2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
21 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
24 * Mark Smith <markzzzsmith@yahoo.com.au>
25 * Use random_ether_addr() for tap MAC address.
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
37 #define DRV_NAME "tun"
38 #define DRV_VERSION "1.6"
39 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
40 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
42 #include <linux/module.h>
43 #include <linux/errno.h>
44 #include <linux/kernel.h>
45 #include <linux/major.h>
46 #include <linux/slab.h>
47 #include <linux/poll.h>
48 #include <linux/fcntl.h>
49 #include <linux/init.h>
50 #include <linux/skbuff.h>
51 #include <linux/netdevice.h>
52 #include <linux/etherdevice.h>
53 #include <linux/miscdevice.h>
54 #include <linux/ethtool.h>
55 #include <linux/rtnetlink.h>
56 #include <linux/compat.h>
58 #include <linux/if_arp.h>
59 #include <linux/if_ether.h>
60 #include <linux/if_tun.h>
61 #include <linux/crc32.h>
62 #include <linux/nsproxy.h>
63 #include <linux/virtio_net.h>
64 #include <linux/rcupdate.h>
65 #include <net/net_namespace.h>
66 #include <net/netns/generic.h>
67 #include <net/rtnetlink.h>
70 #include <asm/system.h>
71 #include <asm/uaccess.h>
73 /* Uncomment to enable debugging */
74 /* #define TUN_DEBUG 1 */
79 #define DBG if(tun->debug)printk
80 #define DBG1 if(debug==2)printk
86 #define FLT_EXACT_COUNT 8
88 unsigned int count; /* Number of addrs. Zero means disabled */
89 u32 mask[2]; /* Mask of the hashed addrs */
90 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
95 struct tun_struct *tun;
102 struct tun_file *tfile;
107 struct net_device *dev;
108 struct fasync_struct *fasync;
110 struct tap_filter txflt;
111 struct socket socket;
123 struct tun_struct *tun;
126 static inline struct tun_sock *tun_sk(struct sock *sk)
128 return container_of(sk, struct tun_sock, sk);
131 static int tun_attach(struct tun_struct *tun, struct file *file)
133 struct tun_file *tfile = file->private_data;
138 netif_tx_lock_bh(tun->dev);
151 tun->socket.file = file;
153 sock_hold(tun->socket.sk);
154 atomic_inc(&tfile->count);
157 netif_tx_unlock_bh(tun->dev);
161 static void __tun_detach(struct tun_struct *tun)
163 /* Detach from net device */
164 netif_tx_lock_bh(tun->dev);
166 tun->socket.file = NULL;
167 netif_tx_unlock_bh(tun->dev);
169 /* Drop read queue */
170 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
172 /* Drop the extra count on the net device */
176 static void tun_detach(struct tun_struct *tun)
183 static struct tun_struct *__tun_get(struct tun_file *tfile)
185 struct tun_struct *tun = NULL;
187 if (atomic_inc_not_zero(&tfile->count))
193 static struct tun_struct *tun_get(struct file *file)
195 return __tun_get(file->private_data);
198 static void tun_put(struct tun_struct *tun)
200 struct tun_file *tfile = tun->tfile;
202 if (atomic_dec_and_test(&tfile->count))
203 tun_detach(tfile->tun);
207 static void addr_hash_set(u32 *mask, const u8 *addr)
209 int n = ether_crc(ETH_ALEN, addr) >> 26;
210 mask[n >> 5] |= (1 << (n & 31));
213 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
215 int n = ether_crc(ETH_ALEN, addr) >> 26;
216 return mask[n >> 5] & (1 << (n & 31));
219 static int update_filter(struct tap_filter *filter, void __user *arg)
221 struct { u8 u[ETH_ALEN]; } *addr;
222 struct tun_filter uf;
223 int err, alen, n, nexact;
225 if (copy_from_user(&uf, arg, sizeof(uf)))
234 alen = ETH_ALEN * uf.count;
235 addr = kmalloc(alen, GFP_KERNEL);
239 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
244 /* The filter is updated without holding any locks. Which is
245 * perfectly safe. We disable it first and in the worst
246 * case we'll accept a few undesired packets. */
250 /* Use first set of addresses as an exact filter */
251 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
252 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
256 /* Remaining multicast addresses are hashed,
257 * unicast will leave the filter disabled. */
258 memset(filter->mask, 0, sizeof(filter->mask));
259 for (; n < uf.count; n++) {
260 if (!is_multicast_ether_addr(addr[n].u)) {
261 err = 0; /* no filter */
264 addr_hash_set(filter->mask, addr[n].u);
267 /* For ALLMULTI just set the mask to all ones.
