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 * Brian Braunstein <linuxkernel@bristyle.com> 2007/03/23
22 * Fixed hw address handling. Now net_device.dev_addr is kept consistent
23 * with tun.dev_addr when the address is set by this module.
25 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
26 * Add TUNSETLINK ioctl to set the link encapsulation
28 * Mark Smith <markzzzsmith@yahoo.com.au>
29 * Use random_ether_addr() for tap MAC address.
31 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
32 * Fixes in packet dropping, queue length setting and queue wakeup.
33 * Increased default tx queue length.
37 * Daniel Podlejski <underley@underley.eu.org>
38 * Modifications for 2.3.99-pre5 kernel.
41 #define DRV_NAME "tun"
42 #define DRV_VERSION "1.6"
43 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
44 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
46 #include <linux/module.h>
47 #include <linux/errno.h>
48 #include <linux/kernel.h>
49 #include <linux/major.h>
50 #include <linux/slab.h>
51 #include <linux/poll.h>
52 #include <linux/fcntl.h>
53 #include <linux/init.h>
54 #include <linux/skbuff.h>
55 #include <linux/netdevice.h>
56 #include <linux/etherdevice.h>
57 #include <linux/miscdevice.h>
58 #include <linux/ethtool.h>
59 #include <linux/rtnetlink.h>
61 #include <linux/if_arp.h>
62 #include <linux/if_ether.h>
63 #include <linux/if_tun.h>
64 #include <linux/crc32.h>
65 #include <net/net_namespace.h>
67 #include <asm/system.h>
68 #include <asm/uaccess.h>
74 /* Network device part of the driver */
76 static LIST_HEAD(tun_dev_list);
77 static const struct ethtool_ops tun_ethtool_ops;
79 /* Net device open. */
80 static int tun_net_open(struct net_device *dev)
82 netif_start_queue(dev);
86 /* Net device close. */
87 static int tun_net_close(struct net_device *dev)
89 netif_stop_queue(dev);
93 /* Net device start xmit */
94 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
96 struct tun_struct *tun = netdev_priv(dev);
98 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
100 /* Drop packet if interface is not attached */
104 /* Packet dropping */
105 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
106 if (!(tun->flags & TUN_ONE_QUEUE)) {
107 /* Normal queueing mode. */
108 /* Packet scheduler handles dropping of further packets. */
109 netif_stop_queue(dev);
111 /* We won't see all dropped packets individually, so overrun
112 * error is more appropriate. */
113 dev->stats.tx_fifo_errors++;
115 /* Single queue mode.
116 * Driver handles dropping of all packets itself. */
122 skb_queue_tail(&tun->readq, skb);
123 dev->trans_start = jiffies;
125 /* Notify and wake up reader process */
126 if (tun->flags & TUN_FASYNC)
127 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
128 wake_up_interruptible(&tun->read_wait);
132 dev->stats.tx_dropped++;
137 /** Add the specified Ethernet address to this multicast filter. */
139 add_multi(u32* filter, const u8* addr)
141 int bit_nr = ether_crc(ETH_ALEN, addr) >> 26;
142 filter[bit_nr >> 5] |= 1 << (bit_nr & 31);
145 /** Remove the specified Ethernet addres from this multicast filter. */
147 del_multi(u32* filter, const u8* addr)
149 int bit_nr = ether_crc(ETH_ALEN, addr) >> 26;
150 filter[bit_nr >> 5] &= ~(1 << (bit_nr & 31));
153 /** Update the list of multicast groups to which the network device belongs.
