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);
160 atomic_inc(&tfile->count);
163 netif_tx_unlock_bh(tun->dev);
167 static void __tun_detach(struct tun_struct *tun)
169 /* Detach from net device */
170 netif_tx_lock_bh(tun->dev);
172 netif_tx_unlock_bh(tun->dev);
174 /* Drop read queue */
175 skb_queue_purge(&tun->readq);
177 /* Drop the extra count on the net device */
181 static void tun_detach(struct tun_struct *tun)
188 static struct tun_struct *__tun_get(struct tun_file *tfile)
190 struct tun_struct *tun = NULL;
192 if (atomic_inc_not_zero(&tfile->count))
198 static struct tun_struct *tun_get(struct file *file)
200 return __tun_get(file->private_data);
203 static void tun_put(struct tun_struct *tun)
205 struct tun_file *tfile = tun->tfile;
207 if (atomic_dec_and_test(&tfile->count))
208 tun_detach(tfile->tun);
212 static void addr_hash_set(u32 *mask, const u8 *addr)
214 int n = ether_crc(ETH_ALEN, addr) >> 26;
215 mask[n >> 5] |= (1 << (n & 31));
218 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
220 int n = ether_crc(ETH_ALEN, addr) >> 26;
221 return mask[n >> 5] & (1 << (n & 31));
224 static int update_filter(struct tap_filter *filter, void __user *arg)
226 struct { u8 u[ETH_ALEN]; } *addr;
227 struct tun_filter uf;
228 int err, alen, n, nexact;
230 if (copy_from_user(&uf, arg, sizeof(uf)))
239 alen = ETH_ALEN * uf.count;
240 addr = kmalloc(alen, GFP_KERNEL);
244 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
249 /* The filter is updated without holding any locks. Which is
250 * perfectly safe. We disable it first and in the worst
251 * case we'll accept a few undesired packets. */
255 /* Use first set of addresses as an exact filter */
256 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
257 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
261 /* Remaining multicast addresses are hashed,
262 * unicast will leave the filter disabled. */
263 memset(filter->mask, 0, sizeof(filter->mask));
264 for (; n < uf.count; n++) {
265 if (!is_multicast_ether_addr(addr[n].u)) {
266 err = 0; /* no filter */
269 addr_hash_set(filter->mask, addr[n].u);
272 /* For ALLMULTI just set the mask to all ones.
273 * This overrides the mask populated above. */
274 if ((uf.flags & TUN_FLT_ALLMULTI))
275 memset(filter->mask, ~0, sizeof(filter->mask));
277 /* Now enable the filter */
279 filter->count = nexact;
281 /* Return the number of exact filters */
289 /* Returns: 0 - drop, !=0 - accept */
290 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
292 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
294 struct ethhdr *eh = (struct ethhdr *) skb->data;
298 for (i = 0; i < filter->count; i++)
299 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
302 /* Inexact match (multicast only) */
303 if (is_multicast_ether_addr(eh->h_dest))
304 return addr_hash_test(filter->mask, eh->h_dest);
310 * Checks whether the packet is accepted or not.
311 * Returns: 0 - drop, !=0 - accept
313 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
318 return run_filter(filter, skb);
321 /* Network device part of the driver */
323 static const struct ethtool_ops tun_ethtool_ops;
325 /* Net device detach from fd. */
326 static void tun_net_uninit(struct net_device *dev)
328 struct tun_struct *tun = netdev_priv(dev);
329 struct tun_file *tfile = tun->tfile;
331 /* Inform the methods they need to stop using the dev.
334 wake_up_all(&tfile->read_wait);
335 if (atomic_dec_and_test(&tfile->count))
340 static void tun_free_netdev(struct net_device *dev)
342 struct tun_struct *tun = netdev_priv(dev);
347 /* Net device open. */
348 static int tun_net_open(struct net_device *dev)
350 netif_start_queue(dev);
354 /* Net device close. */
355 static int tun_net_close(struct net_device *dev)
357 netif_stop_queue(dev);
361 /* Net device start xmit */
362 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
364 struct tun_struct *tun = netdev_priv(dev);
366 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
368 /* Drop packet if interface is not attached */
372 /* Drop if the filter does not like it.
