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 <asm/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 int 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 |
300 if (sock_inode_cachep == NULL)
305 static const struct super_operations sockfs_ops = {
306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
312 * sockfs_dname() is called from d_path().
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 d_inode(dentry)->i_ino);
320 static const struct dentry_operations sockfs_dentry_operations = {
321 .d_dname = sockfs_dname,
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
331 static struct vfsmount *sock_mnt __read_mostly;
333 static struct file_system_type sock_fs_type = {
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
340 * Obtains the first available file descriptor and sets it up for use.
342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
358 struct qstr name = { .name = "" };
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
372 path.mnt = mntget(sock_mnt);
374 d_instantiate(path.dentry, SOCK_INODE(sock));
376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
379 /* drop dentry, keep inode */
380 ihold(d_inode(path.dentry));
386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387 file->private_data = sock;
390 EXPORT_SYMBOL(sock_alloc_file);
392 static int sock_map_fd(struct socket *sock, int flags)
394 struct file *newfile;
395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
399 newfile = sock_alloc_file(sock, flags, NULL);
400 if (likely(!IS_ERR(newfile))) {
401 fd_install(fd, newfile);
406 return PTR_ERR(newfile);
409 struct socket *sock_from_file(struct file *file, int *err)
411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
417 EXPORT_SYMBOL(sock_from_file);
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket *sockfd_lookup(int fd, int *err)
443 sock = sock_from_file(file, err);
448 EXPORT_SYMBOL(sockfd_lookup);
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452 struct fd f = fdget(fd);
457 sock = sock_from_file(f.file, err);
459 *fput_needed = f.flags & FDPUT_FPUT;
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
473 if (!strcmp(name, XATTR_NAME_SOCKPROTONAME)) {
475 if (dentry->d_name.len + 1 > size)
477 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
479 return dentry->d_name.len + 1;
484 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
490 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
500 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
505 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
512 static const struct inode_operations sockfs_inode_ops = {
513 .getxattr = sockfs_getxattr,
514 .listxattr = sockfs_listxattr,
518 * sock_alloc - allocate a socket
520 * Allocate a new inode and socket object. The two are bound together
521 * and initialised. The socket is then returned. If we are out of inodes
525 static struct socket *sock_alloc(void)
530 inode = new_inode_pseudo(sock_mnt->mnt_sb);
534 sock = SOCKET_I(inode);
536 kmemcheck_annotate_bitfield(sock, type);
537 inode->i_ino = get_next_ino();
538 inode->i_mode = S_IFSOCK | S_IRWXUGO;
539 inode->i_uid = current_fsuid();
540 inode->i_gid = current_fsgid();
541 inode->i_op = &sockfs_inode_ops;
543 this_cpu_add(sockets_in_use, 1);
548 * sock_release - close a socket
549 * @sock: socket to close
551 * The socket is released from the protocol stack if it has a release
552 * callback, and the inode is then released if the socket is bound to
553 * an inode not a file.
556 void sock_release(struct socket *sock)
559 struct module *owner = sock->ops->owner;
561 sock->ops->release(sock);
566 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
567 pr_err("%s: fasync list not empty!\n", __func__);
569 this_cpu_sub(sockets_in_use, 1);
571 iput(SOCK_INODE(sock));
576 EXPORT_SYMBOL(sock_release);
578 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
580 u8 flags = *tx_flags;
582 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
583 flags |= SKBTX_HW_TSTAMP;
585 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
586 flags |= SKBTX_SW_TSTAMP;
588 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
589 flags |= SKBTX_SCHED_TSTAMP;
591 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
592 flags |= SKBTX_ACK_TSTAMP;
596 EXPORT_SYMBOL(__sock_tx_timestamp);
598 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
600 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
601 BUG_ON(ret == -EIOCBQUEUED);
605 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
607 int err = security_socket_sendmsg(sock, msg,
610 return err ?: sock_sendmsg_nosec(sock, msg);
612 EXPORT_SYMBOL(sock_sendmsg);
614 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
615 struct kvec *vec, size_t num, size_t size)
617 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
618 return sock_sendmsg(sock, msg);
620 EXPORT_SYMBOL(kernel_sendmsg);
623 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
625 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
628 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
629 struct scm_timestamping tss;
631 struct skb_shared_hwtstamps *shhwtstamps =
634 /* Race occurred between timestamp enabling and packet
635 receiving. Fill in the current time for now. */
636 if (need_software_tstamp && skb->tstamp.tv64 == 0)
637 __net_timestamp(skb);
639 if (need_software_tstamp) {
640 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
642 skb_get_timestamp(skb, &tv);
643 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
647 skb_get_timestampns(skb, &ts);
648 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
653 memset(&tss, 0, sizeof(tss));
654 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
655 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
658 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
659 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
662 put_cmsg(msg, SOL_SOCKET,
663 SCM_TIMESTAMPING, sizeof(tss), &tss);
665 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
667 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
672 if (!sock_flag(sk, SOCK_WIFI_STATUS))
674 if (!skb->wifi_acked_valid)
677 ack = skb->wifi_acked;
679 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
681 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
683 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
686 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
687 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
688 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
691 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
694 sock_recv_timestamp(msg, sk, skb);
695 sock_recv_drops(msg, sk, skb);
697 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
699 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
700 size_t size, int flags)
702 return sock->ops->recvmsg(sock, msg, size, flags);
705 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
708 int err = security_socket_recvmsg(sock, msg, size, flags);
710 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
712 EXPORT_SYMBOL(sock_recvmsg);
715 * kernel_recvmsg - Receive a message from a socket (kernel space)
716 * @sock: The socket to receive the message from
717 * @msg: Received message
718 * @vec: Input s/g array for message data
719 * @num: Size of input s/g array
720 * @size: Number of bytes to read
721 * @flags: Message flags (MSG_DONTWAIT, etc...)
723 * On return the msg structure contains the scatter/gather array passed in the
724 * vec argument. The array is modified so that it consists of the unfilled
725 * portion of the original array.
727 * The returned value is the total number of bytes received, or an error.
