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
];
91 struct tun_struct
*tun
;
96 struct tun_file
*tfile
;
101 wait_queue_head_t read_wait
;
102 struct sk_buff_head readq
;
104 struct net_device
*dev
;
105 struct fasync_struct
*fasync
;
107 struct tap_filter txflt
;
114 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
116 struct tun_file
*tfile
= file
->private_data
;
117 const struct cred
*cred
= current_cred();
122 /* Check permissions */
123 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
124 (tun
->group
!= -1 && cred
->egid
!= tun
->group
)) &&
125 !capable(CAP_NET_ADMIN
))
128 netif_tx_lock_bh(tun
->dev
);
143 netif_tx_unlock_bh(tun
->dev
);
147 static void __tun_detach(struct tun_struct
*tun
)
149 struct tun_file
*tfile
= tun
->tfile
;
151 /* Detach from net device */
152 netif_tx_lock_bh(tun
->dev
);
155 netif_tx_unlock_bh(tun
->dev
);
157 /* Drop read queue */
158 skb_queue_purge(&tun
->readq
);
161 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
166 static struct tun_struct
*tun_get(struct file
*file
)
168 return __tun_get(file
->private_data
);
171 static void tun_put(struct tun_struct
*tun
)
177 static void addr_hash_set(u32
*mask
, const u8
*addr
)
179 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
180 mask
[n
>> 5] |= (1 << (n
& 31));
183 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
185 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
186 return mask
[n
>> 5] & (1 << (n
& 31));
189 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
191 struct { u8 u
[ETH_ALEN
]; } *addr
;
192 struct tun_filter uf
;
193 int err
, alen
, n
, nexact
;
195 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
204 alen
= ETH_ALEN
* uf
.count
;
205 addr
= kmalloc(alen
, GFP_KERNEL
);
209 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
214 /* The filter is updated without holding any locks. Which is
215 * perfectly safe. We disable it first and in the worst
216 * case we'll accept a few undesired packets. */
220 /* Use first set of addresses as an exact filter */
221 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
222 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
226 /* The rest is hashed */
227 memset(filter
->mask
, 0, sizeof(filter
->mask
));
228 for (; n
< uf
.count
; n
++)
229 addr_hash_set(filter
->mask
, addr
[n
].u
);
231 /* For ALLMULTI just set the mask to all ones.
232 * This overrides the mask populated above. */
233 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
234 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
236 /* Now enable the filter */
238 filter
->count
= nexact
;
240 /* Return the number of exact filters */
248 /* Returns: 0 - drop, !=0 - accept */
249 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
251 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
253 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
257 for (i
= 0; i
< filter
->count
; i
++)
258 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
261 /* Inexact match (multicast only) */
262 if (is_multicast_ether_addr(eh
->h_dest
))
263 return addr_hash_test(filter
->mask
, eh
->h_dest
);
269 * Checks whether the packet is accepted or not.
270 * Returns: 0 - drop, !=0 - accept
272 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
277 return run_filter(filter
, skb
);
280 /* Network device part of the driver */
282 static const struct ethtool_ops tun_ethtool_ops
;
284 /* Net device open. */
285 static int tun_net_open(struct net_device
*dev
)
287 netif_start_queue(dev
);
291 /* Net device close. */
292 static int tun_net_close(struct net_device
*dev
)
294 netif_stop_queue(dev
);
298 /* Net device start xmit */
299 static int tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
301 struct tun_struct
*tun
= netdev_priv(dev
);
303 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
305 /* Drop packet if interface is not attached */
309 /* Drop if the filter does not like it.
310 * This is a noop if the filter is disabled.
311 * Filter can be enabled only for the TAP devices. */
312 if (!check_filter(&tun
->txflt
, skb
))
315 if (skb_queue_len(&tun
->readq
) >= dev
->tx_queue_len
) {
316 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
317 /* Normal queueing mode. */
318 /* Packet scheduler handles dropping of further packets. */
319 netif_stop_queue(dev
);
321 /* We won't see all dropped packets individually, so overrun
322 * error is more appropriate. */
323 dev
->stats
.tx_fifo_errors
++;
325 /* Single queue mode.
326 * Driver handles dropping of all packets itself. */
332 skb_queue_tail(&tun
->readq
, skb
);
333 dev
->trans_start
= jiffies
;
335 /* Notify and wake up reader process */
336 if (tun
->flags
& TUN_FASYNC
)
337 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
338 wake_up_interruptible(&tun
->read_wait
);
342 dev
->stats
.tx_dropped
++;
347 static void tun_net_mclist(struct net_device
*dev
)
350 * This callback is supposed to deal with mc filter in
351 * _rx_ path and has nothing to do with the _tx_ path.