268 * This overrides the mask populated above. */
269 if ((uf.flags & TUN_FLT_ALLMULTI))
270 memset(filter->mask, ~0, sizeof(filter->mask));
272 /* Now enable the filter */
274 filter->count = nexact;
276 /* Return the number of exact filters */
284 /* Returns: 0 - drop, !=0 - accept */
285 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
287 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
289 struct ethhdr *eh = (struct ethhdr *) skb->data;
293 for (i = 0; i < filter->count; i++)
294 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
297 /* Inexact match (multicast only) */
298 if (is_multicast_ether_addr(eh->h_dest))
299 return addr_hash_test(filter->mask, eh->h_dest);
305 * Checks whether the packet is accepted or not.
306 * Returns: 0 - drop, !=0 - accept
308 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
313 return run_filter(filter, skb);
316 /* Network device part of the driver */
318 static const struct ethtool_ops tun_ethtool_ops;
320 /* Net device detach from fd. */
321 static void tun_net_uninit(struct net_device *dev)
323 struct tun_struct *tun = netdev_priv(dev);
324 struct tun_file *tfile = tun->tfile;
326 /* Inform the methods they need to stop using the dev.
329 wake_up_all(&tun->wq.wait);
330 if (atomic_dec_and_test(&tfile->count))
335 static void tun_free_netdev(struct net_device *dev)
337 struct tun_struct *tun = netdev_priv(dev);
339 sock_put(tun->socket.sk);
342 /* Net device open. */
343 static int tun_net_open(struct net_device *dev)
345 netif_start_queue(dev);
349 /* Net device close. */
350 static int tun_net_close(struct net_device *dev)
352 netif_stop_queue(dev);
356 /* Net device start xmit */
357 static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
359 struct tun_struct *tun = netdev_priv(dev);
361 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
363 /* Drop packet if interface is not attached */
367 /* Drop if the filter does not like it.
368 * This is a noop if the filter is disabled.
369 * Filter can be enabled only for the TAP devices. */
370 if (!check_filter(&tun->txflt, skb))
373 if (tun->socket.sk->sk_filter &&
374 sk_filter(tun->socket.sk, skb))
377 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
378 if (!(tun->flags & TUN_ONE_QUEUE)) {
379 /* Normal queueing mode. */
380 /* Packet scheduler handles dropping of further packets. */
381 netif_stop_queue(dev);
383 /* We won't see all dropped packets individually, so overrun
384 * error is more appropriate. */
385 dev->stats.tx_fifo_errors++;
387 /* Single queue mode.
388 * Driver handles dropping of all packets itself. */
393 /* Orphan the skb - required as we might hang on to it
394 * for indefinite time. */
398 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
400 /* Notify and wake up reader process */
401 if (tun->flags & TUN_FASYNC)
402 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
403 wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
404 POLLRDNORM | POLLRDBAND);
408 dev->stats.tx_dropped++;
413 static void tun_net_mclist(struct net_device *dev)
416 * This callback is supposed to deal with mc filter in
417 * _rx_ path and has nothing to do with the _tx_ path.
418 * In rx path we always accept everything userspace gives us.
