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/smp_lock.h>
48 #include <linux/poll.h>
49 #include <linux/fcntl.h>
50 #include <linux/init.h>
51 #include <linux/skbuff.h>
52 #include <linux/netdevice.h>
53 #include <linux/etherdevice.h>
54 #include <linux/miscdevice.h>
55 #include <linux/ethtool.h>
56 #include <linux/rtnetlink.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 <net/net_namespace.h>
65 #include <net/netns/generic.h>
66 #include <net/rtnetlink.h>
69 #include <asm/system.h>
70 #include <asm/uaccess.h>
72 /* Uncomment to enable debugging */
73 /* #define TUN_DEBUG 1 */
78 #define DBG if(tun->debug)printk
79 #define DBG1 if(debug==2)printk
85 #define FLT_EXACT_COUNT 8
87 unsigned int count; /* Number of addrs. Zero means disabled */
88 u32 mask[2]; /* Mask of the hashed addrs */
89 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
94 struct tun_struct *tun;
96 wait_queue_head_t read_wait;
102 struct tun_file *tfile;
107 struct sk_buff_head readq;
109 struct net_device *dev;
110 struct fasync_struct *fasync;
112 struct tap_filter txflt;
114 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;
134 const struct cred *cred = current_cred();
139 /* Check permissions */
140 if (((tun->owner != -1 && cred->euid != tun->owner) ||
141 (tun->group != -1 && !in_egroup_p(tun->group))) &&
142 !capable(CAP_NET_ADMIN))
145 netif_tx_lock_bh(tun->dev);
159 atomic_inc(&tfile->count);
162 netif_tx_unlock_bh(tun->dev);
166 static void __tun_detach(struct tun_struct *tun)
168 struct tun_file *tfile = tun->tfile;
170 /* Detach from net device */
171 netif_tx_lock_bh(tun->dev);
174 netif_tx_unlock_bh(tun->dev);
176 /* Drop read queue */
177 skb_queue_purge(&tun->readq);
179 /* Drop the extra count on the net device */
183 static void tun_detach(struct tun_struct *tun)
190 static struct tun_struct *__tun_get(struct tun_file *tfile)
192 struct tun_struct *tun = NULL;
194 if (atomic_inc_not_zero(&tfile->count))
200 static struct tun_struct *tun_get(struct file *file)
202 return __tun_get(file->private_data);
205 static void tun_put(struct tun_struct *tun)
207 struct tun_file *tfile = tun->tfile;
209 if (atomic_dec_and_test(&tfile->count))
210 tun_detach(tfile->tun);
214 static void addr_hash_set(u32 *mask, const u8 *addr)
216 int n = ether_crc(ETH_ALEN, addr) >> 26;
217 mask[n >> 5] |= (1 << (n & 31));
220 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
222 int n = ether_crc(ETH_ALEN, addr) >> 26;
223 return mask[n >> 5] & (1 << (n & 31));
226 static int update_filter(struct tap_filter *filter, void __user *arg)
228 struct { u8 u[ETH_ALEN]; } *addr;
229 struct tun_filter uf;
230 int err, alen, n, nexact;
232 if (copy_from_user(&uf, arg, sizeof(uf)))
241 alen = ETH_ALEN * uf.count;
242 addr = kmalloc(alen, GFP_KERNEL);
246 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
251 /* The filter is updated without holding any locks. Which is
252 * perfectly safe. We disable it first and in the worst
253 * case we'll accept a few undesired packets. */
257 /* Use first set of addresses as an exact filter */
258 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
259 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
263 /* The rest is hashed */
264 memset(filter->mask, 0, sizeof(filter->mask));
265 for (; n < uf.count; n++)
266 addr_hash_set(filter->mask, addr[n].u);
268 /* For ALLMULTI just set the mask to all ones.
269 * This overrides the mask populated above. */
270 if ((uf.flags & TUN_FLT_ALLMULTI))
271 memset(filter->mask, ~0, sizeof(filter->mask));
273 /* Now enable the filter */
275 filter->count = nexact;
277 /* Return the number of exact filters */
285 /* Returns: 0 - drop, !=0 - accept */
286 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
288 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
290 struct ethhdr *eh = (struct ethhdr *) skb->data;
294 for (i = 0; i < filter->count; i++)
295 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
298 /* Inexact match (multicast only) */
299 if (is_multicast_ether_addr(eh->h_dest))
300 return addr_hash_test(filter->mask, eh->h_dest);
306 * Checks whether the packet is accepted or not.
