
2461 struct sockaddr __user *uaddr;
2462 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2464 msg_sys->msg_name = &addr;
2466 if (MSG_CMSG_COMPAT & flags)
2467 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2469 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2473 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2474 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2476 /* We assume all kernel code knows the size of sockaddr_storage */
2477 msg_sys->msg_namelen = 0;
2479 if (sock->file->f_flags & O_NONBLOCK)
2480 flags |= MSG_DONTWAIT;
2481 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2486 if (uaddr != NULL) {
2487 err = move_addr_to_user(&addr,
2488 msg_sys->msg_namelen, uaddr,
2493 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2497 if (MSG_CMSG_COMPAT & flags)
2498 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2499 &msg_compat->msg_controllen);
2501 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2502 &msg->msg_controllen);
2513 * BSD recvmsg interface
2516 long __sys_recvmsg_sock(struct socket *sock, struct user_msghdr __user *msg,
2519 struct msghdr msg_sys;
2521 return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2524 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2525 bool forbid_cmsg_compat)
2527 int fput_needed, err;
2528 struct msghdr msg_sys;
2529 struct socket *sock;
2531 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2534 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2538 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2540 fput_light(sock->file, fput_needed);
2545 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2546 unsigned int, flags)
2548 return __sys_recvmsg(fd, msg, flags, true);
2552 * Linux recvmmsg interface
2555 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2556 unsigned int vlen, unsigned int flags,
2557 struct timespec64 *timeout)
2559 int fput_needed, err, datagrams;
2560 struct socket *sock;
2561 struct mmsghdr __user *entry;
2562 struct compat_mmsghdr __user *compat_entry;
2563 struct msghdr msg_sys;
2564 struct timespec64 end_time;
2565 struct timespec64 timeout64;
2568 poll_select_set_timeout(&end_time, timeout->tv_sec,
2574 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2578 if (likely(!(flags & MSG_ERRQUEUE))) {
2579 err = sock_error(sock->sk);
2587 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2589 while (datagrams < vlen) {
2591 * No need to ask LSM for more than the first datagram.
2593 if (MSG_CMSG_COMPAT & flags) {
2594 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2595 &msg_sys, flags & ~MSG_WAITFORONE,
2599 err = __put_user(err, &compat_entry->msg_len);
2602 err = ___sys_recvmsg(sock,
2603 (struct user_msghdr __user *)entry,
2604 &msg_sys, flags & ~MSG_WAITFORONE,
2608 err = put_user(err, &entry->msg_len);
2616 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2617 if (flags & MSG_WAITFORONE)
2618 flags |= MSG_DONTWAIT;
2621 ktime_get_ts64(&timeout64);
2622 *timeout = timespec64_sub(end_time, timeout64);
2623 if (timeout->tv_sec < 0) {
2624 timeout->tv_sec = timeout->tv_nsec = 0;
2628 /* Timeout, return less than vlen datagrams */
2629 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2633 /* Out of band data, return right away */
2634 if (msg_sys.msg_flags & MSG_OOB)
2642 if (datagrams == 0) {
2648 * We may return less entries than requested (vlen) if the
2649 * sock is non block and there aren't enough datagrams...
2651 if (err != -EAGAIN) {
2653 * ... or if recvmsg returns an error after we
2654 * received some datagrams, where we record the
2655 * error to return on the next call or if the
2656 * app asks about it using getsockopt(SO_ERROR).
2658 sock->sk->sk_err = -err;
2661 fput_light(sock->file, fput_needed);
2666 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2667 unsigned int vlen, unsigned int flags,
2668 struct __kernel_timespec __user *timeout,
2669 struct old_timespec32 __user *timeout32)
2672 struct timespec64 timeout_sys;
2674 if (timeout && get_timespec64(&timeout_sys, timeout))
2677 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2680 if (!timeout && !timeout32)
2681 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2683 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2688 if (timeout && put_timespec64(&timeout_sys, timeout))
2689 datagrams = -EFAULT;
2691 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2692 datagrams = -EFAULT;
2697 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2698 unsigned int, vlen, unsigned int, flags,
2699 struct __kernel_timespec __user *, timeout)
2701 if (flags & MSG_CMSG_COMPAT)
2704 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2707 #ifdef CONFIG_COMPAT_32BIT_TIME
2708 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2709 unsigned int, vlen, unsigned int, flags,
2710 struct old_timespec32 __user *, timeout)
2712 if (flags & MSG_CMSG_COMPAT)
2715 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2719 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2720 /* Argument list sizes for sys_socketcall */
2721 #define AL(x) ((x) * sizeof(unsigned long))
2722 static const unsigned char nargs[21] = {
2723 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2724 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2725 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2732 * System call vectors.