154 * This list is used to filter packets being sent from the character device to
155 * the network device. */
157 tun_net_mclist(struct net_device *dev)
159 struct tun_struct *tun = netdev_priv(dev);
160 const struct dev_mc_list *mclist;
162 DBG(KERN_DEBUG "%s: tun_net_mclist: mc_count %d\n",
163 dev->name, dev->mc_count);
164 memset(tun->chr_filter, 0, sizeof tun->chr_filter);
165 for (i = 0, mclist = dev->mc_list; i < dev->mc_count && mclist != NULL;
166 i++, mclist = mclist->next) {
167 add_multi(tun->net_filter, mclist->dmi_addr);
168 DBG(KERN_DEBUG "%s: tun_net_mclist: %x:%x:%x:%x:%x:%x\n",
170 mclist->dmi_addr[0], mclist->dmi_addr[1], mclist->dmi_addr[2],
171 mclist->dmi_addr[3], mclist->dmi_addr[4], mclist->dmi_addr[5]);
176 #define MAX_MTU 65535
179 tun_net_change_mtu(struct net_device *dev, int new_mtu)
181 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
187 /* Initialize net device. */
188 static void tun_net_init(struct net_device *dev)
190 struct tun_struct *tun = netdev_priv(dev);
192 switch (tun->flags & TUN_TYPE_MASK) {
194 /* Point-to-Point TUN Device */
195 dev->hard_header_len = 0;
198 dev->change_mtu = tun_net_change_mtu;
200 /* Zero header length */
201 dev->type = ARPHRD_NONE;
202 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
203 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
207 /* Ethernet TAP Device */
208 dev->set_multicast_list = tun_net_mclist;
211 dev->change_mtu = tun_net_change_mtu;
213 /* random address already created for us by tun_set_iff, use it */
214 memcpy(dev->dev_addr, tun->dev_addr, min(sizeof(tun->dev_addr), sizeof(dev->dev_addr)) );
216 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
221 /* Character device part */
224 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
226 struct tun_struct *tun = file->private_data;
227 unsigned int mask = POLLOUT | POLLWRNORM;
232 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
234 poll_wait(file, &tun->read_wait, wait);
236 if (!skb_queue_empty(&tun->readq))
237 mask |= POLLIN | POLLRDNORM;
242 /* Get packet from user space buffer */
243 static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, size_t count)
245 struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) };
247 size_t len = count, align = 0;
249 if (!(tun->flags & TUN_NO_PI)) {
250 if ((len -= sizeof(pi)) > count)
253 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
257 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV)
258 align = NET_IP_ALIGN;
260 if (!(skb = alloc_skb(len + align, GFP_KERNEL))) {
261 tun->dev->stats.rx_dropped++;
266 skb_reserve(skb, align);
267 if (memcpy_fromiovec(skb_put(skb, len), iv, len)) {
268 tun->dev->stats.rx_dropped++;
273 switch (tun->flags & TUN_TYPE_MASK) {
275 skb_reset_mac_header(skb);
276 skb->protocol = pi.proto;
280 skb->protocol = eth_type_trans(skb, tun->dev);
284 if (tun->flags & TUN_NOCHECKSUM)
285 skb->ip_summed = CHECKSUM_UNNECESSARY;
288 tun->dev->last_rx = jiffies;
290 tun->dev->stats.rx_packets++;
291 tun->dev->stats.rx_bytes += len;
296 static inline size_t iov_total(const struct iovec *iv, unsigned long count)
301 for (i = 0, len = 0; i < count; i++)
302 len += iv[i].iov_len;
307 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
308 unsigned long count, loff_t pos)
310 struct tun_struct *tun = iocb->ki_filp->private_data;
315 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
317 return tun_get_user(tun, (struct iovec *) iv, iov_total(iv, count));
320 /* Put packet to the user space buffer */
321 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
323 struct iovec *iv, int len)
325 struct tun_pi pi = { 0, skb->protocol };
328 if (!(tun->flags & TUN_NO_PI)) {
329 if ((len -= sizeof(pi)) < 0)
332 if (len < skb->len) {
333 /* Packet will be striped */
334 pi.flags |= TUN_PKT_STRIP;
337 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
342 len = min_t(int, skb->len, len);
344 skb_copy_datagram_iovec(skb, 0, iv, len);
347 tun->dev->stats.tx_packets++;
348 tun->dev->stats.tx_bytes += len;
353 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
354 unsigned long count, loff_t pos)
356 struct file *file = iocb->ki_filp;
357 struct tun_struct *tun = file->private_data;
358 DECLARE_WAITQUEUE(wait, current);
360 ssize_t len, ret = 0;
365 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
367 len = iov_total(iv, count);
371 add_wait_queue(&tun->read_wait, &wait);
373 const u8 ones[ ETH_ALEN] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
377 current->state = TASK_INTERRUPTIBLE;
379 /* Read frames from the queue */
380 if (!(skb=skb_dequeue(&tun->readq))) {
381 if (file->f_flags & O_NONBLOCK) {
385 if (signal_pending(current)) {
390 /* Nothing to read, let's sleep */
394 netif_wake_queue(tun->dev);
396 /** Decide whether to accept this packet. This code is designed to
397 * behave identically to an Ethernet interface. Accept the packet if
398 * - we are promiscuous.
