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/smp_lock.h>
48 #include <linux/poll.h>
49 #include <linux/fcntl.h>
50 #include <linux/init.h>
51 #include <linux/skbuff.h>
52 #include <linux/netdevice.h>
53 #include <linux/etherdevice.h>
54 #include <linux/miscdevice.h>
55 #include <linux/ethtool.h>
56 #include <linux/rtnetlink.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>
67 #include <asm/system.h>
68 #include <asm/uaccess.h>
70 /* Uncomment to enable debugging */
71 /* #define TUN_DEBUG 1 */
76 #define DBG if(tun->debug)printk
77 #define DBG1 if(debug==2)printk
83 #define FLT_EXACT_COUNT 8
85 unsigned int count
; /* Number of addrs. Zero means disabled */
86 u32 mask
[2]; /* Mask of the hashed addrs */
87 unsigned char addr
[FLT_EXACT_COUNT
][ETH_ALEN
];
92 struct tun_struct
*tun
;
94 wait_queue_head_t read_wait
;
98 struct tun_file
*tfile
;
103 struct sk_buff_head readq
;
105 struct net_device
*dev
;
106 struct fasync_struct
*fasync
;
108 struct tap_filter txflt
;
115 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
117 struct tun_file
*tfile
= file
->private_data
;
118 const struct cred
*cred
= current_cred();
123 /* Check permissions */
124 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
125 (tun
->group
!= -1 && cred
->egid
!= tun
->group
)) &&
126 !capable(CAP_NET_ADMIN
))
129 netif_tx_lock_bh(tun
->dev
);
143 atomic_inc(&tfile
->count
);
146 netif_tx_unlock_bh(tun
->dev
);
150 static void __tun_detach(struct tun_struct
*tun
)
152 struct tun_file
*tfile
= tun
->tfile
;
154 /* Detach from net device */
155 netif_tx_lock_bh(tun
->dev
);
158 netif_tx_unlock_bh(tun
->dev
);
160 /* Drop read queue */
161 skb_queue_purge(&tun
->readq
);
163 /* Drop the extra count on the net device */
167 static void tun_detach(struct tun_struct
*tun
)
174 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
176 struct tun_struct
*tun
= NULL
;
178 if (atomic_inc_not_zero(&tfile
->count
))
184 static struct tun_struct
*tun_get(struct file
*file
)
186 return __tun_get(file
->private_data
);
189 static void tun_put(struct tun_struct
*tun
)
191 struct tun_file
*tfile
= tun
->tfile
;
193 if (atomic_dec_and_test(&tfile
->count
))
194 tun_detach(tfile
->tun
);
198 static void addr_hash_set(u32
*mask
, const u8
*addr
)
200 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
201 mask
[n
>> 5] |= (1 << (n
& 31));
204 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
206 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
207 return mask
[n
>> 5] & (1 << (n
& 31));
210 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
212 struct { u8 u
[ETH_ALEN
]; } *addr
;
213 struct tun_filter uf
;
214 int err
, alen
, n
, nexact
;
216 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
225 alen
= ETH_ALEN
* uf
.count
;
226 addr
= kmalloc(alen
, GFP_KERNEL
);
230 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
235 /* The filter is updated without holding any locks. Which is
236 * perfectly safe. We disable it first and in the worst
237 * case we'll accept a few undesired packets. */
241 /* Use first set of addresses as an exact filter */
242 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
243 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
247 /* The rest is hashed */
248 memset(filter
->mask
, 0, sizeof(filter
->mask
));
249 for (; n
< uf
.count
; n
++)
250 addr_hash_set(filter
->mask
, addr
[n
].u
);
252 /* For ALLMULTI just set the mask to all ones.
253 * This overrides the mask populated above. */
254 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
255 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
257 /* Now enable the filter */
259 filter
->count
= nexact
;
261 /* Return the number of exact filters */
269 /* Returns: 0 - drop, !=0 - accept */
270 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
272 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
274 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
278 for (i
= 0; i
< filter
->count
; i
++)
279 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
282 /* Inexact match (multicast only) */
283 if (is_multicast_ether_addr(eh
->h_dest
))
284 return addr_hash_test(filter
->mask
, eh
->h_dest
);
290 * Checks whether the packet is accepted or not.
