2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
21 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
24 * Mark Smith <markzzzsmith@yahoo.com.au>
25 * Use random_ether_addr() for tap MAC address.
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
37 #define DRV_NAME "tun"
38 #define DRV_VERSION "1.6"
39 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
40 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
42 #include <linux/module.h>
43 #include <linux/errno.h>
44 #include <linux/kernel.h>
45 #include <linux/major.h>
46 #include <linux/slab.h>
47 #include <linux/poll.h>
48 #include <linux/fcntl.h>
49 #include <linux/init.h>
50 #include <linux/skbuff.h>
51 #include <linux/netdevice.h>
52 #include <linux/etherdevice.h>
53 #include <linux/miscdevice.h>
54 #include <linux/ethtool.h>
55 #include <linux/rtnetlink.h>
56 #include <linux/compat.h>
58 #include <linux/if_arp.h>
59 #include <linux/if_ether.h>
60 #include <linux/if_tun.h>
61 #include <linux/crc32.h>
62 #include <linux/nsproxy.h>
63 #include <linux/virtio_net.h>
64 #include <linux/rcupdate.h>
65 #include <net/net_namespace.h>
66 #include <net/netns/generic.h>
67 #include <net/rtnetlink.h>
70 #include <asm/system.h>
71 #include <asm/uaccess.h>
73 /* Uncomment to enable debugging */
74 /* #define TUN_DEBUG 1 */
79 #define DBG if(tun->debug)printk
80 #define DBG1 if(debug==2)printk
86 #define FLT_EXACT_COUNT 8
88 unsigned int count
; /* Number of addrs. Zero means disabled */
89 u32 mask
[2]; /* Mask of the hashed addrs */
90 unsigned char addr
[FLT_EXACT_COUNT
][ETH_ALEN
];
95 struct tun_struct
*tun
;
102 struct tun_file
*tfile
;
107 struct net_device
*dev
;
108 struct fasync_struct
*fasync
;
110 struct tap_filter txflt
;
111 struct socket socket
;
123 struct tun_struct
*tun
;
126 static inline struct tun_sock
*tun_sk(struct sock
*sk
)
128 return container_of(sk
, struct tun_sock
, sk
);
131 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
133 struct tun_file
*tfile
= file
->private_data
;
138 netif_tx_lock_bh(tun
->dev
);
151 tun
->socket
.file
= file
;
153 sock_hold(tun
->socket
.sk
);
154 atomic_inc(&tfile
->count
);
157 netif_tx_unlock_bh(tun
->dev
);
161 static void __tun_detach(struct tun_struct
*tun
)
163 /* Detach from net device */
164 netif_tx_lock_bh(tun
->dev
);
166 tun
->socket
.file
= NULL
;
167 netif_tx_unlock_bh(tun
->dev
);
169 /* Drop read queue */
170 skb_queue_purge(&tun
->socket
.sk
->sk_receive_queue
);
172 /* Drop the extra count on the net device */
176 static void tun_detach(struct tun_struct
*tun
)
183 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
185 struct tun_struct
*tun
= NULL
;
187 if (atomic_inc_not_zero(&tfile
->count
))
193 static struct tun_struct
*tun_get(struct file
*file
)
195 return __tun_get(file
->private_data
);
198 static void tun_put(struct tun_struct
*tun
)
200 struct tun_file
*tfile
= tun
->tfile
;
202 if (atomic_dec_and_test(&tfile
->count
))
203 tun_detach(tfile
->tun
);
207 static void addr_hash_set(u32
*mask
, const u8
*addr
)
209 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
210 mask
[n
>> 5] |= (1 << (n
& 31));
213 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
215 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
216 return mask
[n
>> 5] & (1 << (n
& 31));
219 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
221 struct { u8 u
[ETH_ALEN
]; } *addr
;
222 struct tun_filter uf
;
223 int err
, alen
, n
, nexact
;
225 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
234 alen
= ETH_ALEN
* uf
.count
;
235 addr
= kmalloc(alen
, GFP_KERNEL
);
239 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
244 /* The filter is updated without holding any locks. Which is
245 * perfectly safe. We disable it first and in the worst
246 * case we'll accept a few undesired packets. */
250 /* Use first set of addresses as an exact filter */
251 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
252 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
256 /* Remaining multicast addresses are hashed,
257 * unicast will leave the filter disabled. */
258 memset(filter
->mask
, 0, sizeof(filter
->mask
));
259 for (; n
< uf
.count
; n
++) {
260 if (!is_multicast_ether_addr(addr
[n
].u
)) {
261 err
= 0; /* no filter */
264 addr_hash_set(filter
->mask
, addr
[n
].u
);
267 /* For ALLMULTI just set the mask to all ones.
