2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92 #include <linux/nospec.h>
94 #include <linux/uaccess.h>
95 #include <asm/unistd.h>
97 #include <net/compat.h>
99 #include <net/cls_cgroup.h>
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <linux/atalk.h>
109 #include <net/busy_poll.h>
110 #include <linux/errqueue.h>
112 #ifdef CONFIG_NET_RX_BUSY_POLL
113 unsigned int sysctl_net_busy_read __read_mostly;
114 unsigned int sysctl_net_busy_poll __read_mostly;
117 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
118 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 static int sock_close(struct inode *inode, struct file *file);
122 static unsigned int sock_poll(struct file *file,
123 struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 static long compat_sock_ioctl(struct file *file,
127 unsigned int cmd, unsigned long arg);
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131 int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133 struct pipe_inode_info *pipe, size_t len,
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
141 static const struct file_operations socket_file_ops = {
142 .owner = THIS_MODULE,
144 .read_iter = sock_read_iter,
145 .write_iter = sock_write_iter,
147 .unlocked_ioctl = sock_ioctl,
149 .compat_ioctl = compat_sock_ioctl,
152 .release = sock_close,
153 .fasync = sock_fasync,
154 .sendpage = sock_sendpage,
155 .splice_write = generic_splice_sendpage,
156 .splice_read = sock_splice_read,
160 * The protocol list. Each protocol is registered in here.
163 static DEFINE_SPINLOCK(net_family_lock);
164 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
167 * Statistics counters of the socket lists
170 static DEFINE_PER_CPU(int, sockets_in_use);
174 * Move socket addresses back and forth across the kernel/user
175 * divide and look after the messy bits.
179 * move_addr_to_kernel - copy a socket address into kernel space
180 * @uaddr: Address in user space
181 * @kaddr: Address in kernel space
182 * @ulen: Length in user space
184 * The address is copied into kernel space. If the provided address is
185 * too long an error code of -EINVAL is returned. If the copy gives
186 * invalid addresses -EFAULT is returned. On a success 0 is returned.
189 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195 if (copy_from_user(kaddr, uaddr, ulen))
197 return audit_sockaddr(ulen, kaddr);
201 * move_addr_to_user - copy an address to user space
202 * @kaddr: kernel space address
203 * @klen: length of address in kernel
204 * @uaddr: user space address
205 * @ulen: pointer to user length field
207 * The value pointed to by ulen on entry is the buffer length available.
208 * This is overwritten with the buffer space used. -EINVAL is returned
209 * if an overlong buffer is specified or a negative buffer size. -EFAULT
210 * is returned if either the buffer or the length field are not
212 * After copying the data up to the limit the user specifies, the true
213 * length of the data is written over the length limit the user
214 * specified. Zero is returned for a success.
217 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
218 void __user *uaddr, int __user *ulen)
223 BUG_ON(klen > sizeof(struct sockaddr_storage));
224 err = get_user(len, ulen);
232 if (audit_sockaddr(klen, kaddr))
234 if (copy_to_user(uaddr, kaddr, len))
238 * "fromlen shall refer to the value before truncation.."
241 return __put_user(klen, ulen);
244 static struct kmem_cache *sock_inode_cachep __read_mostly;
246 static struct inode *sock_alloc_inode(struct super_block *sb)
248 struct socket_alloc *ei;
249 struct socket_wq *wq;
251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256 kmem_cache_free(sock_inode_cachep, ei);
259 init_waitqueue_head(&wq->wait);
260 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
270 return &ei->vfs_inode;
273 static void sock_destroy_inode(struct inode *inode)
275 struct socket_alloc *ei;
276 struct socket_wq *wq;
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
279 wq = rcu_dereference_protected(ei->socket.wq, 1);
281 kmem_cache_free(sock_inode_cachep, ei);
284 static void init_once(void *foo)
286 struct socket_alloc *ei = (struct socket_alloc *)foo;
288 inode_init_once(&ei->vfs_inode);
291 static void init_inodecache(void)
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294 sizeof(struct socket_alloc),
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
298 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
300 BUG_ON(sock_inode_cachep == NULL);
303 static const struct super_operations sockfs_ops = {
304 .alloc_inode = sock_alloc_inode,
305 .destroy_inode = sock_destroy_inode,
306 .statfs = simple_statfs,
310 * sockfs_dname() is called from d_path().
312 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
315 d_inode(dentry)->i_ino);
318 static const struct dentry_operations sockfs_dentry_operations = {
319 .d_dname = sockfs_dname,
322 static int sockfs_xattr_get(const struct xattr_handler *handler,
323 struct dentry *dentry, struct inode *inode,
324 const char *suffix, void *value, size_t size)
327 if (dentry->d_name.len + 1 > size)
329 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
331 return dentry->d_name.len + 1;
334 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
335 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
336 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
338 static const struct xattr_handler sockfs_xattr_handler = {
339 .name = XATTR_NAME_SOCKPROTONAME,
340 .get = sockfs_xattr_get,
343 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
344 struct dentry *dentry, struct inode *inode,
345 const char *suffix, const void *value,
346 size_t size, int flags)
348 /* Handled by LSM. */
352 static const struct xattr_handler sockfs_security_xattr_handler = {
353 .prefix = XATTR_SECURITY_PREFIX,
354 .set = sockfs_security_xattr_set,
357 static const struct xattr_handler *sockfs_xattr_handlers[] = {
358 &sockfs_xattr_handler,
359 &sockfs_security_xattr_handler,
363 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
364 int flags, const char *dev_name, void *data)
366 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
367 sockfs_xattr_handlers,
368 &sockfs_dentry_operations, SOCKFS_MAGIC);
371 static struct vfsmount *sock_mnt __read_mostly;
373 static struct file_system_type sock_fs_type = {
375 .mount = sockfs_mount,
376 .kill_sb = kill_anon_super,
380 * Obtains the first available file descriptor and sets it up for use.
382 * These functions create file structures and maps them to fd space
383 * of the current process. On success it returns file descriptor
384 * and file struct implicitly stored in sock->file.
385 * Note that another thread may close file descriptor before we return
386 * from this function. We use the fact that now we do not refer
387 * to socket after mapping. If one day we will need it, this
388 * function will increment ref. count on file by 1.
390 * In any case returned fd MAY BE not valid!
391 * This race condition is unavoidable
392 * with shared fd spaces, we cannot solve it inside kernel,
393 * but we take care of internal coherence yet.
396 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
398 struct qstr name = { .name = "" };
404 name.len = strlen(name.name);
405 } else if (sock->sk) {
406 name.name = sock->sk->sk_prot_creator->name;
407 name.len = strlen(name.name);
409 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
410 if (unlikely(!path.dentry))
411 return ERR_PTR(-ENOMEM);
412 path.mnt = mntget(sock_mnt);
414 d_instantiate(path.dentry, SOCK_INODE(sock));
416 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
419 /* drop dentry, keep inode */
420 ihold(d_inode(path.dentry));
426 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
427 file->private_data = sock;
430 EXPORT_SYMBOL(sock_alloc_file);
432 static int sock_map_fd(struct socket *sock, int flags)
434 struct file *newfile;
435 int fd = get_unused_fd_flags(flags);
436 if (unlikely(fd < 0))
439 newfile = sock_alloc_file(sock, flags, NULL);
440 if (likely(!IS_ERR(newfile))) {
441 fd_install(fd, newfile);
446 return PTR_ERR(newfile);
449 struct socket *sock_from_file(struct file *file, int *err)
451 if (file->f_op == &socket_file_ops)
452 return file->private_data; /* set in sock_map_fd */
457 EXPORT_SYMBOL(sock_from_file);
460 * sockfd_lookup - Go from a file number to its socket slot
462 * @err: pointer to an error code return
464 * The file handle passed in is locked and the socket it is bound
465 * to is returned. If an error occurs the err pointer is overwritten
466 * with a negative errno code and NULL is returned. The function checks
467 * for both invalid handles and passing a handle which is not a socket.