291 * Returns: 0 - drop, !=0 - accept
293 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
298 return run_filter(filter
, skb
);
301 /* Network device part of the driver */
303 static const struct ethtool_ops tun_ethtool_ops
;
305 /* Net device detach from fd. */
306 static void tun_net_uninit(struct net_device
*dev
)
308 struct tun_struct
*tun
= netdev_priv(dev
);
309 struct tun_file
*tfile
= tun
->tfile
;
311 /* Inform the methods they need to stop using the dev.
314 wake_up_all(&tfile
->read_wait
);
315 if (atomic_dec_and_test(&tfile
->count
))
320 /* Net device open. */
321 static int tun_net_open(struct net_device
*dev
)
323 netif_start_queue(dev
);
327 /* Net device close. */
328 static int tun_net_close(struct net_device
*dev
)
330 netif_stop_queue(dev
);
334 /* Net device start xmit */
335 static int tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
337 struct tun_struct
*tun
= netdev_priv(dev
);
339 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
341 /* Drop packet if interface is not attached */
345 /* Drop if the filter does not like it.
346 * This is a noop if the filter is disabled.
347 * Filter can be enabled only for the TAP devices. */
348 if (!check_filter(&tun
->txflt
, skb
))
351 if (skb_queue_len(&tun
->readq
) >= dev
->tx_queue_len
) {
352 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
353 /* Normal queueing mode. */
354 /* Packet scheduler handles dropping of further packets. */
355 netif_stop_queue(dev
);
357 /* We won't see all dropped packets individually, so overrun
358 * error is more appropriate. */
359 dev
->stats
.tx_fifo_errors
++;
361 /* Single queue mode.
362 * Driver handles dropping of all packets itself. */
368 skb_queue_tail(&tun
->readq
, skb
);
369 dev
->trans_start
= jiffies
;
371 /* Notify and wake up reader process */
372 if (tun
->flags
& TUN_FASYNC
)
373 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
374 wake_up_interruptible(&tun
->tfile
->read_wait
);
378 dev
->stats
.tx_dropped
++;
383 static void tun_net_mclist(struct net_device
*dev
)
386 * This callback is supposed to deal with mc filter in
387 * _rx_ path and has nothing to do with the _tx_ path.
388 * In rx path we always accept everything userspace gives us.
394 #define MAX_MTU 65535
397 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
399 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
405 static const struct net_device_ops tun_netdev_ops
= {
406 .ndo_uninit
= tun_net_uninit
,
407 .ndo_open
= tun_net_open
,
408 .ndo_stop
= tun_net_close
,
409 .ndo_start_xmit
= tun_net_xmit
,
410 .ndo_change_mtu
= tun_net_change_mtu
,
413 static const struct net_device_ops tap_netdev_ops
= {
414 .ndo_uninit
= tun_net_uninit
,
415 .ndo_open
= tun_net_open
,
416 .ndo_stop
= tun_net_close
,
417 .ndo_start_xmit
= tun_net_xmit
,
418 .ndo_change_mtu
= tun_net_change_mtu
,
419 .ndo_set_multicast_list
= tun_net_mclist
,
420 .ndo_set_mac_address
= eth_mac_addr
,
421 .ndo_validate_addr
= eth_validate_addr
,
424 /* Initialize net device. */
425 static void tun_net_init(struct net_device
*dev
)
427 struct tun_struct
*tun
= netdev_priv(dev
);
429 switch (tun
->flags
& TUN_TYPE_MASK
) {
431 dev
->netdev_ops
= &tun_netdev_ops
;
433 /* Point-to-Point TUN Device */
434 dev
->hard_header_len
= 0;
438 /* Zero header length */
439 dev
->type
= ARPHRD_NONE
;
440 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
441 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
445 dev
->netdev_ops
= &tap_netdev_ops
;
446 /* Ethernet TAP Device */
449 random_ether_addr(dev
->dev_addr
);
451 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
456 /* Character device part */
459 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
461 struct tun_file
*tfile
= file
->private_data
;
462 struct tun_struct
*tun
= __tun_get(tfile
);
463 unsigned int mask
= POLLOUT
| POLLWRNORM
;
468 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
470 poll_wait(file
, &tfile
->read_wait
, wait
);
472 if (!skb_queue_empty(&tun
->readq
))
473 mask
|= POLLIN
| POLLRDNORM
;
475 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
)
482 /* prepad is the amount to reserve at front. len is length after that.