268 * This overrides the mask populated above. */
269 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
270 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
272 /* Now enable the filter */
274 filter
->count
= nexact
;
276 /* Return the number of exact filters */
284 /* Returns: 0 - drop, !=0 - accept */
285 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
287 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
289 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
293 for (i
= 0; i
< filter
->count
; i
++)
294 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
297 /* Inexact match (multicast only) */
298 if (is_multicast_ether_addr(eh
->h_dest
))
299 return addr_hash_test(filter
->mask
, eh
->h_dest
);
305 * Checks whether the packet is accepted or not.
306 * Returns: 0 - drop, !=0 - accept
308 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
313 return run_filter(filter
, skb
);
316 /* Network device part of the driver */
318 static const struct ethtool_ops tun_ethtool_ops
;
320 /* Net device detach from fd. */
321 static void tun_net_uninit(struct net_device
*dev
)
323 struct tun_struct
*tun
= netdev_priv(dev
);
324 struct tun_file
*tfile
= tun
->tfile
;
326 /* Inform the methods they need to stop using the dev.
329 wake_up_all(&tun
->wq
.wait
);
330 if (atomic_dec_and_test(&tfile
->count
))
335 static void tun_free_netdev(struct net_device
*dev
)
337 struct tun_struct
*tun
= netdev_priv(dev
);
339 sock_put(tun
->socket
.sk
);
342 /* Net device open. */
343 static int tun_net_open(struct net_device
*dev
)
345 netif_start_queue(dev
);
349 /* Net device close. */
350 static int tun_net_close(struct net_device
*dev
)
352 netif_stop_queue(dev
);
356 /* Net device start xmit */
357 static netdev_tx_t
tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
359 struct tun_struct
*tun
= netdev_priv(dev
);
361 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
363 /* Drop packet if interface is not attached */
367 /* Drop if the filter does not like it.
368 * This is a noop if the filter is disabled.
369 * Filter can be enabled only for the TAP devices. */
370 if (!check_filter(&tun
->txflt
, skb
))
373 if (tun
->socket
.sk
->sk_filter
&&
374 sk_filter(tun
->socket
.sk
, skb
))
377 if (skb_queue_len(&tun
->socket
.sk
->sk_receive_queue
) >= dev
->tx_queue_len
) {
378 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
379 /* Normal queueing mode. */
380 /* Packet scheduler handles dropping of further packets. */
381 netif_stop_queue(dev
);
383 /* We won't see all dropped packets individually, so overrun
384 * error is more appropriate. */
385 dev
->stats
.tx_fifo_errors
++;
387 /* Single queue mode.
388 * Driver handles dropping of all packets itself. */
393 /* Orphan the skb - required as we might hang on to it
394 * for indefinite time. */
398 skb_queue_tail(&tun
->socket
.sk
->sk_receive_queue
, skb
);
400 /* Notify and wake up reader process */
401 if (tun
->flags
& TUN_FASYNC
)
402 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
403 wake_up_interruptible_poll(&tun
->wq
.wait
, POLLIN
|
404 POLLRDNORM
| POLLRDBAND
);
408 dev
->stats
.tx_dropped
++;
413 static void tun_net_mclist(struct net_device
*dev
)
416 * This callback is supposed to deal with mc filter in
417 * _rx_ path and has nothing to do with the _tx_ path.
418 * In rx path we always accept everything userspace gives us.
