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 eth_random_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/uaccess.h>
74 /* Uncomment to enable debugging */
75 /* #define TUN_DEBUG 1 */
80 #define tun_debug(level, tun, fmt, args...) \
83 netdev_printk(level, tun->dev, fmt, ##args); \
85 #define DBG1(level, fmt, args...) \
88 printk(level fmt, ##args); \
91 #define tun_debug(level, tun, fmt, args...) \
94 netdev_printk(level, tun->dev, fmt, ##args); \
96 #define DBG1(level, fmt, args...) \
99 printk(level fmt, ##args); \
103 #define FLT_EXACT_COUNT 8
105 unsigned int count; /* Number of addrs. Zero means disabled */
106 u32 mask[2]; /* Mask of the hashed addrs */
107 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
112 struct tun_struct *tun;
119 struct tun_file *tfile;
124 struct net_device *dev;
125 netdev_features_t set_features;
126 #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
127 NETIF_F_TSO6|NETIF_F_UFO)
128 struct fasync_struct *fasync;
130 struct tap_filter txflt;
131 struct socket socket;
143 struct tun_struct *tun;
146 static inline struct tun_sock *tun_sk(struct sock *sk)
148 return container_of(sk, struct tun_sock, sk);
151 static int tun_attach(struct tun_struct *tun, struct file *file)
153 struct tun_file *tfile = file->private_data;
158 netif_tx_lock_bh(tun->dev);
171 tun->socket.file = file;
172 netif_carrier_on(tun->dev);
174 sock_hold(tun->socket.sk);
175 atomic_inc(&tfile->count);
178 netif_tx_unlock_bh(tun->dev);
182 static void __tun_detach(struct tun_struct *tun)
184 /* Detach from net device */
185 netif_tx_lock_bh(tun->dev);
186 netif_carrier_off(tun->dev);
188 tun->socket.file = NULL;
189 netif_tx_unlock_bh(tun->dev);
191 /* Drop read queue */
192 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
194 /* Drop the extra count on the net device */
198 static void tun_detach(struct tun_struct *tun)
205 static struct tun_struct *__tun_get(struct tun_file *tfile)
207 struct tun_struct *tun = NULL;
209 if (atomic_inc_not_zero(&tfile->count))
215 static struct tun_struct *tun_get(struct file *file)
217 return __tun_get(file->private_data);
220 static void tun_put(struct tun_struct *tun)
222 struct tun_file *tfile = tun->tfile;
224 if (atomic_dec_and_test(&tfile->count))
225 tun_detach(tfile->tun);
229 static void addr_hash_set(u32 *mask, const u8 *addr)
231 int n = ether_crc(ETH_ALEN, addr) >> 26;
232 mask[n >> 5] |= (1 << (n & 31));
235 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
237 int n = ether_crc(ETH_ALEN, addr) >> 26;
238 return mask[n >> 5] & (1 << (n & 31));
241 static int update_filter(struct tap_filter *filter, void __user *arg)
243 struct { u8 u[ETH_ALEN]; } *addr;
244 struct tun_filter uf;
245 int err, alen, n, nexact;
247 if (copy_from_user(&uf, arg, sizeof(uf)))
256 alen = ETH_ALEN * uf.count;
257 addr = kmalloc(alen, GFP_KERNEL);
261 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
266 /* The filter is updated without holding any locks. Which is
267 * perfectly safe. We disable it first and in the worst
268 * case we'll accept a few undesired packets. */
272 /* Use first set of addresses as an exact filter */
273 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
274 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
278 /* Remaining multicast addresses are hashed,
279 * unicast will leave the filter disabled. */
280 memset(filter->mask, 0, sizeof(filter->mask));
281 for (; n < uf.count; n++) {
282 if (!is_multicast_ether_addr(addr[n].u)) {
283 err = 0; /* no filter */
286 addr_hash_set(filter->mask, addr[n].u);
289 /* For ALLMULTI just set the mask to all ones.