729 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
730 struct kvec *vec, size_t num, size_t size, int flags)
732 mm_segment_t oldfs = get_fs();
735 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
737 result = sock_recvmsg(sock, msg, size, flags);
741 EXPORT_SYMBOL(kernel_recvmsg);
743 static ssize_t sock_sendpage(struct file *file, struct page *page,
744 int offset, size_t size, loff_t *ppos, int more)
749 sock = file->private_data;
751 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
752 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
755 return kernel_sendpage(sock, page, offset, size, flags);
758 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
759 struct pipe_inode_info *pipe, size_t len,
762 struct socket *sock = file->private_data;
764 if (unlikely(!sock->ops->splice_read))
767 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
770 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
772 struct file *file = iocb->ki_filp;
773 struct socket *sock = file->private_data;
774 struct msghdr msg = {.msg_iter = *to,
778 if (file->f_flags & O_NONBLOCK)
779 msg.msg_flags = MSG_DONTWAIT;
781 if (iocb->ki_pos != 0)
784 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
787 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
792 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
794 struct file *file = iocb->ki_filp;
795 struct socket *sock = file->private_data;
796 struct msghdr msg = {.msg_iter = *from,
800 if (iocb->ki_pos != 0)
803 if (file->f_flags & O_NONBLOCK)
804 msg.msg_flags = MSG_DONTWAIT;
806 if (sock->type == SOCK_SEQPACKET)
807 msg.msg_flags |= MSG_EOR;
809 res = sock_sendmsg(sock, &msg);
810 *from = msg.msg_iter;
815 * Atomic setting of ioctl hooks to avoid race
816 * with module unload.
819 static DEFINE_MUTEX(br_ioctl_mutex);
820 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
822 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
824 mutex_lock(&br_ioctl_mutex);
825 br_ioctl_hook = hook;
826 mutex_unlock(&br_ioctl_mutex);
828 EXPORT_SYMBOL(brioctl_set);
830 static DEFINE_MUTEX(vlan_ioctl_mutex);
831 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
833 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
835 mutex_lock(&vlan_ioctl_mutex);
836 vlan_ioctl_hook = hook;
837 mutex_unlock(&vlan_ioctl_mutex);
839 EXPORT_SYMBOL(vlan_ioctl_set);
841 static DEFINE_MUTEX(dlci_ioctl_mutex);
842 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
844 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
846 mutex_lock(&dlci_ioctl_mutex);
847 dlci_ioctl_hook = hook;
848 mutex_unlock(&dlci_ioctl_mutex);
850 EXPORT_SYMBOL(dlci_ioctl_set);
852 static long sock_do_ioctl(struct net *net, struct socket *sock,
853 unsigned int cmd, unsigned long arg)
856 void __user *argp = (void __user *)arg;
858 err = sock->ops->ioctl(sock, cmd, arg);
861 * If this ioctl is unknown try to hand it down
864 if (err == -ENOIOCTLCMD)
865 err = dev_ioctl(net, cmd, argp);
871 * With an ioctl, arg may well be a user mode pointer, but we don't know
872 * what to do with it - that's up to the protocol still.
875 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
879 void __user *argp = (void __user *)arg;
883 sock = file->private_data;
886 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
887 err = dev_ioctl(net, cmd, argp);
889 #ifdef CONFIG_WEXT_CORE
890 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
891 err = dev_ioctl(net, cmd, argp);
898 if (get_user(pid, (int __user *)argp))
900 f_setown(sock->file, pid, 1);
905 err = put_user(f_getown(sock->file),
914 request_module("bridge");
916 mutex_lock(&br_ioctl_mutex);
918 err = br_ioctl_hook(net, cmd, argp);
919 mutex_unlock(&br_ioctl_mutex);
924 if (!vlan_ioctl_hook)
925 request_module("8021q");
927 mutex_lock(&vlan_ioctl_mutex);
929 err = vlan_ioctl_hook(net, argp);
930 mutex_unlock(&vlan_ioctl_mutex);
935 if (!dlci_ioctl_hook)
936 request_module("dlci");
938 mutex_lock(&dlci_ioctl_mutex);
940 err = dlci_ioctl_hook(cmd, argp);
941 mutex_unlock(&dlci_ioctl_mutex);
944 err = sock_do_ioctl(net, sock, cmd, arg);
950 int sock_create_lite(int family, int type, int protocol, struct socket **res)
953 struct socket *sock = NULL;
955 err = security_socket_create(family, type, protocol, 1);
966 err = security_socket_post_create(sock, family, type, protocol, 1);
978 EXPORT_SYMBOL(sock_create_lite);
980 /* No kernel lock held - perfect */
981 static unsigned int sock_poll(struct file *file, poll_table *wait)
983 unsigned int busy_flag = 0;
987 * We can't return errors to poll, so it's either yes or no.
989 sock = file->private_data;
991 if (sk_can_busy_loop(sock->sk)) {
992 /* this socket can poll_ll so tell the system call */
993 busy_flag = POLL_BUSY_LOOP;
995 /* once, only if requested by syscall */
996 if (wait && (wait->_key & POLL_BUSY_LOOP))
997 sk_busy_loop(sock->sk, 1);
1000 return busy_flag | sock->ops->poll(file, sock, wait);
1003 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1005 struct socket *sock = file->private_data;
1007 return sock->ops->mmap(file, sock, vma);
1010 static int sock_close(struct inode *inode, struct file *filp)
1012 sock_release(SOCKET_I(inode));
1017 * Update the socket async list
1019 * Fasync_list locking strategy.
1021 * 1. fasync_list is modified only under process context socket lock
1022 * i.e. under semaphore.
1023 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1024 * or under socket lock
1027 static int sock_fasync(int fd, struct file *filp, int on)
1029 struct socket *sock = filp->private_data;
1030 struct sock *sk = sock->sk;
1031 struct socket_wq *wq;
1037 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1038 fasync_helper(fd, filp, on, &wq->fasync_list);
1040 if (!wq->fasync_list)
1041 sock_reset_flag(sk, SOCK_FASYNC);
1043 sock_set_flag(sk, SOCK_FASYNC);
1049 /* This function may be called only under rcu_lock */
1051 int sock_wake_async(struct socket_wq *wq, int how, int band)
1053 if (!wq || !wq->fasync_list)
1057 case SOCK_WAKE_WAITD:
1058 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1061 case SOCK_WAKE_SPACE:
1062 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1067 kill_fasync(&wq->fasync_list, SIGIO, band);
1070 kill_fasync(&wq->fasync_list, SIGURG, band);
1075 EXPORT_SYMBOL(sock_wake_async);
1077 int __sock_create(struct net *net, int family, int type, int protocol,
1078 struct socket **res, int kern)
1081 struct socket *sock;
1082 const struct net_proto_family *pf;
1085 * Check protocol is in range
1087 if (family < 0 || family >= NPROTO)
1088 return -EAFNOSUPPORT;
1089 if (type < 0 || type >= SOCK_MAX)
1094 This uglymoron is moved from INET layer to here to avoid
1095 deadlock in module load.
