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);
464 static const struct net_device_ops tun_netdev_ops = {
465 .ndo_uninit = tun_net_uninit,
466 .ndo_open = tun_net_open,
467 .ndo_stop = tun_net_close,
468 .ndo_start_xmit = tun_net_xmit,
469 .ndo_change_mtu = tun_net_change_mtu,
470 .ndo_fix_features = tun_net_fix_features,
473 static const struct net_device_ops tap_netdev_ops = {
474 .ndo_uninit = tun_net_uninit,
475 .ndo_open = tun_net_open,
476 .ndo_stop = tun_net_close,
477 .ndo_start_xmit = tun_net_xmit,
478 .ndo_change_mtu = tun_net_change_mtu,
479 .ndo_fix_features = tun_net_fix_features,
480 .ndo_set_multicast_list = tun_net_mclist,
481 .ndo_set_mac_address = eth_mac_addr,
482 .ndo_validate_addr = eth_validate_addr,
485 /* Initialize net device. */
486 static void tun_net_init(struct net_device *dev)
488 struct tun_struct *tun = netdev_priv(dev);
490 switch (tun->flags & TUN_TYPE_MASK) {
492 dev->netdev_ops = &tun_netdev_ops;
494 /* Point-to-Point TUN Device */
495 dev->hard_header_len = 0;
499 /* Zero header length */
500 dev->type = ARPHRD_NONE;
501 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
502 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
506 dev->netdev_ops = &tap_netdev_ops;
507 /* Ethernet TAP Device */
510 random_ether_addr(dev->dev_addr);
512 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
517 /* Character device part */
520 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
522 struct tun_file *tfile = file->private_data;
523 struct tun_struct *tun = __tun_get(tfile);
525 unsigned int mask = 0;
532 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
534 poll_wait(file, &tun->wq.wait, wait);
536 if (!skb_queue_empty(&sk->sk_receive_queue))
537 mask |= POLLIN | POLLRDNORM;
539 if (sock_writeable(sk) ||
540 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
542 mask |= POLLOUT | POLLWRNORM;
544 if (tun->dev->reg_state != NETREG_REGISTERED)
551 /* prepad is the amount to reserve at front. len is length after that.
552 * linear is a hint as to how much to copy (usually headers). */
553 static struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
554 size_t prepad, size_t len,
555 size_t linear, int noblock)
557 struct sock *sk = tun->socket.sk;
561 sock_update_classid(sk);
563 /* Under a page? Don't bother with paged skb. */
564 if (prepad + len < PAGE_SIZE || !linear)
567 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
572 skb_reserve(skb, prepad);
573 skb_put(skb, linear);
574 skb->data_len = len - linear;
575 skb->len += len - linear;
580 /* Get packet from user space buffer */
581 static ssize_t tun_get_user(struct tun_struct *tun,
582 const struct iovec *iv, size_t count,
585 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
587 size_t len = count, align = NET_SKB_PAD;
588 struct virtio_net_hdr gso = { 0 };
591 if (!(tun->flags & TUN_NO_PI)) {
592 if ((len -= sizeof(pi)) > count)
595 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
597 offset += sizeof(pi);
600 if (tun->flags & TUN_VNET_HDR) {
601 if ((len -= tun->vnet_hdr_sz) > count)
604 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
607 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
608 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
609 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
611 if (gso.hdr_len > len)
613 offset += tun->vnet_hdr_sz;
616 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
617 align += NET_IP_ALIGN;
618 if (unlikely(len < ETH_HLEN ||
619 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
623 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
625 if (PTR_ERR(skb) != -EAGAIN)
626 tun->dev->stats.rx_dropped++;
630 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
631 tun->dev->stats.rx_dropped++;
636 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
637 if (!skb_partial_csum_set(skb, gso.csum_start,
639 tun->dev->stats.rx_frame_errors++;
645 switch (tun->flags & TUN_TYPE_MASK) {
647 if (tun->flags & TUN_NO_PI) {
648 switch (skb->data[0] & 0xf0) {
650 pi.proto = htons(ETH_P_IP);
653 pi.proto = htons(ETH_P_IPV6);
656 tun->dev->stats.rx_dropped++;
662 skb_reset_mac_header(skb);
663 skb->protocol = pi.proto;
667 skb->protocol = eth_type_trans(skb, tun->dev);
671 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
673 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
674 case VIRTIO_NET_HDR_GSO_TCPV4:
675 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