290 * This overrides the mask populated above. */
291 if ((uf.flags & TUN_FLT_ALLMULTI))
292 memset(filter->mask, ~0, sizeof(filter->mask));
294 /* Now enable the filter */
296 filter->count = nexact;
298 /* Return the number of exact filters */
306 /* Returns: 0 - drop, !=0 - accept */
307 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
309 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
311 struct ethhdr *eh = (struct ethhdr *) skb->data;
315 for (i = 0; i < filter->count; i++)
316 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
319 /* Inexact match (multicast only) */
320 if (is_multicast_ether_addr(eh->h_dest))
321 return addr_hash_test(filter->mask, eh->h_dest);
327 * Checks whether the packet is accepted or not.
328 * Returns: 0 - drop, !=0 - accept
330 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
335 return run_filter(filter, skb);
338 /* Network device part of the driver */
340 static const struct ethtool_ops tun_ethtool_ops;
342 /* Net device detach from fd. */
343 static void tun_net_uninit(struct net_device *dev)
345 struct tun_struct *tun = netdev_priv(dev);
346 struct tun_file *tfile = tun->tfile;
348 /* Inform the methods they need to stop using the dev.
351 wake_up_all(&tun->wq.wait);
352 if (atomic_dec_and_test(&tfile->count))
357 static void tun_free_netdev(struct net_device *dev)
359 struct tun_struct *tun = netdev_priv(dev);
361 BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED, &tun->socket.flags));
363 sk_release_kernel(tun->socket.sk);
366 /* Net device open. */
367 static int tun_net_open(struct net_device *dev)
369 netif_start_queue(dev);
373 /* Net device close. */
374 static int tun_net_close(struct net_device *dev)
376 netif_stop_queue(dev);
380 /* Net device start xmit */
381 static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
383 struct tun_struct *tun = netdev_priv(dev);
385 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
387 /* Drop packet if interface is not attached */
391 /* Drop if the filter does not like it.
392 * This is a noop if the filter is disabled.
393 * Filter can be enabled only for the TAP devices. */
394 if (!check_filter(&tun->txflt, skb))
397 if (tun->socket.sk->sk_filter &&
398 sk_filter(tun->socket.sk, skb))
401 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
402 if (!(tun->flags & TUN_ONE_QUEUE)) {
403 /* Normal queueing mode. */
404 /* Packet scheduler handles dropping of further packets. */
405 netif_stop_queue(dev);
407 /* We won't see all dropped packets individually, so overrun
408 * error is more appropriate. */
409 dev->stats.tx_fifo_errors++;
411 /* Single queue mode.
412 * Driver handles dropping of all packets itself. */
417 /* Orphan the skb - required as we might hang on to it
418 * for indefinite time. */
419 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
424 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
426 /* Notify and wake up reader process */
427 if (tun->flags & TUN_FASYNC)
428 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
429 wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
430 POLLRDNORM | POLLRDBAND);
434 dev->stats.tx_dropped++;
439 static void tun_net_mclist(struct net_device *dev)
442 * This callback is supposed to deal with mc filter in
443 * _rx_ path and has nothing to do with the _tx_ path.
444 * In rx path we always accept everything userspace gives us.
449 #define MAX_MTU 65535
452 tun_net_change_mtu(struct net_device *dev, int new_mtu)
454 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
460 static netdev_features_t tun_net_fix_features(struct net_device *dev,
461 netdev_features_t features)
463 struct tun_struct *tun = netdev_priv(dev);
465 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
467 #ifdef CONFIG_NET_POLL_CONTROLLER
468 static void tun_poll_controller(struct net_device *dev)
471 * Tun only receives frames when:
472 * 1) the char device endpoint gets data from user space
473 * 2) the tun socket gets a sendmsg call from user space
474 * Since both of those are syncronous operations, we are guaranteed
475 * never to have pending data when we poll for it
476 * so theres nothing to do here but return.