1097 if (family == PF_INET && type == SOCK_PACKET) {
1101 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1107 err = security_socket_create(family, type, protocol, kern);
1112 * Allocate the socket and allow the family to set things up. if
1113 * the protocol is 0, the family is instructed to select an appropriate
1116 sock = sock_alloc();
1118 net_warn_ratelimited("socket: no more sockets\n");
1119 return -ENFILE; /* Not exactly a match, but its the
1120 closest posix thing */
1125 #ifdef CONFIG_MODULES
1126 /* Attempt to load a protocol module if the find failed.
1128 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1129 * requested real, full-featured networking support upon configuration.
1130 * Otherwise module support will break!
1132 if (rcu_access_pointer(net_families[family]) == NULL)
1133 request_module("net-pf-%d", family);
1137 pf = rcu_dereference(net_families[family]);
1138 err = -EAFNOSUPPORT;
1143 * We will call the ->create function, that possibly is in a loadable
1144 * module, so we have to bump that loadable module refcnt first.
1146 if (!try_module_get(pf->owner))
1149 /* Now protected by module ref count */
1152 err = pf->create(net, sock, protocol, kern);
1154 goto out_module_put;
1157 * Now to bump the refcnt of the [loadable] module that owns this
1158 * socket at sock_release time we decrement its refcnt.
1160 if (!try_module_get(sock->ops->owner))
1161 goto out_module_busy;
1164 * Now that we're done with the ->create function, the [loadable]
1165 * module can have its refcnt decremented
1167 module_put(pf->owner);
1168 err = security_socket_post_create(sock, family, type, protocol, kern);
1170 goto out_sock_release;
1176 err = -EAFNOSUPPORT;
1179 module_put(pf->owner);
1186 goto out_sock_release;
1188 EXPORT_SYMBOL(__sock_create);
1190 int sock_create(int family, int type, int protocol, struct socket **res)
1192 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1194 EXPORT_SYMBOL(sock_create);
1196 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1198 return __sock_create(net, family, type, protocol, res, 1);
1200 EXPORT_SYMBOL(sock_create_kern);
1202 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1205 struct socket *sock;
1208 /* Check the SOCK_* constants for consistency. */
1209 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1210 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1211 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1212 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1214 flags = type & ~SOCK_TYPE_MASK;
1215 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1217 type &= SOCK_TYPE_MASK;
1219 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1220 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1222 retval = sock_create(family, type, protocol, &sock);
1226 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1231 /* It may be already another descriptor 8) Not kernel problem. */
1240 * Create a pair of connected sockets.
1243 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1244 int __user *, usockvec)
1246 struct socket *sock1, *sock2;
1248 struct file *newfile1, *newfile2;
1251 flags = type & ~SOCK_TYPE_MASK;
1252 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1254 type &= SOCK_TYPE_MASK;
1256 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1257 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1260 * Obtain the first socket and check if the underlying protocol
1261 * supports the socketpair call.
1264 err = sock_create(family, type, protocol, &sock1);
1268 err = sock_create(family, type, protocol, &sock2);
1272 err = sock1->ops->socketpair(sock1, sock2);
1274 goto out_release_both;
1276 fd1 = get_unused_fd_flags(flags);
1277 if (unlikely(fd1 < 0)) {
1279 goto out_release_both;
1282 fd2 = get_unused_fd_flags(flags);
1283 if (unlikely(fd2 < 0)) {
1285 goto out_put_unused_1;
1288 newfile1 = sock_alloc_file(sock1, flags, NULL);
1289 if (IS_ERR(newfile1)) {
1290 err = PTR_ERR(newfile1);
1291 goto out_put_unused_both;
1294 newfile2 = sock_alloc_file(sock2, flags, NULL);
1295 if (IS_ERR(newfile2)) {
1296 err = PTR_ERR(newfile2);
1300 err = put_user(fd1, &usockvec[0]);
1304 err = put_user(fd2, &usockvec[1]);
1308 audit_fd_pair(fd1, fd2);
1310 fd_install(fd1, newfile1);
1311 fd_install(fd2, newfile2);
1312 /* fd1 and fd2 may be already another descriptors.
1313 * Not kernel problem.
1329 sock_release(sock2);
1332 out_put_unused_both:
1337 sock_release(sock2);
1339 sock_release(sock1);
1345 * Bind a name to a socket. Nothing much to do here since it's
1346 * the protocol's responsibility to handle the local address.
1348 * We move the socket address to kernel space before we call
1349 * the protocol layer (having also checked the address is ok).
1352 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1354 struct socket *sock;
1355 struct sockaddr_storage address;
1356 int err, fput_needed;
1358 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1360 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1362 err = security_socket_bind(sock,
1363 (struct sockaddr *)&address,
1366 err = sock->ops->bind(sock,
1370 fput_light(sock->file, fput_needed);
1376 * Perform a listen. Basically, we allow the protocol to do anything
1377 * necessary for a listen, and if that works, we mark the socket as
1378 * ready for listening.
1381 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1383 struct socket *sock;
1384 int err, fput_needed;
1387 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1389 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1390 if ((unsigned int)backlog > somaxconn)
1391 backlog = somaxconn;
1393 err = security_socket_listen(sock, backlog);
1395 err = sock->ops->listen(sock, backlog);
1397 fput_light(sock->file, fput_needed);
1403 * For accept, we attempt to create a new socket, set up the link
1404 * with the client, wake up the client, then return the new
1405 * connected fd. We collect the address of the connector in kernel
1406 * space and move it to user at the very end. This is unclean because
1407 * we open the socket then return an error.
1409 * 1003.1g adds the ability to recvmsg() to query connection pending
1410 * status to recvmsg. We need to add that support in a way thats
1411 * clean when we restucture accept also.
1414 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1415 int __user *, upeer_addrlen, int, flags)
1417 struct socket *sock, *newsock;
1418 struct file *newfile;
1419 int err, len, newfd, fput_needed;
1420 struct sockaddr_storage address;
1422 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1425 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1426 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1428 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1433 newsock = sock_alloc();
1437 newsock->type = sock->type;
1438 newsock->ops = sock->ops;
1441 * We don't need try_module_get here, as the listening socket (sock)
1442 * has the protocol module (sock->ops->owner) held.
1444 __module_get(newsock->ops->owner);
1446 newfd = get_unused_fd_flags(flags);
1447 if (unlikely(newfd < 0)) {
1449 sock_release(newsock);
1452 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1453 if (IS_ERR(newfile)) {
1454 err = PTR_ERR(newfile);
1455 put_unused_fd(newfd);
1456 sock_release(newsock);
1460 err = security_socket_accept(sock, newsock);
1464 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1468 if (upeer_sockaddr) {
1469 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1471 err = -ECONNABORTED;
1474 err = move_addr_to_user(&address,
1475 len, upeer_sockaddr, upeer_addrlen);
1480 /* File flags are not inherited via accept() unlike another OSes. */
1482 fd_install(newfd, newfile);
1486 fput_light(sock->file, fput_needed);
1491 put_unused_fd(newfd);
1495 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1496 int __user *, upeer_addrlen)
1498 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1502 * Attempt to connect to a socket with the server address. The address
1503 * is in user space so we verify it is OK and move it to kernel space.