677 case VIRTIO_NET_HDR_GSO_TCPV6:
678 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
680 case VIRTIO_NET_HDR_GSO_UDP:
681 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
684 tun->dev->stats.rx_frame_errors++;
689 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
690 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
692 skb_shinfo(skb)->gso_size = gso.gso_size;
693 if (skb_shinfo(skb)->gso_size == 0) {
694 tun->dev->stats.rx_frame_errors++;
699 /* Header must be checked, and gso_segs computed. */
700 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
701 skb_shinfo(skb)->gso_segs = 0;
706 tun->dev->stats.rx_packets++;
707 tun->dev->stats.rx_bytes += len;
712 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
713 unsigned long count, loff_t pos)
715 struct file *file = iocb->ki_filp;
716 struct tun_struct *tun = tun_get(file);
722 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
724 result = tun_get_user(tun, iv, iov_length(iv, count),
725 file->f_flags & O_NONBLOCK);
731 /* Put packet to the user space buffer */
732 static ssize_t tun_put_user(struct tun_struct *tun,
734 const struct iovec *iv, int len)
736 struct tun_pi pi = { 0, skb->protocol };
739 if (!(tun->flags & TUN_NO_PI)) {
740 if ((len -= sizeof(pi)) < 0)
743 if (len < skb->len) {
744 /* Packet will be striped */
745 pi.flags |= TUN_PKT_STRIP;
748 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
753 if (tun->flags & TUN_VNET_HDR) {
754 struct virtio_net_hdr gso = { 0 }; /* no info leak */
755 if ((len -= tun->vnet_hdr_sz) < 0)
758 if (skb_is_gso(skb)) {
759 struct skb_shared_info *sinfo = skb_shinfo(skb);
761 /* This is a hint as to how much should be linear. */
762 gso.hdr_len = skb_headlen(skb);
763 gso.gso_size = sinfo->gso_size;
764 if (sinfo->gso_type & SKB_GSO_TCPV4)
765 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
766 else if (sinfo->gso_type & SKB_GSO_TCPV6)
767 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
768 else if (sinfo->gso_type & SKB_GSO_UDP)
769 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
771 pr_err("unexpected GSO type: "
772 "0x%x, gso_size %d, hdr_len %d\n",
773 sinfo->gso_type, gso.gso_size,
775 print_hex_dump(KERN_ERR, "tun: ",
778 min((int)gso.hdr_len, 64), true);
782 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
783 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
785 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
787 if (skb->ip_summed == CHECKSUM_PARTIAL) {
788 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
789 gso.csum_start = skb_checksum_start_offset(skb);
790 gso.csum_offset = skb->csum_offset;
791 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
792 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
793 } /* else everything is zero */
795 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
798 total += tun->vnet_hdr_sz;
801 len = min_t(int, skb->len, len);
803 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
806 tun->dev->stats.tx_packets++;
807 tun->dev->stats.tx_bytes += len;
812 static ssize_t tun_do_read(struct tun_struct *tun,
813 struct kiocb *iocb, const struct iovec *iv,
814 ssize_t len, int noblock)
816 DECLARE_WAITQUEUE(wait, current);
820 tun_debug(KERN_INFO, tun, "tun_chr_read\n");
822 if (unlikely(!noblock))
823 add_wait_queue(&tun->wq.wait, &wait);
825 current->state = TASK_INTERRUPTIBLE;
827 /* Read frames from the queue */
828 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
833 if (signal_pending(current)) {
837 if (tun->dev->reg_state != NETREG_REGISTERED) {
842 /* Nothing to read, let's sleep */
846 netif_wake_queue(tun->dev);
848 ret = tun_put_user(tun, skb, iv, len);
853 current->state = TASK_RUNNING;
854 if (unlikely(!noblock))
855 remove_wait_queue(&tun->wq.wait, &wait);
860 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
861 unsigned long count, loff_t pos)
863 struct file *file = iocb->ki_filp;
864 struct tun_file *tfile = file->private_data;
865 struct tun_struct *tun = __tun_get(tfile);
870 len = iov_length(iv, count);
876 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
877 ret = min_t(ssize_t, ret, len);
883 static void tun_setup(struct net_device *dev)
885 struct tun_struct *tun = netdev_priv(dev);
890 dev->ethtool_ops = &tun_ethtool_ops;
891 dev->destructor = tun_free_netdev;
894 /* Trivial set of netlink ops to allow deleting tun or tap
895 * device with netlink.
