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 <net/net_namespace.h>
65 #include <net/netns/generic.h>
66 #include <net/rtnetlink.h>
69 #include <asm/system.h>
70 #include <asm/uaccess.h>
72 /* Uncomment to enable debugging */
73 /* #define TUN_DEBUG 1 */
78 #define DBG if(tun->debug)printk
79 #define DBG1 if(debug==2)printk
85 #define FLT_EXACT_COUNT 8
87 unsigned int count
; /* Number of addrs. Zero means disabled */
88 u32 mask
[2]; /* Mask of the hashed addrs */
89 unsigned char addr
[FLT_EXACT_COUNT
][ETH_ALEN
];
94 struct tun_struct
*tun
;
101 struct tun_file
*tfile
;
106 struct net_device
*dev
;
107 struct fasync_struct
*fasync
;
109 struct tap_filter txflt
;
110 struct socket socket
;
119 struct tun_struct
*tun
;
122 static inline struct tun_sock
*tun_sk(struct sock
*sk
)
124 return container_of(sk
, struct tun_sock
, sk
);
127 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
129 struct tun_file
*tfile
= file
->private_data
;
134 netif_tx_lock_bh(tun
->dev
);
147 tun
->socket
.file
= file
;
149 sock_hold(tun
->socket
.sk
);
150 atomic_inc(&tfile
->count
);
153 netif_tx_unlock_bh(tun
->dev
);
157 static void __tun_detach(struct tun_struct
*tun
)
159 /* Detach from net device */
160 netif_tx_lock_bh(tun
->dev
);
162 tun
->socket
.file
= NULL
;
163 netif_tx_unlock_bh(tun
->dev
);
165 /* Drop read queue */
166 skb_queue_purge(&tun
->socket
.sk
->sk_receive_queue
);
168 /* Drop the extra count on the net device */
172 static void tun_detach(struct tun_struct
*tun
)
179 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
181 struct tun_struct
*tun
= NULL
;
183 if (atomic_inc_not_zero(&tfile
->count
))
189 static struct tun_struct
*tun_get(struct file
*file
)
191 return __tun_get(file
->private_data
);
194 static void tun_put(struct tun_struct
*tun
)
196 struct tun_file
*tfile
= tun
->tfile
;
198 if (atomic_dec_and_test(&tfile
->count
))
199 tun_detach(tfile
->tun
);
203 static void addr_hash_set(u32
*mask
, const u8
*addr
)
205 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
206 mask
[n
>> 5] |= (1 << (n
& 31));
209 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
211 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
212 return mask
[n
>> 5] & (1 << (n
& 31));
215 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
217 struct { u8 u
[ETH_ALEN
]; } *addr
;
218 struct tun_filter uf
;
219 int err
, alen
, n
, nexact
;
221 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
230 alen
= ETH_ALEN
* uf
.count
;
231 addr
= kmalloc(alen
, GFP_KERNEL
);
235 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
240 /* The filter is updated without holding any locks. Which is
241 * perfectly safe. We disable it first and in the worst
242 * case we'll accept a few undesired packets. */
246 /* Use first set of addresses as an exact filter */
247 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
248 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
252 /* Remaining multicast addresses are hashed,
253 * unicast will leave the filter disabled. */
254 memset(filter
->mask
, 0, sizeof(filter
->mask
));
255 for (; n
< uf
.count
; n
++) {
256 if (!is_multicast_ether_addr(addr
[n
].u
)) {
257 err
= 0; /* no filter */
260 addr_hash_set(filter
->mask
, addr
[n
].u
);
263 /* For ALLMULTI just set the mask to all ones.
264 * This overrides the mask populated above. */
265 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
266 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
268 /* Now enable the filter */
270 filter
->count
= nexact
;
272 /* Return the number of exact filters */
280 /* Returns: 0 - drop, !=0 - accept */
281 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
283 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
285 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
289 for (i
= 0; i
< filter
->count
; i
++)
290 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
293 /* Inexact match (multicast only) */
294 if (is_multicast_ether_addr(eh
->h_dest
))
295 return addr_hash_test(filter
->mask
, eh
->h_dest
);
301 * Checks whether the packet is accepted or not.