1505 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1508 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1509 * other SEQPACKET protocols that take time to connect() as it doesn't
1510 * include the -EINPROGRESS status for such sockets.
1513 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1516 struct socket *sock;
1517 struct sockaddr_storage address;
1518 int err, fput_needed;
1520 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1523 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1528 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1532 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1533 sock->file->f_flags);
1535 fput_light(sock->file, fput_needed);
1541 * Get the local address ('name') of a socket object. Move the obtained
1542 * name to user space.
1545 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1546 int __user *, usockaddr_len)
1548 struct socket *sock;
1549 struct sockaddr_storage address;
1550 int len, err, fput_needed;
1552 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1556 err = security_socket_getsockname(sock);
1560 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1563 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1566 fput_light(sock->file, fput_needed);
1572 * Get the remote address ('name') of a socket object. Move the obtained
1573 * name to user space.
1576 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1577 int __user *, usockaddr_len)
1579 struct socket *sock;
1580 struct sockaddr_storage address;
1581 int len, err, fput_needed;
1583 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1585 err = security_socket_getpeername(sock);
1587 fput_light(sock->file, fput_needed);
1592 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1595 err = move_addr_to_user(&address, len, usockaddr,
1597 fput_light(sock->file, fput_needed);
1603 * Send a datagram to a given address. We move the address into kernel
1604 * space and check the user space data area is readable before invoking
1608 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1609 unsigned int, flags, struct sockaddr __user *, addr,
1612 struct socket *sock;
1613 struct sockaddr_storage address;
1619 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1622 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1626 msg.msg_name = NULL;
1627 msg.msg_control = NULL;
1628 msg.msg_controllen = 0;
1629 msg.msg_namelen = 0;
1631 err = move_addr_to_kernel(addr, addr_len, &address);
1634 msg.msg_name = (struct sockaddr *)&address;
1635 msg.msg_namelen = addr_len;
1637 if (sock->file->f_flags & O_NONBLOCK)
1638 flags |= MSG_DONTWAIT;
1639 msg.msg_flags = flags;
1640 err = sock_sendmsg(sock, &msg);
1643 fput_light(sock->file, fput_needed);
1649 * Send a datagram down a socket.
1652 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1653 unsigned int, flags)
1655 return sys_sendto(fd, buff, len, flags, NULL, 0);
1659 * Receive a frame from the socket and optionally record the address of the
1660 * sender. We verify the buffers are writable and if needed move the
1661 * sender address from kernel to user space.
1664 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1665 unsigned int, flags, struct sockaddr __user *, addr,
1666 int __user *, addr_len)
1668 struct socket *sock;
1671 struct sockaddr_storage address;
1675 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1678 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1682 msg.msg_control = NULL;
1683 msg.msg_controllen = 0;
1684 /* Save some cycles and don't copy the address if not needed */
1685 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1686 /* We assume all kernel code knows the size of sockaddr_storage */
1687 msg.msg_namelen = 0;
1688 msg.msg_iocb = NULL;
1690 if (sock->file->f_flags & O_NONBLOCK)
1691 flags |= MSG_DONTWAIT;
1692 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1694 if (err >= 0 && addr != NULL) {
1695 err2 = move_addr_to_user(&address,
1696 msg.msg_namelen, addr, addr_len);
1701 fput_light(sock->file, fput_needed);
1707 * Receive a datagram from a socket.
1710 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1711 unsigned int, flags)
1713 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1717 * Set a socket option. Because we don't know the option lengths we have
1718 * to pass the user mode parameter for the protocols to sort out.
1721 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1722 char __user *, optval, int, optlen)
1724 int err, fput_needed;
1725 struct socket *sock;
1730 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1732 err = security_socket_setsockopt(sock, level, optname);
1736 if (level == SOL_SOCKET)
1738 sock_setsockopt(sock, level, optname, optval,
1742 sock->ops->setsockopt(sock, level, optname, optval,
1745 fput_light(sock->file, fput_needed);
1751 * Get a socket option. Because we don't know the option lengths we have
1752 * to pass a user mode parameter for the protocols to sort out.
1755 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1756 char __user *, optval, int __user *, optlen)
1758 int err, fput_needed;
1759 struct socket *sock;
1761 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1763 err = security_socket_getsockopt(sock, level, optname);
1767 if (level == SOL_SOCKET)
1769 sock_getsockopt(sock, level, optname, optval,
1773 sock->ops->getsockopt(sock, level, optname, optval,
1776 fput_light(sock->file, fput_needed);
1782 * Shutdown a socket.
1785 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1787 int err, fput_needed;
1788 struct socket *sock;
1790 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1792 err = security_socket_shutdown(sock, how);
1794 err = sock->ops->shutdown(sock, how);
1795 fput_light(sock->file, fput_needed);
1800 /* A couple of helpful macros for getting the address of the 32/64 bit
1801 * fields which are the same type (int / unsigned) on our platforms.
1803 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1804 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1805 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1807 struct used_address {
1808 struct sockaddr_storage name;
1809 unsigned int name_len;
1812 static int copy_msghdr_from_user(struct msghdr *kmsg,
1813 struct user_msghdr __user *umsg,
1814 struct sockaddr __user **save_addr,
1817 struct sockaddr __user *uaddr;
1818 struct iovec __user *uiov;
1822 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1823 __get_user(uaddr, &umsg->msg_name) ||
1824 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1825 __get_user(uiov, &umsg->msg_iov) ||
1826 __get_user(nr_segs, &umsg->msg_iovlen) ||
1827 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1828 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1829 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1833 kmsg->msg_namelen = 0;
1835 if (kmsg->msg_namelen < 0)
1838 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1839 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1844 if (uaddr && kmsg->msg_namelen) {
1846 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1852 kmsg->msg_name = NULL;
1853 kmsg->msg_namelen = 0;
1856 if (nr_segs > UIO_MAXIOV)
1859 kmsg->msg_iocb = NULL;
1861 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1862 UIO_FASTIOV, iov, &kmsg->msg_iter);
1865 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1866 struct msghdr *msg_sys, unsigned int flags,
1867 struct used_address *used_address)
1869 struct compat_msghdr __user *msg_compat =
1870 (struct compat_msghdr __user *)msg;
1871 struct sockaddr_storage address;
1872 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1873 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1874 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1875 /* 20 is size of ipv6_pktinfo */
1876 unsigned char *ctl_buf = ctl;
1880 msg_sys->msg_name = &address;
1882 if (MSG_CMSG_COMPAT & flags)
1883 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1885 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1891 if (msg_sys->msg_controllen > INT_MAX)
1893 ctl_len = msg_sys->msg_controllen;
1894 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1896 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1900 ctl_buf = msg_sys->msg_control;
1901 ctl_len = msg_sys->msg_controllen;
1902 } else if (ctl_len) {
1903 if (ctl_len > sizeof(ctl)) {
1904 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1905 if (ctl_buf == NULL)
1910 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1911 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1912 * checking falls down on this.