352 * In rx path we always accept everything userspace gives us.
358 #define MAX_MTU 65535
361 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
363 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
369 static const struct net_device_ops tun_netdev_ops
= {
370 .ndo_open
= tun_net_open
,
371 .ndo_stop
= tun_net_close
,
372 .ndo_start_xmit
= tun_net_xmit
,
373 .ndo_change_mtu
= tun_net_change_mtu
,
376 static const struct net_device_ops tap_netdev_ops
= {
377 .ndo_open
= tun_net_open
,
378 .ndo_stop
= tun_net_close
,
379 .ndo_start_xmit
= tun_net_xmit
,
380 .ndo_change_mtu
= tun_net_change_mtu
,
381 .ndo_set_multicast_list
= tun_net_mclist
,
382 .ndo_set_mac_address
= eth_mac_addr
,
383 .ndo_validate_addr
= eth_validate_addr
,
386 /* Initialize net device. */
387 static void tun_net_init(struct net_device
*dev
)
389 struct tun_struct
*tun
= netdev_priv(dev
);
391 switch (tun
->flags
& TUN_TYPE_MASK
) {
393 dev
->netdev_ops
= &tun_netdev_ops
;
395 /* Point-to-Point TUN Device */
396 dev
->hard_header_len
= 0;
400 /* Zero header length */
401 dev
->type
= ARPHRD_NONE
;
402 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
403 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
407 dev
->netdev_ops
= &tap_netdev_ops
;
408 /* Ethernet TAP Device */
411 random_ether_addr(dev
->dev_addr
);
413 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
418 /* Character device part */
421 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
423 struct tun_struct
*tun
= tun_get(file
);
424 unsigned int mask
= POLLOUT
| POLLWRNORM
;
429 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
431 poll_wait(file
, &tun
->read_wait
, wait
);
433 if (!skb_queue_empty(&tun
->readq
))
434 mask
|= POLLIN
| POLLRDNORM
;
440 /* prepad is the amount to reserve at front. len is length after that.
441 * linear is a hint as to how much to copy (usually headers). */
442 static struct sk_buff
*tun_alloc_skb(size_t prepad
, size_t len
, size_t linear
,
448 skb
= alloc_skb(prepad
+ len
, gfp
|__GFP_NOWARN
);
450 skb_reserve(skb
, prepad
);
455 /* Under a page? Don't bother with paged skb. */
456 if (prepad
+ len
< PAGE_SIZE
)
459 /* Start with a normal skb, and add pages. */
460 skb
= alloc_skb(prepad
+ linear
, gfp
);
464 skb_reserve(skb
, prepad
);
465 skb_put(skb
, linear
);
469 for (i
= 0; i
< MAX_SKB_FRAGS
; i
++) {
470 skb_frag_t
*f
= &skb_shinfo(skb
)->frags
[i
];
472 f
->page
= alloc_page(gfp
|__GFP_ZERO
);
479 skb
->data_len
+= PAGE_SIZE
;
480 skb
->len
+= PAGE_SIZE
;
481 skb
->truesize
+= PAGE_SIZE
;
482 skb_shinfo(skb
)->nr_frags
++;
484 if (len
< PAGE_SIZE
) {
491 /* Too large, or alloc fail? */
500 /* Get packet from user space buffer */
501 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
, struct iovec
*iv
, size_t count
)
503 struct tun_pi pi
= { 0, __constant_htons(ETH_P_IP
) };
505 size_t len
= count
, align
= 0;
506 struct virtio_net_hdr gso
= { 0 };
508 if (!(tun
->flags
& TUN_NO_PI
)) {
509 if ((len
-= sizeof(pi
)) > count
)
512 if(memcpy_fromiovec((void *)&pi
, iv
, sizeof(pi
)))
516 if (tun
->flags
& TUN_VNET_HDR
) {
517 if ((len
-= sizeof(gso
)) > count
)
520 if (memcpy_fromiovec((void *)&gso
, iv
, sizeof(gso
)))
523 if (gso
.hdr_len
> len
)
527 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
528 align
= NET_IP_ALIGN
;
529 if (unlikely(len
< ETH_HLEN
))
533 if (!(skb
= tun_alloc_skb(align
, len
, gso
.hdr_len
, GFP_KERNEL
))) {
534 tun
->dev
->stats
.rx_dropped
++;
538 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, len
)) {
539 tun
->dev
->stats
.rx_dropped
++;
544 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
545 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
547 tun
->dev
->stats
.rx_frame_errors
++;
551 } else if (tun
->flags
& TUN_NOCHECKSUM
)
552 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
554 switch (tun
->flags
& TUN_TYPE_MASK
) {
556 if (tun
->flags