477 * We need this though so netpoll recognizes us as an interface that
478 * supports polling, which enables bridge devices in virt setups to
479 * still use netconsole
484 static const struct net_device_ops tun_netdev_ops = {
485 .ndo_uninit = tun_net_uninit,
486 .ndo_open = tun_net_open,
487 .ndo_stop = tun_net_close,
488 .ndo_start_xmit = tun_net_xmit,
489 .ndo_change_mtu = tun_net_change_mtu,
490 .ndo_fix_features = tun_net_fix_features,
491 #ifdef CONFIG_NET_POLL_CONTROLLER
492 .ndo_poll_controller = tun_poll_controller,
496 static const struct net_device_ops tap_netdev_ops = {
497 .ndo_uninit = tun_net_uninit,
498 .ndo_open = tun_net_open,
499 .ndo_stop = tun_net_close,
500 .ndo_start_xmit = tun_net_xmit,
501 .ndo_change_mtu = tun_net_change_mtu,
502 .ndo_fix_features = tun_net_fix_features,
503 .ndo_set_rx_mode = tun_net_mclist,
504 .ndo_set_mac_address = eth_mac_addr,
505 .ndo_validate_addr = eth_validate_addr,
506 #ifdef CONFIG_NET_POLL_CONTROLLER
507 .ndo_poll_controller = tun_poll_controller,
511 /* Initialize net device. */
512 static void tun_net_init(struct net_device *dev)
514 struct tun_struct *tun = netdev_priv(dev);
516 switch (tun->flags & TUN_TYPE_MASK) {
518 dev->netdev_ops = &tun_netdev_ops;
520 /* Point-to-Point TUN Device */
521 dev->hard_header_len = 0;
525 /* Zero header length */
526 dev->type = ARPHRD_NONE;
527 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
528 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
532 dev->netdev_ops = &tap_netdev_ops;
533 /* Ethernet TAP Device */
535 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
537 eth_hw_addr_random(dev);
539 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
544 /* Character device part */
547 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
549 struct tun_file *tfile = file->private_data;
550 struct tun_struct *tun = __tun_get(tfile);
552 unsigned int mask = 0;
559 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
561 poll_wait(file, &tun->wq.wait, wait);
563 if (!skb_queue_empty(&sk->sk_receive_queue))
564 mask |= POLLIN | POLLRDNORM;
566 if (sock_writeable(sk) ||
567 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
569 mask |= POLLOUT | POLLWRNORM;
571 if (tun->dev->reg_state != NETREG_REGISTERED)
578 /* prepad is the amount to reserve at front. len is length after that.
579 * linear is a hint as to how much to copy (usually headers). */
580 static struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
581 size_t prepad, size_t len,
582 size_t linear, int noblock)
584 struct sock *sk = tun->socket.sk;
588 sock_update_classid(sk);
590 /* Under a page? Don't bother with paged skb. */
591 if (prepad + len < PAGE_SIZE || !linear)
594 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
599 skb_reserve(skb, prepad);
600 skb_put(skb, linear);
601 skb->data_len = len - linear;
602 skb->len += len - linear;
607 /* Get packet from user space buffer */
608 static ssize_t tun_get_user(struct tun_struct *tun,
609 const struct iovec *iv, size_t count,
612 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
614 size_t len = count, align = NET_SKB_PAD;
615 struct virtio_net_hdr gso = { 0 };
618 if (!(tun->flags & TUN_NO_PI)) {
619 if ((len -= sizeof(pi)) > count)
622 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
624 offset += sizeof(pi);
627 if (tun->flags & TUN_VNET_HDR) {
628 if ((len -= tun->vnet_hdr_sz) > count)
631 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
634 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
635 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
636 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
638 if (gso.hdr_len > len)
640 offset += tun->vnet_hdr_sz;
643 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
644 align += NET_IP_ALIGN;
645 if (unlikely(len < ETH_HLEN ||
646 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
650 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
652 if (PTR_ERR(skb) != -EAGAIN)
653 tun->dev->stats.rx_dropped++;
657 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
658 tun->dev->stats.rx_dropped++;
663 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
664 if (!skb_partial_csum_set(skb, gso.csum_start,
666 tun->dev->stats.rx_frame_errors++;
672 switch (tun->flags & TUN_TYPE_MASK) {
674 if (tun->flags & TUN_NO_PI) {
675 switch (skb->data[0] & 0xf0) {
677 pi.proto = htons(ETH_P_IP);
680 pi.proto = htons(ETH_P_IPV6);
683 tun->dev->stats.rx_dropped++;
689 skb_reset_mac_header(skb);
690 skb->protocol = pi.proto;
694 skb->protocol = eth_type_trans(skb, tun->dev);