423 #define MAX_MTU 65535
426 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
428 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
434 static const struct net_device_ops tun_netdev_ops
= {
435 .ndo_uninit
= tun_net_uninit
,
436 .ndo_open
= tun_net_open
,
437 .ndo_stop
= tun_net_close
,
438 .ndo_start_xmit
= tun_net_xmit
,
439 .ndo_change_mtu
= tun_net_change_mtu
,
442 static const struct net_device_ops tap_netdev_ops
= {
443 .ndo_uninit
= tun_net_uninit
,
444 .ndo_open
= tun_net_open
,
445 .ndo_stop
= tun_net_close
,
446 .ndo_start_xmit
= tun_net_xmit
,
447 .ndo_change_mtu
= tun_net_change_mtu
,
448 .ndo_set_multicast_list
= tun_net_mclist
,
449 .ndo_set_mac_address
= eth_mac_addr
,
450 .ndo_validate_addr
= eth_validate_addr
,
453 /* Initialize net device. */
454 static void tun_net_init(struct net_device
*dev
)
456 struct tun_struct
*tun
= netdev_priv(dev
);
458 switch (tun
->flags
& TUN_TYPE_MASK
) {
460 dev
->netdev_ops
= &tun_netdev_ops
;
462 /* Point-to-Point TUN Device */
463 dev
->hard_header_len
= 0;
467 /* Zero header length */
468 dev
->type
= ARPHRD_NONE
;
469 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
470 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
474 dev
->netdev_ops
= &tap_netdev_ops
;
475 /* Ethernet TAP Device */
478 random_ether_addr(dev
->dev_addr
);
480 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
485 /* Character device part */
488 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
490 struct tun_file
*tfile
= file
->private_data
;
491 struct tun_struct
*tun
= __tun_get(tfile
);
493 unsigned int mask
= 0;
500 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
502 poll_wait(file
, &tun
->wq
.wait
, wait
);
504 if (!skb_queue_empty(&sk
->sk_receive_queue
))
505 mask
|= POLLIN
| POLLRDNORM
;
507 if (sock_writeable(sk
) ||
508 (!test_and_set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
) &&
510 mask
|= POLLOUT
| POLLWRNORM
;
512 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
)
519 /* prepad is the amount to reserve at front. len is length after that.
520 * linear is a hint as to how much to copy (usually headers). */
521 static inline struct sk_buff
*tun_alloc_skb(struct tun_struct
*tun
,
522 size_t prepad
, size_t len
,
523 size_t linear
, int noblock
)
525 struct sock
*sk
= tun
->socket
.sk
;
529 /* Under a page? Don't bother with paged skb. */
530 if (prepad
+ len
< PAGE_SIZE
|| !linear
)
533 skb
= sock_alloc_send_pskb(sk
, prepad
+ linear
, len
- linear
, noblock
,
538 skb_reserve(skb
, prepad
);
539 skb_put(skb
, linear
);
540 skb
->data_len
= len
- linear
;
541 skb
->len
+= len
- linear
;
546 /* Get packet from user space buffer */
547 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
,
548 const struct iovec
*iv
, size_t count
,
551 struct tun_pi pi
= { 0, cpu_to_be16(ETH_P_IP
) };
553 size_t len
= count
, align
= 0;
554 struct virtio_net_hdr gso
= { 0 };
557 if (!(tun
->flags
& TUN_NO_PI
)) {
558 if ((len
-= sizeof(pi
)) > count
)
561 if (memcpy_fromiovecend((void *)&pi
, iv
, 0, sizeof(pi
)))
563 offset
+= sizeof(pi
);
566 if (tun
->flags
& TUN_VNET_HDR
) {
567 if ((len
-= tun
->vnet_hdr_sz
) > count
)
570 if (memcpy_fromiovecend((void *)&gso
, iv
, offset
, sizeof(gso
)))
573 if ((gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) &&
574 gso
.csum_start
+ gso
.csum_offset
+ 2 > gso
.hdr_len
)
575 gso
.hdr_len
= gso
.csum_start
+ gso
.csum_offset
+ 2;
577 if (gso
.hdr_len
> len
)
579 offset
+= tun
->vnet_hdr_sz
;
582 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
583 align
= NET_IP_ALIGN
;
584 if (unlikely(len
< ETH_HLEN
||
585 (gso
.hdr_len
&& gso
.hdr_len
< ETH_HLEN
)))
589 skb
= tun_alloc_skb(tun
, align
, len
, gso
.hdr_len
, noblock
);
591 if (PTR_ERR(skb
) != -EAGAIN
)
592 tun
->dev
->stats
.rx_dropped
++;
596 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, offset
, len
)) {
597 tun
->dev
->stats
.rx_dropped
++;
602 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
603 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
605 tun
->dev
->stats
.rx_frame_errors
++;
609 } else if (tun
->flags
& TUN_NOCHECKSUM
)
610 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
612 switch (tun
->flags
& TUN_TYPE_MASK
) {
614 if (tun
->flags
& TUN_NO_PI
) {
615 switch (skb
->data
[0] & 0xf0) {
617 pi
.proto
= htons(ETH_P_IP
);
620 pi
.proto
= htons(ETH_P_IPV6
);
623 tun
->dev
->stats
.rx_dropped
++;
629 skb_reset_mac_header(skb
);
630 skb
->protocol
= pi
.proto
;
634 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