& TUN_NO_PI
) {
557 switch (skb
->data
[0] & 0xf0) {
559 pi
.proto
= htons(ETH_P_IP
);
562 pi
.proto
= htons(ETH_P_IPV6
);
565 tun
->dev
->stats
.rx_dropped
++;
571 skb_reset_mac_header(skb
);
572 skb
->protocol
= pi
.proto
;
576 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
580 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
582 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
583 case VIRTIO_NET_HDR_GSO_TCPV4
:
584 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
586 case VIRTIO_NET_HDR_GSO_TCPV6
:
587 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
590 tun
->dev
->stats
.rx_frame_errors
++;
595 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
596 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
598 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
599 if (skb_shinfo(skb
)->gso_size
== 0) {
600 tun
->dev
->stats
.rx_frame_errors
++;
605 /* Header must be checked, and gso_segs computed. */
606 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
607 skb_shinfo(skb
)->gso_segs
= 0;
612 tun
->dev
->stats
.rx_packets
++;
613 tun
->dev
->stats
.rx_bytes
+= len
;
618 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
619 unsigned long count
, loff_t pos
)
621 struct tun_struct
*tun
= tun_get(iocb
->ki_filp
);
627 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
629 result
= tun_get_user(tun
, (struct iovec
*) iv
, iov_length(iv
, count
));
635 /* Put packet to the user space buffer */
636 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
638 struct iovec
*iv
, int len
)
640 struct tun_pi pi
= { 0, skb
->protocol
};
643 if (!(tun
->flags
& TUN_NO_PI
)) {
644 if ((len
-= sizeof(pi
)) < 0)
647 if (len
< skb
->len
) {
648 /* Packet will be striped */
649 pi
.flags
|= TUN_PKT_STRIP
;
652 if (memcpy_toiovec(iv
, (void *) &pi
, sizeof(pi
)))
657 if (tun
->flags
& TUN_VNET_HDR
) {
658 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
659 if ((len
-= sizeof(gso
)) < 0)
662 if (skb_is_gso(skb
)) {
663 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
665 /* This is a hint as to how much should be linear. */
666 gso
.hdr_len
= skb_headlen(skb
);
667 gso
.gso_size
= sinfo
->gso_size
;
668 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
669 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
670 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
671 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
674 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
675 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
677 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
679 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
680 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
681 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
682 gso
.csum_offset
= skb
->csum_offset
;
683 } /* else everything is zero */
685 if (unlikely(memcpy_toiovec(iv
, (void *)&gso
, sizeof(gso
))))
687 total
+= sizeof(gso
);
690 len
= min_t(int, skb
->len
, len
);
692 skb_copy_datagram_iovec(skb
, 0, iv
, len
);
695 tun
->dev
->stats
.tx_packets
++;
696 tun
->dev
->stats
.tx_bytes
+= len
;
701 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
702 unsigned long count
, loff_t pos
)
704 struct file
*file
= iocb
->ki_filp
;
705 struct tun_struct
*tun
= tun_get(file
);
706 DECLARE_WAITQUEUE(wait
, current
);
708 ssize_t len
, ret
= 0;
713 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
715 len
= iov_length(iv
, count
);
721 add_wait_queue(&tun
->read_wait
, &wait
);
723 current
->state
= TASK_INTERRUPTIBLE
;
725 /* Read frames from the queue */
726 if (!(skb
=skb_dequeue(&tun
->readq
))) {
727 if (file
->f_flags
& O_NONBLOCK
) {
731 if (signal_pending(current
)) {
736 /* Nothing to read, let's sleep */
740 netif_wake_queue(tun
->dev
);
742 ret
= tun_put_user(tun
, skb
, (struct iovec
*) iv
, len
);
747 current
->state
= TASK_RUNNING
;
748 remove_wait_queue(&tun
