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 pr_fmt(fmt) KBUILD_MODNAME ": " fmt
39 #define DRV_NAME "tun"
40 #define DRV_VERSION "1.6"
41 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
42 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
44 #include <linux/module.h>
45 #include <linux/errno.h>
46 #include <linux/kernel.h>
47 #include <linux/major.h>
48 #include <linux/slab.h>
49 #include <linux/poll.h>
50 #include <linux/fcntl.h>
51 #include <linux/init.h>
52 #include <linux/skbuff.h>
53 #include <linux/netdevice.h>
54 #include <linux/etherdevice.h>
55 #include <linux/miscdevice.h>
56 #include <linux/ethtool.h>
57 #include <linux/rtnetlink.h>
58 #include <linux/compat.h>
60 #include <linux/if_arp.h>
61 #include <linux/if_ether.h>
62 #include <linux/if_tun.h>
63 #include <linux/crc32.h>
64 #include <linux/nsproxy.h>
65 #include <linux/virtio_net.h>
66 #include <linux/rcupdate.h>
67 #include <net/net_namespace.h>
68 #include <net/netns/generic.h>
69 #include <net/rtnetlink.h>
72 #include <asm/system.h>
73 #include <asm/uaccess.h>
75 /* Uncomment to enable debugging */
76 /* #define TUN_DEBUG 1 */
81 #define tun_debug(level, tun, fmt, args...) \
84 netdev_printk(level, tun->dev, fmt, ##args); \
86 #define DBG1(level, fmt, args...) \
89 printk(level fmt, ##args); \
92 #define tun_debug(level, tun, fmt, args...) \
95 netdev_printk(level, tun->dev, fmt, ##args); \
97 #define DBG1(level, fmt, args...) \
100 printk(level fmt, ##args); \
104 #define FLT_EXACT_COUNT 8
106 unsigned int count; /* Number of addrs. Zero means disabled */
107 u32 mask[2]; /* Mask of the hashed addrs */
108 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
113 struct tun_struct *tun;
120 struct tun_file *tfile;
125 struct net_device *dev;
127 #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
128 NETIF_F_TSO6|NETIF_F_UFO)
129 struct fasync_struct *fasync;
131 struct tap_filter txflt;
132 struct socket socket;
144 struct tun_struct *tun;
147 static inline struct tun_sock *tun_sk(struct sock *sk)
149 return container_of(sk, struct tun_sock, sk);
152 static int tun_attach(struct tun_struct *tun, struct file *file)
154 struct tun_file *tfile = file->private_data;
159 netif_tx_lock_bh(tun->dev);
172 tun->socket.file = file;
173 netif_carrier_on(tun->dev);
175 sock_hold(tun->socket.sk);
176 atomic_inc(&tfile->count);
179 netif_tx_unlock_bh(tun->dev);
183 static void __tun_detach(struct tun_struct *tun)
185 /* Detach from net device */
186 netif_tx_lock_bh(tun->dev);
187 netif_carrier_off(tun->dev);
189 tun->socket.file = NULL;
190 netif_tx_unlock_bh(tun->dev);
192 /* Drop read queue */
193 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
195 /* Drop the extra count on the net device */
199 static void tun_detach(struct tun_struct *tun)
206 static struct tun_struct *__tun_get(struct tun_file *tfile)
208 struct tun_struct *tun = NULL;
210 if (atomic_inc_not_zero(&tfile->count))
216 static struct tun_struct *tun_get(struct file *file)
218 return __tun_get(file->private_data);
221 static void tun_put(struct tun_struct *tun)
223 struct tun_file *tfile = tun->tfile;
225 if (atomic_dec_and_test(&tfile->count))
226 tun_detach(tfile->tun);
230 static void addr_hash_set(u32 *mask, const u8 *addr)
232 int n = ether_crc(ETH_ALEN, addr) >> 26;
233 mask[n >> 5] |= (1 << (n & 31));
236 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
238 int n = ether_crc(ETH_ALEN, addr) >> 26;
239 return mask[n >> 5] & (1 << (n & 31));
242 static int update_filter(struct tap_filter *filter, void __user *arg)
244 struct { u8 u[ETH_ALEN]; } *addr;
245 struct tun_filter uf;
246 int err, alen, n, nexact;
248 if (copy_from_user(&uf, arg, sizeof(uf)))
257 alen = ETH_ALEN * uf.count;
258 addr = kmalloc(alen, GFP_KERNEL);
262 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
267 /* The filter is updated without holding any locks. Which is
268 * perfectly safe. We disable it first and in the worst
269 * case we'll accept a few undesired packets. */
273 /* Use first set of addresses as an exact filter */
274 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
275 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
279 /* Remaining multicast addresses are hashed,
280 * unicast will leave the filter disabled. */
281 memset(filter->mask, 0, sizeof(filter->mask));
282 for (; n < uf.count; n++) {
283 if (!is_multicast_ether_addr(addr[n].u)) {
284 err = 0; /* no filter */
287 addr_hash_set(filter->mask, addr[n].u);
290 /* For ALLMULTI just set the mask to all ones.