638 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
640 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
641 case VIRTIO_NET_HDR_GSO_TCPV4
:
642 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
644 case VIRTIO_NET_HDR_GSO_TCPV6
:
645 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
647 case VIRTIO_NET_HDR_GSO_UDP
:
648 skb_shinfo(skb
)->gso_type
= SKB_GSO_UDP
;
651 tun
->dev
->stats
.rx_frame_errors
++;
656 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
657 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
659 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
660 if (skb_shinfo(skb
)->gso_size
== 0) {
661 tun
->dev
->stats
.rx_frame_errors
++;
666 /* Header must be checked, and gso_segs computed. */
667 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
668 skb_shinfo(skb
)->gso_segs
= 0;
673 tun
->dev
->stats
.rx_packets
++;
674 tun
->dev
->stats
.rx_bytes
+= len
;
679 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
680 unsigned long count
, loff_t pos
)
682 struct file
*file
= iocb
->ki_filp
;
683 struct tun_struct
*tun
= tun_get(file
);
689 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
691 result
= tun_get_user(tun
, iv
, iov_length(iv
, count
),
692 file
->f_flags
& O_NONBLOCK
);
698 /* Put packet to the user space buffer */
699 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
701 const struct iovec
*iv
, int len
)
703 struct tun_pi pi
= { 0, skb
->protocol
};
706 if (!(tun
->flags
& TUN_NO_PI
)) {
707 if ((len
-= sizeof(pi
)) < 0)
710 if (len
< skb
->len
) {
711 /* Packet will be striped */
712 pi
.flags
|= TUN_PKT_STRIP
;
715 if (memcpy_toiovecend(iv
, (void *) &pi
, 0, sizeof(pi
)))
720 if (tun
->flags
& TUN_VNET_HDR
) {
721 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
722 if ((len
-= tun
->vnet_hdr_sz
) < 0)
725 if (skb_is_gso(skb
)) {
726 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
728 /* This is a hint as to how much should be linear. */
729 gso
.hdr_len
= skb_headlen(skb
);
730 gso
.gso_size
= sinfo
->gso_size
;
731 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
732 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
733 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
734 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
735 else if (sinfo
->gso_type
& SKB_GSO_UDP
)
736 gso
.gso_type
= VIRTIO_NET_HDR_GSO_UDP
;
739 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
740 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
742 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
744 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
745 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
746 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
747 gso
.csum_offset
= skb
->csum_offset
;
748 } /* else everything is zero */
750 if (unlikely(memcpy_toiovecend(iv
, (void *)&gso
, total
,
753 total
+= tun
->vnet_hdr_sz
;
756 len
= min_t(int, skb
->len
, len
);
758 skb_copy_datagram_const_iovec(skb
, 0, iv
, total
, len
);
761 tun
->dev
->stats
.tx_packets
++;
762 tun
->dev
->stats
.tx_bytes
+= len
;
767 static ssize_t
tun_do_read(struct tun_struct
*tun
,
768 struct kiocb
*iocb
, const struct iovec
*iv
,
769 ssize_t len
, int noblock
)
771 DECLARE_WAITQUEUE(wait
, current
);
775 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
777 add_wait_queue(&tun
->wq
.wait
, &wait
);
779 current
->state
= TASK_INTERRUPTIBLE
;
781 /* Read frames from the queue */
782 if (!(skb
=skb_dequeue(&tun
->socket
.sk
->sk_receive_queue
))) {
787 if (signal_pending(current
)) {
791 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
) {
796 /* Nothing to read, let's sleep */
800 netif_wake_queue(tun
->dev
);
802 ret
= tun_put_user(tun
, skb
, iv
, len
);
807 current
->state
= TASK_RUNNING
;
808 remove_wait_queue(&tun
->wq
.wait
, &wait
);
813 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
814 unsigned long count
, loff_t pos
)
816 struct file
*file
= iocb
->ki_filp
;
817 struct tun_file
*tfile
= file
->private_data
;
818 struct tun_struct
*tun
= __tun_get(tfile
);
823 len
= iov_length(iv
, count
);
829 ret
= tun_do_read(tun
, iocb
, iv
, len
, file
->f_flags
& O_NONBLOCK
);
830 ret
= min_t(ssize_t
, ret
, len
);
836 static void tun_setup(struct net_device
*dev
)
838 struct tun_struct
*tun
= netdev_priv(dev
);
843 dev
->ethtool_ops
= &tun_ethtool_ops
;
844 dev
->destructor
= tun_free_netdev
;
847 /* Trivial set of netlink ops to allow deleting tun or tap
848 * device with netlink.