373 * This is a noop if the filter is disabled.
374 * Filter can be enabled only for the TAP devices. */
375 if (!check_filter(&tun->txflt, skb))
378 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
379 if (!(tun->flags & TUN_ONE_QUEUE)) {
380 /* Normal queueing mode. */
381 /* Packet scheduler handles dropping of further packets. */
382 netif_stop_queue(dev);
384 /* We won't see all dropped packets individually, so overrun
385 * error is more appropriate. */
386 dev->stats.tx_fifo_errors++;
388 /* Single queue mode.
389 * Driver handles dropping of all packets itself. */
395 skb_queue_tail(&tun->readq, skb);
396 dev->trans_start = jiffies;
398 /* Notify and wake up reader process */
399 if (tun->flags & TUN_FASYNC)
400 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
401 wake_up_interruptible(&tun->tfile->read_wait);
405 dev->stats.tx_dropped++;
410 static void tun_net_mclist(struct net_device *dev)
413 * This callback is supposed to deal with mc filter in
414 * _rx_ path and has nothing to do with the _tx_ path.
415 * In rx path we always accept everything userspace gives us.
421 #define MAX_MTU 65535
424 tun_net_change_mtu(struct net_device *dev, int new_mtu)
426 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
432 static const struct net_device_ops tun_netdev_ops = {
433 .ndo_uninit = tun_net_uninit,
434 .ndo_open = tun_net_open,
435 .ndo_stop = tun_net_close,
436 .ndo_start_xmit = tun_net_xmit,
437 .ndo_change_mtu = tun_net_change_mtu,
440 static const struct net_device_ops tap_netdev_ops = {
441 .ndo_uninit = tun_net_uninit,
442 .ndo_open = tun_net_open,
443 .ndo_stop = tun_net_close,
444 .ndo_start_xmit = tun_net_xmit,
445 .ndo_change_mtu = tun_net_change_mtu,
446 .ndo_set_multicast_list = tun_net_mclist,
447 .ndo_set_mac_address = eth_mac_addr,
448 .ndo_validate_addr = eth_validate_addr,
451 /* Initialize net device. */
452 static void tun_net_init(struct net_device *dev)
454 struct tun_struct *tun = netdev_priv(dev);
456 switch (tun->flags & TUN_TYPE_MASK) {
458 dev->netdev_ops = &tun_netdev_ops;
460 /* Point-to-Point TUN Device */
461 dev->hard_header_len = 0;
465 /* Zero header length */
466 dev->type = ARPHRD_NONE;
467 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
468 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
472 dev->netdev_ops = &tap_netdev_ops;
473 /* Ethernet TAP Device */
476 random_ether_addr(dev->dev_addr);
478 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
483 /* Character device part */
486 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
488 struct tun_file *tfile = file->private_data;
489 struct tun_struct *tun = __tun_get(tfile);
490 struct sock *sk = tun->sk;
491 unsigned int mask = 0;
496 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
498 poll_wait(file, &tfile->read_wait, wait);
500 if (!skb_queue_empty(&tun->readq))
501 mask |= POLLIN | POLLRDNORM;
503 if (sock_writeable(sk) ||
504 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
506 mask |= POLLOUT | POLLWRNORM;
508 if (tun->dev->reg_state != NETREG_REGISTERED)
515 /* prepad is the amount to reserve at front. len is length after that.