307 * Returns: 0 - drop, !=0 - accept
309 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
314 return run_filter(filter, skb);
317 /* Network device part of the driver */
319 static const struct ethtool_ops tun_ethtool_ops;
321 /* Net device detach from fd. */
322 static void tun_net_uninit(struct net_device *dev)
324 struct tun_struct *tun = netdev_priv(dev);
325 struct tun_file *tfile = tun->tfile;
327 /* Inform the methods they need to stop using the dev.
330 wake_up_all(&tfile->read_wait);
331 if (atomic_dec_and_test(&tfile->count))
336 /* Net device open. */
337 static int tun_net_open(struct net_device *dev)
339 netif_start_queue(dev);
343 /* Net device close. */
344 static int tun_net_close(struct net_device *dev)
346 netif_stop_queue(dev);
350 /* Net device start xmit */
351 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
353 struct tun_struct *tun = netdev_priv(dev);
355 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
357 /* Drop packet if interface is not attached */
361 /* Drop if the filter does not like it.
362 * This is a noop if the filter is disabled.
363 * Filter can be enabled only for the TAP devices. */
364 if (!check_filter(&tun->txflt, skb))
367 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
368 if (!(tun->flags & TUN_ONE_QUEUE)) {
369 /* Normal queueing mode. */
370 /* Packet scheduler handles dropping of further packets. */
371 netif_stop_queue(dev);
373 /* We won't see all dropped packets individually, so overrun
374 * error is more appropriate. */
375 dev->stats.tx_fifo_errors++;
377 /* Single queue mode.
378 * Driver handles dropping of all packets itself. */
384 skb_queue_tail(&tun->readq, skb);
385 dev->trans_start = jiffies;
387 /* Notify and wake up reader process */
388 if (tun->flags & TUN_FASYNC)
389 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
390 wake_up_interruptible(&tun->tfile->read_wait);
394 dev->stats.tx_dropped++;
399 static void tun_net_mclist(struct net_device *dev)
402 * This callback is supposed to deal with mc filter in
403 * _rx_ path and has nothing to do with the _tx_ path.
404 * In rx path we always accept everything userspace gives us.
410 #define MAX_MTU 65535
413 tun_net_change_mtu(struct net_device *dev, int new_mtu)
415 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
421 static const struct net_device_ops tun_netdev_ops = {
422 .ndo_uninit = tun_net_uninit,
423 .ndo_open = tun_net_open,
424 .ndo_stop = tun_net_close,
425 .ndo_start_xmit = tun_net_xmit,
426 .ndo_change_mtu = tun_net_change_mtu,
429 static const struct net_device_ops tap_netdev_ops = {
430 .ndo_uninit = tun_net_uninit,
431 .ndo_open = tun_net_open,
432 .ndo_stop = tun_net_close,
433 .ndo_start_xmit = tun_net_xmit,
434 .ndo_change_mtu = tun_net_change_mtu,
435 .ndo_set_multicast_list = tun_net_mclist,
436 .ndo_set_mac_address = eth_mac_addr,
437 .ndo_validate_addr = eth_validate_addr,
440 /* Initialize net device. */
441 static void tun_net_init(struct net_device *dev)
443 struct tun_struct *tun = netdev_priv(dev);
445 switch (tun->flags & TUN_TYPE_MASK) {
447 dev->netdev_ops = &tun_netdev_ops;
449 /* Point-to-Point TUN Device */
450 dev->hard_header_len = 0;
454 /* Zero header length */
455 dev->type = ARPHRD_NONE;
456 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
457 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
461 dev->netdev_ops = &tap_netdev_ops;
462 /* Ethernet TAP Device */
465 random_ether_addr(dev->dev_addr);
467 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
472 /* Character device part */
475 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
477 struct tun_file *tfile = file->private_data;
478 struct tun_struct *tun = __tun_get(tfile);
479 struct sock *sk = tun->sk;
480 unsigned int mask = 0;
485 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
487 poll_wait(file, &tfile->read_wait, wait);
489 if (!skb_queue_empty(&tun->readq))
490 mask |= POLLIN | POLLRDNORM;
492 if (sock_writeable(sk) ||
493 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
495 mask |= POLLOUT | POLLWRNORM;
497 if (tun->dev->reg_state != NETREG_REGISTERED)
504 /* prepad is the amount to reserve at front. len is length after that.