698 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
700 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
701 case VIRTIO_NET_HDR_GSO_TCPV4:
702 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
704 case VIRTIO_NET_HDR_GSO_TCPV6:
705 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
707 case VIRTIO_NET_HDR_GSO_UDP:
708 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
711 tun->dev->stats.rx_frame_errors++;
716 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
717 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
719 skb_shinfo(skb)->gso_size = gso.gso_size;
720 if (skb_shinfo(skb)->gso_size == 0) {
721 tun->dev->stats.rx_frame_errors++;
726 /* Header must be checked, and gso_segs computed. */
727 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
728 skb_shinfo(skb)->gso_segs = 0;
733 tun->dev->stats.rx_packets++;
734 tun->dev->stats.rx_bytes += len;
739 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
740 unsigned long count, loff_t pos)
742 struct file *file = iocb->ki_filp;
743 struct tun_struct *tun = tun_get(file);
749 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
751 result = tun_get_user(tun, iv, iov_length(iv, count),
752 file->f_flags & O_NONBLOCK);
758 /* Put packet to the user space buffer */
759 static ssize_t tun_put_user(struct tun_struct *tun,
761 const struct iovec *iv, int len)
763 struct tun_pi pi = { 0, skb->protocol };
766 if (!(tun->flags & TUN_NO_PI)) {
767 if ((len -= sizeof(pi)) < 0)
770 if (len < skb->len) {
771 /* Packet will be striped */
772 pi.flags |= TUN_PKT_STRIP;
775 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
780 if (tun->flags & TUN_VNET_HDR) {
781 struct virtio_net_hdr gso = { 0 }; /* no info leak */
782 if ((len -= tun->vnet_hdr_sz) < 0)
785 if (skb_is_gso(skb)) {
786 struct skb_shared_info *sinfo = skb_shinfo(skb);
788 /* This is a hint as to how much should be linear. */
789 gso.hdr_len = skb_headlen(skb);
790 gso.gso_size = sinfo->gso_size;
791 if (sinfo->gso_type & SKB_GSO_TCPV4)
792 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
793 else if (sinfo->gso_type & SKB_GSO_TCPV6)
794 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
795 else if (sinfo->gso_type & SKB_GSO_UDP)
796 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
798 pr_err("unexpected GSO type: "
799 "0x%x, gso_size %d, hdr_len %d\n",
800 sinfo->gso_type, gso.gso_size,
802 print_hex_dump(KERN_ERR, "tun: ",
805 min((int)gso.hdr_len, 64), true);
809 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
810 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
812 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
814 if (skb->ip_summed == CHECKSUM_PARTIAL) {
815 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
816 gso.csum_start = skb_checksum_start_offset(skb);
817 gso.csum_offset = skb->csum_offset;
818 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
819 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
820 } /* else everything is zero */
822 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
825 total += tun->vnet_hdr_sz;
828 len = min_t(int, skb->len, len);
830 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
833 tun->dev->stats.tx_packets++;
834 tun->dev->stats.tx_bytes += len;
839 static ssize_t tun_do_read(struct tun_struct *tun,
840 struct kiocb *iocb, const struct iovec *iv,
841 ssize_t len, int noblock)
843 DECLARE_WAITQUEUE(wait, current);
847 tun_debug(KERN_INFO, tun, "tun_chr_read\n");
849 if (unlikely(!noblock))
850 add_wait_queue(&tun->wq.wait, &wait);
852 current->state = TASK_INTERRUPTIBLE;
854 /* Read frames from the queue */
855 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
860 if (signal_pending(current)) {
864 if (tun->dev->reg_state != NETREG_REGISTERED) {
869 /* Nothing to read, let's sleep */
873 netif_wake_queue(tun->dev);
875 ret = tun_put_user(tun, skb, iv, len);
880 current->state = TASK_RUNNING;
881 if (unlikely(!noblock))
882 remove_wait_queue(&tun->wq.wait, &wait);
887 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
888 unsigned long count, loff_t pos)
890 struct file *file = iocb->ki_filp;
891 struct tun_file *tfile = file->private_data;
892 struct tun_struct *tun = __tun_get(tfile);
897 len = iov_length(iv, count);
903 ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
904 ret = min_t(ssize_t, ret, len);
910 static void tun_setup(struct net_device *dev)
912 struct tun_struct *tun = netdev_priv(dev);
917 dev->ethtool_ops = &tun_ethtool_ops;
918 dev->destructor = tun_free_netdev;
921 /* Trivial set of netlink ops to allow deleting tun or tap
922 * device with netlink.