291 * This overrides the mask populated above. */
292 if ((uf.flags & TUN_FLT_ALLMULTI))
293 memset(filter->mask, ~0, sizeof(filter->mask));
295 /* Now enable the filter */
297 filter->count = nexact;
299 /* Return the number of exact filters */
307 /* Returns: 0 - drop, !=0 - accept */
308 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
310 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
312 struct ethhdr *eh = (struct ethhdr *) skb->data;
316 for (i = 0; i < filter->count; i++)
317 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
320 /* Inexact match (multicast only) */
321 if (is_multicast_ether_addr(eh->h_dest))
322 return addr_hash_test(filter->mask, eh->h_dest);
328 * Checks whether the packet is accepted or not.
329 * Returns: 0 - drop, !=0 - accept
331 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
336 return run_filter(filter, skb);
339 /* Network device part of the driver */
341 static const struct ethtool_ops tun_ethtool_ops;
343 /* Net device detach from fd. */
344 static void tun_net_uninit(struct net_device *dev)
346 struct tun_struct *tun = netdev_priv(dev);
347 struct tun_file *tfile = tun->tfile;
349 /* Inform the methods they need to stop using the dev.
352 wake_up_all(&tun->wq.wait);
353 if (atomic_dec_and_test(&tfile->count))
358 static void tun_free_netdev(struct net_device *dev)
360 struct tun_struct *tun = netdev_priv(dev);
362 sock_put(tun->socket.sk);
365 /* Net device open. */
366 static int tun_net_open(struct net_device *dev)
368 netif_start_queue(dev);
372 /* Net device close. */
373 static int tun_net_close(struct net_device *dev)
375 netif_stop_queue(dev);
379 /* Net device start xmit */
380 static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
382 struct tun_struct *tun = netdev_priv(dev);
384 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
386 /* Drop packet if interface is not attached */
390 /* Drop if the filter does not like it.
391 * This is a noop if the filter is disabled.
392 * Filter can be enabled only for the TAP devices. */
393 if (!check_filter(&tun->txflt, skb))
396 if (tun->socket.sk->sk_filter &&
397 sk_filter(tun->socket.sk, skb))
400 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
401 if (!(tun->flags & TUN_ONE_QUEUE)) {
402 /* Normal queueing mode. */
403 /* Packet scheduler handles dropping of further packets. */
404 netif_stop_queue(dev);
406 /* We won't see all dropped packets individually, so overrun
407 * error is more appropriate. */
408 dev->stats.tx_fifo_errors++;
410 /* Single queue mode.
411 * Driver handles dropping of all packets itself. */
416 /* Orphan the skb - required as we might hang on to it
417 * for indefinite time. */
421 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
423 /* Notify and wake up reader process */
424 if (tun->flags & TUN_FASYNC)
425 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
426 wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
427 POLLRDNORM | POLLRDBAND);
431 dev->stats.tx_dropped++;
436 static void tun_net_mclist(struct net_device *dev)
439 * This callback is supposed to deal with mc filter in
440 * _rx_ path and has nothing to do with the _tx_ path.
441 * In rx path we always accept everything userspace gives us.
