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.
424 #define MAX_MTU 65535
427 tun_net_change_mtu(struct net_device *dev, int new_mtu)
429 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
435 static const struct net_device_ops tun_netdev_ops = {
436 .ndo_uninit = tun_net_uninit,
437 .ndo_open = tun_net_open,
438 .ndo_stop = tun_net_close,
439 .ndo_start_xmit = tun_net_xmit,
440 .ndo_change_mtu = tun_net_change_mtu,
443 static const struct net_device_ops tap_netdev_ops = {
444 .ndo_uninit = tun_net_uninit,
445 .ndo_open = tun_net_open,
446 .ndo_stop = tun_net_close,
447 .ndo_start_xmit = tun_net_xmit,
448 .ndo_change_mtu = tun_net_change_mtu,
449 .ndo_set_multicast_list = tun_net_mclist,
450 .ndo_set_mac_address = eth_mac_addr,
451 .ndo_validate_addr = eth_validate_addr,
454 /* Initialize net device. */
455 static void tun_net_init(struct net_device *dev)
457 struct tun_struct *tun = netdev_priv(dev);
459 switch (tun->flags & TUN_TYPE_MASK) {
461 dev->netdev_ops = &tun_netdev_ops;
463 /* Point-to-Point TUN Device */
464 dev->hard_header_len = 0;
468 /* Zero header length */
469 dev->type = ARPHRD_NONE;
470 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
471 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
475 dev->netdev_ops = &tap_netdev_ops;
476 /* Ethernet TAP Device */
479 random_ether_addr(dev->dev_addr);
481 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
486 /* Character device part */
489 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
491 struct tun_file *tfile = file->private_data;
492 struct tun_struct *tun = __tun_get(tfile);
494 unsigned int mask = 0;
501 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
503 poll_wait(file, &tun->wq.wait, wait);
505 if (!skb_queue_empty(&sk->sk_receive_queue))
506 mask |= POLLIN | POLLRDNORM;
508 if (sock_writeable(sk) ||
509 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
511 mask |= POLLOUT | POLLWRNORM;
513 if (tun->dev->reg_state != NETREG_REGISTERED)
520 /* prepad is the amount to reserve at front. len is length after that.
521 * linear is a hint as to how much to copy (usually headers). */
522 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
523 size_t prepad, size_t len,
524 size_t linear, int noblock)
526 struct sock *sk = tun->socket.sk;
530 /* Under a page? Don't bother with paged skb. */
531 if (prepad + len < PAGE_SIZE || !linear)
534 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
539 skb_reserve(skb, prepad);
540 skb_put(skb, linear);
541 skb->data_len = len - linear;
542 skb->len += len - linear;
547 /* Get packet from user space buffer */
548 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
549 const struct iovec *iv, size_t count,
552 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
554 size_t len = count, align = 0;
555 struct virtio_net_hdr gso = { 0 };
558 if (!(tun->flags & TUN_NO_PI)) {
559 if ((len -= sizeof(pi)) > count)
562 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
564 offset += sizeof(pi);
567 if (tun->flags & TUN_VNET_HDR) {
568 if ((len -= tun->vnet_hdr_sz) > count)
571 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
574 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
575 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
576 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
578 if (gso.hdr_len > len)
580 offset += tun->vnet_hdr_sz;
583 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
584 align = NET_IP_ALIGN;
585 if (unlikely(len < ETH_HLEN ||
586 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
590 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
592 if (PTR_ERR(skb) != -EAGAIN)
593 tun->dev->stats.rx_dropped++;
597 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
598 tun->dev->stats.rx_dropped++;
603 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
604 if (!skb_partial_csum_set(skb, gso.csum_start,
606 tun->dev->stats.rx_frame_errors++;
610 } else if (tun->flags & TUN_NOCHECKSUM)
611 skb->ip_summed = CHECKSUM_UNNECESSARY;
613 switch (tun->flags & TUN_TYPE_MASK) {
615 if (tun->flags & TUN_NO_PI) {
616 switch (skb->data[0] & 0xf0) {
618 pi.proto = htons(ETH_P_IP);
621 pi.proto = htons(ETH_P_IPV6);
624 tun->dev->stats.rx_dropped++;
630 skb_reset_mac_header(skb);
631 skb->protocol = pi.proto;