850 static int tun_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
855 static struct rtnl_link_ops tun_link_ops __read_mostly
= {
857 .priv_size
= sizeof(struct tun_struct
),
859 .validate
= tun_validate
,
862 static void tun_sock_write_space(struct sock
*sk
)
864 struct tun_struct
*tun
;
865 wait_queue_head_t
*wqueue
;
867 if (!sock_writeable(sk
))
870 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
))
873 wqueue
= sk_sleep(sk
);
874 if (wqueue
&& waitqueue_active(wqueue
))
875 wake_up_interruptible_sync_poll(wqueue
, POLLOUT
|
876 POLLWRNORM
| POLLWRBAND
);
878 tun
= tun_sk(sk
)->tun
;
879 kill_fasync(&tun
->fasync
, SIGIO
, POLL_OUT
);
882 static void tun_sock_destruct(struct sock
*sk
)
884 free_netdev(tun_sk(sk
)->tun
->dev
);
887 static int tun_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
888 struct msghdr
*m
, size_t total_len
)
890 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
891 return tun_get_user(tun
, m
->msg_iov
, total_len
,
892 m
->msg_flags
& MSG_DONTWAIT
);
895 static int tun_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
896 struct msghdr
*m
, size_t total_len
,
899 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
901 if (flags
& ~(MSG_DONTWAIT
|MSG_TRUNC
))
903 ret
= tun_do_read(tun
, iocb
, m
->msg_iov
, total_len
,
904 flags
& MSG_DONTWAIT
);
905 if (ret
> total_len
) {
906 m
->msg_flags
|= MSG_TRUNC
;
907 ret
= flags
& MSG_TRUNC
? ret
: total_len
;
912 /* Ops structure to mimic raw sockets with tun */
913 static const struct proto_ops tun_socket_ops
= {
914 .sendmsg
= tun_sendmsg
,
915 .recvmsg
= tun_recvmsg
,
918 static struct proto tun_proto
= {
920 .owner
= THIS_MODULE
,
921 .obj_size
= sizeof(struct tun_sock
),
924 static int tun_flags(struct tun_struct
*tun
)
928 if (tun
->flags
& TUN_TUN_DEV
)
933 if (tun
->flags
& TUN_NO_PI
)
936 if (tun
->flags
& TUN_ONE_QUEUE
)
937 flags
|= IFF_ONE_QUEUE
;
939 if (tun
->flags
& TUN_VNET_HDR
)
940 flags
|= IFF_VNET_HDR
;
945 static ssize_t
tun_show_flags(struct device
*dev
, struct device_attribute
*attr
,
948 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
949 return sprintf(buf
, "0x%x\n", tun_flags(tun
));
952 static ssize_t
tun_show_owner(struct device
*dev
, struct device_attribute
*attr
,
955 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
956 return sprintf(buf
, "%d\n", tun
->owner
);
959 static ssize_t
tun_show_group(struct device
*dev
, struct device_attribute
*attr
,
962 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
963 return sprintf(buf
, "%d\n", tun
->group
);
966 static DEVICE_ATTR(tun_flags
, 0444, tun_show_flags
, NULL
);
967 static DEVICE_ATTR(owner
, 0444, tun_show_owner
, NULL
);
968 static DEVICE_ATTR(group
, 0444, tun_show_group
, NULL
);
970 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
973 struct tun_struct
*tun
;
974 struct net_device
*dev
;
977 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
979 const struct cred
*cred
= current_cred();
981 if (ifr
->ifr_flags
& IFF_TUN_EXCL
)
983 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
984 tun
= netdev_priv(dev
);
985 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
986 tun
= netdev_priv(dev
);
990 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
991 (tun
->group
!= -1 && !in_egroup_p(tun
->group
))) &&
992 !capable(CAP_NET_ADMIN
))
994 err
= security_tun_dev_attach(tun
->socket
.sk
);
998 err
= tun_attach(tun
, file
);
1004 unsigned long flags
= 0;
1006 if (!capable(CAP_NET_ADMIN
))
1008 err
= security_tun_dev_create();
1013 if (ifr
->ifr_flags
& IFF_TUN
) {
1015 flags
|= TUN_TUN_DEV
;
1017 } else if (ifr
->ifr_flags