302 * Returns: 0 - drop, !=0 - accept
304 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
309 return run_filter(filter
, skb
);
312 /* Network device part of the driver */
314 static const struct ethtool_ops tun_ethtool_ops
;
316 /* Net device detach from fd. */
317 static void tun_net_uninit(struct net_device
*dev
)
319 struct tun_struct
*tun
= netdev_priv(dev
);
320 struct tun_file
*tfile
= tun
->tfile
;
322 /* Inform the methods they need to stop using the dev.
325 wake_up_all(&tun
->socket
.wait
);
326 if (atomic_dec_and_test(&tfile
->count
))
331 static void tun_free_netdev(struct net_device
*dev
)
333 struct tun_struct
*tun
= netdev_priv(dev
);
335 sock_put(tun
->socket
.sk
);
338 /* Net device open. */
339 static int tun_net_open(struct net_device
*dev
)
341 netif_start_queue(dev
);
345 /* Net device close. */
346 static int tun_net_close(struct net_device
*dev
)
348 netif_stop_queue(dev
);
352 /* Net device start xmit */
353 static netdev_tx_t
tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
355 struct tun_struct
*tun
= netdev_priv(dev
);
357 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
359 /* Drop packet if interface is not attached */
363 /* Drop if the filter does not like it.
364 * This is a noop if the filter is disabled.
365 * Filter can be enabled only for the TAP devices. */
366 if (!check_filter(&tun
->txflt
, skb
))
369 if (skb_queue_len(&tun
->socket
.sk
->sk_receive_queue
) >= dev
->tx_queue_len
) {
370 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
371 /* Normal queueing mode. */
372 /* Packet scheduler handles dropping of further packets. */
373 netif_stop_queue(dev
);
375 /* We won't see all dropped packets individually, so overrun
376 * error is more appropriate. */
377 dev
->stats
.tx_fifo_errors
++;
379 /* Single queue mode.
380 * Driver handles dropping of all packets itself. */
386 skb_queue_tail(&tun
->socket
.sk
->sk_receive_queue
, skb
);
387 dev
->trans_start
= jiffies
;
389 /* Notify and wake up reader process */
390 if (tun
->flags
& TUN_FASYNC
)
391 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
392 wake_up_interruptible_poll(&tun
->socket
.wait
, POLLIN
|
393 POLLRDNORM
| POLLRDBAND
);
397 dev
->stats
.tx_dropped
++;
402 static void tun_net_mclist(struct net_device
*dev
)
405 * This callback is supposed to deal with mc filter in
406 * _rx_ path and has nothing to do with the _tx_ path.
407 * In rx path we always accept everything userspace gives us.
413 #define MAX_MTU 65535
416 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
418 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
424 static const struct net_device_ops tun_netdev_ops
= {
425 .ndo_uninit
= tun_net_uninit
,
426 .ndo_open
= tun_net_open
,
427 .ndo_stop
= tun_net_close
,
428 .ndo_start_xmit
= tun_net_xmit
,
429 .ndo_change_mtu
= tun_net_change_mtu
,
432 static const struct net_device_ops tap_netdev_ops
= {
433 .ndo_uninit
= tun_net_uninit
,
434 .ndo_open
= tun_net_open
,
435 .ndo_stop
= tun_net_close
,
436 .ndo_start_xmit
= tun_net_xmit
,
437 .ndo_change_mtu
= tun_net_change_mtu
,
438 .ndo_set_multicast_list
= tun_net_mclist
,
439 .ndo_set_mac_address
= eth_mac_addr
,
440 .ndo_validate_addr
= eth_validate_addr
,
443 /* Initialize net device. */
444 static void tun_net_init(struct net_device
*dev
)
446 struct tun_struct
*tun
= netdev_priv(dev
);
448 switch (tun
->flags
& TUN_TYPE_MASK
) {
450 dev
->netdev_ops
= &tun_netdev_ops
;
452 /* Point-to-Point TUN Device */
453 dev
->hard_header_len
= 0;
457 /* Zero header length */
458 dev
->type
= ARPHRD_NONE
;
459 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
460 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
464 dev
->netdev_ops
= &tap_netdev_ops
;
465 /* Ethernet TAP Device */
468 random_ether_addr(dev
->dev_addr
);
470 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
475 /* Character device part */
478 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
480 struct tun_file
*tfile
= file
->private_data
;
481 struct tun_struct
*tun
= __tun_get(tfile
);
483 unsigned int mask
= 0;
490 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
492 poll_wait(file
, &tun
->socket
.wait
, wait
);
494 if (!skb_queue_empty(&sk
->sk_receive_queue
))
495 mask
|= POLLIN
| POLLRDNORM
;
497 if (sock_writeable(sk
) ||
498 (!test_and_set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
) &&
500 mask
|= POLLOUT
| POLLWRNORM
;
502 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
)
509 /* prepad is the amount to reserve at front. len is length after that.