897 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
902 static struct rtnl_link_ops tun_link_ops __read_mostly = {
904 .priv_size = sizeof(struct tun_struct),
906 .validate = tun_validate,
909 static void tun_sock_write_space(struct sock *sk)
911 struct tun_struct *tun;
912 wait_queue_head_t *wqueue;
914 if (!sock_writeable(sk))
917 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
920 wqueue = sk_sleep(sk);
921 if (wqueue && waitqueue_active(wqueue))
922 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
923 POLLWRNORM | POLLWRBAND);
925 tun = tun_sk(sk)->tun;
926 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
929 static void tun_sock_destruct(struct sock *sk)
931 free_netdev(tun_sk(sk)->tun->dev);
934 static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
935 struct msghdr *m, size_t total_len)
937 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
938 return tun_get_user(tun, m->msg_iov, total_len,
939 m->msg_flags & MSG_DONTWAIT);
942 static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
943 struct msghdr *m, size_t total_len,
946 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
948 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
950 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
951 flags & MSG_DONTWAIT);
952 if (ret > total_len) {
953 m->msg_flags |= MSG_TRUNC;
954 ret = flags & MSG_TRUNC ? ret : total_len;
959 /* Ops structure to mimic raw sockets with tun */
960 static const struct proto_ops tun_socket_ops = {
961 .sendmsg = tun_sendmsg,
962 .recvmsg = tun_recvmsg,
965 static struct proto tun_proto = {
967 .owner = THIS_MODULE,
968 .obj_size = sizeof(struct tun_sock),
971 static int tun_flags(struct tun_struct *tun)
975 if (tun->flags & TUN_TUN_DEV)
980 if (tun->flags & TUN_NO_PI)
983 if (tun->flags & TUN_ONE_QUEUE)
984 flags |= IFF_ONE_QUEUE;
986 if (tun->flags & TUN_VNET_HDR)
987 flags |= IFF_VNET_HDR;
992 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
995 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
996 return sprintf(buf, "0x%x\n", tun_flags(tun));
999 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1002 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1003 return sprintf(buf, "%d\n", tun->owner);
1006 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1009 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1010 return sprintf(buf, "%d\n", tun->group);
1013 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1014 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1015 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1017 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
1020 struct tun_struct *tun;
1021 struct net_device *dev;
1024 dev = __dev_get_by_name(net, ifr->ifr_name);
1026 const struct cred *cred = current_cred();
1028 if (ifr->ifr_flags & IFF_TUN_EXCL)
1030 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1031 tun = netdev_priv(dev);
1032 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1033 tun = netdev_priv(dev);
1037 if (((tun->owner != -1 && cred->euid != tun->owner) ||
1038 (tun->group != -1 && !in_egroup_p(tun->group))) &&
1039 !capable(CAP_NET_ADMIN))
1041 err = security_tun_dev_attach(tun->socket.sk);
1045 err = tun_attach(tun, file);
1051 unsigned long flags = 0;