446 #define MAX_MTU 65535
449 tun_net_change_mtu(struct net_device *dev, int new_mtu)
451 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
457 static u32 tun_net_fix_features(struct net_device *dev, u32 features)
459 struct tun_struct *tun = netdev_priv(dev);
461 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
463 #ifdef CONFIG_NET_POLL_CONTROLLER
464 static void tun_poll_controller(struct net_device *dev)
467 * Tun only receives frames when:
468 * 1) the char device endpoint gets data from user space
469 * 2) the tun socket gets a sendmsg call from user space
470 * Since both of those are syncronous operations, we are guaranteed
471 * never to have pending data when we poll for it
472 * so theres nothing to do here but return.
473 * We need this though so netpoll recognizes us as an interface that
474 * supports polling, which enables bridge devices in virt setups to
475 * still use netconsole
480 static const struct net_device_ops tun_netdev_ops = {
481 .ndo_uninit = tun_net_uninit,
482 .ndo_open = tun_net_open,
483 .ndo_stop = tun_net_close,
484 .ndo_start_xmit = tun_net_xmit,
485 .ndo_change_mtu = tun_net_change_mtu,
486 .ndo_fix_features = tun_net_fix_features,
487 #ifdef CONFIG_NET_POLL_CONTROLLER
488 .ndo_poll_controller = tun_poll_controller,
492 static const struct net_device_ops tap_netdev_ops = {
493 .ndo_uninit = tun_net_uninit,
494 .ndo_open = tun_net_open,
495 .ndo_stop = tun_net_close,
496 .ndo_start_xmit = tun_net_xmit,
497 .ndo_change_mtu = tun_net_change_mtu,
498 .ndo_fix_features = tun_net_fix_features,
499 .ndo_set_multicast_list = tun_net_mclist,
500 .ndo_set_mac_address = eth_mac_addr,
501 .ndo_validate_addr = eth_validate_addr,
502 #ifdef CONFIG_NET_POLL_CONTROLLER
503 .ndo_poll_controller = tun_poll_controller,
507 /* Initialize net device. */
508 static void tun_net_init(struct net_device *dev)
510 struct tun_struct *tun = netdev_priv(dev);
512 switch (tun->flags & TUN_TYPE_MASK) {
514 dev->netdev_ops = &tun_netdev_ops;
516 /* Point-to-Point TUN Device */
517 dev->hard_header_len = 0;
521 /* Zero header length */
522 dev->type = ARPHRD_NONE;
523 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
524 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
528 dev->netdev_ops = &tap_netdev_ops;
529 /* Ethernet TAP Device */
532 random_ether_addr(dev->dev_addr);
534 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
539 /* Character device part */
542 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
544 struct tun_file *tfile = file->private_data;
545 struct tun_struct *tun = __tun_get(tfile);
547 unsigned int mask = 0;
554 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
556 poll_wait(file, &tun->wq.wait, wait);
558 if (!skb_queue_empty(&sk->sk_receive_queue))
559 mask |= POLLIN | POLLRDNORM;
561 if (sock_writeable(sk) ||
562 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
564 mask |= POLLOUT | POLLWRNORM;
566 if (tun->dev->reg_state != NETREG_REGISTERED)
573 /* prepad is the amount to reserve at front. len is length after that.