1914 if (copy_from_user(ctl_buf,
1915 (void __user __force *)msg_sys->msg_control,
1918 msg_sys->msg_control = ctl_buf;
1920 msg_sys->msg_flags = flags;
1922 if (sock->file->f_flags & O_NONBLOCK)
1923 msg_sys->msg_flags |= MSG_DONTWAIT;
1925 * If this is sendmmsg() and current destination address is same as
1926 * previously succeeded address, omit asking LSM's decision.
1927 * used_address->name_len is initialized to UINT_MAX so that the first
1928 * destination address never matches.
1930 if (used_address && msg_sys->msg_name &&
1931 used_address->name_len == msg_sys->msg_namelen &&
1932 !memcmp(&used_address->name, msg_sys->msg_name,
1933 used_address->name_len)) {
1934 err = sock_sendmsg_nosec(sock, msg_sys);
1937 err = sock_sendmsg(sock, msg_sys);
1939 * If this is sendmmsg() and sending to current destination address was
1940 * successful, remember it.
1942 if (used_address && err >= 0) {
1943 used_address->name_len = msg_sys->msg_namelen;
1944 if (msg_sys->msg_name)
1945 memcpy(&used_address->name, msg_sys->msg_name,
1946 used_address->name_len);
1951 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1958 * BSD sendmsg interface
1961 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1963 int fput_needed, err;
1964 struct msghdr msg_sys;
1965 struct socket *sock;
1967 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1971 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1973 fput_light(sock->file, fput_needed);
1978 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1980 if (flags & MSG_CMSG_COMPAT)
1982 return __sys_sendmsg(fd, msg, flags);
1986 * Linux sendmmsg interface
1989 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1992 int fput_needed, err, datagrams;
1993 struct socket *sock;
1994 struct mmsghdr __user *entry;
1995 struct compat_mmsghdr __user *compat_entry;
1996 struct msghdr msg_sys;
1997 struct used_address used_address;
1999 if (vlen > UIO_MAXIOV)
2004 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2008 used_address.name_len = UINT_MAX;
2010 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2013 while (datagrams < vlen) {
2014 if (MSG_CMSG_COMPAT & flags) {
2015 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2016 &msg_sys, flags, &used_address);
2019 err = __put_user(err, &compat_entry->msg_len);
2022 err = ___sys_sendmsg(sock,
2023 (struct user_msghdr __user *)entry,
2024 &msg_sys, flags, &used_address);
2027 err = put_user(err, &entry->msg_len);
2034 if (msg_data_left(&msg_sys))
2038 fput_light(sock->file, fput_needed);
2040 /* We only return an error if no datagrams were able to be sent */
2047 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2048 unsigned int, vlen, unsigned int, flags)
2050 if (flags & MSG_CMSG_COMPAT)
2052 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2055 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2056 struct msghdr *msg_sys, unsigned int flags, int nosec)
2058 struct compat_msghdr __user *msg_compat =
2059 (struct compat_msghdr __user *)msg;
2060 struct iovec iovstack[UIO_FASTIOV];
2061 struct iovec *iov = iovstack;
2062 unsigned long cmsg_ptr;
2066 /* kernel mode address */
2067 struct sockaddr_storage addr;
2069 /* user mode address pointers */
2070 struct sockaddr __user *uaddr;
2071 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2073 msg_sys->msg_name = &addr;
2075 if (MSG_CMSG_COMPAT & flags)
2076 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2078 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2081 total_len = iov_iter_count(&msg_sys->msg_iter);
2083 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2084 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2086 /* We assume all kernel code knows the size of sockaddr_storage */
2087 msg_sys->msg_namelen = 0;
2089 if (sock->file->f_flags & O_NONBLOCK)
2090 flags |= MSG_DONTWAIT;
2091 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2097 if (uaddr != NULL) {
2098 err = move_addr_to_user(&addr,
2099 msg_sys->msg_namelen, uaddr,
2104 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2108 if (MSG_CMSG_COMPAT & flags)
2109 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2110 &msg_compat->msg_controllen);
2112 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2113 &msg->msg_controllen);
2124 * BSD recvmsg interface
2127 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2129 int fput_needed, err;
2130 struct msghdr msg_sys;
2131 struct socket *sock;
2133 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2137 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2139 fput_light(sock->file, fput_needed);
2144 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2145 unsigned int, flags)
2147 if (flags & MSG_CMSG_COMPAT)
2149 return __sys_recvmsg(fd, msg, flags);
2153 * Linux recvmmsg interface
2156 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2157 unsigned int flags, struct timespec *timeout)
2159 int fput_needed, err, datagrams;
2160 struct socket *sock;
2161 struct mmsghdr __user *entry;
2162 struct compat_mmsghdr __user *compat_entry;
2163 struct msghdr msg_sys;
2164 struct timespec end_time;
2167 poll_select_set_timeout(&end_time, timeout->tv_sec,
2173 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2177 err = sock_error(sock->sk);
2184 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2186 while (datagrams < vlen) {
2188 * No need to ask LSM for more than the first datagram.
2190 if (MSG_CMSG_COMPAT & flags) {
2191 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2192 &msg_sys, flags & ~MSG_WAITFORONE,
2196 err = __put_user(err, &compat_entry->msg_len);
2199 err = ___sys_recvmsg(sock,
2200 (struct user_msghdr __user *)entry,
2201 &msg_sys, flags & ~MSG_WAITFORONE,
2205 err = put_user(err, &entry->msg_len);
2213 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2214 if (flags & MSG_WAITFORONE)
2215 flags |= MSG_DONTWAIT;
2218 ktime_get_ts(timeout);
2219 *timeout = timespec_sub(end_time, *timeout);
2220 if (timeout->tv_sec < 0) {
2221 timeout->tv_sec = timeout->tv_nsec = 0;
2225 /* Timeout, return less than vlen datagrams */
2226 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2230 /* Out of band data, return right away */
2231 if (msg_sys.msg_flags & MSG_OOB)
2238 if (datagrams == 0) {
2244 * We may return less entries than requested (vlen) if the
2245 * sock is non block and there aren't enough datagrams...