510 * linear is a hint as to how much to copy (usually headers). */
511 static inline struct sk_buff
*tun_alloc_skb(struct tun_struct
*tun
,
512 size_t prepad
, size_t len
,
513 size_t linear
, int noblock
)
515 struct sock
*sk
= tun
->socket
.sk
;
519 /* Under a page? Don't bother with paged skb. */
520 if (prepad
+ len
< PAGE_SIZE
|| !linear
)
523 skb
= sock_alloc_send_pskb(sk
, prepad
+ linear
, len
- linear
, noblock
,
528 skb_reserve(skb
, prepad
);
529 skb_put(skb
, linear
);
530 skb
->data_len
= len
- linear
;
531 skb
->len
+= len
- linear
;
536 /* Get packet from user space buffer */
537 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
,
538 const struct iovec
*iv
, size_t count
,
541 struct tun_pi pi
= { 0, cpu_to_be16(ETH_P_IP
) };
543 size_t len
= count
, align
= 0;
544 struct virtio_net_hdr gso
= { 0 };
547 if (!(tun
->flags
& TUN_NO_PI
)) {
548 if ((len
-= sizeof(pi
)) > count
)
551 if (memcpy_fromiovecend((void *)&pi
, iv
, 0, sizeof(pi
)))
553 offset
+= sizeof(pi
);
556 if (tun
->flags
& TUN_VNET_HDR
) {
557 if ((len
-= sizeof(gso
)) > count
)
560 if (memcpy_fromiovecend((void *)&gso
, iv
, offset
, sizeof(gso
)))
563 if ((gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) &&
564 gso
.csum_start
+ gso
.csum_offset
+ 2 > gso
.hdr_len
)
565 gso
.hdr_len
= gso
.csum_start
+ gso
.csum_offset
+ 2;
567 if (gso
.hdr_len
> len
)
569 offset
+= sizeof(gso
);
572 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
573 align
= NET_IP_ALIGN
;
574 if (unlikely(len
< ETH_HLEN
||
575 (gso
.hdr_len
&& gso
.hdr_len
< ETH_HLEN
)))
579 skb
= tun_alloc_skb(tun
, align
, len
, gso
.hdr_len
, noblock
);
581 if (PTR_ERR(skb
) != -EAGAIN
)
582 tun
->dev
->stats
.rx_dropped
++;
586 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, offset
, len
)) {
587 tun
->dev
->stats
.rx_dropped
++;
592 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
593 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
595 tun
->dev
->stats
.rx_frame_errors
++;
599 } else if (tun
->flags
& TUN_NOCHECKSUM
)
600 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
602 switch (tun
->flags
& TUN_TYPE_MASK
) {
604 if (tun
->flags
& TUN_NO_PI
) {
605 switch (skb
->data
[0] & 0xf0) {
607 pi
.proto
= htons(ETH_P_IP
);
610 pi
.proto
= htons(ETH_P_IPV6
);
613 tun
->dev
->stats
.rx_dropped
++;
619 skb_reset_mac_header(skb
);
620 skb
->protocol
= pi
.proto
;
624 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
628 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
630 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
631 case VIRTIO_NET_HDR_GSO_TCPV4
:
632 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
634 case VIRTIO_NET_HDR_GSO_TCPV6
:
635 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
637 case VIRTIO_NET_HDR_GSO_UDP
:
638 skb_shinfo(skb
)->gso_type
= SKB_GSO_UDP
;
641 tun
->dev
->stats
.rx_frame_errors
++;
646 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
647 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
649 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
650 if (skb_shinfo(skb
)->gso_size
== 0) {
651 tun
->dev
->stats
.rx_frame_errors
++;
656 /* Header must be checked, and gso_segs computed. */
657 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
658 skb_shinfo(skb
)->gso_segs
= 0;
663 tun
->dev
->stats
.rx_packets
++;
664 tun
->dev
->stats
.rx_bytes
+= len
;
669 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
670 unsigned long count
, loff_t pos