399 * - the packet is addressed to us.
400 * - the packet is broadcast.
401 * - the packet is multicast and
402 * - we are multicast promiscous.
403 * - we belong to the multicast group.
405 skb_copy_from_linear_data(skb, addr, min_t(size_t, sizeof addr,
407 bit_nr = ether_crc(sizeof addr, addr) >> 26;
408 if ((tun->if_flags & IFF_PROMISC) ||
409 memcmp(addr, tun->dev_addr, sizeof addr) == 0 ||
410 memcmp(addr, ones, sizeof addr) == 0 ||
411 (((addr[0] == 1 && addr[1] == 0 && addr[2] == 0x5e) ||
412 (addr[0] == 0x33 && addr[1] == 0x33)) &&
413 ((tun->if_flags & IFF_ALLMULTI) ||
414 (tun->chr_filter[bit_nr >> 5] & (1 << (bit_nr & 31)))))) {
415 DBG(KERN_DEBUG "%s: tun_chr_readv: accepted: %x:%x:%x:%x:%x:%x\n",
416 tun->dev->name, addr[0], addr[1], addr[2],
417 addr[3], addr[4], addr[5]);
418 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
422 DBG(KERN_DEBUG "%s: tun_chr_readv: rejected: %x:%x:%x:%x:%x:%x\n",
423 tun->dev->name, addr[0], addr[1], addr[2],
424 addr[3], addr[4], addr[5]);
430 current->state = TASK_RUNNING;
431 remove_wait_queue(&tun->read_wait, &wait);
436 static void tun_setup(struct net_device *dev)
438 struct tun_struct *tun = netdev_priv(dev);
440 skb_queue_head_init(&tun->readq);
441 init_waitqueue_head(&tun->read_wait);
446 dev->open = tun_net_open;
447 dev->hard_start_xmit = tun_net_xmit;
448 dev->stop = tun_net_close;
449 dev->ethtool_ops = &tun_ethtool_ops;
450 dev->destructor = free_netdev;
453 static struct tun_struct *tun_get_by_name(const char *name)
455 struct tun_struct *tun;
458 list_for_each_entry(tun, &tun_dev_list, list) {
459 if (!strncmp(tun->dev->name, name, IFNAMSIZ))
466 static int tun_set_iff(struct file *file, struct ifreq *ifr)
468 struct tun_struct *tun;
469 struct net_device *dev;
472 tun = tun_get_by_name(ifr->ifr_name);
477 /* Check permissions */
478 if (((tun->owner != -1 &&
479 current->euid != tun->owner) ||
481 current->egid != tun->group)) &&
482 !capable(CAP_NET_ADMIN))
485 else if (__dev_get_by_name(&init_net, ifr->ifr_name))
489 unsigned long flags = 0;
493 if (!capable(CAP_NET_ADMIN))
497 if (ifr->ifr_flags & IFF_TUN) {
499 flags |= TUN_TUN_DEV;
501 } else if (ifr->ifr_flags & IFF_TAP) {
503 flags |= TUN_TAP_DEV;
509 name = ifr->ifr_name;
511 dev = alloc_netdev(sizeof(struct tun_struct), name,
516 tun = netdev_priv(dev);
519 /* Be promiscuous by default to maintain previous behaviour. */
520 tun->if_flags = IFF_PROMISC;
521 /* Generate random Ethernet address. */
522 *(u16 *)tun->dev_addr = htons(0x00FF);
523 get_random_bytes(tun->dev_addr + sizeof(u16), 4);
524 memset(tun->chr_filter, 0, sizeof tun->chr_filter);
528 if (strchr(dev->name, '%')) {
529 err = dev_alloc_name(dev, dev->name);
534 err = register_netdevice(tun->dev);
538 list_add(&tun->list, &tun_dev_list);
541 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
543 if (ifr->ifr_flags & IFF_NO_PI)
544 tun->flags |= TUN_NO_PI;
546 if (ifr->ifr_flags & IFF_ONE_QUEUE)
547 tun->flags |= TUN_ONE_QUEUE;
549 file->private_data = tun;
552 strcpy(ifr->ifr_name, tun->dev->name);
561 static int tun_chr_ioctl(struct inode *inode, struct file *file,
562 unsigned int cmd, unsigned long arg)
564 struct tun_struct *tun = file->private_data;
565 void __user* argp = (void __user*)arg;
568 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