2734 * Argument checking cleaned up. Saved 20% in size.
2735 * This function doesn't need to set the kernel lock because
2736 * it is set by the callees.
2739 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2741 unsigned long a[AUDITSC_ARGS];
2742 unsigned long a0, a1;
2746 if (call < 1 || call > SYS_SENDMMSG)
2748 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2751 if (len > sizeof(a))
2754 /* copy_from_user should be SMP safe. */
2755 if (copy_from_user(a, args, len))
2758 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2767 err = __sys_socket(a0, a1, a[2]);
2770 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2773 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2776 err = __sys_listen(a0, a1);
2779 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2780 (int __user *)a[2], 0);
2782 case SYS_GETSOCKNAME:
2784 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2785 (int __user *)a[2]);
2787 case SYS_GETPEERNAME:
2789 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2790 (int __user *)a[2]);
2792 case SYS_SOCKETPAIR:
2793 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2796 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2800 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2801 (struct sockaddr __user *)a[4], a[5]);
2804 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2808 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2809 (struct sockaddr __user *)a[4],
2810 (int __user *)a[5]);
2813 err = __sys_shutdown(a0, a1);
2815 case SYS_SETSOCKOPT:
2816 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2819 case SYS_GETSOCKOPT:
2821 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2822 (int __user *)a[4]);
2825 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2829 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2833 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2837 if (IS_ENABLED(CONFIG_64BIT) || !IS_ENABLED(CONFIG_64BIT_TIME))
2838 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2840 (struct __kernel_timespec __user *)a[4],
2843 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2845 (struct old_timespec32 __user *)a[4]);
2848 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2849 (int __user *)a[2], a[3]);
2858 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2861 * sock_register - add a socket protocol handler
2862 * @ops: description of protocol
2864 * This function is called by a protocol handler that wants to
2865 * advertise its address family, and have it linked into the
2866 * socket interface. The value ops->family corresponds to the
2867 * socket system call protocol family.
2869 int sock_register(const struct net_proto_family *ops)
2873 if (ops->family >= NPROTO) {
2874 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2878 spin_lock(&net_family_lock);
2879 if (rcu_dereference_protected(net_families[ops->family],
2880 lockdep_is_held(&net_family_lock)))
2883 rcu_assign_pointer(net_families[ops->family], ops);
2886 spin_unlock(&net_family_lock);
2888 pr_info("NET: Registered protocol family %d\n", ops->family);
2891 EXPORT_SYMBOL(sock_register);
2894 * sock_unregister - remove a protocol handler
2895 * @family: protocol family to remove
2897 * This function is called by a protocol handler that wants to
2898 * remove its address family, and have it unlinked from the
2899 * new socket creation.
2901 * If protocol handler is a module, then it can use module reference
2902 * counts to protect against new references. If protocol handler is not
2903 * a module then it needs to provide its own protection in
2904 * the ops->create routine.
2906 void sock_unregister(int family)
2908 BUG_ON(family < 0 || family >= NPROTO);
2910 spin_lock(&net_family_lock);
2911 RCU_INIT_POINTER(net_families[family], NULL);
2912 spin_unlock(&net_family_lock);
2916 pr_info("NET: Unregistered protocol family %d\n", family);
2918 EXPORT_SYMBOL(sock_unregister);
2920 bool sock_is_registered(int family)
2922 return family < NPROTO && rcu_access_pointer(net_families[family]);
2925 static int __init sock_init(void)
2929 * Initialize the network sysctl infrastructure.