)
672 struct file
*file
= iocb
->ki_filp
;
673 struct tun_struct
*tun
= tun_get(file
);
679 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
681 result
= tun_get_user(tun
, iv
, iov_length(iv
, count
),
682 file
->f_flags
& O_NONBLOCK
);
688 /* Put packet to the user space buffer */
689 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
691 const struct iovec
*iv
, int len
)
693 struct tun_pi pi
= { 0, skb
->protocol
};
696 if (!(tun
->flags
& TUN_NO_PI
)) {
697 if ((len
-= sizeof(pi
)) < 0)
700 if (len
< skb
->len
) {
701 /* Packet will be striped */
702 pi
.flags
|= TUN_PKT_STRIP
;
705 if (memcpy_toiovecend(iv
, (void *) &pi
, 0, sizeof(pi
)))
710 if (tun
->flags
& TUN_VNET_HDR
) {
711 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
712 if ((len
-= sizeof(gso
)) < 0)
715 if (skb_is_gso(skb
)) {
716 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
718 /* This is a hint as to how much should be linear. */
719 gso
.hdr_len
= skb_headlen(skb
);
720 gso
.gso_size
= sinfo
->gso_size
;
721 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
722 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
723 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
724 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
725 else if (sinfo
->gso_type
& SKB_GSO_UDP
)
726 gso
.gso_type
= VIRTIO_NET_HDR_GSO_UDP
;
729 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
730 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
732 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
734 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
735 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
736 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
737 gso
.csum_offset
= skb
->csum_offset
;
738 } /* else everything is zero */
740 if (unlikely(memcpy_toiovecend(iv
, (void *)&gso
, total
,
743 total
+= sizeof(gso
);
746 len
= min_t(int, skb
->len
, len
);
748 skb_copy_datagram_const_iovec(skb
, 0, iv
, total
, len
);
751 tun
->dev
->stats
.tx_packets
++;
752 tun
->dev
->stats
.tx_bytes
+= len
;
757 static ssize_t
tun_do_read(struct tun_struct
*tun
,
758 struct kiocb
*iocb
, const struct iovec
*iv
,
759 ssize_t len
, int noblock
)
761 DECLARE_WAITQUEUE(wait
, current
);
765 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
767 add_wait_queue(&tun
->socket
.wait
, &wait
);
769 current
->state
= TASK_INTERRUPTIBLE
;
771 /* Read frames from the queue */
772 if (!(skb
=skb_dequeue(&tun
->socket
.sk
->sk_receive_queue
))) {
777 if (signal_pending(current
)) {
781 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
) {
786 /* Nothing to read, let's sleep */
790 netif_wake_queue(tun
->dev
);
792 ret
= tun_put_user(tun
, skb
, iv
, len
);
797 current
->state
= TASK_RUNNING
;
798 remove_wait_queue(&tun
->socket
.wait
, &wait
);
803 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
804 unsigned long count
, loff_t pos
)
806 struct file
*file
= iocb
->ki_filp
;
807 struct tun_file
*tfile
= file
->private_data
;
808 struct tun_struct
*tun
= __tun_get(tfile
);
813 len
= iov_length(iv
, count
);
819 ret
= tun_do_read(tun
, iocb
, iv
, len
, file
->f_flags
& O_NONBLOCK
);
820 ret
= min_t(ssize_t
, ret
, len
);
826 static void tun_setup(struct net_device
*dev
)
828 struct tun_struct
*tun
= netdev_priv(dev
);
833 dev
->ethtool_ops
= &tun_ethtool_ops
;
834 dev
->destructor
= tun_free_netdev
;
837 /* Trivial set of netlink ops to allow deleting tun or tap
838 * device with netlink.