924 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
929 static struct rtnl_link_ops tun_link_ops __read_mostly = {
931 .priv_size = sizeof(struct tun_struct),
933 .validate = tun_validate,
936 static void tun_sock_write_space(struct sock *sk)
938 struct tun_struct *tun;
939 wait_queue_head_t *wqueue;
941 if (!sock_writeable(sk))
944 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
947 wqueue = sk_sleep(sk);
948 if (wqueue && waitqueue_active(wqueue))
949 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
950 POLLWRNORM | POLLWRBAND);
952 tun = tun_sk(sk)->tun;
953 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
956 static void tun_sock_destruct(struct sock *sk)
958 free_netdev(tun_sk(sk)->tun->dev);
961 static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
962 struct msghdr *m, size_t total_len)
964 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
965 return tun_get_user(tun, m->msg_iov, total_len,
966 m->msg_flags & MSG_DONTWAIT);
969 static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
970 struct msghdr *m, size_t total_len,
973 struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
975 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
977 ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
978 flags & MSG_DONTWAIT);
979 if (ret > total_len) {
980 m->msg_flags |= MSG_TRUNC;
981 ret = flags & MSG_TRUNC ? ret : total_len;
986 static int tun_release(struct socket *sock)
993 /* Ops structure to mimic raw sockets with tun */
994 static const struct proto_ops tun_socket_ops = {
995 .sendmsg = tun_sendmsg,
996 .recvmsg = tun_recvmsg,
997 .release = tun_release,
1000 static struct proto tun_proto = {
1002 .owner = THIS_MODULE,
1003 .obj_size = sizeof(struct tun_sock),
1006 static int tun_flags(struct tun_struct *tun)
1010 if (tun->flags & TUN_TUN_DEV)
1015 if (tun->flags & TUN_NO_PI)
1018 if (tun->flags & TUN_ONE_QUEUE)
1019 flags |= IFF_ONE_QUEUE;
1021 if (tun->flags & TUN_VNET_HDR)
1022 flags |= IFF_VNET_HDR;
1027 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1030 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1031 return sprintf(buf, "0x%x\n", tun_flags(tun));
1034 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1037 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1038 return sprintf(buf, "%d\n", tun->owner);
1041 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1044 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1045 return sprintf(buf, "%d\n", tun->group);
1048 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1049 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1050 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1052 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
1055 struct tun_struct *tun;
1056 struct net_device *dev;
1059 dev = __dev_get_by_name(net, ifr->ifr_name);
1061 const struct cred *cred = current_cred();
1063 if (ifr->ifr_flags & IFF_TUN_EXCL)
1065 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1066 tun = netdev_priv(dev);
1067 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1068 tun = netdev_priv(dev);
1072 if (((tun->owner != -1 && cred->euid != tun->owner) ||
1073 (tun->group != -1 && !in_egroup_p(tun->group))) &&
1074 !capable(CAP_NET_ADMIN))
1076 err = security_tun_dev_attach(tun->socket.sk);
1080 err = tun_attach(tun, file);
1086 unsigned long flags = 0;
1088 if (!capable(CAP_NET_ADMIN))