505 * linear is a hint as to how much to copy (usually headers). */
506 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
507 size_t prepad, size_t len,
508 size_t linear, int noblock)
510 struct sock *sk = tun->sk;
514 /* Under a page? Don't bother with paged skb. */
515 if (prepad + len < PAGE_SIZE)
518 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
523 skb_reserve(skb, prepad);
524 skb_put(skb, linear);
525 skb->data_len = len - linear;
526 skb->len += len - linear;
531 /* Get packet from user space buffer */
532 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
533 struct iovec *iv, size_t count,
536 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
538 size_t len = count, align = 0;
539 struct virtio_net_hdr gso = { 0 };
541 if (!(tun->flags & TUN_NO_PI)) {
542 if ((len -= sizeof(pi)) > count)
545 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
549 if (tun->flags & TUN_VNET_HDR) {
550 if ((len -= sizeof(gso)) > count)
553 if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso)))
556 if (gso.hdr_len > len)
560 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
561 align = NET_IP_ALIGN;
562 if (unlikely(len < ETH_HLEN))
566 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
568 if (PTR_ERR(skb) != -EAGAIN)
569 tun->dev->stats.rx_dropped++;
573 if (skb_copy_datagram_from_iovec(skb, 0, iv, len)) {
574 tun->dev->stats.rx_dropped++;
579 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
580 if (!skb_partial_csum_set(skb, gso.csum_start,
582 tun->dev->stats.rx_frame_errors++;
586 } else if (tun->flags & TUN_NOCHECKSUM)
587 skb->ip_summed = CHECKSUM_UNNECESSARY;
589 switch (tun->flags & TUN_TYPE_MASK) {
591 if (tun->flags & TUN_NO_PI) {
592 switch (skb->data[0] & 0xf0) {
594 pi.proto = htons(ETH_P_IP);
597 pi.proto = htons(ETH_P_IPV6);
600 tun->dev->stats.rx_dropped++;
606 skb_reset_mac_header(skb);
607 skb->protocol = pi.proto;
611 skb->protocol = eth_type_trans(skb, tun->dev);
615 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
617 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
618 case VIRTIO_NET_HDR_GSO_TCPV4:
619 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
621 case VIRTIO_NET_HDR_GSO_TCPV6:
622 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
625 tun->dev->stats.rx_frame_errors++;
630 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
631 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
633 skb_shinfo(skb)->gso_size = gso.gso_size;
634 if (skb_shinfo(skb)->gso_size == 0) {
635 tun->dev->stats.rx_frame_errors++;
640 /* Header must be checked, and gso_segs computed. */
641 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
642 skb_shinfo(skb)->gso_segs = 0;
647 tun->dev->stats.rx_packets++;
648 tun->dev->stats.rx_bytes += len;
653 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
654 unsigned long count, loff_t pos)
656 struct file *file = iocb->ki_filp;
657 struct tun_struct *tun = file->private_data;
663 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
665 result = tun_get_user(tun, (struct iovec *)iv, iov_length(iv, count),
666 file->f_flags & O_NONBLOCK);
672 /* Put packet to the user space buffer */
673 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
675 struct iovec *iv, int len)
677 struct tun_pi pi = { 0, skb->protocol };
680 if (!(tun->flags & TUN_NO_PI)) {
681 if ((len -= sizeof(pi)) < 0)
684 if (len < skb->len) {
685 /* Packet will be striped */
686 pi.flags |= TUN_PKT_STRIP;
689 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
694 if (tun->flags & TUN_VNET_HDR) {
695 struct virtio_net_hdr gso = { 0 }; /* no info leak */
696 if ((len -= sizeof(gso)) < 0)
699 if (skb_is_gso(skb)) {
700 struct skb_shared_info *sinfo = skb_shinfo(skb);
702 /* This is a hint as to how much should be linear. */
703 gso.hdr_len = skb_headlen(skb);
704 gso.gso_size = sinfo->gso_size;
705 if (sinfo->gso_type & SKB_GSO_TCPV4)
706 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
707 else if (sinfo->gso_type & SKB_GSO_TCPV6)
708 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
711 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
712 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