& IFF_TAP
) {
1019 flags
|= TUN_TAP_DEV
;
1025 name
= ifr
->ifr_name
;
1027 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
1032 dev_net_set(dev
, net
);
1033 dev
->rtnl_link_ops
= &tun_link_ops
;
1035 tun
= netdev_priv(dev
);
1038 tun
->txflt
.count
= 0;
1039 tun
->vnet_hdr_sz
= sizeof(struct virtio_net_hdr
);
1042 sk
= sk_alloc(net
, AF_UNSPEC
, GFP_KERNEL
, &tun_proto
);
1046 tun
->socket
.wq
= &tun
->wq
;
1047 init_waitqueue_head(&tun
->wq
.wait
);
1048 tun
->socket
.ops
= &tun_socket_ops
;
1049 sock_init_data(&tun
->socket
, sk
);
1050 sk
->sk_write_space
= tun_sock_write_space
;
1051 sk
->sk_sndbuf
= INT_MAX
;
1053 tun_sk(sk
)->tun
= tun
;
1055 security_tun_dev_post_create(sk
);
1059 if (strchr(dev
->name
, '%')) {
1060 err
= dev_alloc_name(dev
, dev
->name
);
1065 err
= register_netdevice(tun
->dev
);
1069 if (device_create_file(&tun
->dev
->dev
, &dev_attr_tun_flags
) ||
1070 device_create_file(&tun
->dev
->dev
, &dev_attr_owner
) ||
1071 device_create_file(&tun
->dev
->dev
, &dev_attr_group
))
1072 printk(KERN_ERR
"Failed to create tun sysfs files\n");
1074 sk
->sk_destruct
= tun_sock_destruct
;
1076 err
= tun_attach(tun
, file
);
1081 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
1083 if (ifr
->ifr_flags
& IFF_NO_PI
)
1084 tun
->flags
|= TUN_NO_PI
;
1086 tun
->flags
&= ~TUN_NO_PI
;
1088 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
1089 tun
->flags
|= TUN_ONE_QUEUE
;
1091 tun
->flags
&= ~TUN_ONE_QUEUE
;
1093 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
1094 tun
->flags
|= TUN_VNET_HDR
;
1096 tun
->flags
&= ~TUN_VNET_HDR
;
1098 /* Make sure persistent devices do not get stuck in
1101 if (netif_running(tun
->dev
))
1102 netif_wake_queue(tun
->dev
);
1104 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1115 static int tun_get_iff(struct net
*net
, struct tun_struct
*tun
,
1118 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
1120 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1122 ifr
->ifr_flags
= tun_flags(tun
);
1127 /* This is like a cut-down ethtool ops, except done via tun fd so no
1128 * privs required. */
1129 static int set_offload(struct net_device
*dev
, unsigned long arg
)
1131 unsigned int old_features
, features
;
1133 old_features
= dev
->features
;
1134 /* Unset features, set them as we chew on the arg. */
1135 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
1136 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
1139 if (arg
& TUN_F_CSUM
) {
1140 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
1143 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
1144 if (arg
& TUN_F_TSO_ECN
) {
1145 features
|= NETIF_F_TSO_ECN
;
1146 arg
&= ~TUN_F_TSO_ECN
;
1148 if (arg
& TUN_F_TSO4
)
1149 features
|= NETIF_F_TSO
;
1150 if (arg
& TUN_F_TSO6
)
1151 features
|= NETIF_F_TSO6
;
1152 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
1155 if (arg
& TUN_F_UFO
) {
1156 features
|= NETIF_F_UFO
;
1161 /* This gives the user a way to test for new features in future by
1162 * trying to set them. */
1166 dev
->features
= features
;
1167 if (old_features
!= dev
->features
)
1168 netdev_features_change(dev
);
1173 static long __tun_chr_ioctl(struct file
*file
, unsigned int cmd
,
1174 unsigned long arg
, int ifreq_len
)
1176 struct tun_file
*tfile
= file
->private_data
;
1177 struct tun_struct
*tun
;
1178 void __user
* argp
= (void __user
*)arg
;
1179 struct sock_fprog fprog
;
1185 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
1186 if (copy_from_user(&ifr
, argp
, ifreq_len
))
1189 if (cmd
== TUNGETFEATURES
) {
1190 /* Currently this just means: "what IFF flags are valid?".