469 * On a success the socket object pointer is returned.
472 struct socket *sockfd_lookup(int fd, int *err)
483 sock = sock_from_file(file, err);
488 EXPORT_SYMBOL(sockfd_lookup);
490 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
492 struct fd f = fdget(fd);
497 sock = sock_from_file(f.file, err);
499 *fput_needed = f.flags;
507 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
513 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
523 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
528 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
535 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
537 int err = simple_setattr(dentry, iattr);
539 if (!err && (iattr->ia_valid & ATTR_UID)) {
540 struct socket *sock = SOCKET_I(d_inode(dentry));
543 sock->sk->sk_uid = iattr->ia_uid;
551 static const struct inode_operations sockfs_inode_ops = {
552 .listxattr = sockfs_listxattr,
553 .setattr = sockfs_setattr,
557 * sock_alloc - allocate a socket
559 * Allocate a new inode and socket object. The two are bound together
560 * and initialised. The socket is then returned. If we are out of inodes
564 struct socket *sock_alloc(void)
569 inode = new_inode_pseudo(sock_mnt->mnt_sb);
573 sock = SOCKET_I(inode);
575 inode->i_ino = get_next_ino();
576 inode->i_mode = S_IFSOCK | S_IRWXUGO;
577 inode->i_uid = current_fsuid();
578 inode->i_gid = current_fsgid();
579 inode->i_op = &sockfs_inode_ops;
581 this_cpu_add(sockets_in_use, 1);
584 EXPORT_SYMBOL(sock_alloc);
587 * sock_release - close a socket
588 * @sock: socket to close
590 * The socket is released from the protocol stack if it has a release
591 * callback, and the inode is then released if the socket is bound to
592 * an inode not a file.
595 static void __sock_release(struct socket *sock, struct inode *inode)
598 struct module *owner = sock->ops->owner;
602 sock->ops->release(sock);
609 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
610 pr_err("%s: fasync list not empty!\n", __func__);
612 this_cpu_sub(sockets_in_use, 1);
614 iput(SOCK_INODE(sock));
620 void sock_release(struct socket *sock)
622 __sock_release(sock, NULL);
624 EXPORT_SYMBOL(sock_release);
626 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
628 u8 flags = *tx_flags;
630 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
631 flags |= SKBTX_HW_TSTAMP;
633 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
634 flags |= SKBTX_SW_TSTAMP;
636 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
637 flags |= SKBTX_SCHED_TSTAMP;
641 EXPORT_SYMBOL(__sock_tx_timestamp);
643 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
645 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
646 BUG_ON(ret == -EIOCBQUEUED);
650 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
652 int err = security_socket_sendmsg(sock, msg,
655 return err ?: sock_sendmsg_nosec(sock, msg);
657 EXPORT_SYMBOL(sock_sendmsg);
659 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
660 struct kvec *vec, size_t num, size_t size)
662 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
663 return sock_sendmsg(sock, msg);
665 EXPORT_SYMBOL(kernel_sendmsg);
667 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
668 struct kvec *vec, size_t num, size_t size)
670 struct socket *sock = sk->sk_socket;
672 if (!sock->ops->sendmsg_locked)
673 return sock_no_sendmsg_locked(sk, msg, size);
675 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
677 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
679 EXPORT_SYMBOL(kernel_sendmsg_locked);
681 static bool skb_is_err_queue(const struct sk_buff *skb)
683 /* pkt_type of skbs enqueued on the error queue are set to
684 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
685 * in recvmsg, since skbs received on a local socket will never
686 * have a pkt_type of PACKET_OUTGOING.
688 return skb->pkt_type == PACKET_OUTGOING;
691 /* On transmit, software and hardware timestamps are returned independently.
692 * As the two skb clones share the hardware timestamp, which may be updated
693 * before the software timestamp is received, a hardware TX timestamp may be
694 * returned only if there is no software TX timestamp. Ignore false software
695 * timestamps, which may be made in the __sock_recv_timestamp() call when the
696 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
697 * hardware timestamp.
699 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
701 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
704 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
706 struct scm_ts_pktinfo ts_pktinfo;
707 struct net_device *orig_dev;
709 if (!skb_mac_header_was_set(skb))
712 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
715 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
717 ts_pktinfo.if_index = orig_dev->ifindex;
720 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
721 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
722 sizeof(ts_pktinfo), &ts_pktinfo);
726 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
728 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
731 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
732 struct scm_timestamping tss;
733 int empty = 1, false_tstamp = 0;
734 struct skb_shared_hwtstamps *shhwtstamps =
737 /* Race occurred between timestamp enabling and packet
738 receiving. Fill in the current time for now. */
739 if (need_software_tstamp && skb->tstamp == 0) {
740 __net_timestamp(skb);
744 if (need_software_tstamp) {
745 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
747 skb_get_timestamp(skb, &tv);
748 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
752 skb_get_timestampns(skb, &ts);
753 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
758 memset(&tss, 0, sizeof(tss));
759 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
760 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
763 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
764 !skb_is_swtx_tstamp(skb, false_tstamp) &&
765 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
767 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
768 !skb_is_err_queue(skb))
769 put_ts_pktinfo(msg, skb);
772 put_cmsg(msg, SOL_SOCKET,
773 SCM_TIMESTAMPING, sizeof(tss), &tss);
775 if (skb_is_err_queue(skb) && skb->len &&
776 SKB_EXT_ERR(skb)->opt_stats)
777 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
778 skb->len, skb->data);
781 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
783 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
788 if (!sock_flag(sk, SOCK_WIFI_STATUS))
790 if (!skb->wifi_acked_valid)
793 ack = skb->wifi_acked;
795 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
797 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
799 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
802 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
803 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
804 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
807 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
810 sock_recv_timestamp(msg, sk, skb);
811 sock_recv_drops(msg, sk, skb);
813 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
815 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
818 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
821 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
823 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
825 return err ?: sock_recvmsg_nosec(sock, msg, flags);
827 EXPORT_SYMBOL(sock_recvmsg);
830 * kernel_recvmsg - Receive a message from a socket (kernel space)
831 * @sock: The socket to receive the message from
832 * @msg: Received message
833 * @vec: Input s/g array for message data
834 * @num: Size of input s/g array
835 * @size: Number of bytes to read
836 * @flags: Message flags (MSG_DONTWAIT, etc...)
838 * On return the msg structure contains the scatter/gather array passed in the
839 * vec argument. The array is modified so that it consists of the unfilled
840 * portion of the original array.
842 * The returned value is the total number of bytes received, or an error.
844 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
845 struct kvec *vec, size_t num, size_t size, int flags)
847 mm_segment_t oldfs = get_fs();
850 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
852 result = sock_recvmsg(sock, msg, flags);
856 EXPORT_SYMBOL(kernel_recvmsg);
858 static ssize_t sock_sendpage(struct file *file, struct page *page,
859 int offset, size_t size, loff_t *ppos, int more)
864 sock = file->private_data;
866 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
867 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
870 return kernel_sendpage(sock, page, offset, size, flags);
873 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
874 struct pipe_inode_info *pipe, size_t len,
877 struct socket *sock = file->private_data;
879 if (unlikely(!sock->ops->splice_read))
882 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
885 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
887 struct file *file = iocb->ki_filp;
888 struct socket *sock = file->private_data;
889 struct msghdr msg = {.msg_iter = *to,
893 if (file->f_flags & O_NONBLOCK)
894 msg.msg_flags = MSG_DONTWAIT;
896 if (iocb->ki_pos != 0)
899 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
902 res = sock_recvmsg(sock, &msg, msg.msg_flags);
907 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
909 struct file *file = iocb->ki_filp;
910 struct socket *sock = file->private_data;
911 struct msghdr msg = {.msg_iter = *from,
915 if (iocb->ki_pos != 0)
918 if (file->f_flags & O_NONBLOCK)
919 msg.msg_flags = MSG_DONTWAIT;
921 if (sock->type == SOCK_SEQPACKET)
922 msg.msg_flags |= MSG_EOR;
924 res = sock_sendmsg(sock, &msg);
925 *from = msg.msg_iter;
930 * Atomic setting of ioctl hooks to avoid race
931 * with module unload.