574 * linear is a hint as to how much to copy (usually headers). */
575 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
576 size_t prepad, size_t len,
577 size_t linear, int noblock)
579 struct sock *sk = tun->socket.sk;
583 sock_update_classid(sk);
585 /* Under a page? Don't bother with paged skb. */
586 if (prepad + len < PAGE_SIZE || !linear)
589 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
594 skb_reserve(skb, prepad);
595 skb_put(skb, linear);
596 skb->data_len = len - linear;
597 skb->len += len - linear;
602 /* Get packet from user space buffer */
603 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
604 const struct iovec *iv, size_t count,
607 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
609 size_t len = count, align = 0;
610 struct virtio_net_hdr gso = { 0 };
613 if (!(tun->flags & TUN_NO_PI)) {
614 if ((len -= sizeof(pi)) > count)
617 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
619 offset += sizeof(pi);
622 if (tun->flags & TUN_VNET_HDR) {
623 if ((len -= tun->vnet_hdr_sz) > count)
626 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
629 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
630 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
631 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
633 if (gso.hdr_len > len)
635 offset += tun->vnet_hdr_sz;
638 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
639 align = NET_IP_ALIGN;
640 if (unlikely(len < ETH_HLEN ||
641 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
645 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
647 if (PTR_ERR(skb) != -EAGAIN)
648 tun->dev->stats.rx_dropped++;
652 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
653 tun->dev->stats.rx_dropped++;
658 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
659 if (!skb_partial_csum_set(skb, gso.csum_start,
661 tun->dev->stats.rx_frame_errors++;
667 switch (tun->flags & TUN_TYPE_MASK) {
669 if (tun->flags & TUN_NO_PI) {
670 switch (skb->data[0] & 0xf0) {
672 pi.proto = htons(ETH_P_IP);
675 pi.proto = htons(ETH_P_IPV6);
678 tun->dev->stats.rx_dropped++;
684 skb_reset_mac_header(skb);
685 skb->protocol = pi.proto;
689 skb->protocol = eth_type_trans(skb, tun->dev);
693 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
695 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
696 case VIRTIO_NET_HDR_GSO_TCPV4:
697 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
699 case VIRTIO_NET_HDR_GSO_TCPV6:
700 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
702 case VIRTIO_NET_HDR_GSO_UDP:
703 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
706 tun->dev->stats.rx_frame_errors++;
711 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
712 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
714 skb_shinfo(skb)->gso_size = gso.gso_size;
715 if (skb_shinfo(skb)->gso_size == 0) {
716 tun->dev->stats.rx_frame_errors++;
721 /* Header must be checked, and gso_segs computed. */
722 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
723 skb_shinfo(skb)->gso_segs = 0;
728 tun->dev->stats.rx_packets++;
729 tun->dev->stats.rx_bytes += len;
734 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
735 unsigned long count, loff_t pos)
737 struct file *file = iocb->ki_filp;
738 struct tun_struct *tun = tun_get(file);
744 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
746 result = tun_get_user(tun, iv, iov_length(iv, count),
747 file->f_flags & O_NONBLOCK);
753 /* Put packet to the user space buffer */
754 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
756 const struct iovec *iv, int len)
758 struct tun_pi pi = { 0, skb->protocol };
761 if (!(tun->flags & TUN_NO_PI)) {
762 if ((len -= sizeof(pi)) < 0)
765 if (len < skb->len) {
766 /* Packet will be striped */
767 pi.flags |= TUN_PKT_STRIP;
770 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
775 if (tun->flags & TUN_VNET_HDR) {
776 struct virtio_net_hdr gso = { 0 }; /* no info leak */
777 if ((len -= tun->vnet_hdr_sz) < 0)
780 if (skb_is_gso(skb)) {
781 struct skb_shared_info *sinfo = skb_shinfo(skb);
783 /* This is a hint as to how much should be linear. */
784 gso.hdr_len = skb_headlen(skb);
785 gso.gso_size = sinfo->gso_size;
786 if (sinfo->gso_type & SKB_GSO_TCPV4)
787 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
788 else if (sinfo->gso_type & SKB_GSO_TCPV6)
789 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