->read_wait
, &wait
);
755 static void tun_setup(struct net_device
*dev
)
757 struct tun_struct
*tun
= netdev_priv(dev
);
759 skb_queue_head_init(&tun
->readq
);
760 init_waitqueue_head(&tun
->read_wait
);
765 dev
->ethtool_ops
= &tun_ethtool_ops
;
766 dev
->destructor
= free_netdev
;
767 dev
->features
|= NETIF_F_NETNS_LOCAL
;
770 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
772 struct tun_struct
*tun
;
773 struct net_device
*dev
;
776 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
778 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
779 tun
= netdev_priv(dev
);
780 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
781 tun
= netdev_priv(dev
);
785 err
= tun_attach(tun
, file
);
791 unsigned long flags
= 0;
795 if (!capable(CAP_NET_ADMIN
))
799 if (ifr
->ifr_flags
& IFF_TUN
) {
801 flags
|= TUN_TUN_DEV
;
803 } else if (ifr
->ifr_flags
& IFF_TAP
) {
805 flags
|= TUN_TAP_DEV
;
811 name
= ifr
->ifr_name
;
813 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
818 dev_net_set(dev
, net
);
820 tun
= netdev_priv(dev
);
823 tun
->txflt
.count
= 0;
827 if (strchr(dev
->name
, '%')) {
828 err
= dev_alloc_name(dev
, dev
->name
);
833 err
= register_netdevice(tun
->dev
);
837 err
= tun_attach(tun
, file
);
842 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
844 if (ifr
->ifr_flags
& IFF_NO_PI
)
845 tun
->flags
|= TUN_NO_PI
;
847 tun
->flags
&= ~TUN_NO_PI
;
849 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
850 tun
->flags
|= TUN_ONE_QUEUE
;
852 tun
->flags
&= ~TUN_ONE_QUEUE
;
854 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
855 tun
->flags
|= TUN_VNET_HDR
;
857 tun
->flags
&= ~TUN_VNET_HDR
;
859 /* Make sure persistent devices do not get stuck in
862 if (netif_running(tun
->dev
))
863 netif_wake_queue(tun
->dev
);
865 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
874 static int tun_get_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
876 struct tun_struct
*tun
= tun_get(file
);
881 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
883 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
887 if (ifr
->ifr_flags
& TUN_TUN_DEV
)
888 ifr
->ifr_flags
|= IFF_TUN
;
890 ifr
->ifr_flags
|= IFF_TAP
;
892 if (tun
->flags
& TUN_NO_PI
)
893 ifr
->ifr_flags
|= IFF_NO_PI
;
895 if (tun
->flags
& TUN_ONE_QUEUE
)
896 ifr
->ifr_flags
|= IFF_ONE_QUEUE
;
898 if (tun
->flags
& TUN_VNET_HDR
)
899 ifr
->ifr_flags
|= IFF_VNET_HDR
;
905 /* This is like a cut-down ethtool ops, except done via tun fd so no
907 static int set_offload(struct net_device
*dev
, unsigned long arg
)
909 unsigned int old_features
, features
;
911 old_features
= dev
->features
;
912 /* Unset features, set them as we chew on the arg. */
913 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
914 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
));
916 if (arg
& TUN_F_CSUM
) {
917 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
920 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
921 if (arg
& TUN_F_TSO_ECN
) {
922 features
|= NETIF_F_TSO_ECN
;
923 arg
&= ~TUN_F_TSO_ECN
;
925 if (arg
& TUN_F_TSO4
)
926 features
|= NETIF_F_TSO
;
927 if (arg
& TUN_F_TSO6
)
928 features
|= NETIF_F_TSO6
;
929 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
933 /* This gives the user a way to test for new features in future by
934 * trying to set them. */
938 dev
->features
= features
;
939 if (old_features
!= dev
->features
)
940 netdev_features_change(dev
);
945 static int tun_chr_ioctl(struct inode
*inode
, struct file
*file
,
946 unsigned int cmd
, unsigned long arg
)
948 struct tun_file
*tfile
= file
->private_data
;
949 struct tun_struct
*tun
;
950 void __user
* argp
= (void __user
*)arg
;
954 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
955 if (copy_from_user(&ifr
, argp
, sizeof ifr
))
958 if (cmd
== TUNGETFEATURES
) {
959 /* Currently this just means: "what IFF flags are valid?".