934 static DEFINE_MUTEX(br_ioctl_mutex);
935 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
937 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
939 mutex_lock(&br_ioctl_mutex);
940 br_ioctl_hook = hook;
941 mutex_unlock(&br_ioctl_mutex);
943 EXPORT_SYMBOL(brioctl_set);
945 static DEFINE_MUTEX(vlan_ioctl_mutex);
946 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
948 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
950 mutex_lock(&vlan_ioctl_mutex);
951 vlan_ioctl_hook = hook;
952 mutex_unlock(&vlan_ioctl_mutex);
954 EXPORT_SYMBOL(vlan_ioctl_set);
956 static DEFINE_MUTEX(dlci_ioctl_mutex);
957 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
959 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
961 mutex_lock(&dlci_ioctl_mutex);
962 dlci_ioctl_hook = hook;
963 mutex_unlock(&dlci_ioctl_mutex);
965 EXPORT_SYMBOL(dlci_ioctl_set);
967 static long sock_do_ioctl(struct net *net, struct socket *sock,
968 unsigned int cmd, unsigned long arg)
971 void __user *argp = (void __user *)arg;
973 err = sock->ops->ioctl(sock, cmd, arg);
976 * If this ioctl is unknown try to hand it down
979 if (err == -ENOIOCTLCMD)
980 err = dev_ioctl(net, cmd, argp);
986 * With an ioctl, arg may well be a user mode pointer, but we don't know
987 * what to do with it - that's up to the protocol still.
990 static struct ns_common *get_net_ns(struct ns_common *ns)
992 return &get_net(container_of(ns, struct net, ns))->ns;
995 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
999 void __user *argp = (void __user *)arg;
1003 sock = file->private_data;
1006 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1007 err = dev_ioctl(net, cmd, argp);
1009 #ifdef CONFIG_WEXT_CORE
1010 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1011 err = dev_ioctl(net, cmd, argp);
1018 if (get_user(pid, (int __user *)argp))
1020 err = f_setown(sock->file, pid, 1);
1024 err = put_user(f_getown(sock->file),
1025 (int __user *)argp);
1033 request_module("bridge");
1035 mutex_lock(&br_ioctl_mutex);
1037 err = br_ioctl_hook(net, cmd, argp);
1038 mutex_unlock(&br_ioctl_mutex);
1043 if (!vlan_ioctl_hook)
1044 request_module("8021q");
1046 mutex_lock(&vlan_ioctl_mutex);
1047 if (vlan_ioctl_hook)
1048 err = vlan_ioctl_hook(net, argp);
1049 mutex_unlock(&vlan_ioctl_mutex);
1054 if (!dlci_ioctl_hook)
1055 request_module("dlci");
1057 mutex_lock(&dlci_ioctl_mutex);
1058 if (dlci_ioctl_hook)
1059 err = dlci_ioctl_hook(cmd, argp);
1060 mutex_unlock(&dlci_ioctl_mutex);
1064 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1067 err = open_related_ns(&net->ns, get_net_ns);
1070 err = sock_do_ioctl(net, sock, cmd, arg);
1076 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1079 struct socket *sock = NULL;
1081 err = security_socket_create(family, type, protocol, 1);
1085 sock = sock_alloc();
1092 err = security_socket_post_create(sock, family, type, protocol, 1);
1104 EXPORT_SYMBOL(sock_create_lite);
1106 /* No kernel lock held - perfect */
1107 static unsigned int sock_poll(struct file *file, poll_table *wait)
1109 unsigned int busy_flag = 0;
1110 struct socket *sock;
1113 * We can't return errors to poll, so it's either yes or no.
1115 sock = file->private_data;
1117 if (sk_can_busy_loop(sock->sk)) {
1118 /* this socket can poll_ll so tell the system call */
1119 busy_flag = POLL_BUSY_LOOP;
1121 /* once, only if requested by syscall */
1122 if (wait && (wait->_key & POLL_BUSY_LOOP))
1123 sk_busy_loop(sock->sk, 1);
1126 return busy_flag | sock->ops->poll(file, sock, wait);
1129 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1131 struct socket *sock = file->private_data;
1133 return sock->ops->mmap(file, sock, vma);
1136 static int sock_close(struct inode *inode, struct file *filp)
1138 __sock_release(SOCKET_I(inode), inode);
1143 * Update the socket async list
1145 * Fasync_list locking strategy.
1147 * 1. fasync_list is modified only under process context socket lock
1148 * i.e. under semaphore.
1149 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1150 * or under socket lock
1153 static int sock_fasync(int fd, struct file *filp, int on)
1155 struct socket *sock = filp->private_data;
1156 struct sock *sk = sock->sk;
1157 struct socket_wq *wq;
1163 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1164 fasync_helper(fd, filp, on, &wq->fasync_list);
1166 if (!wq->fasync_list)
1167 sock_reset_flag(sk, SOCK_FASYNC);
1169 sock_set_flag(sk, SOCK_FASYNC);
1175 /* This function may be called only under rcu_lock */
1177 int sock_wake_async(struct socket_wq *wq, int how, int band)
1179 if (!wq || !wq->fasync_list)
1183 case SOCK_WAKE_WAITD:
1184 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1187 case SOCK_WAKE_SPACE:
1188 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1193 kill_fasync(&wq->fasync_list, SIGIO, band);
1196 kill_fasync(&wq->fasync_list, SIGURG, band);
1201 EXPORT_SYMBOL(sock_wake_async);
1203 int __sock_create(struct net *net, int family, int type, int protocol,
1204 struct socket **res, int kern)
1207 struct socket *sock;
1208 const struct net_proto_family *pf;
1211 * Check protocol is in range
1213 if (family < 0 || family >= NPROTO)
1214 return -EAFNOSUPPORT;
1215 if (type < 0 || type >= SOCK_MAX)
1220 This uglymoron is moved from INET layer to here to avoid
1221 deadlock in module load.
1223 if (family == PF_INET && type == SOCK_PACKET) {
1224 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1229 err = security_socket_create(family, type, protocol, kern);
1234 * Allocate the socket and allow the family to set things up. if
1235 * the protocol is 0, the family is instructed to select an appropriate
1238 sock = sock_alloc();
1240 net_warn_ratelimited("socket: no more sockets\n");
1241 return -ENFILE; /* Not exactly a match, but its the
1242 closest posix thing */
1247 #ifdef CONFIG_MODULES
1248 /* Attempt to load a protocol module if the find failed.
1250 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1251 * requested real, full-featured networking support upon configuration.
1252 * Otherwise module support will break!
1254 if (rcu_access_pointer(net_families[family]) == NULL)
1255 request_module("net-pf-%d", family);
1259 pf = rcu_dereference(net_families[family]);
1260 err = -EAFNOSUPPORT;
1265 * We will call the ->create function, that possibly is in a loadable
1266 * module, so we have to bump that loadable module refcnt first.
1268 if (!try_module_get(pf->owner))
1271 /* Now protected by module ref count */
1274 err = pf->create(net, sock, protocol, kern);
1276 goto out_module_put;
1279 * Now to bump the refcnt of the [loadable] module that owns this
1280 * socket at sock_release time we decrement its refcnt.