1053 if (!capable(CAP_NET_ADMIN))
1055 err = security_tun_dev_create();
1060 if (ifr->ifr_flags & IFF_TUN) {
1062 flags |= TUN_TUN_DEV;
1064 } else if (ifr->ifr_flags & IFF_TAP) {
1066 flags |= TUN_TAP_DEV;
1072 name = ifr->ifr_name;
1074 dev = alloc_netdev(sizeof(struct tun_struct), name,
1079 dev_net_set(dev, net);
1080 dev->rtnl_link_ops = &tun_link_ops;
1082 tun = netdev_priv(dev);
1085 tun->txflt.count = 0;
1086 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
1089 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1093 tun->socket.wq = &tun->wq;
1094 init_waitqueue_head(&tun->wq.wait);
1095 tun->socket.ops = &tun_socket_ops;
1096 sock_init_data(&tun->socket, sk);
1097 sk->sk_write_space = tun_sock_write_space;
1098 sk->sk_sndbuf = INT_MAX;
1100 tun_sk(sk)->tun = tun;
1102 security_tun_dev_post_create(sk);
1106 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
1108 dev->features = dev->hw_features;
1110 err = register_netdevice(tun->dev);
1114 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1115 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1116 device_create_file(&tun->dev->dev, &dev_attr_group))
1117 pr_err("Failed to create tun sysfs files\n");
1119 sk->sk_destruct = tun_sock_destruct;
1121 err = tun_attach(tun, file);
1126 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
1128 if (ifr->ifr_flags & IFF_NO_PI)
1129 tun->flags |= TUN_NO_PI;
1131 tun->flags &= ~TUN_NO_PI;
1133 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1134 tun->flags |= TUN_ONE_QUEUE;
1136 tun->flags &= ~TUN_ONE_QUEUE;
1138 if (ifr->ifr_flags & IFF_VNET_HDR)
1139 tun->flags |= TUN_VNET_HDR;
1141 tun->flags &= ~TUN_VNET_HDR;
1143 /* Make sure persistent devices do not get stuck in
1146 if (netif_running(tun->dev))
1147 netif_wake_queue(tun->dev);
1149 strcpy(ifr->ifr_name, tun->dev->name);
1160 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1163 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
1165 strcpy(ifr->ifr_name, tun->dev->name);
1167 ifr->ifr_flags = tun_flags(tun);
1172 /* This is like a cut-down ethtool ops, except done via tun fd so no
1173 * privs required. */
1174 static int set_offload(struct tun_struct *tun, unsigned long arg)
1178 if (arg & TUN_F_CSUM) {
1179 features |= NETIF_F_HW_CSUM;
1182 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1183 if (arg & TUN_F_TSO_ECN) {
1184 features |= NETIF_F_TSO_ECN;
1185 arg &= ~TUN_F_TSO_ECN;
1187 if (arg & TUN_F_TSO4)
1188 features |= NETIF_F_TSO;
1189 if (arg & TUN_F_TSO6)
1190 features |= NETIF_F_TSO6;
1191 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1194 if (arg & TUN_F_UFO) {
1195 features |= NETIF_F_UFO;
1200 /* This gives the user a way to test for new features in future by
1201 * trying to set them. */
1205 tun->set_features = features;
1206 netdev_update_features(tun->dev);
1211 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1212 unsigned long arg, int ifreq_len)
1214 struct tun_file *tfile = file->private_data;
1215 struct tun_struct *tun;
1216 void __user* argp = (void __user*)arg;
1217 struct sock_fprog fprog;
1223 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1224 if (copy_from_user(&ifr, argp, ifreq_len))
1227 if (cmd == TUNGETFEATURES) {
1228 /* Currently this just means: "what IFF flags are valid?".