714 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
716 if (skb->ip_summed == CHECKSUM_PARTIAL) {
717 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
718 gso.csum_start = skb->csum_start - skb_headroom(skb);
719 gso.csum_offset = skb->csum_offset;
720 } /* else everything is zero */
722 if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso))))
724 total += sizeof(gso);
727 len = min_t(int, skb->len, len);
729 skb_copy_datagram_iovec(skb, 0, iv, len);
732 tun->dev->stats.tx_packets++;
733 tun->dev->stats.tx_bytes += len;
738 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
739 unsigned long count, loff_t pos)
741 struct file *file = iocb->ki_filp;
742 struct tun_file *tfile = file->private_data;
743 struct tun_struct *tun = __tun_get(tfile);
744 DECLARE_WAITQUEUE(wait, current);
746 ssize_t len, ret = 0;
751 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
753 len = iov_length(iv, count);
759 add_wait_queue(&tfile->read_wait, &wait);
761 current->state = TASK_INTERRUPTIBLE;
763 /* Read frames from the queue */
764 if (!(skb=skb_dequeue(&tun->readq))) {
765 if (file->f_flags & O_NONBLOCK) {
769 if (signal_pending(current)) {
773 if (tun->dev->reg_state != NETREG_REGISTERED) {
778 /* Nothing to read, let's sleep */
782 netif_wake_queue(tun->dev);
784 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
789 current->state = TASK_RUNNING;
790 remove_wait_queue(&tfile->read_wait, &wait);
797 static void tun_setup(struct net_device *dev)
799 struct tun_struct *tun = netdev_priv(dev);
801 skb_queue_head_init(&tun->readq);
806 dev->ethtool_ops = &tun_ethtool_ops;
807 dev->destructor = free_netdev;
810 /* Trivial set of netlink ops to allow deleting tun or tap
811 * device with netlink.
813 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
818 static struct rtnl_link_ops tun_link_ops __read_mostly = {
820 .priv_size = sizeof(struct tun_struct),
822 .validate = tun_validate,
825 static void tun_sock_write_space(struct sock *sk)
827 struct tun_struct *tun;
829 if (!sock_writeable(sk))
832 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
833 wake_up_interruptible_sync(sk->sk_sleep);
835 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
838 tun = container_of(sk, struct tun_sock, sk)->tun;
839 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
842 static void tun_sock_destruct(struct sock *sk)
844 dev_put(container_of(sk, struct tun_sock, sk)->tun->dev);
847 static struct proto tun_proto = {
849 .owner = THIS_MODULE,
850 .obj_size = sizeof(struct tun_sock),
853 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
856 struct tun_struct *tun;
857 struct net_device *dev;
858 struct tun_file *tfile = file->private_data;
861 dev = __dev_get_by_name(net, ifr->ifr_name);
863 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
864 tun = netdev_priv(dev);
865 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
866 tun = netdev_priv(dev);
870 err = tun_attach(tun, file);
876 unsigned long flags = 0;
880 if (!capable(CAP_NET_ADMIN))
884 if (ifr->ifr_flags & IFF_TUN) {
886 flags |= TUN_TUN_DEV;
888 } else if (ifr->ifr_flags & IFF_TAP) {
890 flags |= TUN_TAP_DEV;
896 name = ifr->ifr_name;
898 dev = alloc_netdev(sizeof(struct tun_struct), name,
903 dev_net_set(dev, net);
904 dev->rtnl_link_ops = &tun_link_ops;
906 tun = netdev_priv(dev);
909 tun->txflt.count = 0;
912 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
916 /* This ref count is for tun->sk. */
918 sock_init_data(&tun->socket, sk);
919 sk->sk_write_space = tun_sock_write_space;
920 sk->sk_destruct = tun_sock_destruct;
921 sk->sk_sndbuf = INT_MAX;
922 sk->sk_sleep = &tfile->read_wait;
925 container_of(sk, struct tun_sock, sk)->tun = tun;
929 if (strchr(dev->name, '%')) {
930 err = dev_alloc_name(dev, dev->name);
936 err = register_netdevice(tun->dev);
940 err = tun_attach(tun, file);
945 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
947 if (ifr->ifr_flags & IFF_NO_PI)
948 tun->flags |= TUN_NO_PI;
950 tun->flags &= ~TUN_NO_PI;
952 if (ifr->ifr_flags & IFF_ONE_QUEUE)
953 tun->flags |= TUN_ONE_QUEUE;
955 tun->flags &= ~TUN_ONE_QUEUE;
957 if (ifr->ifr_flags & IFF_VNET_HDR)
958 tun->flags |= TUN_VNET_HDR;