635 skb->protocol = eth_type_trans(skb, tun->dev);
639 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
641 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
642 case VIRTIO_NET_HDR_GSO_TCPV4:
643 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
645 case VIRTIO_NET_HDR_GSO_TCPV6:
646 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
648 case VIRTIO_NET_HDR_GSO_UDP:
649 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
652 tun->dev->stats.rx_frame_errors++;
657 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
658 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
660 skb_shinfo(skb)->gso_size = gso.gso_size;
661 if (skb_shinfo(skb)->gso_size == 0) {
662 tun->dev->stats.rx_frame_errors++;
667 /* Header must be checked, and gso_segs computed. */
668 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
669 skb_shinfo(skb)->gso_segs = 0;
674 tun->dev->stats.rx_packets++;
675 tun->dev->stats.rx_bytes += len;
680 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
681 unsigned long count, loff_t pos)
683 struct file *file = iocb->ki_filp;
684 struct tun_struct *tun = tun_get(file);
690 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
692 result = tun_get_user(tun, iv, iov_length(iv, count),
693 file->f_flags & O_NONBLOCK);
699 /* Put packet to the user space buffer */
700 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
702 const struct iovec *iv, int len)
704 struct tun_pi pi = { 0, skb->protocol };
707 if (!(tun->flags & TUN_NO_PI)) {
708 if ((len -= sizeof(pi)) < 0)
711 if (len < skb->len) {
712 /* Packet will be striped */
713 pi.flags |= TUN_PKT_STRIP;
716 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
721 if (tun->flags & TUN_VNET_HDR) {
722 struct virtio_net_hdr gso = { 0 }; /* no info leak */
723 if ((len -= tun->vnet_hdr_sz) < 0)
726 if (skb_is_gso(skb)) {
727 struct skb_shared_info *sinfo = skb_shinfo(skb);
729 /* This is a hint as to how much should be linear. */
730 gso.hdr_len = skb_headlen(skb);
731 gso.gso_size = sinfo->gso_size;
732 if (sinfo->gso_type & SKB_GSO_TCPV4)
733 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
734 else if (sinfo->gso_type & SKB_GSO_TCPV6)
735 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
736 else if (sinfo->gso_type & SKB_GSO_UDP)
737 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
740 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
741 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
743 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
745 if (skb->ip_summed == CHECKSUM_PARTIAL) {
746 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
747 gso.csum_start = skb->csum_start - skb_headroom(skb);
748 gso.csum_offset = skb->csum_offset;
749 } /* else everything is zero */
751 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
754 total += tun->vnet_hdr_sz;
757 len = min_t(int, skb->len, len);
759 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
762 tun->dev->stats.tx_packets++;
763 tun->dev->stats.tx_bytes += len;
768 static ssize_t tun_do_read(struct tun_struct *tun,
769 struct kiocb *iocb, const struct iovec *iv,
770 ssize_t len, int noblock)
772 DECLARE_WAITQUEUE(wait, current);
776 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
778 add_wait_queue(&tun->wq.wait, &wait);
780 current->state = TASK_INTERRUPTIBLE;
782 /* Read frames from the queue */
783 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
788 if (signal_pending(current)) {
792 if (tun->dev->reg_state != NETREG_REGISTERED) {
797 /* Nothing to read, let's sleep */
801 netif_wake_queue(tun->dev);
803 ret = tun_put_user(tun, skb, iv, len);
808 current->state = TASK_RUNNING;
809 remove_wait_queue(&tun->wq.wait, &wait);
814 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
815 unsigned long count, loff_t pos)
817 struct file *file = iocb->ki_filp;
818 struct tun_file *tfile = file->private_data;
819 struct tun_struct *tun = __tun_get(tfile);
824 len = iov_length(iv, count);
830 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
831 ret = min_t(ssize_t, ret, len);
837 static void tun_setup(struct net_device *dev)
839 struct tun_struct *tun = netdev_priv(dev);
844 dev->ethtool_ops = &tun_ethtool_ops;
845 dev->destructor = tun_free_netdev;
848 /* Trivial set of netlink ops to allow deleting tun or tap
849 * device with netlink.