1282 if (!try_module_get(sock->ops->owner))
1283 goto out_module_busy;
1286 * Now that we're done with the ->create function, the [loadable]
1287 * module can have its refcnt decremented
1289 module_put(pf->owner);
1290 err = security_socket_post_create(sock, family, type, protocol, kern);
1292 goto out_sock_release;
1298 err = -EAFNOSUPPORT;
1301 module_put(pf->owner);
1308 goto out_sock_release;
1310 EXPORT_SYMBOL(__sock_create);
1312 int sock_create(int family, int type, int protocol, struct socket **res)
1314 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1316 EXPORT_SYMBOL(sock_create);
1318 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1320 return __sock_create(net, family, type, protocol, res, 1);
1322 EXPORT_SYMBOL(sock_create_kern);
1324 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1327 struct socket *sock;
1330 /* Check the SOCK_* constants for consistency. */
1331 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1332 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1333 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1334 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1336 flags = type & ~SOCK_TYPE_MASK;
1337 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1339 type &= SOCK_TYPE_MASK;
1341 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1342 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1344 retval = sock_create(family, type, protocol, &sock);
1348 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1353 /* It may be already another descriptor 8) Not kernel problem. */
1362 * Create a pair of connected sockets.
1365 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1366 int __user *, usockvec)
1368 struct socket *sock1, *sock2;
1370 struct file *newfile1, *newfile2;
1373 flags = type & ~SOCK_TYPE_MASK;
1374 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1376 type &= SOCK_TYPE_MASK;
1378 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1379 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1382 * Obtain the first socket and check if the underlying protocol
1383 * supports the socketpair call.
1386 err = sock_create(family, type, protocol, &sock1);
1390 err = sock_create(family, type, protocol, &sock2);
1394 err = sock1->ops->socketpair(sock1, sock2);
1396 goto out_release_both;
1398 fd1 = get_unused_fd_flags(flags);
1399 if (unlikely(fd1 < 0)) {
1401 goto out_release_both;
1404 fd2 = get_unused_fd_flags(flags);
1405 if (unlikely(fd2 < 0)) {
1407 goto out_put_unused_1;
1410 newfile1 = sock_alloc_file(sock1, flags, NULL);
1411 if (IS_ERR(newfile1)) {
1412 err = PTR_ERR(newfile1);
1413 goto out_put_unused_both;
1416 newfile2 = sock_alloc_file(sock2, flags, NULL);
1417 if (IS_ERR(newfile2)) {
1418 err = PTR_ERR(newfile2);
1422 err = put_user(fd1, &usockvec[0]);
1426 err = put_user(fd2, &usockvec[1]);
1430 audit_fd_pair(fd1, fd2);
1432 fd_install(fd1, newfile1);
1433 fd_install(fd2, newfile2);
1434 /* fd1 and fd2 may be already another descriptors.
1435 * Not kernel problem.
1451 sock_release(sock2);
1454 out_put_unused_both:
1459 sock_release(sock2);
1461 sock_release(sock1);
1467 * Bind a name to a socket. Nothing much to do here since it's
1468 * the protocol's responsibility to handle the local address.
1470 * We move the socket address to kernel space before we call
1471 * the protocol layer (having also checked the address is ok).
1474 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1476 struct socket *sock;
1477 struct sockaddr_storage address;
1478 int err, fput_needed;
1480 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1482 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1484 err = security_socket_bind(sock,
1485 (struct sockaddr *)&address,
1488 err = sock->ops->bind(sock,
1492 fput_light(sock->file, fput_needed);
1498 * Perform a listen. Basically, we allow the protocol to do anything
1499 * necessary for a listen, and if that works, we mark the socket as
1500 * ready for listening.
1503 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1505 struct socket *sock;
1506 int err, fput_needed;
1509 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1511 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1512 if ((unsigned int)backlog > somaxconn)
1513 backlog = somaxconn;
1515 err = security_socket_listen(sock, backlog);
1517 err = sock->ops->listen(sock, backlog);
1519 fput_light(sock->file, fput_needed);
1525 * For accept, we attempt to create a new socket, set up the link
1526 * with the client, wake up the client, then return the new
1527 * connected fd. We collect the address of the connector in kernel
1528 * space and move it to user at the very end. This is unclean because
1529 * we open the socket then return an error.
1531 * 1003.1g adds the ability to recvmsg() to query connection pending
1532 * status to recvmsg. We need to add that support in a way thats
1533 * clean when we restucture accept also.
1536 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1537 int __user *, upeer_addrlen, int, flags)
1539 struct socket *sock, *newsock;
1540 struct file *newfile;
1541 int err, len, newfd, fput_needed;
1542 struct sockaddr_storage address;
1544 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1547 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1548 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1550 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1555 newsock = sock_alloc();
1559 newsock->type = sock->type;
1560 newsock->ops = sock->ops;
1563 * We don't need try_module_get here, as the listening socket (sock)
1564 * has the protocol module (sock->ops->owner) held.
1566 __module_get(newsock->ops->owner);
1568 newfd = get_unused_fd_flags(flags);
1569 if (unlikely(newfd < 0)) {
1571 sock_release(newsock);
1574 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1575 if (IS_ERR(newfile)) {
1576 err = PTR_ERR(newfile);
1577 put_unused_fd(newfd);
1578 sock_release(newsock);
1582 err = security_socket_accept(sock, newsock);
1586 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1590 if (upeer_sockaddr) {
1591 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1593 err = -ECONNABORTED;
1596 err = move_addr_to_user(&address,
1597 len, upeer_sockaddr, upeer_addrlen);
1602 /* File flags are not inherited via accept() unlike another OSes. */
1604 fd_install(newfd, newfile);
1608 fput_light(sock->file, fput_needed);
1613 put_unused_fd(newfd);
1617 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1618 int __user *, upeer_addrlen)
1620 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1624 * Attempt to connect to a socket with the server address. The address
1625 * is in user space so we verify it is OK and move it to kernel space.
1627 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1630 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1631 * other SEQPACKET protocols that take time to connect() as it doesn't
1632 * include the -EINPROGRESS status for such sockets.
1635 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1638 struct socket *sock;
1639 struct sockaddr_storage address;
1640 int err, fput_needed;
1642 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1645 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1650 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1654 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1655 sock->file->f_flags);
1657 fput_light(sock->file, fput_needed);
1663 * Get the local address ('name') of a socket object. Move the obtained
1664 * name to user space.
1667 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1668 int __user *, usockaddr_len)
1670 struct socket *sock;
1671 struct sockaddr_storage address;
1672 int len, err, fput_needed;
1674 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1678 err = security_socket_getsockname(sock);
1682 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1685 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1688 fput_light(sock->file, fput_needed);
1694 * Get the remote address ('name') of a socket object. Move the obtained
1695 * name to user space.
1698 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1699 int __user *, usockaddr_len)
1701 struct socket *sock;
1702 struct sockaddr_storage address;
1703 int len, err, fput_needed;
1705 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707 err = security_socket_getpeername(sock);
1709 fput_light(sock->file, fput_needed);
1714 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1717 err = move_addr_to_user(&address, len, usockaddr,
1719 fput_light(sock->file, fput_needed);
1725 * Send a datagram to a given address. We move the address into kernel
1726 * space and check the user space data area is readable before invoking
1730 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1731 unsigned int, flags, struct sockaddr __user *, addr,
1734 struct socket *sock;
1735 struct sockaddr_storage address;
1741 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1744 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1748 msg.msg_name = NULL;
1749 msg.msg_control = NULL;
1750 msg.msg_controllen = 0;
1751 msg.msg_namelen = 0;
1753 err = move_addr_to_kernel(addr, addr_len, &address);
1756 msg.msg_name = (struct sockaddr *)&address;
1757 msg.msg_namelen = addr_len;
1759 if (sock->file->f_flags & O_NONBLOCK)
1760 flags |= MSG_DONTWAIT;
1761 msg.msg_flags = flags;
1762 err = sock_sendmsg(sock, &msg);
1765 fput_light(sock->file, fput_needed);
1771 * Send a datagram down a socket.