423 #define MAX_MTU 65535
426 tun_net_change_mtu(struct net_device *dev, int new_mtu)
428 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
434 static const struct net_device_ops tun_netdev_ops = {
435 .ndo_uninit = tun_net_uninit,
436 .ndo_open = tun_net_open,
437 .ndo_stop = tun_net_close,
438 .ndo_start_xmit = tun_net_xmit,
439 .ndo_change_mtu = tun_net_change_mtu,
442 static const struct net_device_ops tap_netdev_ops = {
443 .ndo_uninit = tun_net_uninit,
444 .ndo_open = tun_net_open,
445 .ndo_stop = tun_net_close,
446 .ndo_start_xmit = tun_net_xmit,
447 .ndo_change_mtu = tun_net_change_mtu,
448 .ndo_set_multicast_list = tun_net_mclist,
449 .ndo_set_mac_address = eth_mac_addr,
450 .ndo_validate_addr = eth_validate_addr,
453 /* Initialize net device. */
454 static void tun_net_init(struct net_device *dev)
456 struct tun_struct *tun = netdev_priv(dev);
458 switch (tun->flags & TUN_TYPE_MASK) {
460 dev->netdev_ops = &tun_netdev_ops;
462 /* Point-to-Point TUN Device */
463 dev->hard_header_len = 0;
467 /* Zero header length */
468 dev->type = ARPHRD_NONE;
469 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
470 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
474 dev->netdev_ops = &tap_netdev_ops;
475 /* Ethernet TAP Device */
478 random_ether_addr(dev->dev_addr);
480 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
485 /* Character device part */
488 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
490 struct tun_file *tfile = file->private_data;
491 struct tun_struct *tun = __tun_get(tfile);
493 unsigned int mask = 0;
500 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
502 poll_wait(file, &tun->wq.wait, wait);
504 if (!skb_queue_empty(&sk->sk_receive_queue))
505 mask |= POLLIN | POLLRDNORM;
507 if (sock_writeable(sk) ||
508 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
510 mask |= POLLOUT | POLLWRNORM;
512 if (tun->dev->reg_state != NETREG_REGISTERED)
519 /* prepad is the amount to reserve at front. len is length after that.
520 * linear is a hint as to how much to copy (usually headers). */
521 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
522 size_t prepad, size_t len,
523 size_t linear, int noblock)
525 struct sock *sk = tun->socket.sk;
529 sock_update_classid(sk);
531 /* Under a page? Don't bother with paged skb. */
532 if (prepad + len < PAGE_SIZE || !linear)
535 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
540 skb_reserve(skb, prepad);
541 skb_put(skb, linear);
542 skb->data_len = len - linear;
543 skb->len += len - linear;
548 /* Get packet from user space buffer */
549 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
550 const struct iovec *iv, size_t count,
553 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
555 size_t len = count, align = 0;
556 struct virtio_net_hdr gso = { 0 };
559 if (!(tun->flags & TUN_NO_PI)) {
560 if ((len -= sizeof(pi)) > count)
563 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
565 offset += sizeof(pi);
568 if (tun->flags & TUN_VNET_HDR) {
569 if ((len -= tun->vnet_hdr_sz) > count)
572 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
575 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
576 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
577 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
579 if (gso.hdr_len > len)
581 offset += tun->vnet_hdr_sz;
584 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
585 align = NET_IP_ALIGN;
586 if (unlikely(len < ETH_HLEN ||
587 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
591 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
593 if (PTR_ERR(skb) != -EAGAIN)
594 tun->dev->stats.rx_dropped++;
598 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
599 tun->dev->stats.rx_dropped++;
604 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
605 if (!skb_partial_csum_set(skb, gso.csum_start,
607 tun->dev->stats.rx_frame_errors++;
611 } else if (tun->flags & TUN_NOCHECKSUM)
612 skb->ip_summed = CHECKSUM_UNNECESSARY;
614 switch (tun->flags & TUN_TYPE_MASK) {
616 if (tun->flags & TUN_NO_PI) {
617 switch (skb->data[0] & 0xf0) {
619 pi.proto = htons(ETH_P_IP);
622 pi.proto = htons(ETH_P_IPV6);
625 tun->dev->stats.rx_dropped++;
631 skb_reset_mac_header(skb);
632 skb->protocol = pi.proto;
636 skb->protocol = eth_type_trans(skb, tun->dev);
640 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