790 else if (sinfo->gso_type & SKB_GSO_UDP)
791 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
793 pr_err("unexpected GSO type: "
794 "0x%x, gso_size %d, hdr_len %d\n",
795 sinfo->gso_type, gso.gso_size,
797 print_hex_dump(KERN_ERR, "tun: ",
800 min((int)gso.hdr_len, 64), true);
804 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
805 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
807 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
809 if (skb->ip_summed == CHECKSUM_PARTIAL) {
810 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
811 gso.csum_start = skb_checksum_start_offset(skb);
812 gso.csum_offset = skb->csum_offset;
813 } /* else everything is zero */
815 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
818 total += tun->vnet_hdr_sz;
821 len = min_t(int, skb->len, len);
823 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
826 tun->dev->stats.tx_packets++;
827 tun->dev->stats.tx_bytes += len;
832 static ssize_t tun_do_read(struct tun_struct *tun,
833 struct kiocb *iocb, const struct iovec *iv,
834 ssize_t len, int noblock)
836 DECLARE_WAITQUEUE(wait, current);
840 tun_debug(KERN_INFO, tun, "tun_chr_read\n");
842 add_wait_queue(&tun->wq.wait, &wait);
844 current->state = TASK_INTERRUPTIBLE;
846 /* Read frames from the queue */
847 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
852 if (signal_pending(current)) {
856 if (tun->dev->reg_state != NETREG_REGISTERED) {
861 /* Nothing to read, let's sleep */
865 netif_wake_queue(tun->dev);
867 ret = tun_put_user(tun, skb, iv, len);
872 current->state = TASK_RUNNING;
873 remove_wait_queue(&tun->wq.wait, &wait);
878 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
879 unsigned long count, loff_t pos)
881 struct file *file = iocb->ki_filp;
882 struct tun_file *tfile = file->private_data;
883 struct tun_struct *tun = __tun_get(tfile);
888 len = iov_length(iv, count);
894 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
895 ret = min_t(ssize_t, ret, len);
901 static void tun_setup(struct net_device *dev)
903 struct tun_struct *tun = netdev_priv(dev);
908 dev->ethtool_ops = &tun_ethtool_ops;
909 dev->destructor = tun_free_netdev;
912 /* Trivial set of netlink ops to allow deleting tun or tap
913 * device with netlink.
915 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
920 static struct rtnl_link_ops tun_link_ops __read_mostly = {
922 .priv_size = sizeof(struct tun_struct),
924 .validate = tun_validate,
927 static void tun_sock_write_space(struct sock *sk)
929 struct tun_struct *tun;
930 wait_queue_head_t *wqueue;
932 if (!sock_writeable(sk))
935 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
938 wqueue = sk_sleep(sk);
939 if (wqueue && waitqueue_active(wqueue))
940 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
941 POLLWRNORM | POLLWRBAND);
943 tun = tun_sk(sk)->tun;
944 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
947 static void tun_sock_destruct(struct sock *sk)
949 free_netdev(tun_sk(sk)->tun->dev);
952 static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
953 struct msghdr *m, size_t total_len)
955 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
956 return tun_get_user(tun, m->msg_iov, total_len,
957 m->msg_flags & MSG_DONTWAIT);
960 static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
961 struct msghdr *m, size_t total_len,
964 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
966 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
968 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
969 flags & MSG_DONTWAIT);
970 if (ret > total_len) {
971 m->msg_flags |= MSG_TRUNC;
972 ret = flags & MSG_TRUNC ? ret : total_len;
977 /* Ops structure to mimic raw sockets with tun */
978 static const struct proto_ops tun_socket_ops = {
979 .sendmsg = tun_sendmsg,
980 .recvmsg = tun_recvmsg,
983 static struct proto tun_proto = {
985 .owner = THIS_MODULE,
986 .obj_size = sizeof(struct tun_sock),
989 static int tun_flags(struct tun_struct *tun)
993 if (tun->flags & TUN_TUN_DEV)
998 if (tun->flags & TUN_NO_PI)
1001 if (tun->flags & TUN_ONE_QUEUE)
1002 flags |= IFF_ONE_QUEUE;
1004 if (tun->flags & TUN_VNET_HDR)
1005 flags |= IFF_VNET_HDR;
1010 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1013 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1014 return sprintf(buf, "0x%x\n", tun_flags(tun));