2247 if (err != -EAGAIN) {
2249 * ... or if recvmsg returns an error after we
2250 * received some datagrams, where we record the
2251 * error to return on the next call or if the
2252 * app asks about it using getsockopt(SO_ERROR).
2254 sock->sk->sk_err = -err;
2257 fput_light(sock->file, fput_needed);
2262 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2263 unsigned int, vlen, unsigned int, flags,
2264 struct timespec __user *, timeout)
2267 struct timespec timeout_sys;
2269 if (flags & MSG_CMSG_COMPAT)
2273 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2275 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2278 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2280 if (datagrams > 0 &&
2281 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2282 datagrams = -EFAULT;
2287 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2288 /* Argument list sizes for sys_socketcall */
2289 #define AL(x) ((x) * sizeof(unsigned long))
2290 static const unsigned char nargs[21] = {
2291 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2292 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2293 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2300 * System call vectors.
2302 * Argument checking cleaned up. Saved 20% in size.
2303 * This function doesn't need to set the kernel lock because
2304 * it is set by the callees.
2307 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2309 unsigned long a[AUDITSC_ARGS];
2310 unsigned long a0, a1;
2314 if (call < 1 || call > SYS_SENDMMSG)
2316 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2319 if (len > sizeof(a))
2322 /* copy_from_user should be SMP safe. */
2323 if (copy_from_user(a, args, len))
2326 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2335 err = sys_socket(a0, a1, a[2]);
2338 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2341 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2344 err = sys_listen(a0, a1);
2347 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2348 (int __user *)a[2], 0);
2350 case SYS_GETSOCKNAME:
2352 sys_getsockname(a0, (struct sockaddr __user *)a1,
2353 (int __user *)a[2]);
2355 case SYS_GETPEERNAME:
2357 sys_getpeername(a0, (struct sockaddr __user *)a1,
2358 (int __user *)a[2]);
2360 case SYS_SOCKETPAIR:
2361 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2364 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2367 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2368 (struct sockaddr __user *)a[4], a[5]);
2371 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2374 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2375 (struct sockaddr __user *)a[4],
2376 (int __user *)a[5]);
2379 err = sys_shutdown(a0, a1);
2381 case SYS_SETSOCKOPT:
2382 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2384 case SYS_GETSOCKOPT:
2386 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2387 (int __user *)a[4]);
2390 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2393 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2396 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2399 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2400 (struct timespec __user *)a[4]);
2403 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2404 (int __user *)a[2], a[3]);
2413 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2416 * sock_register - add a socket protocol handler
2417 * @ops: description of protocol
2419 * This function is called by a protocol handler that wants to
2420 * advertise its address family, and have it linked into the
2421 * socket interface. The value ops->family corresponds to the
2422 * socket system call protocol family.
2424 int sock_register(const struct net_proto_family *ops)
2428 if (ops->family >= NPROTO) {
2429 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2433 spin_lock(&net_family_lock);
2434 if (rcu_dereference_protected(net_families[ops->family],
2435 lockdep_is_held(&net_family_lock)))
2438 rcu_assign_pointer(net_families[ops->family], ops);
2441 spin_unlock(&net_family_lock);
2443 pr_info("NET: Registered protocol family %d\n", ops->family);
2446 EXPORT_SYMBOL(sock_register);
2449 * sock_unregister - remove a protocol handler
2450 * @family: protocol family to remove
2452 * This function is called by a protocol handler that wants to
2453 * remove its address family, and have it unlinked from the
2454 * new socket creation.
2456 * If protocol handler is a module, then it can use module reference
2457 * counts to protect against new references. If protocol handler is not
2458 * a module then it needs to provide its own protection in
2459 * the ops->create routine.
2461 void sock_unregister(int family)
2463 BUG_ON(family < 0 || family >= NPROTO);
2465 spin_lock(&net_family_lock);
2466 RCU_INIT_POINTER(net_families[family], NULL);
2467 spin_unlock(&net_family_lock);
2471 pr_info("NET: Unregistered protocol family %d\n", family);
2473 EXPORT_SYMBOL(sock_unregister);
2475 static int __init sock_init(void)
2479 * Initialize the network sysctl infrastructure.
2481 err = net_sysctl_init();
2486 * Initialize skbuff SLAB cache
2491 * Initialize the protocols module.
2496 err = register_filesystem(&sock_fs_type);
2499 sock_mnt = kern_mount(&sock_fs_type);
2500 if (IS_ERR(sock_mnt)) {
2501 err = PTR_ERR(sock_mnt);
2505 /* The real protocol initialization is performed in later initcalls.
2508 #ifdef CONFIG_NETFILTER
2509 err = netfilter_init();
2514 ptp_classifier_init();
2520 unregister_filesystem(&sock_fs_type);
2525 core_initcall(sock_init); /* early initcall */
2527 static int __init jit_init(void)
2529 #ifdef CONFIG_BPF_JIT_ALWAYS_ON
2534 pure_initcall(jit_init);
2536 #ifdef CONFIG_PROC_FS
2537 void socket_seq_show(struct seq_file *seq)
2542 for_each_possible_cpu(cpu)
2543 counter += per_cpu(sockets_in_use, cpu);
2545 /* It can be negative, by the way. 8) */
2549 seq_printf(seq, "sockets: used %d\n", counter);
2551 #endif /* CONFIG_PROC_FS */
2553 #ifdef CONFIG_COMPAT
2554 static int do_siocgstamp(struct net *net, struct socket *sock,
2555 unsigned int cmd, void __user *up)
2557 mm_segment_t old_fs = get_fs();
2562 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2565 err = compat_put_timeval(&ktv, up);
2570 static int do_siocgstampns(struct net *net, struct socket *sock,
2571 unsigned int cmd, void __user *up)
2573 mm_segment_t old_fs = get_fs();
2574 struct timespec kts;
2578 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2581 err = compat_put_timespec(&kts, up);
2586 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2588 struct ifreq __user *uifr;
2591 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2592 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2595 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2599 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2605 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2607 struct compat_ifconf ifc32;
2609 struct ifconf __user *uifc;
2610 struct compat_ifreq __user *ifr32;
2611 struct ifreq __user *ifr;
2615 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2618 memset(&ifc, 0, sizeof(ifc));
2619 if (ifc32.ifcbuf == 0) {
2623 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2625 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2626 sizeof(struct ifreq);
2627 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2629 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2630 ifr32 = compat_ptr(ifc32.ifcbuf);
2631 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2632 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2638 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2641 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2645 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2649 ifr32 = compat_ptr(ifc32.ifcbuf);
2651 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2652 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2653 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2659 if (ifc32.ifcbuf == 0) {
2660 /* Translate from 64-bit structure multiple to
2664 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2669 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2675 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2677 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2678 bool convert_in = false, convert_out = false;
2679 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2680 struct ethtool_rxnfc __user *rxnfc;
2681 struct ifreq __user *ifr;
2682 u32 rule_cnt = 0, actual_rule_cnt;
2687 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2690 compat_rxnfc = compat_ptr(data);
2692 if (get_user(ethcmd, &compat_rxnfc->cmd))
2695 /* Most ethtool structures are defined without padding.