642 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
643 case VIRTIO_NET_HDR_GSO_TCPV4:
644 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
646 case VIRTIO_NET_HDR_GSO_TCPV6:
647 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
649 case VIRTIO_NET_HDR_GSO_UDP:
650 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
653 tun->dev->stats.rx_frame_errors++;
658 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
659 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
661 skb_shinfo(skb)->gso_size = gso.gso_size;
662 if (skb_shinfo(skb)->gso_size == 0) {
663 tun->dev->stats.rx_frame_errors++;
668 /* Header must be checked, and gso_segs computed. */
669 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
670 skb_shinfo(skb)->gso_segs = 0;
675 tun->dev->stats.rx_packets++;
676 tun->dev->stats.rx_bytes += len;
681 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
682 unsigned long count, loff_t pos)
684 struct file *file = iocb->ki_filp;
685 struct tun_struct *tun = tun_get(file);
691 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
693 result = tun_get_user(tun, iv, iov_length(iv, count),
694 file->f_flags & O_NONBLOCK);
700 /* Put packet to the user space buffer */
701 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
703 const struct iovec *iv, int len)
705 struct tun_pi pi = { 0, skb->protocol };
708 if (!(tun->flags & TUN_NO_PI)) {
709 if ((len -= sizeof(pi)) < 0)
712 if (len < skb->len) {
713 /* Packet will be striped */
714 pi.flags |= TUN_PKT_STRIP;
717 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
722 if (tun->flags & TUN_VNET_HDR) {
723 struct virtio_net_hdr gso = { 0 }; /* no info leak */
724 if ((len -= tun->vnet_hdr_sz) < 0)
727 if (skb_is_gso(skb)) {
728 struct skb_shared_info *sinfo = skb_shinfo(skb);
730 /* This is a hint as to how much should be linear. */
731 gso.hdr_len = skb_headlen(skb);
732 gso.gso_size = sinfo->gso_size;
733 if (sinfo->gso_type & SKB_GSO_TCPV4)
734 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
735 else if (sinfo->gso_type & SKB_GSO_TCPV6)
736 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
737 else if (sinfo->gso_type & SKB_GSO_UDP)
738 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
740 printk(KERN_ERR "tun: unexpected GSO type: "
741 "0x%x, gso_size %d, hdr_len %d\n",
742 sinfo->gso_type, gso.gso_size,
744 print_hex_dump(KERN_ERR, "tun: ",
747 min((int)gso.hdr_len, 64), true);
751 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
752 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
754 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
756 if (skb->ip_summed == CHECKSUM_PARTIAL) {
757 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
758 gso.csum_start = skb->csum_start - skb_headroom(skb);
759 gso.csum_offset = skb->csum_offset;
760 } /* else everything is zero */
762 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
765 total += tun->vnet_hdr_sz;
768 len = min_t(int, skb->len, len);
770 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
773 tun->dev->stats.tx_packets++;
774 tun->dev->stats.tx_bytes += len;
779 static ssize_t tun_do_read(struct tun_struct *tun,
780 struct kiocb *iocb, const struct iovec *iv,
781 ssize_t len, int noblock)
783 DECLARE_WAITQUEUE(wait, current);
787 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
789 add_wait_queue(&tun->wq.wait, &wait);
791 current->state = TASK_INTERRUPTIBLE;
793 /* Read frames from the queue */
794 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
799 if (signal_pending(current)) {
803 if (tun->dev->reg_state != NETREG_REGISTERED) {
808 /* Nothing to read, let's sleep */
812 netif_wake_queue(tun->dev);
814 ret = tun_put_user(tun, skb, iv, len);
819 current->state = TASK_RUNNING;
820 remove_wait_queue(&tun->wq.wait, &wait);
825 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
826 unsigned long count, loff_t pos)
828 struct file *file = iocb->ki_filp;
829 struct tun_file *tfile = file->private_data;
830 struct tun_struct *tun = __tun_get(tfile);
835 len = iov_length(iv, count);
841 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
842 ret = min_t(ssize_t, ret, len);
848 static void tun_setup(struct net_device *dev)
850 struct tun_struct *tun = netdev_priv(dev);
855 dev->ethtool_ops = &tun_ethtool_ops;
856 dev->destructor = tun_free_netdev;
859 /* Trivial set of netlink ops to allow deleting tun or tap
860 * device with netlink.