851 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
856 static struct rtnl_link_ops tun_link_ops __read_mostly = {
858 .priv_size = sizeof(struct tun_struct),
860 .validate = tun_validate,
863 static void tun_sock_write_space(struct sock *sk)
865 struct tun_struct *tun;
866 wait_queue_head_t *wqueue;
868 if (!sock_writeable(sk))
871 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
874 wqueue = sk_sleep(sk);
875 if (wqueue && waitqueue_active(wqueue))
876 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
877 POLLWRNORM | POLLWRBAND);
879 tun = tun_sk(sk)->tun;
880 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
883 static void tun_sock_destruct(struct sock *sk)
885 free_netdev(tun_sk(sk)->tun->dev);
888 static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
889 struct msghdr *m, size_t total_len)
891 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
892 return tun_get_user(tun, m->msg_iov, total_len,
893 m->msg_flags & MSG_DONTWAIT);
896 static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
897 struct msghdr *m, size_t total_len,
900 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
902 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
904 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
905 flags & MSG_DONTWAIT);
906 if (ret > total_len) {
907 m->msg_flags |= MSG_TRUNC;
908 ret = flags & MSG_TRUNC ? ret : total_len;
913 /* Ops structure to mimic raw sockets with tun */
914 static const struct proto_ops tun_socket_ops = {
915 .sendmsg = tun_sendmsg,
916 .recvmsg = tun_recvmsg,
919 static struct proto tun_proto = {
921 .owner = THIS_MODULE,
922 .obj_size = sizeof(struct tun_sock),
925 static int tun_flags(struct tun_struct *tun)
929 if (tun->flags & TUN_TUN_DEV)
934 if (tun->flags & TUN_NO_PI)
937 if (tun->flags & TUN_ONE_QUEUE)
938 flags |= IFF_ONE_QUEUE;
940 if (tun->flags & TUN_VNET_HDR)
941 flags |= IFF_VNET_HDR;
946 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
949 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
950 return sprintf(buf, "0x%x\n", tun_flags(tun));
953 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
956 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
957 return sprintf(buf, "%d\n", tun->owner);
960 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
963 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
964 return sprintf(buf, "%d\n", tun->group);
967 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
968 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
969 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
971 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
974 struct tun_struct *tun;
975 struct net_device *dev;
978 dev = __dev_get_by_name(net, ifr->ifr_name);
980 const struct cred *cred = current_cred();
982 if (ifr->ifr_flags & IFF_TUN_EXCL)
984 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
985 tun = netdev_priv(dev);
986 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
987 tun = netdev_priv(dev);
991 if (((tun->owner != -1 && cred->euid != tun->owner) ||
992 (tun->group != -1 && !in_egroup_p(tun->group))) &&
993 !capable(CAP_NET_ADMIN))
995 err = security_tun_dev_attach(tun->socket.sk);
999 err = tun_attach(tun, file);
1005 unsigned long flags = 0;
1007 if (!capable(CAP_NET_ADMIN))
1009 err = security_tun_dev_create();
1014 if (ifr->ifr_flags & IFF_TUN) {
1016 flags |= TUN_TUN_DEV;
1018 } else if (ifr->ifr_flags & IFF_TAP) {