1090 err = security_tun_dev_create();
1095 if (ifr->ifr_flags & IFF_TUN) {
1097 flags |= TUN_TUN_DEV;
1099 } else if (ifr->ifr_flags & IFF_TAP) {
1101 flags |= TUN_TAP_DEV;
1107 name = ifr->ifr_name;
1109 dev = alloc_netdev(sizeof(struct tun_struct), name,
1114 dev_net_set(dev, net);
1115 dev->rtnl_link_ops = &tun_link_ops;
1117 tun = netdev_priv(dev);
1120 tun->txflt.count = 0;
1121 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
1122 set_bit(SOCK_EXTERNALLY_ALLOCATED, &tun->socket.flags);
1125 sk = sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
1129 sk_change_net(sk, net);
1130 tun->socket.wq = &tun->wq;
1131 init_waitqueue_head(&tun->wq.wait);
1132 tun->socket.ops = &tun_socket_ops;
1133 sock_init_data(&tun->socket, sk);
1134 sk->sk_write_space = tun_sock_write_space;
1135 sk->sk_sndbuf = INT_MAX;
1137 tun_sk(sk)->tun = tun;
1139 security_tun_dev_post_create(sk);
1143 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
1145 dev->features = dev->hw_features;
1147 err = register_netdevice(tun->dev);
1151 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1152 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1153 device_create_file(&tun->dev->dev, &dev_attr_group))
1154 pr_err("Failed to create tun sysfs files\n");
1156 sk->sk_destruct = tun_sock_destruct;
1158 err = tun_attach(tun, file);
1163 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
1165 if (ifr->ifr_flags & IFF_NO_PI)
1166 tun->flags |= TUN_NO_PI;
1168 tun->flags &= ~TUN_NO_PI;
1170 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1171 tun->flags |= TUN_ONE_QUEUE;
1173 tun->flags &= ~TUN_ONE_QUEUE;
1175 if (ifr->ifr_flags & IFF_VNET_HDR)
1176 tun->flags |= TUN_VNET_HDR;
1178 tun->flags &= ~TUN_VNET_HDR;
1180 /* Make sure persistent devices do not get stuck in
1183 if (netif_running(tun->dev))
1184 netif_wake_queue(tun->dev);
1186 strcpy(ifr->ifr_name, tun->dev->name);
1190 tun_free_netdev(dev);
1197 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1200 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
1202 strcpy(ifr->ifr_name, tun->dev->name);
1204 ifr->ifr_flags = tun_flags(tun);
1209 /* This is like a cut-down ethtool ops, except done via tun fd so no
1210 * privs required. */
1211 static int set_offload(struct tun_struct *tun, unsigned long arg)
1213 netdev_features_t features = 0;
1215 if (arg & TUN_F_CSUM) {
1216 features |= NETIF_F_HW_CSUM;
1219 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1220 if (arg & TUN_F_TSO_ECN) {
1221 features |= NETIF_F_TSO_ECN;
1222 arg &= ~TUN_F_TSO_ECN;
1224 if (arg & TUN_F_TSO4)
1225 features |= NETIF_F_TSO;
1226 if (arg & TUN_F_TSO6)
1227 features |= NETIF_F_TSO6;
1228 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1231 if (arg & TUN_F_UFO) {
1232 features |= NETIF_F_UFO;
1237 /* This gives the user a way to test for new features in future by
1238 * trying to set them. */
1242 tun->set_features = features;
1243 netdev_update_features(tun->dev);
1248 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1249 unsigned long arg, int ifreq_len)
1251 struct tun_file *tfile = file->private_data;
1252 struct tun_struct *tun;
1253 void __user* argp = (void __user*)arg;
1254 struct sock_fprog fprog;
1260 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1261 if (copy_from_user(&ifr, argp, ifreq_len))
1264 if (cmd == TUNGETFEATURES) {
1265 /* Currently this just means: "what IFF flags are valid?".