960 tun->flags &= ~TUN_VNET_HDR;
962 /* Make sure persistent devices do not get stuck in
965 if (netif_running(tun->dev))
966 netif_wake_queue(tun->dev);
968 strcpy(ifr->ifr_name, tun->dev->name);
979 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
981 struct tun_struct *tun = tun_get(file);
986 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
988 strcpy(ifr->ifr_name, tun->dev->name);
992 if (ifr->ifr_flags & TUN_TUN_DEV)
993 ifr->ifr_flags |= IFF_TUN;
995 ifr->ifr_flags |= IFF_TAP;
997 if (tun->flags & TUN_NO_PI)
998 ifr->ifr_flags |= IFF_NO_PI;
1000 if (tun->flags & TUN_ONE_QUEUE)
1001 ifr->ifr_flags |= IFF_ONE_QUEUE;
1003 if (tun->flags & TUN_VNET_HDR)
1004 ifr->ifr_flags |= IFF_VNET_HDR;
1010 /* This is like a cut-down ethtool ops, except done via tun fd so no
1011 * privs required. */
1012 static int set_offload(struct net_device *dev, unsigned long arg)
1014 unsigned int old_features, features;
1016 old_features = dev->features;
1017 /* Unset features, set them as we chew on the arg. */
1018 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1019 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
1021 if (arg & TUN_F_CSUM) {
1022 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1025 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1026 if (arg & TUN_F_TSO_ECN) {
1027 features |= NETIF_F_TSO_ECN;
1028 arg &= ~TUN_F_TSO_ECN;
1030 if (arg & TUN_F_TSO4)
1031 features |= NETIF_F_TSO;
1032 if (arg & TUN_F_TSO6)
1033 features |= NETIF_F_TSO6;
1034 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1038 /* This gives the user a way to test for new features in future by
1039 * trying to set them. */
1043 dev->features = features;
1044 if (old_features != dev->features)
1045 netdev_features_change(dev);
1050 static int tun_chr_ioctl(struct inode *inode, struct file *file,
1051 unsigned int cmd, unsigned long arg)
1053 struct tun_file *tfile = file->private_data;
1054 struct tun_struct *tun;
1055 void __user* argp = (void __user*)arg;
1060 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1061 if (copy_from_user(&ifr, argp, sizeof ifr))
1064 if (cmd == TUNGETFEATURES) {
1065 /* Currently this just means: "what IFF flags are valid?".
1066 * This is needed because we never checked for invalid flags on
1068 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1070 (unsigned int __user*)argp);
1073 tun = __tun_get(tfile);
1074 if (cmd == TUNSETIFF && !tun) {
1077 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1080 err = tun_set_iff(tfile->net, file, &ifr);
1086 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1095 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1100 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
1104 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1109 /* Disable/Enable checksum */
1111 tun->flags |= TUN_NOCHECKSUM;
1113 tun->flags &= ~TUN_NOCHECKSUM;
1115 DBG(KERN_INFO "%s: checksum %s\n",
1116 tun->dev->name, arg ? "disabled" : "enabled");
1120 /* Disable/Enable persist mode */
1122 tun->flags |= TUN_PERSIST;
1124 tun->flags &= ~TUN_PERSIST;
1126 DBG(KERN_INFO "%s: persist %s\n",
1127 tun->dev->name, arg ? "enabled" : "disabled");
1131 /* Set owner of the device */
1132 tun->owner = (uid_t) arg;
1134 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1138 /* Set group of the device */
1139 tun->group= (gid_t) arg;
1141 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1145 /* Only allow setting the type when the interface is down */
1147 if (tun->dev->flags & IFF_UP) {
1148 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1152 tun->dev->type = (int) arg;
1153 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1166 ret = set_offload(tun->dev, arg);
1170 case TUNSETTXFILTER:
1171 /* Can be set only for TAPs */
1173 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1176 ret = update_filter(&tun->txflt, (void __user *)arg);
1182 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1183 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1184 if (copy_to_user(argp, &ifr, sizeof ifr))
1189 /* Set hw address */
1190 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1191 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1194 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1199 sndbuf = tun->sk->sk_sndbuf;