2931 err = net_sysctl_init();
2936 * Initialize skbuff SLAB cache
2941 * Initialize the protocols module.
2946 err = register_filesystem(&sock_fs_type);
2949 sock_mnt = kern_mount(&sock_fs_type);
2950 if (IS_ERR(sock_mnt)) {
2951 err = PTR_ERR(sock_mnt);
2955 /* The real protocol initialization is performed in later initcalls.
2958 #ifdef CONFIG_NETFILTER
2959 err = netfilter_init();
2964 ptp_classifier_init();
2970 unregister_filesystem(&sock_fs_type);
2975 core_initcall(sock_init); /* early initcall */
2977 #ifdef CONFIG_PROC_FS
2978 void socket_seq_show(struct seq_file *seq)
2980 seq_printf(seq, "sockets: used %d\n",
2981 sock_inuse_get(seq->private));
2983 #endif /* CONFIG_PROC_FS */
2985 #ifdef CONFIG_COMPAT
2986 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2988 struct compat_ifconf ifc32;
2992 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2995 ifc.ifc_len = ifc32.ifc_len;
2996 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2999 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3004 ifc32.ifc_len = ifc.ifc_len;
3005 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3011 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3013 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3014 bool convert_in = false, convert_out = false;
3015 size_t buf_size = 0;
3016 struct ethtool_rxnfc __user *rxnfc = NULL;
3018 u32 rule_cnt = 0, actual_rule_cnt;
3023 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3026 compat_rxnfc = compat_ptr(data);
3028 if (get_user(ethcmd, &compat_rxnfc->cmd))
3031 /* Most ethtool structures are defined without padding.
3032 * Unfortunately struct ethtool_rxnfc is an exception.
3037 case ETHTOOL_GRXCLSRLALL:
3038 /* Buffer size is variable */
3039 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3041 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3043 buf_size += rule_cnt * sizeof(u32);
3045 case ETHTOOL_GRXRINGS:
3046 case ETHTOOL_GRXCLSRLCNT:
3047 case ETHTOOL_GRXCLSRULE:
3048 case ETHTOOL_SRXCLSRLINS:
3051 case ETHTOOL_SRXCLSRLDEL:
3052 buf_size += sizeof(struct ethtool_rxnfc);
3054 rxnfc = compat_alloc_user_space(buf_size);
3058 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3061 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3064 /* We expect there to be holes between fs.m_ext and
3065 * fs.ring_cookie and at the end of fs, but nowhere else.
3067 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3068 sizeof(compat_rxnfc->fs.m_ext) !=
3069 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3070 sizeof(rxnfc->fs.m_ext));
3072 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3073 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3074 offsetof(struct ethtool_rxnfc, fs.location) -
3075 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3077 if (copy_in_user(rxnfc, compat_rxnfc,
3078 (void __user *)(&rxnfc->fs.m_ext + 1) -
3079 (void __user *)rxnfc) ||
3080 copy_in_user(&rxnfc->fs.ring_cookie,
3081 &compat_rxnfc->fs.ring_cookie,
3082 (void __user *)(&rxnfc->fs.location + 1) -
3083 (void __user *)&rxnfc->fs.ring_cookie))
3085 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3086 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3088 } else if (copy_in_user(&rxnfc->rule_cnt,
3089 &compat_rxnfc->rule_cnt,
3090 sizeof(rxnfc->rule_cnt)))
3094 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3099 if (copy_in_user(compat_rxnfc, rxnfc,
3100 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3101 (const void __user *)rxnfc) ||
3102 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3103 &rxnfc->fs.ring_cookie,
3104 (const void __user *)(&rxnfc->fs.location + 1) -
3105 (const void __user *)&rxnfc->fs.ring_cookie) ||
3106 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3107 sizeof(rxnfc->rule_cnt)))
3110 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3111 /* As an optimisation, we only copy the actual
3112 * number of rules that the underlying
3113 * function returned. Since Mallory might
3114 * change the rule count in user memory, we
3115 * check that it is less than the rule count
3116 * originally given (as the user buffer size),
3117 * which has been range-checked.