1229 * This is needed because we never checked for invalid flags on
1231 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1233 (unsigned int __user*)argp);
1238 tun = __tun_get(tfile);
1239 if (cmd == TUNSETIFF && !tun) {
1240 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1242 ret = tun_set_iff(tfile->net, file, &ifr);
1247 if (copy_to_user(argp, &ifr, ifreq_len))
1256 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
1261 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1265 if (copy_to_user(argp, &ifr, ifreq_len))
1270 /* Disable/Enable checksum */
1272 /* [unimplemented] */
1273 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
1274 arg ? "disabled" : "enabled");
1278 /* Disable/Enable persist mode */
1280 tun->flags |= TUN_PERSIST;
1282 tun->flags &= ~TUN_PERSIST;
1284 tun_debug(KERN_INFO, tun, "persist %s\n",
1285 arg ? "enabled" : "disabled");
1289 /* Set owner of the device */
1290 tun->owner = (uid_t) arg;
1292 tun_debug(KERN_INFO, tun, "owner set to %d\n", tun->owner);
1296 /* Set group of the device */
1297 tun->group= (gid_t) arg;
1299 tun_debug(KERN_INFO, tun, "group set to %d\n", tun->group);
1303 /* Only allow setting the type when the interface is down */
1304 if (tun->dev->flags & IFF_UP) {
1305 tun_debug(KERN_INFO, tun,
1306 "Linktype set failed because interface is up\n");
1309 tun->dev->type = (int) arg;
1310 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1322 ret = set_offload(tun, arg);
1325 case TUNSETTXFILTER:
1326 /* Can be set only for TAPs */
1328 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1330 ret = update_filter(&tun->txflt, (void __user *)arg);
1334 /* Get hw address */
1335 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1336 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1337 if (copy_to_user(argp, &ifr, ifreq_len))
1342 /* Set hw address */
1343 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
1344 ifr.ifr_hwaddr.sa_data);
1346 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1350 sndbuf = tun->socket.sk->sk_sndbuf;
1351 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1356 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1361 tun->socket.sk->sk_sndbuf = sndbuf;
1364 case TUNGETVNETHDRSZ:
1365 vnet_hdr_sz = tun->vnet_hdr_sz;
1366 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1370 case TUNSETVNETHDRSZ:
1371 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1375 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1380 tun->vnet_hdr_sz = vnet_hdr_sz;
1383 case TUNATTACHFILTER:
1384 /* Can be set only for TAPs */
1386 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1389 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1392 ret = sk_attach_filter(&fprog, tun->socket.sk);
1395 case TUNDETACHFILTER:
1396 /* Can be set only for TAPs */
1398 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1400 ret = sk_detach_filter(tun->socket.sk);
1415 static long tun_chr_ioctl(struct file *file,
1416 unsigned int cmd, unsigned long arg)
1418 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1421 #ifdef CONFIG_COMPAT
1422 static long tun_chr_compat_ioctl(struct file *file,
1423 unsigned int cmd, unsigned long arg)
1428 case TUNSETTXFILTER:
1433 arg = (unsigned long)compat_ptr(arg);
1436 arg = (compat_ulong_t)arg;
1441 * compat_ifreq is shorter than ifreq, so we must not access beyond
1442 * the end of that structure. All fields that are used in this
1443 * driver are compatible though, we don't need to convert the
1446 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1448 #endif /* CONFIG_COMPAT */
1450 static int tun_chr_fasync(int fd, struct file *file, int on)
1452 struct tun_struct *tun = tun_get(file);
1458 tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
1460 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1464 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1467 tun->flags |= TUN_FASYNC;