516 * linear is a hint as to how much to copy (usually headers). */
517 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
518 size_t prepad, size_t len,
519 size_t linear, int noblock)
521 struct sock *sk = tun->sk;
525 /* Under a page? Don't bother with paged skb. */
526 if (prepad + len < PAGE_SIZE || !linear)
529 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
534 skb_reserve(skb, prepad);
535 skb_put(skb, linear);
536 skb->data_len = len - linear;
537 skb->len += len - linear;
542 /* Get packet from user space buffer */
543 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
544 struct iovec *iv, size_t count,
547 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
549 size_t len = count, align = 0;
550 struct virtio_net_hdr gso = { 0 };
552 if (!(tun->flags & TUN_NO_PI)) {
553 if ((len -= sizeof(pi)) > count)
556 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
560 if (tun->flags & TUN_VNET_HDR) {
561 if ((len -= sizeof(gso)) > count)
564 if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso)))
567 if (gso.hdr_len > len)
571 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
572 align = NET_IP_ALIGN;
573 if (unlikely(len < ETH_HLEN ||
574 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
578 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
580 if (PTR_ERR(skb) != -EAGAIN)
581 tun->dev->stats.rx_dropped++;
585 if (skb_copy_datagram_from_iovec(skb, 0, iv, len)) {
586 tun->dev->stats.rx_dropped++;
591 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
592 if (!skb_partial_csum_set(skb, gso.csum_start,
594 tun->dev->stats.rx_frame_errors++;
598 } else if (tun->flags & TUN_NOCHECKSUM)
599 skb->ip_summed = CHECKSUM_UNNECESSARY;
601 switch (tun->flags & TUN_TYPE_MASK) {
603 if (tun->flags & TUN_NO_PI) {
604 switch (skb->data[0] & 0xf0) {
606 pi.proto = htons(ETH_P_IP);
609 pi.proto = htons(ETH_P_IPV6);
612 tun->dev->stats.rx_dropped++;
618 skb_reset_mac_header(skb);
619 skb->protocol = pi.proto;
623 skb->protocol = eth_type_trans(skb, tun->dev);
627 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
629 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
630 case VIRTIO_NET_HDR_GSO_TCPV4:
631 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
633 case VIRTIO_NET_HDR_GSO_TCPV6:
634 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
637 tun->dev->stats.rx_frame_errors++;
642 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
643 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
645 skb_shinfo(skb)->gso_size = gso.gso_size;
646 if (skb_shinfo(skb)->gso_size == 0) {
647 tun->dev->stats.rx_frame_errors++;
652 /* Header must be checked, and gso_segs computed. */
653 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
654 skb_shinfo(skb)->gso_segs = 0;
659 tun->dev->stats.rx_packets++;
660 tun->dev->stats.rx_bytes += len;
665 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
666 unsigned long count, loff_t pos)
668 struct file *file = iocb->ki_filp;
669 struct tun_struct *tun = tun_get(file);
675 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
677 result = tun_get_user(tun, (struct iovec *)iv, iov_length(iv, count),
678 file->f_flags & O_NONBLOCK);
684 /* Put packet to the user space buffer */
685 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
687 struct iovec *iv, int len)
689 struct tun_pi pi = { 0, skb->protocol };
692 if (!(tun->flags & TUN_NO_PI)) {
693 if ((len -= sizeof(pi)) < 0)
696 if (len < skb->len) {
697 /* Packet will be striped */
698 pi.flags |= TUN_PKT_STRIP;
701 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
706 if (tun->flags & TUN_VNET_HDR) {
707 struct virtio_net_hdr gso = { 0 }; /* no info leak */
708 if ((len -= sizeof(gso)) < 0)
711 if (skb_is_gso(skb)) {
712 struct skb_shared_info *sinfo = skb_shinfo(skb);
714 /* This is a hint as to how much should be linear. */
715 gso.hdr_len = skb_headlen(skb);
716 gso.gso_size = sinfo->gso_size;
717 if (sinfo->gso_type & SKB_GSO_TCPV4)
718 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
719 else if (sinfo->gso_type & SKB_GSO_TCPV6)
720 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
723 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
724 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
726 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
728 if (skb->ip_summed == CHECKSUM_PARTIAL) {
729 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
730 gso.csum_start = skb->csum_start - skb_headroom(skb);
731 gso.csum_offset = skb->csum_offset;
732 } /* else everything is zero */
734 if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso))))
736 total += sizeof(gso);
739 len = min_t(int, skb->len, len);
741 skb_copy_datagram_iovec(skb, 0, iv, len);
744 tun->dev->stats.tx_packets++;
745 tun->dev->stats.tx_bytes += len;
750 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
751 unsigned long count, loff_t pos)
753 struct file *file = iocb->ki_filp;
754 struct tun_file *tfile = file->private_data;
755 struct tun_struct *tun = __tun_get(tfile);
756 DECLARE_WAITQUEUE(wait, current);
758 ssize_t len, ret = 0;
763 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
765 len = iov_length(iv, count);
771 add_wait_queue(&tfile->read_wait, &wait);
773 current->state = TASK_INTERRUPTIBLE;
775 /* Read frames from the queue */
776 if (!(skb=skb_dequeue(&tun->readq))) {
777 if (file->f_flags & O_NONBLOCK) {
781 if (signal_pending(current)) {
785 if (tun->dev->reg_state != NETREG_REGISTERED) {
790 /* Nothing to read, let's sleep */
794 netif_wake_queue(tun->dev);
796 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
801 current->state = TASK_RUNNING;
802 remove_wait_queue(&tfile->read_wait, &wait);
809 static void tun_setup(struct net_device *dev)
811 struct tun_struct *tun = netdev_priv(dev);
813 skb_queue_head_init(&tun->readq);
818 dev->ethtool_ops = &tun_ethtool_ops;
819 dev->destructor = tun_free_netdev;
822 /* Trivial set of netlink ops to allow deleting tun or tap
823 * device with netlink.