483 * linear is a hint as to how much to copy (usually headers). */
484 static struct sk_buff
*tun_alloc_skb(size_t prepad
, size_t len
, size_t linear
,
490 skb
= alloc_skb(prepad
+ len
, gfp
|__GFP_NOWARN
);
492 skb_reserve(skb
, prepad
);
497 /* Under a page? Don't bother with paged skb. */
498 if (prepad
+ len
< PAGE_SIZE
)
501 /* Start with a normal skb, and add pages. */
502 skb
= alloc_skb(prepad
+ linear
, gfp
);
506 skb_reserve(skb
, prepad
);
507 skb_put(skb
, linear
);
511 for (i
= 0; i
< MAX_SKB_FRAGS
; i
++) {
512 skb_frag_t
*f
= &skb_shinfo(skb
)->frags
[i
];
514 f
->page
= alloc_page(gfp
|__GFP_ZERO
);
521 skb
->data_len
+= PAGE_SIZE
;
522 skb
->len
+= PAGE_SIZE
;
523 skb
->truesize
+= PAGE_SIZE
;
524 skb_shinfo(skb
)->nr_frags
++;
526 if (len
< PAGE_SIZE
) {
533 /* Too large, or alloc fail? */
542 /* Get packet from user space buffer */
543 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
, struct iovec
*iv
, size_t count
)
545 struct tun_pi pi
= { 0, __constant_htons(ETH_P_IP
) };
547 size_t len
= count
, align
= 0;
548 struct virtio_net_hdr gso
= { 0 };
550 if (!(tun
->flags
& TUN_NO_PI
)) {
551 if ((len
-= sizeof(pi
)) > count
)
554 if(memcpy_fromiovec((void *)&pi
, iv
, sizeof(pi
)))
558 if (tun
->flags
& TUN_VNET_HDR
) {
559 if ((len
-= sizeof(gso
)) > count
)
562 if (memcpy_fromiovec((void *)&gso
, iv
, sizeof(gso
)))
565 if (gso
.hdr_len
> len
)
569 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
570 align
= NET_IP_ALIGN
;
571 if (unlikely(len
< ETH_HLEN
))
575 if (!(skb
= tun_alloc_skb(align
, len
, gso
.hdr_len
, GFP_KERNEL
))) {
576 tun
->dev
->stats
.rx_dropped
++;
580 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, len
)) {
581 tun
->dev
->stats
.rx_dropped
++;
586 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
587 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
589 tun
->dev
->stats
.rx_frame_errors
++;
593 } else if (tun
->flags
& TUN_NOCHECKSUM
)
594 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
596 switch (tun
->flags
& TUN_TYPE_MASK
) {
598 if (tun
->flags
& TUN_NO_PI
) {
599 switch (skb
->data
[0] & 0xf0) {
601 pi
.proto
= htons(ETH_P_IP
);
604 pi
.proto
= htons(ETH_P_IPV6
);
607 tun
->dev
->stats
.rx_dropped
++;
613 skb_reset_mac_header(skb
);
614 skb
->protocol
= pi
.proto
;
618 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
622 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
624 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
625 case VIRTIO_NET_HDR_GSO_TCPV4
:
626 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
628 case VIRTIO_NET_HDR_GSO_TCPV6
:
629 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
632 tun
->dev
->stats
.rx_frame_errors
++;
637 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
638 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
640 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
641 if (skb_shinfo(skb
)->gso_size
== 0) {
642 tun
->dev
->stats
.rx_frame_errors
++;
647 /* Header must be checked, and gso_segs computed. */
648 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
649 skb_shinfo(skb
)->gso_segs
= 0;
654 tun
->dev
->stats
.rx_packets
++;
655 tun
->dev
->stats
.rx_bytes
+= len
;
660 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
661 unsigned long count
, loff_t pos
)
663 struct tun_struct
*tun
= tun_get(iocb
->ki_filp
);
669 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
671 result
= tun_get_user(tun
, (struct iovec
*) iv
, iov_length(iv
, count
));
677 /* Put packet to the user space buffer */
678 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
680 struct iovec
*iv
, int len
)
682 struct tun_pi pi
= { 0, skb
->protocol
};
685 if (!(tun
->flags
& TUN_NO_PI
)) {
686 if ((len
-= sizeof(pi
)) < 0)
689 if (len
< skb
->len
) {
690 /* Packet will be striped */
691 pi
.flags
|= TUN_PKT_STRIP
;
694 if (memcpy_toiovec(iv
, (void *) &pi
, sizeof(pi
)))
699 if (tun
->flags
& TUN_VNET_HDR
) {
700 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
701 if ((len
-= sizeof(gso
)) < 0)
704 if (skb_is_gso(skb
)) {
705 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