1266 * This is needed because we never checked for invalid flags on
1268 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1270 (unsigned int __user*)argp);
1275 tun = __tun_get(tfile);
1276 if (cmd == TUNSETIFF && !tun) {
1277 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1279 ret = tun_set_iff(tfile->net, file, &ifr);
1284 if (copy_to_user(argp, &ifr, ifreq_len))
1293 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
1298 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1302 if (copy_to_user(argp, &ifr, ifreq_len))
1307 /* Disable/Enable checksum */
1309 /* [unimplemented] */
1310 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
1311 arg ? "disabled" : "enabled");
1315 /* Disable/Enable persist mode */
1317 tun->flags |= TUN_PERSIST;
1319 tun->flags &= ~TUN_PERSIST;
1321 tun_debug(KERN_INFO, tun, "persist %s\n",
1322 arg ? "enabled" : "disabled");
1326 /* Set owner of the device */
1327 tun->owner = (uid_t) arg;
1329 tun_debug(KERN_INFO, tun, "owner set to %d\n", tun->owner);
1333 /* Set group of the device */
1334 tun->group= (gid_t) arg;
1336 tun_debug(KERN_INFO, tun, "group set to %d\n", tun->group);
1340 /* Only allow setting the type when the interface is down */
1341 if (tun->dev->flags & IFF_UP) {
1342 tun_debug(KERN_INFO, tun,
1343 "Linktype set failed because interface is up\n");
1346 tun->dev->type = (int) arg;
1347 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1359 ret = set_offload(tun, arg);
1362 case TUNSETTXFILTER:
1363 /* Can be set only for TAPs */
1365 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1367 ret = update_filter(&tun->txflt, (void __user *)arg);
1371 /* Get hw address */
1372 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1373 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1374 if (copy_to_user(argp, &ifr, ifreq_len))
1379 /* Set hw address */
1380 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
1381 ifr.ifr_hwaddr.sa_data);
1383 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1387 sndbuf = tun->socket.sk->sk_sndbuf;
1388 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1393 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1398 tun->socket.sk->sk_sndbuf = sndbuf;
1401 case TUNGETVNETHDRSZ:
1402 vnet_hdr_sz = tun->vnet_hdr_sz;
1403 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
1407 case TUNSETVNETHDRSZ:
1408 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
1412 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
1417 tun->vnet_hdr_sz = vnet_hdr_sz;
1420 case TUNATTACHFILTER:
1421 /* Can be set only for TAPs */
1423 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1426 if (copy_from_user(&fprog, argp, sizeof(fprog)))
1429 ret = sk_attach_filter(&fprog, tun->socket.sk);
1432 case TUNDETACHFILTER:
1433 /* Can be set only for TAPs */
1435 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1437 ret = sk_detach_filter(tun->socket.sk);
1452 static long tun_chr_ioctl(struct file *file,
1453 unsigned int cmd, unsigned long arg)
1455 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1458 #ifdef CONFIG_COMPAT
1459 static long tun_chr_compat_ioctl(struct file *file,
1460 unsigned int cmd, unsigned long arg)
1465 case TUNSETTXFILTER:
1470 arg = (unsigned long)compat_ptr(arg);
1473 arg = (compat_ulong_t)arg;
1478 * compat_ifreq is shorter than ifreq, so we must not access beyond
1479 * the end of that structure. All fields that are used in this
1480 * driver are compatible though, we don't need to convert the
1483 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1485 #endif /* CONFIG_COMPAT */
1487 static int tun_chr_fasync(int fd, struct file *file, int on)
1489 struct tun_struct *tun = tun_get(file);
1495 tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
1497 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1501 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1504 tun->flags |= TUN_FASYNC;