825 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
830 static struct rtnl_link_ops tun_link_ops __read_mostly = {
832 .priv_size = sizeof(struct tun_struct),
834 .validate = tun_validate,
837 static void tun_sock_write_space(struct sock *sk)
839 struct tun_struct *tun;
841 if (!sock_writeable(sk))
844 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
845 wake_up_interruptible_sync(sk->sk_sleep);
847 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
850 tun = container_of(sk, struct tun_sock, sk)->tun;
851 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
854 static void tun_sock_destruct(struct sock *sk)
856 free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
859 static struct proto tun_proto = {
861 .owner = THIS_MODULE,
862 .obj_size = sizeof(struct tun_sock),
865 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
868 struct tun_struct *tun;
869 struct net_device *dev;
870 struct tun_file *tfile = file->private_data;
873 dev = __dev_get_by_name(net, ifr->ifr_name);
875 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
876 tun = netdev_priv(dev);
877 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
878 tun = netdev_priv(dev);
882 err = tun_attach(tun, file);
888 unsigned long flags = 0;
892 if (!capable(CAP_NET_ADMIN))
896 if (ifr->ifr_flags & IFF_TUN) {
898 flags |= TUN_TUN_DEV;
900 } else if (ifr->ifr_flags & IFF_TAP) {
902 flags |= TUN_TAP_DEV;
908 name = ifr->ifr_name;
910 dev = alloc_netdev(sizeof(struct tun_struct), name,
915 dev_net_set(dev, net);
916 dev->rtnl_link_ops = &tun_link_ops;
918 tun = netdev_priv(dev);
921 tun->txflt.count = 0;
924 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
928 sock_init_data(&tun->socket, sk);
929 sk->sk_write_space = tun_sock_write_space;
930 sk->sk_sndbuf = INT_MAX;
931 sk->sk_sleep = &tfile->read_wait;
934 container_of(sk, struct tun_sock, sk)->tun = tun;
938 if (strchr(dev->name, '%')) {
939 err = dev_alloc_name(dev, dev->name);
945 err = register_netdevice(tun->dev);
949 sk->sk_destruct = tun_sock_destruct;
951 err = tun_attach(tun, file);
956 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
958 if (ifr->ifr_flags & IFF_NO_PI)
959 tun->flags |= TUN_NO_PI;
961 tun->flags &= ~TUN_NO_PI;
963 if (ifr->ifr_flags & IFF_ONE_QUEUE)
964 tun->flags |= TUN_ONE_QUEUE;
966 tun->flags &= ~TUN_ONE_QUEUE;
968 if (ifr->ifr_flags & IFF_VNET_HDR)
969 tun->flags |= TUN_VNET_HDR;
971 tun->flags &= ~TUN_VNET_HDR;
973 /* Make sure persistent devices do not get stuck in
976 if (netif_running(tun->dev))
977 netif_wake_queue(tun->dev);
979 strcpy(ifr->ifr_name, tun->dev->name);
990 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
992 struct tun_struct *tun = tun_get(file);
997 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
999 strcpy(ifr->ifr_name, tun->dev->name);
1003 if (ifr->ifr_flags & TUN_TUN_DEV)
1004 ifr->ifr_flags |= IFF_TUN;
1006 ifr->ifr_flags |= IFF_TAP;
1008 if (tun->flags & TUN_NO_PI)
1009 ifr->ifr_flags |= IFF_NO_PI;
1011 if (tun->flags & TUN_ONE_QUEUE)
1012 ifr->ifr_flags |= IFF_ONE_QUEUE;
1014 if (tun->flags & TUN_VNET_HDR)
1015 ifr->ifr_flags |= IFF_VNET_HDR;
1021 /* This is like a cut-down ethtool ops, except done via tun fd so no
1022 * privs required. */
1023 static int set_offload(struct net_device *dev, unsigned long arg)
1025 unsigned int old_features, features;
1027 old_features = dev->features;
1028 /* Unset features, set them as we chew on the arg. */
1029 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1030 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
1032 if (arg & TUN_F_CSUM) {