1020 flags |= TUN_TAP_DEV;
1026 name = ifr->ifr_name;
1028 dev = alloc_netdev(sizeof(struct tun_struct), name,
1033 dev_net_set(dev, net);
1034 dev->rtnl_link_ops = &tun_link_ops;
1036 tun = netdev_priv(dev);
1039 tun->txflt.count = 0;
1040 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
1043 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1047 tun->socket.wq = &tun->wq;
1048 init_waitqueue_head(&tun->wq.wait);
1049 tun->socket.ops = &tun_socket_ops;
1050 sock_init_data(&tun->socket, sk);
1051 sk->sk_write_space = tun_sock_write_space;
1052 sk->sk_sndbuf = INT_MAX;
1054 tun_sk(sk)->tun = tun;
1056 security_tun_dev_post_create(sk);
1060 if (strchr(dev->name, '%')) {
1061 err = dev_alloc_name(dev, dev->name);
1066 err = register_netdevice(tun->dev);
1070 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1071 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1072 device_create_file(&tun->dev->dev, &dev_attr_group))
1073 printk(KERN_ERR "Failed to create tun sysfs files\n");
1075 sk->sk_destruct = tun_sock_destruct;
1077 err = tun_attach(tun, file);
1082 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1084 if (ifr->ifr_flags & IFF_NO_PI)
1085 tun->flags |= TUN_NO_PI;
1087 tun->flags &= ~TUN_NO_PI;
1089 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1090 tun->flags |= TUN_ONE_QUEUE;
1092 tun->flags &= ~TUN_ONE_QUEUE;
1094 if (ifr->ifr_flags & IFF_VNET_HDR)
1095 tun->flags |= TUN_VNET_HDR;
1097 tun->flags &= ~TUN_VNET_HDR;
1099 /* Make sure persistent devices do not get stuck in
1102 if (netif_running(tun->dev))
1103 netif_wake_queue(tun->dev);
1105 strcpy(ifr->ifr_name, tun->dev->name);
1116 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1119 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1121 strcpy(ifr->ifr_name, tun->dev->name);
1123 ifr->ifr_flags = tun_flags(tun);
1128 /* This is like a cut-down ethtool ops, except done via tun fd so no
1129 * privs required. */
1130 static int set_offload(struct net_device *dev, unsigned long arg)
1132 unsigned int old_features, features;
1134 old_features = dev->features;
1135 /* Unset features, set them as we chew on the arg. */
1136 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1137 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6
1140 if (arg & TUN_F_CSUM) {
1141 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1144 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1145 if (arg & TUN_F_TSO_ECN) {
1146 features |= NETIF_F_TSO_ECN;
1147 arg &= ~TUN_F_TSO_ECN;
1149 if (arg & TUN_F_TSO4)
1150 features |= NETIF_F_TSO;
1151 if (arg & TUN_F_TSO6)
1152 features |= NETIF_F_TSO6;
1153 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1156 if (arg & TUN_F_UFO) {
1157 features |= NETIF_F_UFO;
1162 /* This gives the user a way to test for new features in future by
1163 * trying to set them. */
1167 dev->features = features;
1168 if (old_features != dev->features)
1169 netdev_features_change(dev);
1174 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1175 unsigned long arg, int ifreq_len)
1177 struct tun_file *tfile = file->private_data;
1178 struct tun_struct *tun;
1179 void __user* argp = (void __user*)arg;
1180 struct sock_fprog fprog;
1186 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1187 if (copy_from_user(&ifr, argp, ifreq_len))
1190 if (cmd == TUNGETFEATURES) {
1191 /* Currently this just means: "what IFF flags are valid?".