1506 tun->flags &= ~TUN_FASYNC;
1513 static int tun_chr_open(struct inode *inode, struct file * file)
1515 struct tun_file *tfile;
1517 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
1519 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1522 atomic_set(&tfile->count, 0);
1524 tfile->net = get_net(current->nsproxy->net_ns);
1525 file->private_data = tfile;
1529 static int tun_chr_close(struct inode *inode, struct file *file)
1531 struct tun_file *tfile = file->private_data;
1532 struct tun_struct *tun;
1534 tun = __tun_get(tfile);
1536 struct net_device *dev = tun->dev;
1538 tun_debug(KERN_INFO, tun, "tun_chr_close\n");
1542 /* If desirable, unregister the netdevice. */
1543 if (!(tun->flags & TUN_PERSIST)) {
1545 if (dev->reg_state == NETREG_REGISTERED)
1546 unregister_netdevice(dev);
1553 sock_put(tun->socket.sk);
1555 put_net(tfile->net);
1561 static const struct file_operations tun_fops = {
1562 .owner = THIS_MODULE,
1563 .llseek = no_llseek,
1564 .read = do_sync_read,
1565 .aio_read = tun_chr_aio_read,
1566 .write = do_sync_write,
1567 .aio_write = tun_chr_aio_write,
1568 .poll = tun_chr_poll,
1569 .unlocked_ioctl = tun_chr_ioctl,
1570 #ifdef CONFIG_COMPAT
1571 .compat_ioctl = tun_chr_compat_ioctl,
1573 .open = tun_chr_open,
1574 .release = tun_chr_close,
1575 .fasync = tun_chr_fasync
1578 static struct miscdevice tun_miscdev = {
1581 .nodename = "net/tun",
1585 /* ethtool interface */
1587 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1590 cmd->advertising = 0;
1591 ethtool_cmd_speed_set(cmd, SPEED_10);
1592 cmd->duplex = DUPLEX_FULL;
1593 cmd->port = PORT_TP;
1594 cmd->phy_address = 0;
1595 cmd->transceiver = XCVR_INTERNAL;
1596 cmd->autoneg = AUTONEG_DISABLE;
1602 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1604 struct tun_struct *tun = netdev_priv(dev);
1606 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
1607 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
1609 switch (tun->flags & TUN_TYPE_MASK) {
1611 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
1614 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
1619 static u32 tun_get_msglevel(struct net_device *dev)
1622 struct tun_struct *tun = netdev_priv(dev);
1629 static void tun_set_msglevel(struct net_device *dev, u32 value)
1632 struct tun_struct *tun = netdev_priv(dev);
1637 static const struct ethtool_ops tun_ethtool_ops = {
1638 .get_settings = tun_get_settings,
1639 .get_drvinfo = tun_get_drvinfo,
1640 .get_msglevel = tun_get_msglevel,
1641 .set_msglevel = tun_set_msglevel,
1642 .get_link = ethtool_op_get_link,
1646 static int __init tun_init(void)
1650 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1651 pr_info("%s\n", DRV_COPYRIGHT);
1653 ret = rtnl_link_register(&tun_link_ops);
1655 pr_err("Can't register link_ops\n");
1659 ret = misc_register(&tun_miscdev);
1661 pr_err("Can't register misc device %d\n", TUN_MINOR);
1666 rtnl_link_unregister(&tun_link_ops);
1671 static void tun_cleanup(void)
1673 misc_deregister(&tun_miscdev);
1674 rtnl_link_unregister(&tun_link_ops);
1677 /* Get an underlying socket object from tun file. Returns error unless file is
1678 * attached to a device. The returned object works like a packet socket, it
1679 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1680 * holding a reference to the file for as long as the socket is in use. */
1681 struct socket *tun_get_socket(struct file *file)
1683 struct tun_struct *tun;
1684 if (file->f_op != &tun_fops)
1685 return ERR_PTR(-EINVAL);
1686 tun = tun_get(file);
1688 return ERR_PTR(-EBADFD);
1690 return &tun->socket;
1692 EXPORT_SYMBOL_GPL(tun_get_socket);
1694 module_init(tun_init);
1695 module_exit(tun_cleanup);
1696 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1697 MODULE_AUTHOR(DRV_COPYRIGHT);
1698 MODULE_LICENSE("GPL");
1699 MODULE_ALIAS_MISCDEV(TUN_MINOR);
1700 MODULE_ALIAS("devname:net/tun");