862 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
867 static struct rtnl_link_ops tun_link_ops __read_mostly = {
869 .priv_size = sizeof(struct tun_struct),
871 .validate = tun_validate,
874 static void tun_sock_write_space(struct sock *sk)
876 struct tun_struct *tun;
877 wait_queue_head_t *wqueue;
879 if (!sock_writeable(sk))
882 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
885 wqueue = sk_sleep(sk);
886 if (wqueue && waitqueue_active(wqueue))
887 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
888 POLLWRNORM | POLLWRBAND);
890 tun = tun_sk(sk)->tun;
891 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
894 static void tun_sock_destruct(struct sock *sk)
896 free_netdev(tun_sk(sk)->tun->dev);
899 static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
900 struct msghdr *m, size_t total_len)
902 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
903 return tun_get_user(tun, m->msg_iov, total_len,
904 m->msg_flags & MSG_DONTWAIT);
907 static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
908 struct msghdr *m, size_t total_len,
911 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
913 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
915 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
916 flags & MSG_DONTWAIT);
917 if (ret > total_len) {
918 m->msg_flags |= MSG_TRUNC;
919 ret = flags & MSG_TRUNC ? ret : total_len;
924 /* Ops structure to mimic raw sockets with tun */
925 static const struct proto_ops tun_socket_ops = {
926 .sendmsg = tun_sendmsg,
927 .recvmsg = tun_recvmsg,
930 static struct proto tun_proto = {
932 .owner = THIS_MODULE,
933 .obj_size = sizeof(struct tun_sock),
936 static int tun_flags(struct tun_struct *tun)
940 if (tun->flags & TUN_TUN_DEV)
945 if (tun->flags & TUN_NO_PI)
948 if (tun->flags & TUN_ONE_QUEUE)
949 flags |= IFF_ONE_QUEUE;
951 if (tun->flags & TUN_VNET_HDR)
952 flags |= IFF_VNET_HDR;
957 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
960 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
961 return sprintf(buf, "0x%x\n", tun_flags(tun));
964 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
967 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
968 return sprintf(buf, "%d\n", tun->owner);
971 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
974 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
975 return sprintf(buf, "%d\n", tun->group);
978 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
979 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
980 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
982 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
985 struct tun_struct *tun;
986 struct net_device *dev;
989 dev = __dev_get_by_name(net, ifr->ifr_name);
991 const struct cred *cred = current_cred();
993 if (ifr->ifr_flags & IFF_TUN_EXCL)
995 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
996 tun = netdev_priv(dev);
997 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
998 tun = netdev_priv(dev);
1002 if (((tun->owner != -1 && cred->euid != tun->owner) ||
1003 (tun->group != -1 && !in_egroup_p(tun->group))) &&
1004 !capable(CAP_NET_ADMIN))
1006 err = security_tun_dev_attach(tun->socket.sk);
1010 err = tun_attach(tun, file);
1016 unsigned long flags = 0;
1018 if (!capable(CAP_NET_ADMIN))
1020 err = security_tun_dev_create();
1025 if (ifr->ifr_flags & IFF_TUN) {
1027 flags |= TUN_TUN_DEV;
1029 } else if (ifr->ifr_flags & IFF_TAP) {