3119 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3121 if (actual_rule_cnt < rule_cnt)
3122 rule_cnt = actual_rule_cnt;
3123 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3124 &rxnfc->rule_locs[0],
3125 rule_cnt * sizeof(u32)))
3133 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3135 compat_uptr_t uptr32;
3140 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3143 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3146 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3147 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3149 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3151 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3152 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3158 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3159 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3160 struct compat_ifreq __user *u_ifreq32)
3165 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3167 if (get_user(data32, &u_ifreq32->ifr_data))
3169 ifreq.ifr_data = compat_ptr(data32);
3171 return dev_ioctl(net, cmd, &ifreq, NULL);
3174 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3176 struct compat_ifreq __user *uifr32)
3178 struct ifreq __user *uifr;
3181 /* Handle the fact that while struct ifreq has the same *layout* on
3182 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3183 * which are handled elsewhere, it still has different *size* due to
3184 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3185 * resulting in struct ifreq being 32 and 40 bytes respectively).
3186 * As a result, if the struct happens to be at the end of a page and
3187 * the next page isn't readable/writable, we get a fault. To prevent
3188 * that, copy back and forth to the full size.
3191 uifr = compat_alloc_user_space(sizeof(*uifr));
3192 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3195 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3206 case SIOCGIFBRDADDR:
3207 case SIOCGIFDSTADDR:
3208 case SIOCGIFNETMASK:
3214 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3222 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3223 struct compat_ifreq __user *uifr32)
3226 struct compat_ifmap __user *uifmap32;
3229 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3230 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3231 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3232 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3233 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3234 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3235 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3236 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3240 err = dev_ioctl(net, cmd, &ifr, NULL);
3242 if (cmd == SIOCGIFMAP && !err) {
3243 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3244 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3245 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3246 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3247 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3248 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3249 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3258 struct sockaddr rt_dst; /* target address */
3259 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3260 struct sockaddr rt_genmask; /* target network mask (IP) */
3261 unsigned short rt_flags;
3264 unsigned char rt_tos;
3265 unsigned char rt_class;
3267 short rt_metric; /* +1 for binary compatibility! */
3268 /* char * */ u32 rt_dev; /* forcing the device at add */
3269 u32 rt_mtu; /* per route MTU/Window */
3270 u32 rt_window; /* Window clamping */
3271 unsigned short rt_irtt; /* Initial RTT */
3274 struct in6_rtmsg32 {
3275 struct in6_addr rtmsg_dst;
3276 struct in6_addr rtmsg_src;
3277 struct in6_addr rtmsg_gateway;
3287 static int routing_ioctl(struct net *net, struct socket *sock,
3288 unsigned int cmd, void __user *argp)
3292 struct in6_rtmsg r6;
3296 mm_segment_t old_fs = get_fs();
3298 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3299 struct in6_rtmsg32 __user *ur6 = argp;
3300 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3301 3 * sizeof(struct in6_addr));
3302 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3303 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3304 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3305 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3306 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3307 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3308 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3312 struct rtentry32 __user *ur4 = argp;
3313 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3314 3 * sizeof(struct sockaddr));
3315 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3316 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3317 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3318 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3319 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3320 ret |= get_user(rtdev, &(ur4->rt_dev));