569 if (copy_from_user(&ifr, argp, sizeof ifr))
572 if (cmd == TUNSETIFF && !tun) {
575 ifr.ifr_name[IFNAMSIZ-1] = '\0';
578 err = tun_set_iff(file, &ifr);
584 if (copy_to_user(argp, &ifr, sizeof(ifr)))
592 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
596 /* Disable/Enable checksum */
598 tun->flags |= TUN_NOCHECKSUM;
600 tun->flags &= ~TUN_NOCHECKSUM;
602 DBG(KERN_INFO "%s: checksum %s\n",
603 tun->dev->name, arg ? "disabled" : "enabled");
607 /* Disable/Enable persist mode */
609 tun->flags |= TUN_PERSIST;
611 tun->flags &= ~TUN_PERSIST;
613 DBG(KERN_INFO "%s: persist %s\n",
614 tun->dev->name, arg ? "disabled" : "enabled");
618 /* Set owner of the device */
619 tun->owner = (uid_t) arg;
621 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
625 /* Set group of the device */
626 tun->group= (gid_t) arg;
628 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
632 /* Only allow setting the type when the interface is down */
633 if (tun->dev->flags & IFF_UP) {
634 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
638 tun->dev->type = (int) arg;
639 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
650 ifr.ifr_flags = tun->if_flags;
651 if (copy_to_user( argp, &ifr, sizeof ifr))
656 /** Set the character device's interface flags. Currently only
657 * IFF_PROMISC and IFF_ALLMULTI are used. */
658 tun->if_flags = ifr.ifr_flags;
659 DBG(KERN_INFO "%s: interface flags 0x%lx\n",
660 tun->dev->name, tun->if_flags);
664 /* Note: the actual net device's address may be different */
665 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev_addr,
666 min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
667 if (copy_to_user( argp, &ifr, sizeof ifr))
673 /* try to set the actual net device's hw address */
674 int ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
677 /** Set the character device's hardware address. This is used when
678 * filtering packets being sent from the network device to the character
680 memcpy(tun->dev_addr, ifr.ifr_hwaddr.sa_data,
681 min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
682 DBG(KERN_DEBUG "%s: set hardware address: %x:%x:%x:%x:%x:%x\n",
684 tun->dev_addr[0], tun->dev_addr[1], tun->dev_addr[2],
685 tun->dev_addr[3], tun->dev_addr[4], tun->dev_addr[5]);
692 /** Add the specified group to the character device's multicast filter
694 add_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
695 DBG(KERN_DEBUG "%s: add multi: %x:%x:%x:%x:%x:%x\n",
697 (u8)ifr.ifr_hwaddr.sa_data[0], (u8)ifr.ifr_hwaddr.sa_data[1],
698 (u8)ifr.ifr_hwaddr.sa_data[2], (u8)ifr.ifr_hwaddr.sa_data[3],
699 (u8)ifr.ifr_hwaddr.sa_data[4], (u8)ifr.ifr_hwaddr.sa_data[5]);
703 /** Remove the specified group from the character device's multicast
705 del_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
706 DBG(KERN_DEBUG "%s: del multi: %x:%x:%x:%x:%x:%x\n",
708 (u8)ifr.ifr_hwaddr.sa_data[0], (u8)ifr.ifr_hwaddr.sa_data[1],
709 (u8)ifr.ifr_hwaddr.sa_data[2], (u8)ifr.ifr_hwaddr.sa_data[3],
710 (u8)ifr.ifr_hwaddr.sa_data[4], (u8)ifr.ifr_hwaddr.sa_data[5]);
720 static int tun_chr_fasync(int fd, struct file *file, int on)
722 struct tun_struct *tun = file->private_data;
728 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
730 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
734 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