1031 flags |= TUN_TAP_DEV;
1037 name = ifr->ifr_name;
1039 dev = alloc_netdev(sizeof(struct tun_struct), name,
1044 dev_net_set(dev, net);
1045 dev->rtnl_link_ops = &tun_link_ops;
1047 tun = netdev_priv(dev);
1050 tun->txflt.count = 0;
1051 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
1054 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1058 tun->socket.wq = &tun->wq;
1059 init_waitqueue_head(&tun->wq.wait);
1060 tun->socket.ops = &tun_socket_ops;
1061 sock_init_data(&tun->socket, sk);
1062 sk->sk_write_space = tun_sock_write_space;
1063 sk->sk_sndbuf = INT_MAX;
1065 tun_sk(sk)->tun = tun;
1067 security_tun_dev_post_create(sk);
1071 if (strchr(dev->name, '%')) {
1072 err = dev_alloc_name(dev, dev->name);
1077 err = register_netdevice(tun->dev);
1081 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1082 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1083 device_create_file(&tun->dev->dev, &dev_attr_group))
1084 printk(KERN_ERR "Failed to create tun sysfs files\n");
1086 sk->sk_destruct = tun_sock_destruct;
1088 err = tun_attach(tun, file);
1093 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1095 if (ifr->ifr_flags & IFF_NO_PI)
1096 tun->flags |= TUN_NO_PI;
1098 tun->flags &= ~TUN_NO_PI;
1100 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1101 tun->flags |= TUN_ONE_QUEUE;
1103 tun->flags &= ~TUN_ONE_QUEUE;
1105 if (ifr->ifr_flags & IFF_VNET_HDR)
1106 tun->flags |= TUN_VNET_HDR;
1108 tun->flags &= ~TUN_VNET_HDR;
1110 /* Make sure persistent devices do not get stuck in
1113 if (netif_running(tun->dev))
1114 netif_wake_queue(tun->dev);
1116 strcpy(ifr->ifr_name, tun->dev->name);
1127 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1130 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1132 strcpy(ifr->ifr_name, tun->dev->name);
1134 ifr->ifr_flags = tun_flags(tun);
1139 /* This is like a cut-down ethtool ops, except done via tun fd so no
1140 * privs required. */
1141 static int set_offload(struct net_device *dev, unsigned long arg)
1143 unsigned int old_features, features;
1145 old_features = dev->features;
1146 /* Unset features, set them as we chew on the arg. */
1147 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1148 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6
1151 if (arg & TUN_F_CSUM) {
1152 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1155 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1156 if (arg & TUN_F_TSO_ECN) {
1157 features |= NETIF_F_TSO_ECN;
1158 arg &= ~TUN_F_TSO_ECN;
1160 if (arg & TUN_F_TSO4)
1161 features |= NETIF_F_TSO;
1162 if (arg & TUN_F_TSO6)
1163 features |= NETIF_F_TSO6;
1164 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1167 if (arg & TUN_F_UFO) {
1168 features |= NETIF_F_UFO;
1173 /* This gives the user a way to test for new features in future by
1174 * trying to set them. */
1178 dev->features = features;
1179 if (old_features != dev->features)
1180 netdev_features_change(dev);
1185 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1186 unsigned long arg, int ifreq_len)
1188 struct tun_file *tfile = file->private_data;
1189 struct tun_struct *tun;
1190 void __user* argp = (void __user*)arg;
1191 struct sock_fprog fprog;
1197 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1198 if (copy_from_user(&ifr, argp, ifreq_len))
1201 if (cmd == TUNGETFEATURES) {
1202 /* Currently this just means: "what IFF flags are valid?".