1192 * This is needed because we never checked for invalid flags on
1194 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1196 (unsigned int __user*)argp);
1201 tun = __tun_get(tfile);
1202 if (cmd == TUNSETIFF && !tun) {
1203 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1205 ret = tun_set_iff(tfile->net, file, &ifr);
1210 if (copy_to_user(argp, &ifr, ifreq_len))
1219 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1224 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1228 if (copy_to_user(argp, &ifr, ifreq_len))
1233 /* Disable/Enable checksum */
1235 tun->flags |= TUN_NOCHECKSUM;
1237 tun->flags &= ~TUN_NOCHECKSUM;
1239 DBG(KERN_INFO "%s: checksum %s\n",
1240 tun->dev->name, arg ? "disabled" : "enabled");
1244 /* Disable/Enable persist mode */
1246 tun->flags |= TUN_PERSIST;
1248 tun->flags &= ~TUN_PERSIST;
1250 DBG(KERN_INFO "%s: persist %s\n",
1251 tun->dev->name, arg ? "enabled" : "disabled");
1255 /* Set owner of the device */
1256 tun->owner = (uid_t) arg;
1258 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1262 /* Set group of the device */
1263 tun->group= (gid_t) arg;
1265 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1269 /* Only allow setting the type when the interface is down */
1270 if (tun->dev->flags & IFF_UP) {
1271 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1275 tun->dev->type = (int) arg;
1276 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1287 ret = set_offload(tun->dev, arg);
1290 case TUNSETTXFILTER:
1291 /* Can be set only for TAPs */
1293 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1295 ret = update_filter(&tun->txflt, (void __user *)arg);
1300 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1301 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1302 if (copy_to_user(argp, &ifr, ifreq_len))
1307 /* Set hw address */
1308 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1309 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1311 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1315 sndbuf = tun->socket.sk->sk_sndbuf;
1316 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1321 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1326 tun->socket.sk->sk_sndbuf = sndbuf;
1329 case TUNGETVNETHDRSZ:
1330 vnet_hdr_sz = tun->vnet_hdr_sz;
1331 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1335 case TUNSETVNETHDRSZ:
1336 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1340 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1345 tun->vnet_hdr_sz = vnet_hdr_sz;
1348 case TUNATTACHFILTER:
1349 /* Can be set only for TAPs */
1351 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1354 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1357 ret = sk_attach_filter(&fprog, tun->socket.sk);
1360 case TUNDETACHFILTER:
1361 /* Can be set only for TAPs */
1363 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1365 ret = sk_detach_filter(tun->socket.sk);
1380 static long tun_chr_ioctl(struct file *file,
1381 unsigned int cmd, unsigned long arg)
1383 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1386 #ifdef CONFIG_COMPAT
1387 static long tun_chr_compat_ioctl(struct file *file,
1388 unsigned int cmd, unsigned long arg)
1393 case TUNSETTXFILTER:
1398 arg = (unsigned long)compat_ptr(arg);
1401 arg = (compat_ulong_t)arg;
1406 * compat_ifreq is shorter than ifreq, so we must not access beyond
1407 * the end of that structure. All fields that are used in this
1408 * driver are compatible though, we don't need to convert the
1411 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1413 #endif /* CONFIG_COMPAT */
1415 static int tun_chr_fasync(int fd, struct file *file, int on)
1417 struct tun_struct *tun = tun_get(file);
1423 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1425 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1429 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1432 tun->flags |= TUN_FASYNC;
1434 tun->flags &= ~TUN_FASYNC;
1441 static int tun_chr_open(struct inode *inode, struct file * file)
1443 struct tun_file *tfile;
1445 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1447 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1450 atomic_set(&tfile->count, 0);