1033 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1036 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1037 if (arg & TUN_F_TSO_ECN) {
1038 features |= NETIF_F_TSO_ECN;
1039 arg &= ~TUN_F_TSO_ECN;
1041 if (arg & TUN_F_TSO4)
1042 features |= NETIF_F_TSO;
1043 if (arg & TUN_F_TSO6)
1044 features |= NETIF_F_TSO6;
1045 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1049 /* This gives the user a way to test for new features in future by
1050 * trying to set them. */
1054 dev->features = features;
1055 if (old_features != dev->features)
1056 netdev_features_change(dev);
1061 static int tun_chr_ioctl(struct inode *inode, struct file *file,
1062 unsigned int cmd, unsigned long arg)
1064 struct tun_file *tfile = file->private_data;
1065 struct tun_struct *tun;
1066 void __user* argp = (void __user*)arg;
1071 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1072 if (copy_from_user(&ifr, argp, sizeof ifr))
1075 if (cmd == TUNGETFEATURES) {
1076 /* Currently this just means: "what IFF flags are valid?".
1077 * This is needed because we never checked for invalid flags on
1079 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1081 (unsigned int __user*)argp);
1084 tun = __tun_get(tfile);
1085 if (cmd == TUNSETIFF && !tun) {
1088 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1091 err = tun_set_iff(tfile->net, file, &ifr);
1097 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1106 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1111 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
1115 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1120 /* Disable/Enable checksum */
1122 tun->flags |= TUN_NOCHECKSUM;
1124 tun->flags &= ~TUN_NOCHECKSUM;
1126 DBG(KERN_INFO "%s: checksum %s\n",
1127 tun->dev->name, arg ? "disabled" : "enabled");
1131 /* Disable/Enable persist mode */
1133 tun->flags |= TUN_PERSIST;
1135 tun->flags &= ~TUN_PERSIST;
1137 DBG(KERN_INFO "%s: persist %s\n",
1138 tun->dev->name, arg ? "enabled" : "disabled");
1142 /* Set owner of the device */
1143 tun->owner = (uid_t) arg;
1145 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1149 /* Set group of the device */
1150 tun->group= (gid_t) arg;
1152 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1156 /* Only allow setting the type when the interface is down */
1158 if (tun->dev->flags & IFF_UP) {
1159 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1163 tun->dev->type = (int) arg;
1164 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1177 ret = set_offload(tun->dev, arg);
1181 case TUNSETTXFILTER:
1182 /* Can be set only for TAPs */
1184 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1187 ret = update_filter(&tun->txflt, (void __user *)arg);
1193 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1194 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1195 if (copy_to_user(argp, &ifr, sizeof ifr))
1200 /* Set hw address */
1201 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1202 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1205 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1210 sndbuf = tun->sk->sk_sndbuf;
1211 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1216 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1221 tun->sk->sk_sndbuf = sndbuf;
1233 static int tun_chr_fasync(int fd, struct file *file, int on)
1235 struct tun_struct *tun = tun_get(file);
1241 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1244 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1248 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1251 tun->flags |= TUN_FASYNC;
1253 tun->flags &= ~TUN_FASYNC;