2696 * Unfortunately struct ethtool_rxnfc is an exception.
2701 case ETHTOOL_GRXCLSRLALL:
2702 /* Buffer size is variable */
2703 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2705 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2707 buf_size += rule_cnt * sizeof(u32);
2709 case ETHTOOL_GRXRINGS:
2710 case ETHTOOL_GRXCLSRLCNT:
2711 case ETHTOOL_GRXCLSRULE:
2712 case ETHTOOL_SRXCLSRLINS:
2715 case ETHTOOL_SRXCLSRLDEL:
2716 buf_size += sizeof(struct ethtool_rxnfc);
2721 ifr = compat_alloc_user_space(buf_size);
2722 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2724 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2727 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2728 &ifr->ifr_ifru.ifru_data))
2732 /* We expect there to be holes between fs.m_ext and
2733 * fs.ring_cookie and at the end of fs, but nowhere else.
2735 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2736 sizeof(compat_rxnfc->fs.m_ext) !=
2737 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2738 sizeof(rxnfc->fs.m_ext));
2740 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2741 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2742 offsetof(struct ethtool_rxnfc, fs.location) -
2743 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2745 if (copy_in_user(rxnfc, compat_rxnfc,
2746 (void __user *)(&rxnfc->fs.m_ext + 1) -
2747 (void __user *)rxnfc) ||
2748 copy_in_user(&rxnfc->fs.ring_cookie,
2749 &compat_rxnfc->fs.ring_cookie,
2750 (void __user *)(&rxnfc->fs.location + 1) -
2751 (void __user *)&rxnfc->fs.ring_cookie))
2753 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2754 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2756 } else if (copy_in_user(&rxnfc->rule_cnt,
2757 &compat_rxnfc->rule_cnt,
2758 sizeof(rxnfc->rule_cnt)))
2762 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2767 if (copy_in_user(compat_rxnfc, rxnfc,
2768 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2769 (const void __user *)rxnfc) ||
2770 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2771 &rxnfc->fs.ring_cookie,
2772 (const void __user *)(&rxnfc->fs.location + 1) -
2773 (const void __user *)&rxnfc->fs.ring_cookie) ||
2774 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2775 sizeof(rxnfc->rule_cnt)))
2778 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2779 /* As an optimisation, we only copy the actual
2780 * number of rules that the underlying
2781 * function returned. Since Mallory might
2782 * change the rule count in user memory, we
2783 * check that it is less than the rule count
2784 * originally given (as the user buffer size),
2785 * which has been range-checked.
2787 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2789 if (actual_rule_cnt < rule_cnt)
2790 rule_cnt = actual_rule_cnt;
2791 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2792 &rxnfc->rule_locs[0],
2793 rule_cnt * sizeof(u32)))
2801 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2804 compat_uptr_t uptr32;
2805 struct ifreq __user *uifr;
2807 uifr = compat_alloc_user_space(sizeof(*uifr));
2808 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2811 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2814 uptr = compat_ptr(uptr32);
2816 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2819 return dev_ioctl(net, SIOCWANDEV, uifr);
2822 static int bond_ioctl(struct net *net, unsigned int cmd,
2823 struct compat_ifreq __user *ifr32)
2826 mm_segment_t old_fs;
2830 case SIOCBONDENSLAVE:
2831 case SIOCBONDRELEASE:
2832 case SIOCBONDSETHWADDR:
2833 case SIOCBONDCHANGEACTIVE:
2834 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2839 err = dev_ioctl(net, cmd,
2840 (struct ifreq __user __force *) &kifr);
2845 return -ENOIOCTLCMD;
2849 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2850 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2851 struct compat_ifreq __user *u_ifreq32)
2853 struct ifreq __user *u_ifreq64;
2854 char tmp_buf[IFNAMSIZ];
2855 void __user *data64;
2858 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2861 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2863 data64 = compat_ptr(data32);
2865 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2867 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2870 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2873 return dev_ioctl(net, cmd, u_ifreq64);
2876 static int dev_ifsioc(struct net *net, struct socket *sock,
2877 unsigned int cmd, struct compat_ifreq __user *uifr32)
2879 struct ifreq __user *uifr;
2882 uifr = compat_alloc_user_space(sizeof(*uifr));
2883 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2886 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2897 case SIOCGIFBRDADDR:
2898 case SIOCGIFDSTADDR:
2899 case SIOCGIFNETMASK:
2904 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2912 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2913 struct compat_ifreq __user *uifr32)
2916 struct compat_ifmap __user *uifmap32;
2917 mm_segment_t old_fs;
2920 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2921 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2922 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2923 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2924 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2925 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2926 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2927 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2933 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2936 if (cmd == SIOCGIFMAP && !err) {
2937 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2938 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2939 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2940 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2941 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2942 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2943 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2952 struct sockaddr rt_dst; /* target address */
2953 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2954 struct sockaddr rt_genmask; /* target network mask (IP) */
2955 unsigned short rt_flags;
2958 unsigned char rt_tos;
2959 unsigned char rt_class;
2961 short rt_metric; /* +1 for binary compatibility! */
2962 /* char * */ u32 rt_dev; /* forcing the device at add */
2963 u32 rt_mtu; /* per route MTU/Window */
2964 u32 rt_window; /* Window clamping */
2965 unsigned short rt_irtt; /* Initial RTT */
2968 struct in6_rtmsg32 {
2969 struct in6_addr rtmsg_dst;
2970 struct in6_addr rtmsg_src;
2971 struct in6_addr rtmsg_gateway;
2981 static int routing_ioctl(struct net *net, struct socket *sock,
2982 unsigned int cmd, void __user *argp)
2986 struct in6_rtmsg r6;
2990 mm_segment_t old_fs = get_fs();
2992 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2993 struct in6_rtmsg32 __user *ur6 = argp;