3322 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3323 r4.rt_dev = (char __user __force *)devname;
3337 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3344 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3345 * for some operations; this forces use of the newer bridge-utils that
3346 * use compatible ioctls
3348 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3352 if (get_user(tmp, argp))
3354 if (tmp == BRCTL_GET_VERSION)
3355 return BRCTL_VERSION + 1;
3359 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3360 unsigned int cmd, unsigned long arg)
3362 void __user *argp = compat_ptr(arg);
3363 struct sock *sk = sock->sk;
3364 struct net *net = sock_net(sk);
3366 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3367 return compat_ifr_data_ioctl(net, cmd, argp);
3372 return old_bridge_ioctl(argp);
3374 return compat_dev_ifconf(net, argp);
3376 return ethtool_ioctl(net, argp);
3378 return compat_siocwandev(net, argp);
3381 return compat_sioc_ifmap(net, cmd, argp);
3384 return routing_ioctl(net, sock, cmd, argp);
3385 case SIOCGSTAMP_OLD:
3386 case SIOCGSTAMPNS_OLD:
3387 if (!sock->ops->gettstamp)
3388 return -ENOIOCTLCMD;
3389 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3390 !COMPAT_USE_64BIT_TIME);
3392 case SIOCBONDSLAVEINFOQUERY:
3393 case SIOCBONDINFOQUERY:
3396 return compat_ifr_data_ioctl(net, cmd, argp);
3409 case SIOCGSTAMP_NEW:
3410 case SIOCGSTAMPNS_NEW:
3411 return sock_ioctl(file, cmd, arg);
3428 case SIOCSIFHWBROADCAST:
3430 case SIOCGIFBRDADDR:
3431 case SIOCSIFBRDADDR:
3432 case SIOCGIFDSTADDR:
3433 case SIOCSIFDSTADDR:
3434 case SIOCGIFNETMASK:
3435 case SIOCSIFNETMASK:
3447 case SIOCBONDENSLAVE:
3448 case SIOCBONDRELEASE:
3449 case SIOCBONDSETHWADDR:
3450 case SIOCBONDCHANGEACTIVE:
3451 return compat_ifreq_ioctl(net, sock, cmd, argp);
3457 return sock_do_ioctl(net, sock, cmd, arg);
3460 return -ENOIOCTLCMD;
3463 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3466 struct socket *sock = file->private_data;
3467 int ret = -ENOIOCTLCMD;
3474 if (sock->ops->compat_ioctl)
3475 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3477 if (ret == -ENOIOCTLCMD &&
3478 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3479 ret = compat_wext_handle_ioctl(net, cmd, arg);
3481 if (ret == -ENOIOCTLCMD)
3482 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3489 * kernel_bind - bind an address to a socket (kernel space)
3492 * @addrlen: length of address
3494 * Returns 0 or an error.
3497 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3499 return sock->ops->bind(sock, addr, addrlen);
3501 EXPORT_SYMBOL(kernel_bind);
3504 * kernel_listen - move socket to listening state (kernel space)
3506 * @backlog: pending connections queue size
3508 * Returns 0 or an error.
3511 int kernel_listen(struct socket *sock, int backlog)
3513 return sock->ops->listen(sock, backlog);
3515 EXPORT_SYMBOL(kernel_listen);
3518 * kernel_accept - accept a connection (kernel space)
3519 * @sock: listening socket
3520 * @newsock: new connected socket
3523 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3524 * If it fails, @newsock is guaranteed to be %NULL.
3525 * Returns 0 or an error.
3528 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3530 struct sock *sk = sock->sk;
3533 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3538 err = sock->ops->accept(sock, *newsock, flags, true);
3540 sock_release(*newsock);
3545 (*newsock)->ops = sock->ops;
3546 __module_get((*newsock)->ops->owner);
3551 EXPORT_SYMBOL(kernel_accept);
3554 * kernel_connect - connect a socket (kernel space)
3557 * @addrlen: address length
3558 * @flags: flags (O_NONBLOCK, ...)
3560 * For datagram sockets, @addr is the addres to which datagrams are sent
3561 * by default, and the only address from which datagrams are received.
3562 * For stream sockets, attempts to connect to @addr.
3563 * Returns 0 or an error code.
3566 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3569 return sock->ops->connect(sock, addr, addrlen, flags);
3571 EXPORT_SYMBOL(kernel_connect);
3574 * kernel_getsockname - get the address which the socket is bound (kernel space)
3576 * @addr: address holder
3578 * Fills the @addr pointer with the address which the socket is bound.
3579 * Returns 0 or an error code.
3582 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3584 return sock->ops->getname(sock, addr, 0);
3586 EXPORT_SYMBOL(kernel_getsockname);
3589 * kernel_peername - get the address which the socket is connected (kernel space)
3591 * @addr: address holder
3593 * Fills the @addr pointer with the address which the socket is connected.
3594 * Returns 0 or an error code.
3597 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3599 return sock->ops->getname(sock, addr, 1);
3601 EXPORT_SYMBOL(kernel_getpeername);
3604 * kernel_getsockopt - get a socket option (kernel space)
3606 * @level: API level (SOL_SOCKET, ...)