1774 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1775 unsigned int, flags)
1777 return sys_sendto(fd, buff, len, flags, NULL, 0);
1781 * Receive a frame from the socket and optionally record the address of the
1782 * sender. We verify the buffers are writable and if needed move the
1783 * sender address from kernel to user space.
1786 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1787 unsigned int, flags, struct sockaddr __user *, addr,
1788 int __user *, addr_len)
1790 struct socket *sock;
1793 struct sockaddr_storage address;
1797 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1800 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1804 msg.msg_control = NULL;
1805 msg.msg_controllen = 0;
1806 /* Save some cycles and don't copy the address if not needed */
1807 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1808 /* We assume all kernel code knows the size of sockaddr_storage */
1809 msg.msg_namelen = 0;
1810 msg.msg_iocb = NULL;
1812 if (sock->file->f_flags & O_NONBLOCK)
1813 flags |= MSG_DONTWAIT;
1814 err = sock_recvmsg(sock, &msg, flags);
1816 if (err >= 0 && addr != NULL) {
1817 err2 = move_addr_to_user(&address,
1818 msg.msg_namelen, addr, addr_len);
1823 fput_light(sock->file, fput_needed);
1829 * Receive a datagram from a socket.
1832 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1833 unsigned int, flags)
1835 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1839 * Set a socket option. Because we don't know the option lengths we have
1840 * to pass the user mode parameter for the protocols to sort out.
1843 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1844 char __user *, optval, int, optlen)
1846 int err, fput_needed;
1847 struct socket *sock;
1852 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1854 err = security_socket_setsockopt(sock, level, optname);
1858 if (level == SOL_SOCKET)
1860 sock_setsockopt(sock, level, optname, optval,
1864 sock->ops->setsockopt(sock, level, optname, optval,
1867 fput_light(sock->file, fput_needed);
1873 * Get a socket option. Because we don't know the option lengths we have
1874 * to pass a user mode parameter for the protocols to sort out.
1877 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1878 char __user *, optval, int __user *, optlen)
1880 int err, fput_needed;
1881 struct socket *sock;
1883 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1885 err = security_socket_getsockopt(sock, level, optname);
1889 if (level == SOL_SOCKET)
1891 sock_getsockopt(sock, level, optname, optval,
1895 sock->ops->getsockopt(sock, level, optname, optval,
1898 fput_light(sock->file, fput_needed);
1904 * Shutdown a socket.
1907 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1909 int err, fput_needed;
1910 struct socket *sock;
1912 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1914 err = security_socket_shutdown(sock, how);
1916 err = sock->ops->shutdown(sock, how);
1917 fput_light(sock->file, fput_needed);
1922 /* A couple of helpful macros for getting the address of the 32/64 bit
1923 * fields which are the same type (int / unsigned) on our platforms.
1925 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1926 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1927 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1929 struct used_address {
1930 struct sockaddr_storage name;
1931 unsigned int name_len;
1934 static int copy_msghdr_from_user(struct msghdr *kmsg,
1935 struct user_msghdr __user *umsg,
1936 struct sockaddr __user **save_addr,
1939 struct user_msghdr msg;
1942 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1945 kmsg->msg_control = (void __force *)msg.msg_control;
1946 kmsg->msg_controllen = msg.msg_controllen;
1947 kmsg->msg_flags = msg.msg_flags;
1949 kmsg->msg_namelen = msg.msg_namelen;
1951 kmsg->msg_namelen = 0;
1953 if (kmsg->msg_namelen < 0)
1956 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1957 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1960 *save_addr = msg.msg_name;
1962 if (msg.msg_name && kmsg->msg_namelen) {
1964 err = move_addr_to_kernel(msg.msg_name,
1971 kmsg->msg_name = NULL;
1972 kmsg->msg_namelen = 0;
1975 if (msg.msg_iovlen > UIO_MAXIOV)
1978 kmsg->msg_iocb = NULL;
1980 return import_iovec(save_addr ? READ : WRITE,
1981 msg.msg_iov, msg.msg_iovlen,
1982 UIO_FASTIOV, iov, &kmsg->msg_iter);
1985 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1986 struct msghdr *msg_sys, unsigned int flags,
1987 struct used_address *used_address,
1988 unsigned int allowed_msghdr_flags)
1990 struct compat_msghdr __user *msg_compat =
1991 (struct compat_msghdr __user *)msg;
1992 struct sockaddr_storage address;
1993 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1994 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1995 __aligned(sizeof(__kernel_size_t));
1996 /* 20 is size of ipv6_pktinfo */
1997 unsigned char *ctl_buf = ctl;
2001 msg_sys->msg_name = &address;
2003 if (MSG_CMSG_COMPAT & flags)
2004 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2006 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2012 if (msg_sys->msg_controllen > INT_MAX)
2014 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2015 ctl_len = msg_sys->msg_controllen;
2016 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2018 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2022 ctl_buf = msg_sys->msg_control;
2023 ctl_len = msg_sys->msg_controllen;
2024 } else if (ctl_len) {
2025 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2026 CMSG_ALIGN(sizeof(struct cmsghdr)));
2027 if (ctl_len > sizeof(ctl)) {
2028 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2029 if (ctl_buf == NULL)
2034 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2035 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2036 * checking falls down on this.
2038 if (copy_from_user(ctl_buf,
2039 (void __user __force *)msg_sys->msg_control,
2042 msg_sys->msg_control = ctl_buf;
2044 msg_sys->msg_flags = flags;
2046 if (sock->file->f_flags & O_NONBLOCK)
2047 msg_sys->msg_flags |= MSG_DONTWAIT;
2049 * If this is sendmmsg() and current destination address is same as
2050 * previously succeeded address, omit asking LSM's decision.
2051 * used_address->name_len is initialized to UINT_MAX so that the first
2052 * destination address never matches.
2054 if (used_address && msg_sys->msg_name &&
2055 used_address->name_len == msg_sys->msg_namelen &&
2056 !memcmp(&used_address->name, msg_sys->msg_name,
2057 used_address->name_len)) {
2058 err = sock_sendmsg_nosec(sock, msg_sys);
2061 err = sock_sendmsg(sock, msg_sys);
2063 * If this is sendmmsg() and sending to current destination address was
2064 * successful, remember it.