840 static int tun_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
845 static struct rtnl_link_ops tun_link_ops __read_mostly
= {
847 .priv_size
= sizeof(struct tun_struct
),
849 .validate
= tun_validate
,
852 static void tun_sock_write_space(struct sock
*sk
)
854 struct tun_struct
*tun
;
856 if (!sock_writeable(sk
))
859 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
))
862 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
863 wake_up_interruptible_sync_poll(sk
->sk_sleep
, POLLOUT
|
864 POLLWRNORM
| POLLWRBAND
);
866 tun
= tun_sk(sk
)->tun
;
867 kill_fasync(&tun
->fasync
, SIGIO
, POLL_OUT
);
870 static void tun_sock_destruct(struct sock
*sk
)
872 free_netdev(tun_sk(sk
)->tun
->dev
);
875 static int tun_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
876 struct msghdr
*m
, size_t total_len
)
878 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
879 return tun_get_user(tun
, m
->msg_iov
, total_len
,
880 m
->msg_flags
& MSG_DONTWAIT
);
883 static int tun_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
884 struct msghdr
*m
, size_t total_len
,
887 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
889 if (flags
& ~(MSG_DONTWAIT
|MSG_TRUNC
))
891 ret
= tun_do_read(tun
, iocb
, m
->msg_iov
, total_len
,
892 flags
& MSG_DONTWAIT
);
893 if (ret
> total_len
) {
894 m
->msg_flags
|= MSG_TRUNC
;
895 ret
= flags
& MSG_TRUNC
? ret
: total_len
;
900 /* Ops structure to mimic raw sockets with tun */
901 static const struct proto_ops tun_socket_ops
= {
902 .sendmsg
= tun_sendmsg
,
903 .recvmsg
= tun_recvmsg
,
906 static struct proto tun_proto
= {
908 .owner
= THIS_MODULE
,
909 .obj_size
= sizeof(struct tun_sock
),
912 static int tun_flags(struct tun_struct
*tun
)
916 if (tun
->flags
& TUN_TUN_DEV
)
921 if (tun
->flags
& TUN_NO_PI
)
924 if (tun
->flags
& TUN_ONE_QUEUE
)
925 flags
|= IFF_ONE_QUEUE
;
927 if (tun
->flags
& TUN_VNET_HDR
)
928 flags
|= IFF_VNET_HDR
;
933 static ssize_t
tun_show_flags(struct device
*dev
, struct device_attribute
*attr
,
936 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
937 return sprintf(buf
, "0x%x\n", tun_flags(tun
));
940 static ssize_t
tun_show_owner(struct device
*dev
, struct device_attribute
*attr
,
943 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
944 return sprintf(buf
, "%d\n", tun
->owner
);
947 static ssize_t
tun_show_group(struct device
*dev
, struct device_attribute
*attr
,
950 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
951 return sprintf(buf
, "%d\n", tun
->group
);
954 static DEVICE_ATTR(tun_flags
, 0444, tun_show_flags
, NULL
);
955 static DEVICE_ATTR(owner
, 0444, tun_show_owner
, NULL
);
956 static DEVICE_ATTR(group
, 0444, tun_show_group
, NULL
);
958 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
961 struct tun_struct
*tun
;
962 struct net_device
*dev
;
965 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
967 const struct cred
*cred
= current_cred();
969 if (ifr
->ifr_flags
& IFF_TUN_EXCL
)
971 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
972 tun
= netdev_priv(dev
);
973 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
974 tun
= netdev_priv(dev
);
978 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
979 (tun
->group
!= -1 && !in_egroup_p(tun
->group
))) &&
980 !capable(CAP_NET_ADMIN
))
982 err
= security_tun_dev_attach(tun
->socket
.sk
);
986 err
= tun_attach(tun
, file
);
992 unsigned long flags
= 0;
994 if (!capable(CAP_NET_ADMIN
))
996 err
= security_tun_dev_create();
1001 if (ifr
->ifr_flags
& IFF_TUN
) {
1003 flags
|= TUN_TUN_DEV