3607 * @optname: option tag
3608 * @optval: option value
3609 * @optlen: option length
3611 * Assigns the option length to @optlen.
3612 * Returns 0 or an error.
3615 int kernel_getsockopt(struct socket *sock, int level, int optname,
3616 char *optval, int *optlen)
3618 mm_segment_t oldfs = get_fs();
3619 char __user *uoptval;
3620 int __user *uoptlen;
3623 uoptval = (char __user __force *) optval;
3624 uoptlen = (int __user __force *) optlen;
3627 if (level == SOL_SOCKET)
3628 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3630 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3635 EXPORT_SYMBOL(kernel_getsockopt);
3638 * kernel_setsockopt - set a socket option (kernel space)
3640 * @level: API level (SOL_SOCKET, ...)
3641 * @optname: option tag
3642 * @optval: option value
3643 * @optlen: option length
3645 * Returns 0 or an error.
3648 int kernel_setsockopt(struct socket *sock, int level, int optname,
3649 char *optval, unsigned int optlen)
3651 mm_segment_t oldfs = get_fs();
3652 char __user *uoptval;
3655 uoptval = (char __user __force *) optval;
3658 if (level == SOL_SOCKET)
3659 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3661 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3666 EXPORT_SYMBOL(kernel_setsockopt);
3669 * kernel_sendpage - send a &page through a socket (kernel space)
3672 * @offset: page offset
3673 * @size: total size in bytes
3674 * @flags: flags (MSG_DONTWAIT, ...)
3676 * Returns the total amount sent in bytes or an error.
3679 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3680 size_t size, int flags)
3682 if (sock->ops->sendpage)
3683 return sock->ops->sendpage(sock, page, offset, size, flags);
3685 return sock_no_sendpage(sock, page, offset, size, flags);
3687 EXPORT_SYMBOL(kernel_sendpage);
3690 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3693 * @offset: page offset
3694 * @size: total size in bytes
3695 * @flags: flags (MSG_DONTWAIT, ...)
3697 * Returns the total amount sent in bytes or an error.
3698 * Caller must hold @sk.
3701 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3702 size_t size, int flags)
3704 struct socket *sock = sk->sk_socket;
3706 if (sock->ops->sendpage_locked)
3707 return sock->ops->sendpage_locked(sk, page, offset, size,
3710 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3712 EXPORT_SYMBOL(kernel_sendpage_locked);
3715 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3717 * @how: connection part
3719 * Returns 0 or an error.
3722 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3724 return sock->ops->shutdown(sock, how);
3726 EXPORT_SYMBOL(kernel_sock_shutdown);
3729 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3732 * This routine returns the IP overhead imposed by a socket i.e.
3733 * the length of the underlying IP header, depending on whether
3734 * this is an IPv4 or IPv6 socket and the length from IP options turned
3735 * on at the socket. Assumes that the caller has a lock on the socket.
3738 u32 kernel_sock_ip_overhead(struct sock *sk)
3740 struct inet_sock *inet;
3741 struct ip_options_rcu *opt;
3743 #if IS_ENABLED(CONFIG_IPV6)
3744 struct ipv6_pinfo *np;
3745 struct ipv6_txoptions *optv6 = NULL;
3746 #endif /* IS_ENABLED(CONFIG_IPV6) */
3751 switch (sk->sk_family) {
3754 overhead += sizeof(struct iphdr);
3755 opt = rcu_dereference_protected(inet->inet_opt,
3756 sock_owned_by_user(sk));
3758 overhead += opt->opt.optlen;
3760 #if IS_ENABLED(CONFIG_IPV6)
3763 overhead += sizeof(struct ipv6hdr);
3765 optv6 = rcu_dereference_protected(np->opt,
3766 sock_owned_by_user(sk));
3768 overhead += (optv6->opt_flen + optv6->opt_nflen);
3770 #endif /* IS_ENABLED(CONFIG_IPV6) */
3771 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3775 EXPORT_SYMBOL(kernel_sock_ip_overhead);