1261 static int tun_chr_open(struct inode *inode, struct file * file)
1263 struct tun_file *tfile;
1264 cycle_kernel_lock();
1265 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1267 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1270 atomic_set(&tfile->count, 0);
1272 tfile->net = get_net(current->nsproxy->net_ns);
1273 init_waitqueue_head(&tfile->read_wait);
1274 file->private_data = tfile;
1278 static int tun_chr_close(struct inode *inode, struct file *file)
1280 struct tun_file *tfile = file->private_data;
1281 struct tun_struct *tun = __tun_get(tfile);
1285 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1290 /* If desireable, unregister the netdevice. */
1291 if (!(tun->flags & TUN_PERSIST))
1292 unregister_netdevice(tun->dev);
1301 put_net(tfile->net);
1307 static const struct file_operations tun_fops = {
1308 .owner = THIS_MODULE,
1309 .llseek = no_llseek,
1310 .read = do_sync_read,
1311 .aio_read = tun_chr_aio_read,
1312 .write = do_sync_write,
1313 .aio_write = tun_chr_aio_write,
1314 .poll = tun_chr_poll,
1315 .ioctl = tun_chr_ioctl,
1316 .open = tun_chr_open,
1317 .release = tun_chr_close,
1318 .fasync = tun_chr_fasync
1321 static struct miscdevice tun_miscdev = {
1327 /* ethtool interface */
1329 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1332 cmd->advertising = 0;
1333 cmd->speed = SPEED_10;
1334 cmd->duplex = DUPLEX_FULL;
1335 cmd->port = PORT_TP;
1336 cmd->phy_address = 0;
1337 cmd->transceiver = XCVR_INTERNAL;
1338 cmd->autoneg = AUTONEG_DISABLE;
1344 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1346 struct tun_struct *tun = netdev_priv(dev);
1348 strcpy(info->driver, DRV_NAME);
1349 strcpy(info->version, DRV_VERSION);
1350 strcpy(info->fw_version, "N/A");
1352 switch (tun->flags & TUN_TYPE_MASK) {
1354 strcpy(info->bus_info, "tun");
1357 strcpy(info->bus_info, "tap");
1362 static u32 tun_get_msglevel(struct net_device *dev)
1365 struct tun_struct *tun = netdev_priv(dev);
1372 static void tun_set_msglevel(struct net_device *dev, u32 value)
1375 struct tun_struct *tun = netdev_priv(dev);
1380 static u32 tun_get_link(struct net_device *dev)
1382 struct tun_struct *tun = netdev_priv(dev);
1383 return !!tun->tfile;
1386 static u32 tun_get_rx_csum(struct net_device *dev)
1388 struct tun_struct *tun = netdev_priv(dev);
1389 return (tun->flags & TUN_NOCHECKSUM) == 0;
1392 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1394 struct tun_struct *tun = netdev_priv(dev);
1396 tun->flags &= ~TUN_NOCHECKSUM;
1398 tun->flags |= TUN_NOCHECKSUM;
1402 static const struct ethtool_ops tun_ethtool_ops = {
1403 .get_settings = tun_get_settings,
1404 .get_drvinfo = tun_get_drvinfo,
1405 .get_msglevel = tun_get_msglevel,
1406 .set_msglevel = tun_set_msglevel,
1407 .get_link = tun_get_link,
1408 .get_rx_csum = tun_get_rx_csum,
1409 .set_rx_csum = tun_set_rx_csum
1413 static int __init tun_init(void)
1417 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1418 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1420 ret = rtnl_link_register(&tun_link_ops);
1422 printk(KERN_ERR "tun: Can't register link_ops\n");
1426 ret = misc_register(&tun_miscdev);
1428 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1433 rtnl_link_unregister(&tun_link_ops);
1438 static void tun_cleanup(void)
1440 misc_deregister(&tun_miscdev);
1441 rtnl_link_unregister(&tun_link_ops);
1444 module_init(tun_init);
1445 module_exit(tun_cleanup);
1446 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1447 MODULE_AUTHOR(DRV_COPYRIGHT);
1448 MODULE_LICENSE("GPL");
1449 MODULE_ALIAS_MISCDEV(TUN_MINOR);