2066 if (used_address && err >= 0) {
2067 used_address->name_len = msg_sys->msg_namelen;
2068 if (msg_sys->msg_name)
2069 memcpy(&used_address->name, msg_sys->msg_name,
2070 used_address->name_len);
2075 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2082 * BSD sendmsg interface
2085 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2087 int fput_needed, err;
2088 struct msghdr msg_sys;
2089 struct socket *sock;
2091 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2095 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2097 fput_light(sock->file, fput_needed);
2102 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2104 if (flags & MSG_CMSG_COMPAT)
2106 return __sys_sendmsg(fd, msg, flags);
2110 * Linux sendmmsg interface
2113 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2116 int fput_needed, err, datagrams;
2117 struct socket *sock;
2118 struct mmsghdr __user *entry;
2119 struct compat_mmsghdr __user *compat_entry;
2120 struct msghdr msg_sys;
2121 struct used_address used_address;
2122 unsigned int oflags = flags;
2124 if (vlen > UIO_MAXIOV)
2129 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2133 used_address.name_len = UINT_MAX;
2135 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2139 while (datagrams < vlen) {
2140 if (datagrams == vlen - 1)
2143 if (MSG_CMSG_COMPAT & flags) {
2144 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2145 &msg_sys, flags, &used_address, MSG_EOR);
2148 err = __put_user(err, &compat_entry->msg_len);
2151 err = ___sys_sendmsg(sock,
2152 (struct user_msghdr __user *)entry,
2153 &msg_sys, flags, &used_address, MSG_EOR);
2156 err = put_user(err, &entry->msg_len);
2163 if (msg_data_left(&msg_sys))
2168 fput_light(sock->file, fput_needed);
2170 /* We only return an error if no datagrams were able to be sent */
2177 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2178 unsigned int, vlen, unsigned int, flags)
2180 if (flags & MSG_CMSG_COMPAT)
2182 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2185 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2186 struct msghdr *msg_sys, unsigned int flags, int nosec)
2188 struct compat_msghdr __user *msg_compat =
2189 (struct compat_msghdr __user *)msg;
2190 struct iovec iovstack[UIO_FASTIOV];
2191 struct iovec *iov = iovstack;
2192 unsigned long cmsg_ptr;
2196 /* kernel mode address */
2197 struct sockaddr_storage addr;
2199 /* user mode address pointers */
2200 struct sockaddr __user *uaddr;
2201 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2203 msg_sys->msg_name = &addr;
2205 if (MSG_CMSG_COMPAT & flags)
2206 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2208 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2212 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2213 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2215 /* We assume all kernel code knows the size of sockaddr_storage */
2216 msg_sys->msg_namelen = 0;
2218 if (sock->file->f_flags & O_NONBLOCK)
2219 flags |= MSG_DONTWAIT;
2220 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2225 if (uaddr != NULL) {
2226 err = move_addr_to_user(&addr,
2227 msg_sys->msg_namelen, uaddr,
2232 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2236 if (MSG_CMSG_COMPAT & flags)
2237 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2238 &msg_compat->msg_controllen);
2240 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2241 &msg->msg_controllen);
2252 * BSD recvmsg interface
2255 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2257 int fput_needed, err;
2258 struct msghdr msg_sys;
2259 struct socket *sock;
2261 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2265 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2267 fput_light(sock->file, fput_needed);
2272 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2273 unsigned int, flags)
2275 if (flags & MSG_CMSG_COMPAT)
2277 return __sys_recvmsg(fd, msg, flags);
2281 * Linux recvmmsg interface
2284 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2285 unsigned int flags, struct timespec *timeout)
2287 int fput_needed, err, datagrams;
2288 struct socket *sock;
2289 struct mmsghdr __user *entry;
2290 struct compat_mmsghdr __user *compat_entry;
2291 struct msghdr msg_sys;
2292 struct timespec64 end_time;
2293 struct timespec64 timeout64;
2296 poll_select_set_timeout(&end_time, timeout->tv_sec,
2302 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2306 err = sock_error(sock->sk);
2313 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2315 while (datagrams < vlen) {
2317 * No need to ask LSM for more than the first datagram.
2319 if (MSG_CMSG_COMPAT & flags) {
2320 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2321 &msg_sys, flags & ~MSG_WAITFORONE,
2325 err = __put_user(err, &compat_entry->msg_len);
2328 err = ___sys_recvmsg(sock,
2329 (struct user_msghdr __user *)entry,
2330 &msg_sys, flags & ~MSG_WAITFORONE,
2334 err = put_user(err, &entry->msg_len);
2342 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2343 if (flags & MSG_WAITFORONE)
2344 flags |= MSG_DONTWAIT;
2347 ktime_get_ts64(&timeout64);
2348 *timeout = timespec64_to_timespec(
2349 timespec64_sub(end_time, timeout64));
2350 if (timeout->tv_sec < 0) {
2351 timeout->tv_sec = timeout->tv_nsec = 0;
2355 /* Timeout, return less than vlen datagrams */
2356 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2360 /* Out of band data, return right away */
2361 if (msg_sys.msg_flags & MSG_OOB)
2369 if (datagrams == 0) {
2375 * We may return less entries than requested (vlen) if the
2376 * sock is non block and there aren't enough datagrams...
2378 if (err != -EAGAIN) {
2380 * ... or if recvmsg returns an error after we
2381 * received some datagrams, where we record the
2382 * error to return on the next call or if the
2383 * app asks about it using getsockopt(SO_ERROR).
2385 sock->sk->sk_err = -err;
2388 fput_light(sock->file, fput_needed);
2393 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2394 unsigned int, vlen, unsigned int, flags,
2395 struct timespec __user *, timeout)
2398 struct timespec timeout_sys;
2400 if (flags & MSG_CMSG_COMPAT)
2404 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2406 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2409 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2411 if (datagrams > 0 &&
2412 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2413 datagrams = -EFAULT;
2418 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2419 /* Argument list sizes for sys_socketcall */
2420 #define AL(x) ((x) * sizeof(unsigned long))
2421 static const unsigned char nargs[21] = {
2422 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2423 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2424 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2431 * System call vectors.
2433 * Argument checking cleaned up. Saved 20% in size.
2434 * This function doesn't need to set the kernel lock because
2435 * it is set by the callees.
2438 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2440 unsigned long a[AUDITSC_ARGS];
2441 unsigned long a0, a1;
2445 if (call < 1 || call > SYS_SENDMMSG)
2447 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2450 if (len > sizeof(a))
2453 /* copy_from_user should be SMP safe. */
2454 if (copy_from_user(a, args, len))
2457 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2466 err = sys_socket(a0, a1, a[2]);
2469 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2472 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2475 err = sys_listen(a0, a1);
2478 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2479 (int __user *)a[2], 0);
2481 case SYS_GETSOCKNAME:
2483 sys_getsockname(a0, (struct sockaddr __user *)a1,
2484 (int __user *)a[2]);
2486 case SYS_GETPEERNAME:
2488 sys_getpeername(a0, (struct sockaddr __user *)a1,
2489 (int __user *)a[2]);
2491 case SYS_SOCKETPAIR:
2492 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2495 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2498 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2499 (struct sockaddr __user *)a[4], a[5]);
2502 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2505 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2506 (struct sockaddr __user *)a[4],
2507 (int __user *)a[5]);
2510 err = sys_shutdown(a0, a1);
2512 case SYS_SETSOCKOPT:
2513 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2515 case SYS_GETSOCKOPT:
2517 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2518 (int __user *)a[4]);
2521 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2524 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2527 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2530 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2531 (struct timespec __user *)a[4]);
2534 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2535 (int __user *)a[2], a[3]);
2544 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2547 * sock_register - add a socket protocol handler
2548 * @ops: description of protocol
2550 * This function is called by a protocol handler that wants to
2551 * advertise its address family, and have it linked into the
2552 * socket interface. The value ops->family corresponds to the
2553 * socket system call protocol family.
2555 int sock_register(const struct net_proto_family *ops)
2559 if (ops->family >= NPROTO) {
2560 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2564 spin_lock(&net_family_lock);
2565 if (rcu_dereference_protected(net_families[ops->family],
2566 lockdep_is_held(&net_family_lock)))
2569 rcu_assign_pointer(net_families[ops->family], ops);
2572 spin_unlock(&net_family_lock);
2574 pr_info("NET: Registered protocol family %d\n", ops->family);
2577 EXPORT_SYMBOL(sock_register);
2580 * sock_unregister - remove a protocol handler
2581 * @family: protocol family to remove
2583 * This function is called by a protocol handler that wants to
2584 * remove its address family, and have it unlinked from the
2585 * new socket creation.
2587 * If protocol handler is a module, then it can use module reference
2588 * counts to protect against new references. If protocol handler is not
2589 * a module then it needs to provide its own protection in
2590 * the ops->create routine.
2592 void sock_unregister(int family)
2594 BUG_ON(family < 0 || family >= NPROTO);
2596 spin_lock(&net_family_lock);
2597 RCU_INIT_POINTER(net_families[family], NULL);
2598 spin_unlock(&net_family_lock);
2602 pr_info("NET: Unregistered protocol family %d\n", family);
2604 EXPORT_SYMBOL(sock_unregister);
2606 static int __init sock_init(void)
2610 * Initialize the network sysctl infrastructure.
2612 err = net_sysctl_init();
2617 * Initialize skbuff SLAB cache
2622 * Initialize the protocols module.