1452 tfile->net = get_net(current->nsproxy->net_ns);
1453 file->private_data = tfile;
1457 static int tun_chr_close(struct inode *inode, struct file *file)
1459 struct tun_file *tfile = file->private_data;
1460 struct tun_struct *tun;
1462 tun = __tun_get(tfile);
1464 struct net_device *dev = tun->dev;
1466 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
1470 /* If desirable, unregister the netdevice. */
1471 if (!(tun->flags & TUN_PERSIST)) {
1473 if (dev->reg_state == NETREG_REGISTERED)
1474 unregister_netdevice(dev);
1481 sock_put(tun->socket.sk);
1483 put_net(tfile->net);
1489 static const struct file_operations tun_fops = {
1490 .owner = THIS_MODULE,
1491 .llseek = no_llseek,
1492 .read = do_sync_read,
1493 .aio_read = tun_chr_aio_read,
1494 .write = do_sync_write,
1495 .aio_write = tun_chr_aio_write,
1496 .poll = tun_chr_poll,
1497 .unlocked_ioctl = tun_chr_ioctl,
1498 #ifdef CONFIG_COMPAT
1499 .compat_ioctl = tun_chr_compat_ioctl,
1501 .open = tun_chr_open,
1502 .release = tun_chr_close,
1503 .fasync = tun_chr_fasync
1506 static struct miscdevice tun_miscdev = {
1509 .nodename = "net/tun",
1513 /* ethtool interface */
1515 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1518 cmd->advertising = 0;
1519 cmd->speed = SPEED_10;
1520 cmd->duplex = DUPLEX_FULL;
1521 cmd->port = PORT_TP;
1522 cmd->phy_address = 0;
1523 cmd->transceiver = XCVR_INTERNAL;
1524 cmd->autoneg = AUTONEG_DISABLE;
1530 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1532 struct tun_struct *tun = netdev_priv(dev);
1534 strcpy(info->driver, DRV_NAME);
1535 strcpy(info->version, DRV_VERSION);
1536 strcpy(info->fw_version, "N/A");
1538 switch (tun->flags & TUN_TYPE_MASK) {
1540 strcpy(info->bus_info, "tun");
1543 strcpy(info->bus_info, "tap");
1548 static u32 tun_get_msglevel(struct net_device *dev)
1551 struct tun_struct *tun = netdev_priv(dev);
1558 static void tun_set_msglevel(struct net_device *dev, u32 value)
1561 struct tun_struct *tun = netdev_priv(dev);
1566 static u32 tun_get_link(struct net_device *dev)
1568 struct tun_struct *tun = netdev_priv(dev);
1569 return !!tun->tfile;
1572 static u32 tun_get_rx_csum(struct net_device *dev)
1574 struct tun_struct *tun = netdev_priv(dev);
1575 return (tun->flags & TUN_NOCHECKSUM) == 0;
1578 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1580 struct tun_struct *tun = netdev_priv(dev);
1582 tun->flags &= ~TUN_NOCHECKSUM;
1584 tun->flags |= TUN_NOCHECKSUM;
1588 static const struct ethtool_ops tun_ethtool_ops = {
1589 .get_settings = tun_get_settings,
1590 .get_drvinfo = tun_get_drvinfo,
1591 .get_msglevel = tun_get_msglevel,
1592 .set_msglevel = tun_set_msglevel,
1593 .get_link = tun_get_link,
1594 .get_rx_csum = tun_get_rx_csum,
1595 .set_rx_csum = tun_set_rx_csum
1599 static int __init tun_init(void)
1603 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1604 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1606 ret = rtnl_link_register(&tun_link_ops);
1608 printk(KERN_ERR "tun: Can't register link_ops\n");
1612 ret = misc_register(&tun_miscdev);
1614 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1619 rtnl_link_unregister(&tun_link_ops);
1624 static void tun_cleanup(void)
1626 misc_deregister(&tun_miscdev);
1627 rtnl_link_unregister(&tun_link_ops);
1630 /* Get an underlying socket object from tun file. Returns error unless file is
1631 * attached to a device. The returned object works like a packet socket, it
1632 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1633 * holding a reference to the file for as long as the socket is in use. */
1634 struct socket *tun_get_socket(struct file *file)
1636 struct tun_struct *tun;
1637 if (file->f_op != &tun_fops)
1638 return ERR_PTR(-EINVAL);
1639 tun = tun_get(file);
1641 return ERR_PTR(-EBADFD);
1643 return &tun->socket;
1645 EXPORT_SYMBOL_GPL(tun_get_socket);
1647 module_init(tun_init);
1648 module_exit(tun_cleanup);
1649 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1650 MODULE_AUTHOR(DRV_COPYRIGHT);
1651 MODULE_LICENSE("GPL");
1652 MODULE_ALIAS_MISCDEV(TUN_MINOR);