1203 * This is needed because we never checked for invalid flags on
1205 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1207 (unsigned int __user*)argp);
1212 tun = __tun_get(tfile);
1213 if (cmd == TUNSETIFF && !tun) {
1214 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1216 ret = tun_set_iff(tfile->net, file, &ifr);
1221 if (copy_to_user(argp, &ifr, ifreq_len))
1230 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1235 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1239 if (copy_to_user(argp, &ifr, ifreq_len))
1244 /* Disable/Enable checksum */
1246 tun->flags |= TUN_NOCHECKSUM;
1248 tun->flags &= ~TUN_NOCHECKSUM;
1250 DBG(KERN_INFO "%s: checksum %s\n",
1251 tun->dev->name, arg ? "disabled" : "enabled");
1255 /* Disable/Enable persist mode */
1257 tun->flags |= TUN_PERSIST;
1259 tun->flags &= ~TUN_PERSIST;
1261 DBG(KERN_INFO "%s: persist %s\n",
1262 tun->dev->name, arg ? "enabled" : "disabled");
1266 /* Set owner of the device */
1267 tun->owner = (uid_t) arg;
1269 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1273 /* Set group of the device */
1274 tun->group= (gid_t) arg;
1276 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1280 /* Only allow setting the type when the interface is down */
1281 if (tun->dev->flags & IFF_UP) {
1282 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1286 tun->dev->type = (int) arg;
1287 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1298 ret = set_offload(tun->dev, arg);
1301 case TUNSETTXFILTER:
1302 /* Can be set only for TAPs */
1304 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1306 ret = update_filter(&tun->txflt, (void __user *)arg);
1311 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1312 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1313 if (copy_to_user(argp, &ifr, ifreq_len))
1318 /* Set hw address */
1319 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1320 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1322 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1326 sndbuf = tun->socket.sk->sk_sndbuf;
1327 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1332 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1337 tun->socket.sk->sk_sndbuf = sndbuf;
1340 case TUNGETVNETHDRSZ:
1341 vnet_hdr_sz = tun->vnet_hdr_sz;
1342 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1346 case TUNSETVNETHDRSZ:
1347 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1351 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1356 tun->vnet_hdr_sz = vnet_hdr_sz;
1359 case TUNATTACHFILTER:
1360 /* Can be set only for TAPs */
1362 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1365 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1368 ret = sk_attach_filter(&fprog, tun->socket.sk);
1371 case TUNDETACHFILTER:
1372 /* Can be set only for TAPs */
1374 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1376 ret = sk_detach_filter(tun->socket.sk);
1391 static long tun_chr_ioctl(struct file *file,
1392 unsigned int cmd, unsigned long arg)
1394 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1397 #ifdef CONFIG_COMPAT
1398 static long tun_chr_compat_ioctl(struct file *file,
1399 unsigned int cmd, unsigned long arg)
1404 case TUNSETTXFILTER:
1409 arg = (unsigned long)compat_ptr(arg);
1412 arg = (compat_ulong_t)arg;
1417 * compat_ifreq is shorter than ifreq, so we must not access beyond
1418 * the end of that structure. All fields that are used in this
1419 * driver are compatible though, we don't need to convert the
1422 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1424 #endif /* CONFIG_COMPAT */
1426 static int tun_chr_fasync(int fd, struct file *file, int on)
1428 struct tun_struct *tun = tun_get(file);
1434 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1436 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1440 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1443 tun->flags |= TUN_FASYNC;
1445 tun->flags &= ~TUN_FASYNC;
1452 static int tun_chr_open(struct inode *inode, struct file * file)
1454 struct tun_file *tfile;
1456 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1458 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1461 atomic_set(&tfile->count, 0);