1469 tun->flags &= ~TUN_FASYNC;
1476 static int tun_chr_open(struct inode *inode, struct file * file)
1478 struct tun_file *tfile;
1480 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
1482 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1485 atomic_set(&tfile->count, 0);
1487 tfile->net = get_net(current->nsproxy->net_ns);
1488 file->private_data = tfile;
1492 static int tun_chr_close(struct inode *inode, struct file *file)
1494 struct tun_file *tfile = file->private_data;
1495 struct tun_struct *tun;
1497 tun = __tun_get(tfile);
1499 struct net_device *dev = tun->dev;
1501 tun_debug(KERN_INFO, tun, "tun_chr_close\n");
1505 /* If desirable, unregister the netdevice. */
1506 if (!(tun->flags & TUN_PERSIST)) {
1508 if (dev->reg_state == NETREG_REGISTERED)
1509 unregister_netdevice(dev);
1516 sock_put(tun->socket.sk);
1518 put_net(tfile->net);
1524 static const struct file_operations tun_fops = {
1525 .owner = THIS_MODULE,
1526 .llseek = no_llseek,
1527 .read = do_sync_read,
1528 .aio_read = tun_chr_aio_read,
1529 .write = do_sync_write,
1530 .aio_write = tun_chr_aio_write,
1531 .poll = tun_chr_poll,
1532 .unlocked_ioctl = tun_chr_ioctl,
1533 #ifdef CONFIG_COMPAT
1534 .compat_ioctl = tun_chr_compat_ioctl,
1536 .open = tun_chr_open,
1537 .release = tun_chr_close,
1538 .fasync = tun_chr_fasync
1541 static struct miscdevice tun_miscdev = {
1544 .nodename = "net/tun",
1548 /* ethtool interface */
1550 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1553 cmd->advertising = 0;
1554 ethtool_cmd_speed_set(cmd, SPEED_10);
1555 cmd->duplex = DUPLEX_FULL;
1556 cmd->port = PORT_TP;
1557 cmd->phy_address = 0;
1558 cmd->transceiver = XCVR_INTERNAL;
1559 cmd->autoneg = AUTONEG_DISABLE;
1565 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1567 struct tun_struct *tun = netdev_priv(dev);
1569 strcpy(info->driver, DRV_NAME);
1570 strcpy(info->version, DRV_VERSION);
1571 strcpy(info->fw_version, "N/A");
1573 switch (tun->flags & TUN_TYPE_MASK) {
1575 strcpy(info->bus_info, "tun");
1578 strcpy(info->bus_info, "tap");
1583 static u32 tun_get_msglevel(struct net_device *dev)
1586 struct tun_struct *tun = netdev_priv(dev);
1593 static void tun_set_msglevel(struct net_device *dev, u32 value)
1596 struct tun_struct *tun = netdev_priv(dev);
1601 static const struct ethtool_ops tun_ethtool_ops = {
1602 .get_settings = tun_get_settings,
1603 .get_drvinfo = tun_get_drvinfo,
1604 .get_msglevel = tun_get_msglevel,
1605 .set_msglevel = tun_set_msglevel,
1606 .get_link = ethtool_op_get_link,
1610 static int __init tun_init(void)
1614 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1615 pr_info("%s\n", DRV_COPYRIGHT);
1617 ret = rtnl_link_register(&tun_link_ops);
1619 pr_err("Can't register link_ops\n");
1623 ret = misc_register(&tun_miscdev);
1625 pr_err("Can't register misc device %d\n", TUN_MINOR);
1630 rtnl_link_unregister(&tun_link_ops);
1635 static void tun_cleanup(void)
1637 misc_deregister(&tun_miscdev);
1638 rtnl_link_unregister(&tun_link_ops);
1641 /* Get an underlying socket object from tun file. Returns error unless file is
1642 * attached to a device. The returned object works like a packet socket, it
1643 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1644 * holding a reference to the file for as long as the socket is in use. */
1645 struct socket *tun_get_socket(struct file *file)
1647 struct tun_struct *tun;
1648 if (file->f_op != &tun_fops)
1649 return ERR_PTR(-EINVAL);
1650 tun = tun_get(file);
1652 return ERR_PTR(-EBADFD);
1654 return &tun->socket;
1656 EXPORT_SYMBOL_GPL(tun_get_socket);
1658 module_init(tun_init);
1659 module_exit(tun_cleanup);
1660 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1661 MODULE_AUTHOR(DRV_COPYRIGHT);
1662 MODULE_LICENSE("GPL");
1663 MODULE_ALIAS_MISCDEV(TUN_MINOR);
1664 MODULE_ALIAS("devname:net/tun");