1191 * This is needed because we never checked for invalid flags on
1193 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
1195 (unsigned int __user
*)argp
);
1200 tun
= __tun_get(tfile
);
1201 if (cmd
== TUNSETIFF
&& !tun
) {
1202 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
1204 ret
= tun_set_iff(tfile
->net
, file
, &ifr
);
1209 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1218 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
1223 ret
= tun_get_iff(current
->nsproxy
->net_ns
, tun
, &ifr
);
1227 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1232 /* Disable/Enable checksum */
1234 tun
->flags
|= TUN_NOCHECKSUM
;
1236 tun
->flags
&= ~TUN_NOCHECKSUM
;
1238 DBG(KERN_INFO
"%s: checksum %s\n",
1239 tun
->dev
->name
, arg
? "disabled" : "enabled");
1243 /* Disable/Enable persist mode */
1245 tun
->flags
|= TUN_PERSIST
;
1247 tun
->flags
&= ~TUN_PERSIST
;
1249 DBG(KERN_INFO
"%s: persist %s\n",
1250 tun
->dev
->name
, arg
? "enabled" : "disabled");
1254 /* Set owner of the device */
1255 tun
->owner
= (uid_t
) arg
;
1257 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1261 /* Set group of the device */
1262 tun
->group
= (gid_t
) arg
;
1264 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1268 /* Only allow setting the type when the interface is down */
1269 if (tun
->dev
->flags
& IFF_UP
) {
1270 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1274 tun
->dev
->type
= (int) arg
;
1275 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1286 ret
= set_offload(tun
->dev
, arg
);
1289 case TUNSETTXFILTER
:
1290 /* Can be set only for TAPs */
1292 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1294 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1299 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1300 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1301 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1306 /* Set hw address */
1307 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1308 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1310 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1314 sndbuf
= tun
->socket
.sk
->sk_sndbuf
;
1315 if (copy_to_user(argp
, &sndbuf
, sizeof(sndbuf
)))
1320 if (copy_from_user(&sndbuf
, argp
, sizeof(sndbuf
))) {
1325 tun
->socket
.sk
->sk_sndbuf
= sndbuf
;
1328 case TUNGETVNETHDRSZ
:
1329 vnet_hdr_sz
= tun
->vnet_hdr_sz
;
1330 if (copy_to_user(argp
, &vnet_hdr_sz
, sizeof(vnet_hdr_sz
)))
1334 case TUNSETVNETHDRSZ
:
1335 if (copy_from_user(&vnet_hdr_sz
, argp
, sizeof(vnet_hdr_sz
))) {
1339 if (vnet_hdr_sz
< (int)sizeof(struct virtio_net_hdr
)) {
1344 tun
->vnet_hdr_sz
= vnet_hdr_sz
;
1347 case TUNATTACHFILTER
:
1348 /* Can be set only for TAPs */
1350 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1353 if (copy_from_user(&fprog
, argp
, sizeof(fprog
)))
1356 ret
= sk_attach_filter(&fprog
, tun
->socket
.sk
);
1359 case TUNDETACHFILTER
:
1360 /* Can be set only for TAPs */
1362 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1364 ret
= sk_detach_filter(tun
->socket
.sk
);
1379 static long tun_chr_ioctl(struct file
*file
,
1380 unsigned int cmd
, unsigned long arg
)
1382 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof (struct ifreq
));
1385 #ifdef CONFIG_COMPAT
1386 static long tun_chr_compat_ioctl(struct file
*file
,
1387 unsigned int cmd
, unsigned long arg
)
1392 case TUNSETTXFILTER
:
1397 arg
= (unsigned long)compat_ptr(arg
);
1400 arg
= (compat_ulong_t
)arg
;
1405 * compat_ifreq is shorter than ifreq, so we must not access beyond
1406 * the end of that structure. All fields that are used in this
1407 * driver are compatible though, we don't need to convert the
1410 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof(struct compat_ifreq
));
1412 #endif /* CONFIG_COMPAT */
1414 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1416 struct tun_struct
*tun
= tun_get(file
);
1422 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1424 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1428 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1431 tun
->flags
|= TUN_FASYNC
;
1433 tun
->flags
&= ~TUN_FASYNC
;
1440 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1442 struct tun_file
*tfile
;