1463 tfile->net = get_net(current->nsproxy->net_ns);
1464 file->private_data = tfile;
1468 static int tun_chr_close(struct inode *inode, struct file *file)
1470 struct tun_file *tfile = file->private_data;
1471 struct tun_struct *tun;
1473 tun = __tun_get(tfile);
1475 struct net_device *dev = tun->dev;
1477 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
1481 /* If desirable, unregister the netdevice. */
1482 if (!(tun->flags & TUN_PERSIST)) {
1484 if (dev->reg_state == NETREG_REGISTERED)
1485 unregister_netdevice(dev);
1492 sock_put(tun->socket.sk);
1494 put_net(tfile->net);
1500 static const struct file_operations tun_fops = {
1501 .owner = THIS_MODULE,
1502 .llseek = no_llseek,
1503 .read = do_sync_read,
1504 .aio_read = tun_chr_aio_read,
1505 .write = do_sync_write,
1506 .aio_write = tun_chr_aio_write,
1507 .poll = tun_chr_poll,
1508 .unlocked_ioctl = tun_chr_ioctl,
1509 #ifdef CONFIG_COMPAT
1510 .compat_ioctl = tun_chr_compat_ioctl,
1512 .open = tun_chr_open,
1513 .release = tun_chr_close,
1514 .fasync = tun_chr_fasync
1517 static struct miscdevice tun_miscdev = {
1520 .nodename = "net/tun",
1524 /* ethtool interface */
1526 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1529 cmd->advertising = 0;
1530 cmd->speed = SPEED_10;
1531 cmd->duplex = DUPLEX_FULL;
1532 cmd->port = PORT_TP;
1533 cmd->phy_address = 0;
1534 cmd->transceiver = XCVR_INTERNAL;
1535 cmd->autoneg = AUTONEG_DISABLE;
1541 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1543 struct tun_struct *tun = netdev_priv(dev);
1545 strcpy(info->driver, DRV_NAME);
1546 strcpy(info->version, DRV_VERSION);
1547 strcpy(info->fw_version, "N/A");
1549 switch (tun->flags & TUN_TYPE_MASK) {
1551 strcpy(info->bus_info, "tun");
1554 strcpy(info->bus_info, "tap");
1559 static u32 tun_get_msglevel(struct net_device *dev)
1562 struct tun_struct *tun = netdev_priv(dev);
1569 static void tun_set_msglevel(struct net_device *dev, u32 value)
1572 struct tun_struct *tun = netdev_priv(dev);
1577 static u32 tun_get_link(struct net_device *dev)
1579 struct tun_struct *tun = netdev_priv(dev);
1580 return !!tun->tfile;
1583 static u32 tun_get_rx_csum(struct net_device *dev)
1585 struct tun_struct *tun = netdev_priv(dev);
1586 return (tun->flags & TUN_NOCHECKSUM) == 0;
1589 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1591 struct tun_struct *tun = netdev_priv(dev);
1593 tun->flags &= ~TUN_NOCHECKSUM;
1595 tun->flags |= TUN_NOCHECKSUM;
1599 static const struct ethtool_ops tun_ethtool_ops = {
1600 .get_settings = tun_get_settings,
1601 .get_drvinfo = tun_get_drvinfo,
1602 .get_msglevel = tun_get_msglevel,
1603 .set_msglevel = tun_set_msglevel,
1604 .get_link = tun_get_link,
1605 .get_rx_csum = tun_get_rx_csum,
1606 .set_rx_csum = tun_set_rx_csum
1610 static int __init tun_init(void)
1614 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1615 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1617 ret = rtnl_link_register(&tun_link_ops);
1619 printk(KERN_ERR "tun: Can't register link_ops\n");
1623 ret = misc_register(&tun_miscdev);
1625 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1630 rtnl_link_unregister(&tun_link_ops);
1635 static void tun_cleanup(void)
1637 misc_deregister(&tun_miscdev);
1638 rtnl_link_unregister(&tun_link_ops);
1641 /* Get an underlying socket object from tun file. Returns error unless file is
1642 * attached to a device. The returned object works like a packet socket, it
1643 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1644 * holding a reference to the file for as long as the socket is in use. */
1645 struct socket *tun_get_socket(struct file *file)
1647 struct tun_struct *tun;
1648 if (file->f_op != &tun_fops)
1649 return ERR_PTR(-EINVAL);
1650 tun = tun_get(file);
1652 return ERR_PTR(-EBADFD);
1654 return &tun->socket;
1656 EXPORT_SYMBOL_GPL(tun_get_socket);
1658 module_init(tun_init);
1659 module_exit(tun_cleanup);
1660 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1661 MODULE_AUTHOR(DRV_COPYRIGHT);
1662 MODULE_LICENSE("GPL");
1663 MODULE_ALIAS_MISCDEV(TUN_MINOR);
1664 MODULE_ALIAS("devname:net/tun");