707 /* This is a hint as to how much should be linear. */
708 gso
.hdr_len
= skb_headlen(skb
);
709 gso
.gso_size
= sinfo
->gso_size
;
710 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
711 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
712 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
713 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
716 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
717 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
719 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
721 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
722 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
723 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
724 gso
.csum_offset
= skb
->csum_offset
;
725 } /* else everything is zero */
727 if (unlikely(memcpy_toiovec(iv
, (void *)&gso
, sizeof(gso
))))
729 total
+= sizeof(gso
);
732 len
= min_t(int, skb
->len
, len
);
734 skb_copy_datagram_iovec(skb
, 0, iv
, len
);
737 tun
->dev
->stats
.tx_packets
++;
738 tun
->dev
->stats
.tx_bytes
+= len
;
743 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
744 unsigned long count
, loff_t pos
)
746 struct file
*file
= iocb
->ki_filp
;
747 struct tun_file
*tfile
= file
->private_data
;
748 struct tun_struct
*tun
= __tun_get(tfile
);
749 DECLARE_WAITQUEUE(wait
, current
);
751 ssize_t len
, ret
= 0;
756 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
758 len
= iov_length(iv
, count
);
764 add_wait_queue(&tfile
->read_wait
, &wait
);
766 current
->state
= TASK_INTERRUPTIBLE
;
768 /* Read frames from the queue */
769 if (!(skb
=skb_dequeue(&tun
->readq
))) {
770 if (file
->f_flags
& O_NONBLOCK
) {
774 if (signal_pending(current
)) {
778 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
) {
783 /* Nothing to read, let's sleep */
787 netif_wake_queue(tun
->dev
);
789 ret
= tun_put_user(tun
, skb
, (struct iovec
*) iv
, len
);
794 current
->state
= TASK_RUNNING
;
795 remove_wait_queue(&tfile
->read_wait
, &wait
);
802 static void tun_setup(struct net_device
*dev
)
804 struct tun_struct
*tun
= netdev_priv(dev
);
806 skb_queue_head_init(&tun
->readq
);
811 dev
->ethtool_ops
= &tun_ethtool_ops
;
812 dev
->destructor
= free_netdev
;
815 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
817 struct tun_struct
*tun
;
818 struct net_device
*dev
;
821 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
823 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
824 tun
= netdev_priv(dev
);
825 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
826 tun
= netdev_priv(dev
);
830 err
= tun_attach(tun
, file
);
836 unsigned long flags
= 0;
840 if (!capable(CAP_NET_ADMIN
))
844 if (ifr
->ifr_flags
& IFF_TUN
) {
846 flags
|= TUN_TUN_DEV
;
848 } else if (ifr
->ifr_flags
& IFF_TAP
) {
850 flags
|= TUN_TAP_DEV
;
856 name
= ifr
->ifr_name
;
858 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
863 dev_net_set(dev
, net
);
865 tun
= netdev_priv(dev
);
868 tun
->txflt
.count
= 0;
872 if (strchr(dev
->name
, '%')) {
873 err
= dev_alloc_name(dev
, dev
->name
);
878 err
= register_netdevice(tun
->dev
);
882 err
= tun_attach(tun
, file
);
887 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
889 if (ifr
->ifr_flags
& IFF_NO_PI
)
890 tun
->flags
|= TUN_NO_PI
;
892 tun
->flags
&= ~TUN_NO_PI
;
894 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
895 tun
->flags
|= TUN_ONE_QUEUE
;
897 tun
->flags
&= ~TUN_ONE_QUEUE
;
899 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
900 tun
->flags
|= TUN_VNET_HDR
;
902 tun
->flags
&= ~TUN_VNET_HDR
;
904 /* Make sure persistent devices do not get stuck in
907 if (netif_running(tun
->dev
))
908 netif_wake_queue(tun
->dev
);
910 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
919 static int tun_get_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
921 struct tun_struct
*tun
= tun_get(file
);
926 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
928 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
932 if (ifr
->ifr_flags
& TUN_TUN_DEV
)
933 ifr
->ifr_flags
|= IFF_TUN
;
935 ifr
->ifr_flags
|= IFF_TAP
;
937 if (tun