1017 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1020 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1021 return sprintf(buf, "%d\n", tun->owner);
1024 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1027 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1028 return sprintf(buf, "%d\n", tun->group);
1031 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1032 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1033 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1035 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
1038 struct tun_struct *tun;
1039 struct net_device *dev;
1042 dev = __dev_get_by_name(net, ifr->ifr_name);
1044 const struct cred *cred = current_cred();
1046 if (ifr->ifr_flags & IFF_TUN_EXCL)
1048 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1049 tun = netdev_priv(dev);
1050 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1051 tun = netdev_priv(dev);
1055 if (((tun->owner != -1 && cred->euid != tun->owner) ||
1056 (tun->group != -1 && !in_egroup_p(tun->group))) &&
1057 !capable(CAP_NET_ADMIN))
1059 err = security_tun_dev_attach(tun->socket.sk);
1063 err = tun_attach(tun, file);
1069 unsigned long flags = 0;
1071 if (!capable(CAP_NET_ADMIN))
1073 err = security_tun_dev_create();
1078 if (ifr->ifr_flags & IFF_TUN) {
1080 flags |= TUN_TUN_DEV;
1082 } else if (ifr->ifr_flags & IFF_TAP) {
1084 flags |= TUN_TAP_DEV;
1090 name = ifr->ifr_name;
1092 dev = alloc_netdev(sizeof(struct tun_struct), name,
1097 dev_net_set(dev, net);
1098 dev->rtnl_link_ops = &tun_link_ops;
1100 tun = netdev_priv(dev);
1103 tun->txflt.count = 0;
1104 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
1107 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1111 tun->socket.wq = &tun->wq;
1112 init_waitqueue_head(&tun->wq.wait);
1113 tun->socket.ops = &tun_socket_ops;
1114 sock_init_data(&tun->socket, sk);
1115 sk->sk_write_space = tun_sock_write_space;
1116 sk->sk_sndbuf = INT_MAX;
1118 tun_sk(sk)->tun = tun;
1120 security_tun_dev_post_create(sk);
1124 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
1126 dev->features = dev->hw_features;
1128 err = register_netdevice(tun->dev);
1132 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1133 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1134 device_create_file(&tun->dev->dev, &dev_attr_group))
1135 pr_err("Failed to create tun sysfs files\n");
1137 sk->sk_destruct = tun_sock_destruct;
1139 err = tun_attach(tun, file);
1144 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
1146 if (ifr->ifr_flags & IFF_NO_PI)
1147 tun->flags |= TUN_NO_PI;
1149 tun->flags &= ~TUN_NO_PI;
1151 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1152 tun->flags |= TUN_ONE_QUEUE;
1154 tun->flags &= ~TUN_ONE_QUEUE;
1156 if (ifr->ifr_flags & IFF_VNET_HDR)
1157 tun->flags |= TUN_VNET_HDR;
1159 tun->flags &= ~TUN_VNET_HDR;
1161 /* Make sure persistent devices do not get stuck in
1164 if (netif_running(tun->dev))
1165 netif_wake_queue(tun->dev);
1167 strcpy(ifr->ifr_name, tun->dev->name);
1178 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1181 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
1183 strcpy(ifr->ifr_name, tun->dev->name);
1185 ifr->ifr_flags = tun_flags(tun);
1190 /* This is like a cut-down ethtool ops, except done via tun fd so no
1191 * privs required. */
1192 static int set_offload(struct tun_struct *tun, unsigned long arg)
1196 if (arg & TUN_F_CSUM) {
1197 features |= NETIF_F_HW_CSUM;
1200 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1201 if (arg & TUN_F_TSO_ECN) {
1202 features |= NETIF_F_TSO_ECN;
1203 arg &= ~TUN_F_TSO_ECN;
1205 if (arg & TUN_F_TSO4)
1206 features |= NETIF_F_TSO;
1207 if (arg & TUN_F_TSO6)
1208 features |= NETIF_F_TSO6;
1209 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1212 if (arg & TUN_F_UFO) {
1213 features |= NETIF_F_UFO;
1218 /* This gives the user a way to test for new features in future by
1219 * trying to set them. */
1223 tun->set_features = features;
1224 netdev_update_features(tun->dev);
1229 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1230 unsigned long arg, int ifreq_len)
1232 struct tun_file *tfile = file->private_data;
1233 struct tun_struct *tun;
1234 void __user* argp = (void __user*)arg;
1235 struct sock_fprog fprog;
1241 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1242 if (copy_from_user(&ifr, argp, ifreq_len))
1245 if (cmd == TUNGETFEATURES) {
1246 /* Currently this just means: "what IFF flags are valid?".