2627 err = register_filesystem(&sock_fs_type);
2630 sock_mnt = kern_mount(&sock_fs_type);
2631 if (IS_ERR(sock_mnt)) {
2632 err = PTR_ERR(sock_mnt);
2636 /* The real protocol initialization is performed in later initcalls.
2639 #ifdef CONFIG_NETFILTER
2640 err = netfilter_init();
2645 ptp_classifier_init();
2651 unregister_filesystem(&sock_fs_type);
2656 core_initcall(sock_init); /* early initcall */
2658 static int __init jit_init(void)
2660 #ifdef CONFIG_BPF_JIT_ALWAYS_ON
2665 pure_initcall(jit_init);
2667 #ifdef CONFIG_PROC_FS
2668 void socket_seq_show(struct seq_file *seq)
2673 for_each_possible_cpu(cpu)
2674 counter += per_cpu(sockets_in_use, cpu);
2676 /* It can be negative, by the way. 8) */
2680 seq_printf(seq, "sockets: used %d\n", counter);
2682 #endif /* CONFIG_PROC_FS */
2684 #ifdef CONFIG_COMPAT
2685 static int do_siocgstamp(struct net *net, struct socket *sock,
2686 unsigned int cmd, void __user *up)
2688 mm_segment_t old_fs = get_fs();
2693 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2696 err = compat_put_timeval(&ktv, up);
2701 static int do_siocgstampns(struct net *net, struct socket *sock,
2702 unsigned int cmd, void __user *up)
2704 mm_segment_t old_fs = get_fs();
2705 struct timespec kts;
2709 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2712 err = compat_put_timespec(&kts, up);
2717 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2719 struct ifreq __user *uifr;
2722 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2723 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2726 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2730 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2736 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2738 struct compat_ifconf ifc32;
2740 struct ifconf __user *uifc;
2741 struct compat_ifreq __user *ifr32;
2742 struct ifreq __user *ifr;
2746 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2749 memset(&ifc, 0, sizeof(ifc));
2750 if (ifc32.ifcbuf == 0) {
2754 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2756 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2757 sizeof(struct ifreq);
2758 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2760 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2761 ifr32 = compat_ptr(ifc32.ifcbuf);
2762 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2763 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2769 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2772 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2776 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2780 ifr32 = compat_ptr(ifc32.ifcbuf);
2782 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2783 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2784 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2790 if (ifc32.ifcbuf == 0) {
2791 /* Translate from 64-bit structure multiple to
2795 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2800 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2806 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2808 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2809 bool convert_in = false, convert_out = false;
2810 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2811 struct ethtool_rxnfc __user *rxnfc;
2812 struct ifreq __user *ifr;
2813 u32 rule_cnt = 0, actual_rule_cnt;
2818 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2821 compat_rxnfc = compat_ptr(data);
2823 if (get_user(ethcmd, &compat_rxnfc->cmd))
2826 /* Most ethtool structures are defined without padding.
2827 * Unfortunately struct ethtool_rxnfc is an exception.
2832 case ETHTOOL_GRXCLSRLALL:
2833 /* Buffer size is variable */
2834 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2836 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2838 buf_size += rule_cnt * sizeof(u32);
2840 case ETHTOOL_GRXRINGS:
2841 case ETHTOOL_GRXCLSRLCNT:
2842 case ETHTOOL_GRXCLSRULE:
2843 case ETHTOOL_SRXCLSRLINS:
2846 case ETHTOOL_SRXCLSRLDEL:
2847 buf_size += sizeof(struct ethtool_rxnfc);
2852 ifr = compat_alloc_user_space(buf_size);
2853 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2855 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2858 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2859 &ifr->ifr_ifru.ifru_data))
2863 /* We expect there to be holes between fs.m_ext and
2864 * fs.ring_cookie and at the end of fs, but nowhere else.
2866 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2867 sizeof(compat_rxnfc->fs.m_ext) !=
2868 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2869 sizeof(rxnfc->fs.m_ext));
2871 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2872 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2873 offsetof(struct ethtool_rxnfc, fs.location) -
2874 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2876 if (copy_in_user(rxnfc, compat_rxnfc,
2877 (void __user *)(&rxnfc->fs.m_ext + 1) -
2878 (void __user *)rxnfc) ||
2879 copy_in_user(&rxnfc->fs.ring_cookie,
2880 &compat_rxnfc->fs.ring_cookie,
2881 (void __user *)(&rxnfc->fs.location + 1) -
2882 (void __user *)&rxnfc->fs.ring_cookie))
2884 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2885 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2887 } else if (copy_in_user(&rxnfc->rule_cnt,
2888 &compat_rxnfc->rule_cnt,
2889 sizeof(rxnfc->rule_cnt)))
2893 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2898 if (copy_in_user(compat_rxnfc, rxnfc,
2899 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2900 (const void __user *)rxnfc) ||
2901 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2902 &rxnfc->fs.ring_cookie,
2903 (const void __user *)(&rxnfc->fs.location + 1) -
2904 (const void __user *)&rxnfc->fs.ring_cookie) ||
2905 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2906 sizeof(rxnfc->rule_cnt)))
2909 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2910 /* As an optimisation, we only copy the actual
2911 * number of rules that the underlying
2912 * function returned. Since Mallory might
2913 * change the rule count in user memory, we
2914 * check that it is less than the rule count
2915 * originally given (as the user buffer size),
2916 * which has been range-checked.
2918 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2920 if (actual_rule_cnt < rule_cnt)
2921 rule_cnt = actual_rule_cnt;
2922 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2923 &rxnfc->rule_locs[0],
2924 rule_cnt * sizeof(u32)))
2932 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2935 compat_uptr_t uptr32;
2936 struct ifreq __user *uifr;
2938 uifr = compat_alloc_user_space(sizeof(*uifr));
2939 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2942 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2945 uptr = compat_ptr(uptr32);
2947 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2950 return dev_ioctl(net, SIOCWANDEV, uifr);
2953 static int bond_ioctl(struct net *net, unsigned int cmd,
2954 struct compat_ifreq __user *ifr32)
2957 mm_segment_t old_fs;
2961 case SIOCBONDENSLAVE:
2962 case SIOCBONDRELEASE:
2963 case SIOCBONDSETHWADDR:
2964 case SIOCBONDCHANGEACTIVE:
2965 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2970 err = dev_ioctl(net, cmd,
2971 (struct ifreq __user __force *) &kifr);
2976 return -ENOIOCTLCMD;
2980 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2981 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2982 struct compat_ifreq __user *u_ifreq32)
2984 struct ifreq __user *u_ifreq64;
2985 char tmp_buf[IFNAMSIZ];
2986 void __user *data64;
2989 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2992 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2994 data64 = compat_ptr(data32);
2996 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2998 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3001 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3004 return dev_ioctl(net, cmd, u_ifreq64);
3007 static int dev_ifsioc(struct net *net, struct socket *sock,
3008 unsigned int cmd, struct compat_ifreq __user *uifr32)
3010 struct ifreq __user *uifr;
3013 uifr = compat_alloc_user_space(sizeof(*uifr));
3014 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3017 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3028 case SIOCGIFBRDADDR:
3029 case SIOCGIFDSTADDR:
3030 case SIOCGIFNETMASK:
3035 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3043 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3044 struct compat_ifreq __user *uifr32)
3047 struct compat_ifmap __user *uifmap32;
3048 mm_segment_t old_fs;
3051 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3052 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3053 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3054 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3055 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3056 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3057 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3058 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3064 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3067 if (cmd == SIOCGIFMAP && !err) {