->flags
& TUN_NO_PI
)
938 ifr
->ifr_flags
|= IFF_NO_PI
;
940 if (tun
->flags
& TUN_ONE_QUEUE
)
941 ifr
->ifr_flags
|= IFF_ONE_QUEUE
;
943 if (tun
->flags
& TUN_VNET_HDR
)
944 ifr
->ifr_flags
|= IFF_VNET_HDR
;
950 /* This is like a cut-down ethtool ops, except done via tun fd so no
952 static int set_offload(struct net_device
*dev
, unsigned long arg
)
954 unsigned int old_features
, features
;
956 old_features
= dev
->features
;
957 /* Unset features, set them as we chew on the arg. */
958 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
959 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
));
961 if (arg
& TUN_F_CSUM
) {
962 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
965 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
966 if (arg
& TUN_F_TSO_ECN
) {
967 features
|= NETIF_F_TSO_ECN
;
968 arg
&= ~TUN_F_TSO_ECN
;
970 if (arg
& TUN_F_TSO4
)
971 features
|= NETIF_F_TSO
;
972 if (arg
& TUN_F_TSO6
)
973 features
|= NETIF_F_TSO6
;
974 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
978 /* This gives the user a way to test for new features in future by
979 * trying to set them. */
983 dev
->features
= features
;
984 if (old_features
!= dev
->features
)
985 netdev_features_change(dev
);
990 static int tun_chr_ioctl(struct inode
*inode
, struct file
*file
,
991 unsigned int cmd
, unsigned long arg
)
993 struct tun_file
*tfile
= file
->private_data
;
994 struct tun_struct
*tun
;
995 void __user
* argp
= (void __user
*)arg
;
999 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
1000 if (copy_from_user(&ifr
, argp
, sizeof ifr
))
1003 if (cmd
== TUNGETFEATURES
) {
1004 /* Currently this just means: "what IFF flags are valid?".
1005 * This is needed because we never checked for invalid flags on
1007 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
1009 (unsigned int __user
*)argp
);
1012 tun
= __tun_get(tfile
);
1013 if (cmd
== TUNSETIFF
&& !tun
) {
1016 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
1019 err
= tun_set_iff(tfile
->net
, file
, &ifr
);
1025 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
1034 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
1039 ret
= tun_get_iff(current
->nsproxy
->net_ns
, file
, &ifr
);
1043 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
1048 /* Disable/Enable checksum */
1050 tun
->flags
|= TUN_NOCHECKSUM
;
1052 tun
->flags
&= ~TUN_NOCHECKSUM
;
1054 DBG(KERN_INFO
"%s: checksum %s\n",
1055 tun
->dev
->name
, arg
? "disabled" : "enabled");
1059 /* Disable/Enable persist mode */
1061 tun
->flags
|= TUN_PERSIST
;
1063 tun
->flags
&= ~TUN_PERSIST
;
1065 DBG(KERN_INFO
"%s: persist %s\n",
1066 tun
->dev
->name
, arg
? "enabled" : "disabled");
1070 /* Set owner of the device */
1071 tun
->owner
= (uid_t
) arg
;
1073 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1077 /* Set group of the device */
1078 tun
->group
= (gid_t
) arg
;
1080 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1084 /* Only allow setting the type when the interface is down */
1086 if (tun
->dev
->flags
& IFF_UP
) {
1087 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1091 tun
->dev
->type
= (int) arg
;
1092 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1105 ret
= set_offload(tun
->dev
, arg
);
1109 case TUNSETTXFILTER
:
1110 /* Can be set only for TAPs */
1112 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1115 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1121 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1122 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1123 if (copy_to_user(argp
, &ifr
, sizeof ifr
))
1128 /* Set hw address */
1129 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1130 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1133 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1145 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1147 struct tun_struct
*tun
= tun_get(file
);
1153 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1156 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1160 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1163 tun
->flags
|= TUN_FASYNC
;
1165 tun