1247 * This is needed because we never checked for invalid flags on
1249 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1251 (unsigned int __user*)argp);
1256 tun = __tun_get(tfile);
1257 if (cmd == TUNSETIFF && !tun) {
1258 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1260 ret = tun_set_iff(tfile->net, file, &ifr);
1265 if (copy_to_user(argp, &ifr, ifreq_len))
1274 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
1279 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1283 if (copy_to_user(argp, &ifr, ifreq_len))
1288 /* Disable/Enable checksum */
1290 /* [unimplemented] */
1291 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
1292 arg ? "disabled" : "enabled");
1296 /* Disable/Enable persist mode */
1298 tun->flags |= TUN_PERSIST;
1300 tun->flags &= ~TUN_PERSIST;
1302 tun_debug(KERN_INFO, tun, "persist %s\n",
1303 arg ? "enabled" : "disabled");
1307 /* Set owner of the device */
1308 tun->owner = (uid_t) arg;
1310 tun_debug(KERN_INFO, tun, "owner set to %d\n", tun->owner);
1314 /* Set group of the device */
1315 tun->group= (gid_t) arg;
1317 tun_debug(KERN_INFO, tun, "group set to %d\n", tun->group);
1321 /* Only allow setting the type when the interface is down */
1322 if (tun->dev->flags & IFF_UP) {
1323 tun_debug(KERN_INFO, tun,
1324 "Linktype set failed because interface is up\n");
1327 tun->dev->type = (int) arg;
1328 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1340 ret = set_offload(tun, arg);
1343 case TUNSETTXFILTER:
1344 /* Can be set only for TAPs */
1346 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1348 ret = update_filter(&tun->txflt, (void __user *)arg);
1352 /* Get hw address */
1353 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1354 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1355 if (copy_to_user(argp, &ifr, ifreq_len))
1360 /* Set hw address */
1361 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
1362 ifr.ifr_hwaddr.sa_data);
1364 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1368 sndbuf = tun->socket.sk->sk_sndbuf;
1369 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1374 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1379 tun->socket.sk->sk_sndbuf = sndbuf;
1382 case TUNGETVNETHDRSZ:
1383 vnet_hdr_sz = tun->vnet_hdr_sz;
1384 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1388 case TUNSETVNETHDRSZ:
1389 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1393 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1398 tun->vnet_hdr_sz = vnet_hdr_sz;
1401 case TUNATTACHFILTER:
1402 /* Can be set only for TAPs */
1404 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1407 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1410 ret = sk_attach_filter(&fprog, tun->socket.sk);
1413 case TUNDETACHFILTER:
1414 /* Can be set only for TAPs */
1416 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1418 ret = sk_detach_filter(tun->socket.sk);
1433 static long tun_chr_ioctl(struct file *file,
1434 unsigned int cmd, unsigned long arg)
1436 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1439 #ifdef CONFIG_COMPAT
1440 static long tun_chr_compat_ioctl(struct file *file,
1441 unsigned int cmd, unsigned long arg)
1446 case TUNSETTXFILTER:
1451 arg = (unsigned long)compat_ptr(arg);
1454 arg = (compat_ulong_t)arg;
1459 * compat_ifreq is shorter than ifreq, so we must not access beyond
1460 * the end of that structure. All fields that are used in this
1461 * driver are compatible though, we don't need to convert the
1464 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1466 #endif /* CONFIG_COMPAT */
1468 static int tun_chr_fasync(int fd, struct file *file, int on)
1470 struct tun_struct *tun = tun_get(file);
1476 tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
1478 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1482 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1485 tun->flags |= TUN_FASYNC;