737 tun->flags |= TUN_FASYNC;
739 tun->flags &= ~TUN_FASYNC;
744 static int tun_chr_open(struct inode *inode, struct file * file)
746 DBG1(KERN_INFO "tunX: tun_chr_open\n");
747 file->private_data = NULL;
751 static int tun_chr_close(struct inode *inode, struct file *file)
753 struct tun_struct *tun = file->private_data;
758 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
760 tun_chr_fasync(-1, file, 0);
764 /* Detach from net device */
765 file->private_data = NULL;
768 /* Drop read queue */
769 skb_queue_purge(&tun->readq);
771 if (!(tun->flags & TUN_PERSIST)) {
772 list_del(&tun->list);
773 unregister_netdevice(tun->dev);
781 static const struct file_operations tun_fops = {
782 .owner = THIS_MODULE,
784 .read = do_sync_read,
785 .aio_read = tun_chr_aio_read,
786 .write = do_sync_write,
787 .aio_write = tun_chr_aio_write,
788 .poll = tun_chr_poll,
789 .ioctl = tun_chr_ioctl,
790 .open = tun_chr_open,
791 .release = tun_chr_close,
792 .fasync = tun_chr_fasync
795 static struct miscdevice tun_miscdev = {
801 /* ethtool interface */
803 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
806 cmd->advertising = 0;
807 cmd->speed = SPEED_10;
808 cmd->duplex = DUPLEX_FULL;
810 cmd->phy_address = 0;
811 cmd->transceiver = XCVR_INTERNAL;
812 cmd->autoneg = AUTONEG_DISABLE;
818 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
820 struct tun_struct *tun = netdev_priv(dev);
822 strcpy(info->driver, DRV_NAME);
823 strcpy(info->version, DRV_VERSION);
824 strcpy(info->fw_version, "N/A");
826 switch (tun->flags & TUN_TYPE_MASK) {
828 strcpy(info->bus_info, "tun");
831 strcpy(info->bus_info, "tap");
836 static u32 tun_get_msglevel(struct net_device *dev)
839 struct tun_struct *tun = netdev_priv(dev);
846 static void tun_set_msglevel(struct net_device *dev, u32 value)
849 struct tun_struct *tun = netdev_priv(dev);
854 static u32 tun_get_link(struct net_device *dev)
856 struct tun_struct *tun = netdev_priv(dev);
857 return tun->attached;
860 static u32 tun_get_rx_csum(struct net_device *dev)
862 struct tun_struct *tun = netdev_priv(dev);
863 return (tun->flags & TUN_NOCHECKSUM) == 0;
866 static int tun_set_rx_csum(struct net_device *dev, u32 data)
868 struct tun_struct *tun = netdev_priv(dev);
870 tun->flags &= ~TUN_NOCHECKSUM;
872 tun->flags |= TUN_NOCHECKSUM;
876 static const struct ethtool_ops tun_ethtool_ops = {
877 .get_settings = tun_get_settings,
878 .get_drvinfo = tun_get_drvinfo,
879 .get_msglevel = tun_get_msglevel,
880 .set_msglevel = tun_set_msglevel,
881 .get_link = tun_get_link,
882 .get_rx_csum = tun_get_rx_csum,
883 .set_rx_csum = tun_set_rx_csum
886 static int __init tun_init(void)
890 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
891 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
893 ret = misc_register(&tun_miscdev);
895 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
899 static void tun_cleanup(void)
901 struct tun_struct *tun, *nxt;
903 misc_deregister(&tun_miscdev);
906 list_for_each_entry_safe(tun, nxt, &tun_dev_list, list) {
907 DBG(KERN_INFO "%s cleaned up\n", tun->dev->name);
908 unregister_netdevice(tun->dev);
914 module_init(tun_init);
915 module_exit(tun_cleanup);
916 MODULE_DESCRIPTION(DRV_DESCRIPTION);
917 MODULE_AUTHOR(DRV_COPYRIGHT);
918 MODULE_LICENSE("GPL");
919 MODULE_ALIAS_MISCDEV(TUN_MINOR);