->flags
&= ~TUN_FASYNC
;
1173 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1175 struct tun_file
*tfile
;
1176 cycle_kernel_lock();
1177 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1179 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1182 atomic_set(&tfile
->count
, 0);
1184 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1185 init_waitqueue_head(&tfile
->read_wait
);
1186 file
->private_data
= tfile
;
1190 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1192 struct tun_file
*tfile
= file
->private_data
;
1193 struct tun_struct
*tun
= __tun_get(tfile
);
1197 DBG(KERN_INFO
"%s: tun_chr_close\n", tun
->dev
->name
);
1202 /* If desireable, unregister the netdevice. */
1203 if (!(tun
->flags
& TUN_PERSIST
))
1204 unregister_netdevice(tun
->dev
);
1209 put_net(tfile
->net
);
1215 static const struct file_operations tun_fops
= {
1216 .owner
= THIS_MODULE
,
1217 .llseek
= no_llseek
,
1218 .read
= do_sync_read
,
1219 .aio_read
= tun_chr_aio_read
,
1220 .write
= do_sync_write
,
1221 .aio_write
= tun_chr_aio_write
,
1222 .poll
= tun_chr_poll
,
1223 .ioctl
= tun_chr_ioctl
,
1224 .open
= tun_chr_open
,
1225 .release
= tun_chr_close
,
1226 .fasync
= tun_chr_fasync
1229 static struct miscdevice tun_miscdev
= {
1235 /* ethtool interface */
1237 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1240 cmd
->advertising
= 0;
1241 cmd
->speed
= SPEED_10
;
1242 cmd
->duplex
= DUPLEX_FULL
;
1243 cmd
->port
= PORT_TP
;
1244 cmd
->phy_address
= 0;
1245 cmd
->transceiver
= XCVR_INTERNAL
;
1246 cmd
->autoneg
= AUTONEG_DISABLE
;
1252 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1254 struct tun_struct
*tun
= netdev_priv(dev
);
1256 strcpy(info
->driver
, DRV_NAME
);
1257 strcpy(info
->version
, DRV_VERSION
);
1258 strcpy(info
->fw_version
, "N/A");
1260 switch (tun
->flags
& TUN_TYPE_MASK
) {
1262 strcpy(info
->bus_info
, "tun");
1265 strcpy(info
->bus_info
, "tap");
1270 static u32
tun_get_msglevel(struct net_device
*dev
)
1273 struct tun_struct
*tun
= netdev_priv(dev
);
1280 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1283 struct tun_struct
*tun
= netdev_priv(dev
);
1288 static u32
tun_get_link(struct net_device
*dev
)
1290 struct tun_struct
*tun
= netdev_priv(dev
);
1291 return !!tun
->tfile
;
1294 static u32
tun_get_rx_csum(struct net_device
*dev
)
1296 struct tun_struct
*tun
= netdev_priv(dev
);
1297 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1300 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1302 struct tun_struct
*tun
= netdev_priv(dev
);
1304 tun
->flags
&= ~TUN_NOCHECKSUM
;
1306 tun
->flags
|= TUN_NOCHECKSUM
;
1310 static const struct ethtool_ops tun_ethtool_ops
= {
1311 .get_settings
= tun_get_settings
,
1312 .get_drvinfo
= tun_get_drvinfo
,
1313 .get_msglevel
= tun_get_msglevel
,
1314 .set_msglevel
= tun_set_msglevel
,
1315 .get_link
= tun_get_link
,
1316 .get_rx_csum
= tun_get_rx_csum
,
1317 .set_rx_csum
= tun_set_rx_csum
1320 static int tun_init_net(struct net
*net
)
1325 static void tun_exit_net(struct net
*net
)
1327 struct net_device
*dev
, *next
;
1330 for_each_netdev_safe(net
, dev
, next
) {
1331 if (dev
->ethtool_ops
!= &tun_ethtool_ops
)
1333 DBG(KERN_INFO
"%s cleaned up\n", dev
->name
);
1334 unregister_netdevice(dev
);
1339 static struct pernet_operations tun_net_ops
= {
1340 .init
= tun_init_net
,
1341 .exit
= tun_exit_net
,
1344 static int __init
tun_init(void)
1348 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1349 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1351 ret
= register_pernet_device(&tun_net_ops
);
1353 printk(KERN_ERR
"tun: Can't register pernet ops\n");
1357 ret
= misc_register(&tun_miscdev
);
1359 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1365 unregister_pernet_device(&tun_net_ops
);
1370 static void tun_cleanup(void)
1372 misc_deregister(&tun_miscdev
);
1373 unregister_pernet_device(&tun_net_ops
);
1376 module_init(tun_init
);
1377 module_exit(tun_cleanup
);
1378 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1379 MODULE_AUTHOR(DRV_COPYRIGHT
);
1380 MODULE_LICENSE("GPL");
1381 MODULE_ALIAS_MISCDEV(TUN_MINOR
);