3068 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3069 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3070 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3071 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3072 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3073 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3074 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3083 struct sockaddr rt_dst; /* target address */
3084 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3085 struct sockaddr rt_genmask; /* target network mask (IP) */
3086 unsigned short rt_flags;
3089 unsigned char rt_tos;
3090 unsigned char rt_class;
3092 short rt_metric; /* +1 for binary compatibility! */
3093 /* char * */ u32 rt_dev; /* forcing the device at add */
3094 u32 rt_mtu; /* per route MTU/Window */
3095 u32 rt_window; /* Window clamping */
3096 unsigned short rt_irtt; /* Initial RTT */
3099 struct in6_rtmsg32 {
3100 struct in6_addr rtmsg_dst;
3101 struct in6_addr rtmsg_src;
3102 struct in6_addr rtmsg_gateway;
3112 static int routing_ioctl(struct net *net, struct socket *sock,
3113 unsigned int cmd, void __user *argp)
3117 struct in6_rtmsg r6;
3121 mm_segment_t old_fs = get_fs();
3123 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3124 struct in6_rtmsg32 __user *ur6 = argp;
3125 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3126 3 * sizeof(struct in6_addr));
3127 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3128 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3129 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3130 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3131 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3132 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3133 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3137 struct rtentry32 __user *ur4 = argp;
3138 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3139 3 * sizeof(struct sockaddr));
3140 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3141 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3142 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3143 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3144 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3145 ret |= get_user(rtdev, &(ur4->rt_dev));
3147 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3148 r4.rt_dev = (char __user __force *)devname;
3162 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3169 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3170 * for some operations; this forces use of the newer bridge-utils that
3171 * use compatible ioctls
3173 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3177 if (get_user(tmp, argp))
3179 if (tmp == BRCTL_GET_VERSION)
3180 return BRCTL_VERSION + 1;
3184 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3185 unsigned int cmd, unsigned long arg)
3187 void __user *argp = compat_ptr(arg);
3188 struct sock *sk = sock->sk;
3189 struct net *net = sock_net(sk);
3191 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3192 return compat_ifr_data_ioctl(net, cmd, argp);
3197 return old_bridge_ioctl(argp);
3199 return dev_ifname32(net, argp);
3201 return dev_ifconf(net, argp);
3203 return ethtool_ioctl(net, argp);
3205 return compat_siocwandev(net, argp);
3208 return compat_sioc_ifmap(net, cmd, argp);
3209 case SIOCBONDENSLAVE:
3210 case SIOCBONDRELEASE:
3211 case SIOCBONDSETHWADDR:
3212 case SIOCBONDCHANGEACTIVE:
3213 return bond_ioctl(net, cmd, argp);
3216 return routing_ioctl(net, sock, cmd, argp);
3218 return do_siocgstamp(net, sock, cmd, argp);
3220 return do_siocgstampns(net, sock, cmd, argp);
3221 case SIOCBONDSLAVEINFOQUERY:
3222 case SIOCBONDINFOQUERY:
3225 return compat_ifr_data_ioctl(net, cmd, argp);
3238 return sock_ioctl(file, cmd, arg);
3255 case SIOCSIFHWBROADCAST:
3257 case SIOCGIFBRDADDR:
3258 case SIOCSIFBRDADDR:
3259 case SIOCGIFDSTADDR:
3260 case SIOCSIFDSTADDR:
3261 case SIOCGIFNETMASK:
3262 case SIOCSIFNETMASK:
3273 return dev_ifsioc(net, sock, cmd, argp);
3279 return sock_do_ioctl(net, sock, cmd, arg);
3282 return -ENOIOCTLCMD;
3285 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3288 struct socket *sock = file->private_data;
3289 int ret = -ENOIOCTLCMD;
3296 if (sock->ops->compat_ioctl)
3297 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3299 if (ret == -ENOIOCTLCMD &&
3300 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3301 ret = compat_wext_handle_ioctl(net, cmd, arg);
3303 if (ret == -ENOIOCTLCMD)
3304 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3310 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3312 return sock->ops->bind(sock, addr, addrlen);
3314 EXPORT_SYMBOL(kernel_bind);
3316 int kernel_listen(struct socket *sock, int backlog)
3318 return sock->ops->listen(sock, backlog);
3320 EXPORT_SYMBOL(kernel_listen);
3322 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3324 struct sock *sk = sock->sk;
3327 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3332 err = sock->ops->accept(sock, *newsock, flags, true);
3334 sock_release(*newsock);
3339 (*newsock)->ops = sock->ops;
3340 __module_get((*newsock)->ops->owner);
3345 EXPORT_SYMBOL(kernel_accept);
3347 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3350 return sock->ops->connect(sock, addr, addrlen, flags);
3352 EXPORT_SYMBOL(kernel_connect);
3354 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3357 return sock->ops->getname(sock, addr, addrlen, 0);
3359 EXPORT_SYMBOL(kernel_getsockname);
3361 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3364 return sock->ops->getname(sock, addr, addrlen, 1);
3366 EXPORT_SYMBOL(kernel_getpeername);
3368 int kernel_getsockopt(struct socket *sock, int level, int optname,
3369 char *optval, int *optlen)
3371 mm_segment_t oldfs = get_fs();
3372 char __user *uoptval;
3373 int __user *uoptlen;
3376 uoptval = (char __user __force *) optval;
3377 uoptlen = (int __user __force *) optlen;
3380 if (level == SOL_SOCKET)
3381 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3383 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3388 EXPORT_SYMBOL(kernel_getsockopt);
3390 int kernel_setsockopt(struct socket *sock, int level, int optname,
3391 char *optval, unsigned int optlen)
3393 mm_segment_t oldfs = get_fs();
3394 char __user *uoptval;
3397 uoptval = (char __user __force *) optval;
3400 if (level == SOL_SOCKET)
3401 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3403 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3408 EXPORT_SYMBOL(kernel_setsockopt);
3410 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3411 size_t size, int flags)
3413 if (sock->ops->sendpage)
3414 return sock->ops->sendpage(sock, page, offset, size, flags);
3416 return sock_no_sendpage(sock, page, offset, size, flags);
3418 EXPORT_SYMBOL(kernel_sendpage);
3420 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3421 size_t size, int flags)
3423 struct socket *sock = sk->sk_socket;
3425 if (sock->ops->sendpage_locked)
3426 return sock->ops->sendpage_locked(sk, page, offset, size,
3429 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3431 EXPORT_SYMBOL(kernel_sendpage_locked);
3433 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3435 mm_segment_t oldfs = get_fs();
3439 err = sock->ops->ioctl(sock, cmd, arg);
3444 EXPORT_SYMBOL(kernel_sock_ioctl);
3446 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3448 return sock->ops->shutdown(sock, how);
3450 EXPORT_SYMBOL(kernel_sock_shutdown);
3452 /* This routine returns the IP overhead imposed by a socket i.e.
3453 * the length of the underlying IP header, depending on whether
3454 * this is an IPv4 or IPv6 socket and the length from IP options turned
3455 * on at the socket. Assumes that the caller has a lock on the socket.
3457 u32 kernel_sock_ip_overhead(struct sock *sk)
3459 struct inet_sock *inet;
3460 struct ip_options_rcu *opt;
3462 #if IS_ENABLED(CONFIG_IPV6)
3463 struct ipv6_pinfo *np;
3464 struct ipv6_txoptions *optv6 = NULL;
3465 #endif /* IS_ENABLED(CONFIG_IPV6) */
3470 switch (sk->sk_family) {
3473 overhead += sizeof(struct iphdr);
3474 opt = rcu_dereference_protected(inet->inet_opt,
3475 sock_owned_by_user(sk));
3477 overhead += opt->opt.optlen;
3479 #if IS_ENABLED(CONFIG_IPV6)
3482 overhead += sizeof(struct ipv6hdr);
3484 optv6 = rcu_dereference_protected(np->opt,
3485 sock_owned_by_user(sk));
3487 overhead += (optv6->opt_flen + optv6->opt_nflen);
3489 #endif /* IS_ENABLED(CONFIG_IPV6) */
3490 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3494 EXPORT_SYMBOL(kernel_sock_ip_overhead);