;
1005 } else if (ifr
->ifr_flags
& IFF_TAP
) {
1007 flags
|= TUN_TAP_DEV
;
1013 name
= ifr
->ifr_name
;
1015 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
1020 dev_net_set(dev
, net
);
1021 dev
->rtnl_link_ops
= &tun_link_ops
;
1023 tun
= netdev_priv(dev
);
1026 tun
->txflt
.count
= 0;
1029 sk
= sk_alloc(net
, AF_UNSPEC
, GFP_KERNEL
, &tun_proto
);
1033 init_waitqueue_head(&tun
->socket
.wait
);
1034 tun
->socket
.ops
= &tun_socket_ops
;
1035 sock_init_data(&tun
->socket
, sk
);
1036 sk
->sk_write_space
= tun_sock_write_space
;
1037 sk
->sk_sndbuf
= INT_MAX
;
1039 tun_sk(sk
)->tun
= tun
;
1041 security_tun_dev_post_create(sk
);
1045 if (strchr(dev
->name
, '%')) {
1046 err
= dev_alloc_name(dev
, dev
->name
);
1051 err
= register_netdevice(tun
->dev
);
1055 if (device_create_file(&tun
->dev
->dev
, &dev_attr_tun_flags
) ||
1056 device_create_file(&tun
->dev
->dev
, &dev_attr_owner
) ||
1057 device_create_file(&tun
->dev
->dev
, &dev_attr_group
))
1058 printk(KERN_ERR
"Failed to create tun sysfs files\n");
1060 sk
->sk_destruct
= tun_sock_destruct
;
1062 err
= tun_attach(tun
, file
);
1067 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
1069 if (ifr
->ifr_flags
& IFF_NO_PI
)
1070 tun
->flags
|= TUN_NO_PI
;
1072 tun
->flags
&= ~TUN_NO_PI
;
1074 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
1075 tun
->flags
|= TUN_ONE_QUEUE
;
1077 tun
->flags
&= ~TUN_ONE_QUEUE
;
1079 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
1080 tun
->flags
|= TUN_VNET_HDR
;
1082 tun
->flags
&= ~TUN_VNET_HDR
;
1084 /* Make sure persistent devices do not get stuck in
1087 if (netif_running(tun
->dev
))
1088 netif_wake_queue(tun
->dev
);
1090 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1101 static int tun_get_iff(struct net
*net
, struct tun_struct
*tun
,
1104 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
1106 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1108 ifr
->ifr_flags
= tun_flags(tun
);
1113 /* This is like a cut-down ethtool ops, except done via tun fd so no
1114 * privs required. */
1115 static int set_offload(struct net_device
*dev
, unsigned long arg
)
1117 unsigned int old_features
, features
;
1119 old_features
= dev
->features
;
1120 /* Unset features, set them as we chew on the arg. */
1121 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
1122 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
1125 if (arg
& TUN_F_CSUM
) {
1126 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
1129 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
1130 if (arg
& TUN_F_TSO_ECN
) {
1131 features
|= NETIF_F_TSO_ECN
;
1132 arg
&= ~TUN_F_TSO_ECN
;
1134 if (arg
& TUN_F_TSO4
)
1135 features
|= NETIF_F_TSO
;
1136 if (arg
& TUN_F_TSO6
)
1137 features
|= NETIF_F_TSO6
;
1138 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
1141 if (arg
& TUN_F_UFO
) {
1142 features
|= NETIF_F_UFO
;
1147 /* This gives the user a way to test for new features in future by
1148 * trying to set them. */
1152 dev
->features
= features
;
1153 if (old_features
!= dev
->features
)
1154 netdev_features_change(dev
);
1159 static long __tun_chr_ioctl(struct file
*file
, unsigned int cmd
,
1160 unsigned long arg
, int ifreq_len
)
1162 struct tun_file
*tfile
= file
->private_data
;
1163 struct tun_struct
*tun
;
1164 void __user
* argp
= (void __user
*)arg
;
1169 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
1170 if (copy_from_user(&ifr
, argp
, ifreq_len
))
1173 if (cmd
== TUNGETFEATURES
) {
1174 /* Currently this just means: "what IFF flags are valid?".