1487 tun->flags &= ~TUN_FASYNC;
1494 static int tun_chr_open(struct inode *inode, struct file * file)
1496 struct tun_file *tfile;
1498 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
1500 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1503 atomic_set(&tfile->count, 0);
1505 tfile->net = get_net(current->nsproxy->net_ns);
1506 file->private_data = tfile;
1510 static int tun_chr_close(struct inode *inode, struct file *file)
1512 struct tun_file *tfile = file->private_data;
1513 struct tun_struct *tun;
1515 tun = __tun_get(tfile);
1517 struct net_device *dev = tun->dev;
1519 tun_debug(KERN_INFO, tun, "tun_chr_close\n");
1523 /* If desirable, unregister the netdevice. */
1524 if (!(tun->flags & TUN_PERSIST)) {
1526 if (dev->reg_state == NETREG_REGISTERED)
1527 unregister_netdevice(dev);
1534 sock_put(tun->socket.sk);
1536 put_net(tfile->net);
1542 static const struct file_operations tun_fops = {
1543 .owner = THIS_MODULE,
1544 .llseek = no_llseek,
1545 .read = do_sync_read,
1546 .aio_read = tun_chr_aio_read,
1547 .write = do_sync_write,
1548 .aio_write = tun_chr_aio_write,
1549 .poll = tun_chr_poll,
1550 .unlocked_ioctl = tun_chr_ioctl,
1551 #ifdef CONFIG_COMPAT
1552 .compat_ioctl = tun_chr_compat_ioctl,
1554 .open = tun_chr_open,
1555 .release = tun_chr_close,
1556 .fasync = tun_chr_fasync
1559 static struct miscdevice tun_miscdev = {
1562 .nodename = "net/tun",
1566 /* ethtool interface */
1568 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1571 cmd->advertising = 0;
1572 ethtool_cmd_speed_set(cmd, SPEED_10);
1573 cmd->duplex = DUPLEX_FULL;
1574 cmd->port = PORT_TP;
1575 cmd->phy_address = 0;
1576 cmd->transceiver = XCVR_INTERNAL;
1577 cmd->autoneg = AUTONEG_DISABLE;
1583 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1585 struct tun_struct *tun = netdev_priv(dev);
1587 strcpy(info->driver, DRV_NAME);
1588 strcpy(info->version, DRV_VERSION);
1589 strcpy(info->fw_version, "N/A");
1591 switch (tun->flags & TUN_TYPE_MASK) {
1593 strcpy(info->bus_info, "tun");
1596 strcpy(info->bus_info, "tap");
1601 static u32 tun_get_msglevel(struct net_device *dev)
1604 struct tun_struct *tun = netdev_priv(dev);
1611 static void tun_set_msglevel(struct net_device *dev, u32 value)
1614 struct tun_struct *tun = netdev_priv(dev);
1619 static const struct ethtool_ops tun_ethtool_ops = {
1620 .get_settings = tun_get_settings,
1621 .get_drvinfo = tun_get_drvinfo,
1622 .get_msglevel = tun_get_msglevel,
1623 .set_msglevel = tun_set_msglevel,
1624 .get_link = ethtool_op_get_link,
1628 static int __init tun_init(void)
1632 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1633 pr_info("%s\n", DRV_COPYRIGHT);
1635 ret = rtnl_link_register(&tun_link_ops);
1637 pr_err("Can't register link_ops\n");
1641 ret = misc_register(&tun_miscdev);
1643 pr_err("Can't register misc device %d\n", TUN_MINOR);
1648 rtnl_link_unregister(&tun_link_ops);
1653 static void tun_cleanup(void)
1655 misc_deregister(&tun_miscdev);
1656 rtnl_link_unregister(&tun_link_ops);
1659 /* Get an underlying socket object from tun file. Returns error unless file is
1660 * attached to a device. The returned object works like a packet socket, it
1661 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1662 * holding a reference to the file for as long as the socket is in use. */
1663 struct socket *tun_get_socket(struct file *file)
1665 struct tun_struct *tun;
1666 if (file->f_op != &tun_fops)
1667 return ERR_PTR(-EINVAL);
1668 tun = tun_get(file);
1670 return ERR_PTR(-EBADFD);
1672 return &tun->socket;
1674 EXPORT_SYMBOL_GPL(tun_get_socket);
1676 module_init(tun_init);
1677 module_exit(tun_cleanup);
1678 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1679 MODULE_AUTHOR(DRV_COPYRIGHT);
1680 MODULE_LICENSE("GPL");
1681 MODULE_ALIAS_MISCDEV(TUN_MINOR);
1682 MODULE_ALIAS("devname:net/tun");