1444 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1446 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1449 atomic_set(&tfile
->count
, 0);
1451 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1452 file
->private_data
= tfile
;
1456 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1458 struct tun_file
*tfile
= file
->private_data
;
1459 struct tun_struct
*tun
;
1461 tun
= __tun_get(tfile
);
1463 struct net_device
*dev
= tun
->dev
;
1465 DBG(KERN_INFO
"%s: tun_chr_close\n", dev
->name
);
1469 /* If desirable, unregister the netdevice. */
1470 if (!(tun
->flags
& TUN_PERSIST
)) {
1472 if (dev
->reg_state
== NETREG_REGISTERED
)
1473 unregister_netdevice(dev
);
1480 sock_put(tun
->socket
.sk
);
1482 put_net(tfile
->net
);
1488 static const struct file_operations tun_fops
= {
1489 .owner
= THIS_MODULE
,
1490 .llseek
= no_llseek
,
1491 .read
= do_sync_read
,
1492 .aio_read
= tun_chr_aio_read
,
1493 .write
= do_sync_write
,
1494 .aio_write
= tun_chr_aio_write
,
1495 .poll
= tun_chr_poll
,
1496 .unlocked_ioctl
= tun_chr_ioctl
,
1497 #ifdef CONFIG_COMPAT
1498 .compat_ioctl
= tun_chr_compat_ioctl
,
1500 .open
= tun_chr_open
,
1501 .release
= tun_chr_close
,
1502 .fasync
= tun_chr_fasync
1505 static struct miscdevice tun_miscdev
= {
1508 .nodename
= "net/tun",
1512 /* ethtool interface */
1514 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1517 cmd
->advertising
= 0;
1518 cmd
->speed
= SPEED_10
;
1519 cmd
->duplex
= DUPLEX_FULL
;
1520 cmd
->port
= PORT_TP
;
1521 cmd
->phy_address
= 0;
1522 cmd
->transceiver
= XCVR_INTERNAL
;
1523 cmd
->autoneg
= AUTONEG_DISABLE
;
1529 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1531 struct tun_struct
*tun
= netdev_priv(dev
);
1533 strcpy(info
->driver
, DRV_NAME
);
1534 strcpy(info
->version
, DRV_VERSION
);
1535 strcpy(info
->fw_version
, "N/A");
1537 switch (tun
->flags
& TUN_TYPE_MASK
) {
1539 strcpy(info
->bus_info
, "tun");
1542 strcpy(info
->bus_info
, "tap");
1547 static u32
tun_get_msglevel(struct net_device
*dev
)
1550 struct tun_struct
*tun
= netdev_priv(dev
);
1557 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1560 struct tun_struct
*tun
= netdev_priv(dev
);
1565 static u32
tun_get_link(struct net_device
*dev
)
1567 struct tun_struct
*tun
= netdev_priv(dev
);
1568 return !!tun
->tfile
;
1571 static u32
tun_get_rx_csum(struct net_device
*dev
)
1573 struct tun_struct
*tun
= netdev_priv(dev
);
1574 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1577 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1579 struct tun_struct
*tun
= netdev_priv(dev
);
1581 tun
->flags
&= ~TUN_NOCHECKSUM
;
1583 tun
->flags
|= TUN_NOCHECKSUM
;
1587 static const struct ethtool_ops tun_ethtool_ops
= {
1588 .get_settings
= tun_get_settings
,
1589 .get_drvinfo
= tun_get_drvinfo
,
1590 .get_msglevel
= tun_get_msglevel
,
1591 .set_msglevel
= tun_set_msglevel
,
1592 .get_link
= tun_get_link
,
1593 .get_rx_csum
= tun_get_rx_csum
,
1594 .set_rx_csum
= tun_set_rx_csum
1598 static int __init
tun_init(void)
1602 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1603 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1605 ret
= rtnl_link_register(&tun_link_ops
);
1607 printk(KERN_ERR
"tun: Can't register link_ops\n");
1611 ret
= misc_register(&tun_miscdev
);
1613 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1618 rtnl_link_unregister(&tun_link_ops
);
1623 static void tun_cleanup(void)
1625 misc_deregister(&tun_miscdev
);
1626 rtnl_link_unregister(&tun_link_ops
);
1629 /* Get an underlying socket object from tun file. Returns error unless file is
1630 * attached to a device. The returned object works like a packet socket, it
1631 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1632 * holding a reference to the file for as long as the socket is in use. */
1633 struct socket
*tun_get_socket(struct file
*file
)
1635 struct tun_struct
*tun
;
1636 if (file
->f_op
!= &tun_fops
)
1637 return ERR_PTR(-EINVAL
);
1638 tun
= tun_get(file
);
1640 return ERR_PTR(-EBADFD
);
1642 return &tun
->socket
;
1644 EXPORT_SYMBOL_GPL(tun_get_socket
);
1646 module_init(tun_init
);
1647 module_exit(tun_cleanup
);
1648 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1649 MODULE_AUTHOR(DRV_COPYRIGHT
);
1650 MODULE_LICENSE("GPL");
1651 MODULE_ALIAS_MISCDEV(TUN_MINOR
);
1652 MODULE_ALIAS("devname:net/tun");