2994 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2995 3 * sizeof(struct in6_addr));
2996 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2997 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2998 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2999 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3000 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3001 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3002 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3006 struct rtentry32 __user *ur4 = argp;
3007 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3008 3 * sizeof(struct sockaddr));
3009 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3010 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3011 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3012 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3013 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3014 ret |= get_user(rtdev, &(ur4->rt_dev));
3016 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3017 r4.rt_dev = (char __user __force *)devname;
3031 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3038 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3039 * for some operations; this forces use of the newer bridge-utils that
3040 * use compatible ioctls
3042 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3046 if (get_user(tmp, argp))
3048 if (tmp == BRCTL_GET_VERSION)
3049 return BRCTL_VERSION + 1;
3053 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3054 unsigned int cmd, unsigned long arg)
3056 void __user *argp = compat_ptr(arg);
3057 struct sock *sk = sock->sk;
3058 struct net *net = sock_net(sk);
3060 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3061 return compat_ifr_data_ioctl(net, cmd, argp);
3066 return old_bridge_ioctl(argp);
3068 return dev_ifname32(net, argp);
3070 return dev_ifconf(net, argp);
3072 return ethtool_ioctl(net, argp);
3074 return compat_siocwandev(net, argp);
3077 return compat_sioc_ifmap(net, cmd, argp);
3078 case SIOCBONDENSLAVE:
3079 case SIOCBONDRELEASE:
3080 case SIOCBONDSETHWADDR:
3081 case SIOCBONDCHANGEACTIVE:
3082 return bond_ioctl(net, cmd, argp);
3085 return routing_ioctl(net, sock, cmd, argp);
3087 return do_siocgstamp(net, sock, cmd, argp);
3089 return do_siocgstampns(net, sock, cmd, argp);
3090 case SIOCBONDSLAVEINFOQUERY:
3091 case SIOCBONDINFOQUERY:
3094 return compat_ifr_data_ioctl(net, cmd, argp);
3106 return sock_ioctl(file, cmd, arg);
3123 case SIOCSIFHWBROADCAST:
3125 case SIOCGIFBRDADDR:
3126 case SIOCSIFBRDADDR:
3127 case SIOCGIFDSTADDR:
3128 case SIOCSIFDSTADDR:
3129 case SIOCGIFNETMASK:
3130 case SIOCSIFNETMASK:
3141 return dev_ifsioc(net, sock, cmd, argp);
3148 return sock_do_ioctl(net, sock, cmd, arg);
3151 return -ENOIOCTLCMD;
3154 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3157 struct socket *sock = file->private_data;
3158 int ret = -ENOIOCTLCMD;
3165 if (sock->ops->compat_ioctl)
3166 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3168 if (ret == -ENOIOCTLCMD &&
3169 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3170 ret = compat_wext_handle_ioctl(net, cmd, arg);
3172 if (ret == -ENOIOCTLCMD)
3173 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3179 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3181 return sock->ops->bind(sock, addr, addrlen);
3183 EXPORT_SYMBOL(kernel_bind);
3185 int kernel_listen(struct socket *sock, int backlog)
3187 return sock->ops->listen(sock, backlog);
3189 EXPORT_SYMBOL(kernel_listen);
3191 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3193 struct sock *sk = sock->sk;
3196 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3201 err = sock->ops->accept(sock, *newsock, flags);
3203 sock_release(*newsock);
3208 (*newsock)->ops = sock->ops;
3209 __module_get((*newsock)->ops->owner);
3214 EXPORT_SYMBOL(kernel_accept);
3216 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3219 return sock->ops->connect(sock, addr, addrlen, flags);
3221 EXPORT_SYMBOL(kernel_connect);
3223 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3226 return sock->ops->getname(sock, addr, addrlen, 0);
3228 EXPORT_SYMBOL(kernel_getsockname);
3230 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3233 return sock->ops->getname(sock, addr, addrlen, 1);
3235 EXPORT_SYMBOL(kernel_getpeername);
3237 int kernel_getsockopt(struct socket *sock, int level, int optname,
3238 char *optval, int *optlen)
3240 mm_segment_t oldfs = get_fs();
3241 char __user *uoptval;
3242 int __user *uoptlen;
3245 uoptval = (char __user __force *) optval;
3246 uoptlen = (int __user __force *) optlen;
3249 if (level == SOL_SOCKET)
3250 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3252 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3257 EXPORT_SYMBOL(kernel_getsockopt);
3259 int kernel_setsockopt(struct socket *sock, int level, int optname,
3260 char *optval, unsigned int optlen)
3262 mm_segment_t oldfs = get_fs();
3263 char __user *uoptval;
3266 uoptval = (char __user __force *) optval;
3269 if (level == SOL_SOCKET)
3270 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3272 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3277 EXPORT_SYMBOL(kernel_setsockopt);
3279 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3280 size_t size, int flags)
3282 if (sock->ops->sendpage)
3283 return sock->ops->sendpage(sock, page, offset, size, flags);
3285 return sock_no_sendpage(sock, page, offset, size, flags);
3287 EXPORT_SYMBOL(kernel_sendpage);
3289 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3291 mm_segment_t oldfs = get_fs();
3295 err = sock->ops->ioctl(sock, cmd, arg);
3300 EXPORT_SYMBOL(kernel_sock_ioctl);
3302 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3304 return sock->ops->shutdown(sock, how);
3306 EXPORT_SYMBOL(kernel_sock_shutdown);
3308 /* This routine returns the IP overhead imposed by a socket i.e.
3309 * the length of the underlying IP header, depending on whether
3310 * this is an IPv4 or IPv6 socket and the length from IP options turned
3311 * on at the socket. Assumes that the caller has a lock on the socket.
3313 u32 kernel_sock_ip_overhead(struct sock *sk)
3315 struct inet_sock *inet;
3316 struct ip_options_rcu *opt;
3319 #if IS_ENABLED(CONFIG_IPV6)
3320 struct ipv6_pinfo *np;
3321 struct ipv6_txoptions *optv6 = NULL;
3322 #endif /* IS_ENABLED(CONFIG_IPV6) */
3327 owned_by_user = sock_owned_by_user(sk);
3328 switch (sk->sk_family) {
3331 overhead += sizeof(struct iphdr);
3332 opt = rcu_dereference_protected(inet->inet_opt,
3335 overhead += opt->opt.optlen;
3337 #if IS_ENABLED(CONFIG_IPV6)
3340 overhead += sizeof(struct ipv6hdr);
3342 optv6 = rcu_dereference_protected(np->opt,
3345 overhead += (optv6->opt_flen + optv6->opt_nflen);
3347 #endif /* IS_ENABLED(CONFIG_IPV6) */
3348 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3352 EXPORT_SYMBOL(kernel_sock_ip_overhead);