1200 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1205 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1210 tun->sk->sk_sndbuf = sndbuf;
1222 static int tun_chr_fasync(int fd, struct file *file, int on)
1224 struct tun_struct *tun = tun_get(file);
1230 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1233 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1237 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1240 tun->flags |= TUN_FASYNC;
1242 tun->flags &= ~TUN_FASYNC;
1250 static int tun_chr_open(struct inode *inode, struct file * file)
1252 struct tun_file *tfile;
1253 cycle_kernel_lock();
1254 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1256 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1259 atomic_set(&tfile->count, 0);
1261 tfile->net = get_net(current->nsproxy->net_ns);
1262 init_waitqueue_head(&tfile->read_wait);
1263 file->private_data = tfile;
1267 static int tun_chr_close(struct inode *inode, struct file *file)
1269 struct tun_file *tfile = file->private_data;
1270 struct tun_struct *tun = __tun_get(tfile);
1274 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1279 /* If desireable, unregister the netdevice. */
1280 if (!(tun->flags & TUN_PERSIST)) {
1282 unregister_netdevice(tun->dev);
1288 put_net(tfile->net);
1294 static const struct file_operations tun_fops = {
1295 .owner = THIS_MODULE,
1296 .llseek = no_llseek,
1297 .read = do_sync_read,
1298 .aio_read = tun_chr_aio_read,
1299 .write = do_sync_write,
1300 .aio_write = tun_chr_aio_write,
1301 .poll = tun_chr_poll,
1302 .ioctl = tun_chr_ioctl,
1303 .open = tun_chr_open,
1304 .release = tun_chr_close,
1305 .fasync = tun_chr_fasync
1308 static struct miscdevice tun_miscdev = {
1314 /* ethtool interface */
1316 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1319 cmd->advertising = 0;
1320 cmd->speed = SPEED_10;
1321 cmd->duplex = DUPLEX_FULL;
1322 cmd->port = PORT_TP;
1323 cmd->phy_address = 0;
1324 cmd->transceiver = XCVR_INTERNAL;
1325 cmd->autoneg = AUTONEG_DISABLE;
1331 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1333 struct tun_struct *tun = netdev_priv(dev);
1335 strcpy(info->driver, DRV_NAME);
1336 strcpy(info->version, DRV_VERSION);
1337 strcpy(info->fw_version, "N/A");
1339 switch (tun->flags & TUN_TYPE_MASK) {
1341 strcpy(info->bus_info, "tun");
1344 strcpy(info->bus_info, "tap");
1349 static u32 tun_get_msglevel(struct net_device *dev)
1352 struct tun_struct *tun = netdev_priv(dev);
1359 static void tun_set_msglevel(struct net_device *dev, u32 value)
1362 struct tun_struct *tun = netdev_priv(dev);
1367 static u32 tun_get_link(struct net_device *dev)
1369 struct tun_struct *tun = netdev_priv(dev);
1370 return !!tun->tfile;
1373 static u32 tun_get_rx_csum(struct net_device *dev)
1375 struct tun_struct *tun = netdev_priv(dev);
1376 return (tun->flags & TUN_NOCHECKSUM) == 0;
1379 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1381 struct tun_struct *tun = netdev_priv(dev);
1383 tun->flags &= ~TUN_NOCHECKSUM;
1385 tun->flags |= TUN_NOCHECKSUM;
1389 static const struct ethtool_ops tun_ethtool_ops = {
1390 .get_settings = tun_get_settings,
1391 .get_drvinfo = tun_get_drvinfo,
1392 .get_msglevel = tun_get_msglevel,
1393 .set_msglevel = tun_set_msglevel,
1394 .get_link = tun_get_link,
1395 .get_rx_csum = tun_get_rx_csum,
1396 .set_rx_csum = tun_set_rx_csum
1400 static int __init tun_init(void)
1404 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1405 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1407 ret = rtnl_link_register(&tun_link_ops);
1409 printk(KERN_ERR "tun: Can't register link_ops\n");
1413 ret = misc_register(&tun_miscdev);
1415 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1420 rtnl_link_unregister(&tun_link_ops);
1425 static void tun_cleanup(void)
1427 misc_deregister(&tun_miscdev);
1428 rtnl_link_unregister(&tun_link_ops);
1431 module_init(tun_init);
1432 module_exit(tun_cleanup);
1433 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1434 MODULE_AUTHOR(DRV_COPYRIGHT);
1435 MODULE_LICENSE("GPL");
1436 MODULE_ALIAS_MISCDEV(TUN_MINOR);