1175 * This is needed because we never checked for invalid flags on
1177 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
1179 (unsigned int __user
*)argp
);
1184 tun
= __tun_get(tfile
);
1185 if (cmd
== TUNSETIFF
&& !tun
) {
1186 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
1188 ret
= tun_set_iff(tfile
->net
, file
, &ifr
);
1193 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1202 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
1207 ret
= tun_get_iff(current
->nsproxy
->net_ns
, tun
, &ifr
);
1211 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1216 /* Disable/Enable checksum */
1218 tun
->flags
|= TUN_NOCHECKSUM
;
1220 tun
->flags
&= ~TUN_NOCHECKSUM
;
1222 DBG(KERN_INFO
"%s: checksum %s\n",
1223 tun
->dev
->name
, arg
? "disabled" : "enabled");
1227 /* Disable/Enable persist mode */
1229 tun
->flags
|= TUN_PERSIST
;
1231 tun
->flags
&= ~TUN_PERSIST
;
1233 DBG(KERN_INFO
"%s: persist %s\n",
1234 tun
->dev
->name
, arg
? "enabled" : "disabled");
1238 /* Set owner of the device */
1239 tun
->owner
= (uid_t
) arg
;
1241 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1245 /* Set group of the device */
1246 tun
->group
= (gid_t
) arg
;
1248 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1252 /* Only allow setting the type when the interface is down */
1253 if (tun
->dev
->flags
& IFF_UP
) {
1254 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1258 tun
->dev
->type
= (int) arg
;
1259 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1270 ret
= set_offload(tun
->dev
, arg
);
1273 case TUNSETTXFILTER
:
1274 /* Can be set only for TAPs */
1276 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1278 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1283 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1284 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1285 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1290 /* Set hw address */
1291 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1292 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1294 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1298 sndbuf
= tun
->socket
.sk
->sk_sndbuf
;
1299 if (copy_to_user(argp
, &sndbuf
, sizeof(sndbuf
)))
1304 if (copy_from_user(&sndbuf
, argp
, sizeof(sndbuf
))) {
1309 tun
->socket
.sk
->sk_sndbuf
= sndbuf
;
1324 static long tun_chr_ioctl(struct file
*file
,
1325 unsigned int cmd
, unsigned long arg
)
1327 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof (struct ifreq
));
1330 #ifdef CONFIG_COMPAT
1331 static long tun_chr_compat_ioctl(struct file
*file
,
1332 unsigned int cmd
, unsigned long arg
)
1337 case TUNSETTXFILTER
:
1342 arg
= (unsigned long)compat_ptr(arg
);
1345 arg
= (compat_ulong_t
)arg
;
1350 * compat_ifreq is shorter than ifreq, so we must not access beyond
1351 * the end of that structure. All fields that are used in this
1352 * driver are compatible though, we don't need to convert the
1355 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof(struct compat_ifreq
));
1357 #endif /* CONFIG_COMPAT */
1359 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1361 struct tun_struct
*tun
= tun_get(file
);
1367 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1369 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1373 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1376 tun
->flags
|= TUN_FASYNC
;
1378 tun
->flags
&= ~TUN_FASYNC
;
1385 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1387 struct tun_file
*tfile
;
1389 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1391 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1394 atomic_set(&tfile
->count
, 0);
1396 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1397 file
->private_data
= tfile
;
1401 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1403 struct tun_file
*tfile
= file
->private_data
;
1404 struct tun_struct
*tun
;
1406 tun
= __tun_get(tfile
);
1408 struct net_device
*dev
= tun
->dev
;
1410 DBG(KERN_INFO
"%s: tun_chr_close\n", dev
->name
);
1414 /* If desireable, unregister the netdevice. */
1415 if (!(tun
->flags
& TUN_PERSIST
)) {
1417 if (dev
->reg_state
== NETREG_REGISTERED
)
1418 unregister_netdevice(dev
);
1425 sock_put(tun
->socket
.sk
);
1427 put_net(tfile
->net
);
1433 static const struct file_operations tun_fops
= {
1434 .owner
= THIS_MODULE
,
1435 .llseek
= no_llseek
,
1436 .read
= do_sync_read
,
1437 .aio_read
= tun_chr_aio_read
,
1438 .write
= do_sync_write
,
1439 .aio_write
= tun_chr_aio_write
,
1440 .poll
= tun_chr_poll
,
1441 .unlocked_ioctl
= tun_chr_ioctl
,
1442 #ifdef CONFIG_COMPAT
1443 .compat_ioctl
= tun_chr_compat_ioctl
,
1445 .open
= tun_chr_open
,
1446 .release
= tun_chr_close
,
1447 .fasync
= tun_chr_fasync
1450 static struct miscdevice tun_miscdev
= {
1453 .nodename
= "net/tun",
1457 /* ethtool interface */
1459 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1462 cmd
->advertising
= 0;
1463 cmd
->speed
= SPEED_10
;
1464 cmd
->duplex
= DUPLEX_FULL
;
1465 cmd
->port
= PORT_TP
;
1466 cmd
->phy_address
= 0;
1467 cmd
->transceiver
= XCVR_INTERNAL
;
1468 cmd
->autoneg
= AUTONEG_DISABLE
;
1474 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1476 struct tun_struct
*tun
= netdev_priv(dev
);
1478 strcpy(info
->driver
, DRV_NAME
);
1479 strcpy(info
->version
, DRV_VERSION
);
1480 strcpy(info
->fw_version
, "N/A");
1482 switch (tun
->flags
& TUN_TYPE_MASK
) {
1484 strcpy(info
->bus_info
, "tun");
1487 strcpy(info
->bus_info
, "tap");
1492 static u32
tun_get_msglevel(struct net_device
*dev
)
1495 struct tun_struct
*tun
= netdev_priv(dev
);
1502 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1505 struct tun_struct
*tun
= netdev_priv(dev
);
1510 static u32
tun_get_link(struct net_device
*dev
)
1512 struct tun_struct
*tun
= netdev_priv(dev
);
1513 return !!tun
->tfile
;
1516 static u32
tun_get_rx_csum(struct net_device
*dev
)
1518 struct tun_struct
*tun
= netdev_priv(dev
);
1519 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1522 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1524 struct tun_struct
*tun
= netdev_priv(dev
);
1526 tun
->flags
&= ~TUN_NOCHECKSUM
;
1528 tun
->flags
|= TUN_NOCHECKSUM
;
1532 static const struct ethtool_ops tun_ethtool_ops
= {
1533 .get_settings
= tun_get_settings
,
1534 .get_drvinfo
= tun_get_drvinfo
,
1535 .get_msglevel
= tun_get_msglevel
,
1536 .set_msglevel
= tun_set_msglevel
,
1537 .get_link
= tun_get_link
,
1538 .get_rx_csum
= tun_get_rx_csum
,
1539 .set_rx_csum
= tun_set_rx_csum
1543 static int __init
tun_init(void)
1547 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1548 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1550 ret
= rtnl_link_register(&tun_link_ops
);
1552 printk(KERN_ERR
"tun: Can't register link_ops\n");
1556 ret
= misc_register(&tun_miscdev
);
1558 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1563 rtnl_link_unregister(&tun_link_ops
);
1568 static void tun_cleanup(void)
1570 misc_deregister(&tun_miscdev
);
1571 rtnl_link_unregister(&tun_link_ops
);
1574 /* Get an underlying socket object from tun file. Returns error unless file is
1575 * attached to a device. The returned object works like a packet socket, it
1576 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1577 * holding a reference to the file for as long as the socket is in use. */
1578 struct socket
*tun_get_socket(struct file
*file
)
1580 struct tun_struct
*tun
;
1581 if (file
->f_op
!= &tun_fops
)
1582 return ERR_PTR(-EINVAL
);
1583 tun
= tun_get(file
);
1585 return ERR_PTR(-EBADFD
);
1587 return &tun
->socket
;
1589 EXPORT_SYMBOL_GPL(tun_get_socket
);
1591 module_init(tun_init
);
1592 module_exit(tun_cleanup
);
1593 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1594 MODULE_AUTHOR(DRV_COPYRIGHT
);
1595 MODULE_LICENSE("GPL");
1596 MODULE_ALIAS_MISCDEV(TUN_MINOR
);