960 * This is needed because we never checked for invalid flags on
962 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
964 (unsigned int __user
*)argp
);
967 tun
= __tun_get(tfile
);
968 if (cmd
== TUNSETIFF
&& !tun
) {
971 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
974 err
= tun_set_iff(tfile
->net
, file
, &ifr
);
980 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
989 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
994 ret
= tun_get_iff(current
->nsproxy
->net_ns
, file
, &ifr
);
998 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
1003 /* Disable/Enable checksum */
1005 tun
->flags
|= TUN_NOCHECKSUM
;
1007 tun
->flags
&= ~TUN_NOCHECKSUM
;
1009 DBG(KERN_INFO
"%s: checksum %s\n",
1010 tun
->dev
->name
, arg
? "disabled" : "enabled");
1014 /* Disable/Enable persist mode */
1016 tun
->flags
|= TUN_PERSIST
;
1018 tun
->flags
&= ~TUN_PERSIST
;
1020 DBG(KERN_INFO
"%s: persist %s\n",
1021 tun
->dev
->name
, arg
? "enabled" : "disabled");
1025 /* Set owner of the device */
1026 tun
->owner
= (uid_t
) arg
;
1028 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1032 /* Set group of the device */
1033 tun
->group
= (gid_t
) arg
;
1035 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1039 /* Only allow setting the type when the interface is down */
1041 if (tun
->dev
->flags
& IFF_UP
) {
1042 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1046 tun
->dev
->type
= (int) arg
;
1047 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1060 ret
= set_offload(tun
->dev
, arg
);
1064 case TUNSETTXFILTER
:
1065 /* Can be set only for TAPs */
1067 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1070 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1076 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1077 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1078 if (copy_to_user(argp
, &ifr
, sizeof ifr
))
1083 /* Set hw address */
1084 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1085 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1088 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1100 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1102 struct tun_struct
*tun
= tun_get(file
);
1108 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1111 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1115 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1118 tun
->flags
|= TUN_FASYNC
;
1120 tun
->flags
&= ~TUN_FASYNC
;
1128 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1130 struct tun_file
*tfile
;
1131 cycle_kernel_lock();
1132 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1134 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1138 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1139 file
->private_data
= tfile
;
1143 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1145 struct tun_file
*tfile
= file
->private_data
;
1146 struct tun_struct
*tun
= __tun_get(tfile
);
1150 DBG(KERN_INFO
"%s: tun_chr_close\n", tun
->dev
->name
);
1155 /* If desireable, unregister the netdevice. */
1156 if (!(tun
->flags
& TUN_PERSIST
))
1157 unregister_netdevice(tun
->dev
);
1162 put_net(tfile
->net
);
1168 static const struct file_operations tun_fops
= {
1169 .owner
= THIS_MODULE
,
1170 .llseek
= no_llseek
,
1171 .read
= do_sync_read
,
1172 .aio_read
= tun_chr_aio_read
,
1173 .write
= do_sync_write
,
1174 .aio_write
= tun_chr_aio_write
,
1175 .poll
= tun_chr_poll
,
1176 .ioctl
= tun_chr_ioctl
,
1177 .open
= tun_chr_open
,
1178 .release
= tun_chr_close
,
1179 .fasync
= tun_chr_fasync
1182 static struct miscdevice tun_miscdev
= {
1188 /* ethtool interface */
1190 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1193 cmd
->advertising
= 0;
1194 cmd
->speed
= SPEED_10
;
1195 cmd
->duplex
= DUPLEX_FULL
;
1196 cmd
->port
= PORT_TP
;
1197 cmd
->phy_address
= 0;
1198 cmd
->transceiver
= XCVR_INTERNAL
;
1199 cmd
->autoneg
= AUTONEG_DISABLE
;
1205 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1207 struct tun_struct
*tun
= netdev_priv(dev
);
1209 strcpy(info
->driver
, DRV_NAME
);
1210 strcpy(info
->version
, DRV_VERSION
);
1211 strcpy(info
->fw_version
, "N/A");
1213 switch (tun
->flags
& TUN_TYPE_MASK
) {
1215 strcpy(info
->bus_info
, "tun");
1218 strcpy(info
->bus_info
, "tap");
1223 static u32
tun_get_msglevel(struct net_device
*dev
)
1226 struct tun_struct
*tun
= netdev_priv(dev
);
1233 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1236 struct tun_struct
*tun
= netdev_priv(dev
);
1241 static u32
tun_get_link(struct net_device
*dev
)
1243 struct tun_struct
*tun
= netdev_priv(dev
);
1244 return !!tun
->tfile
;
1247 static u32
tun_get_rx_csum(struct net_device
*dev
)
1249 struct tun_struct
*tun
= netdev_priv(dev
);
1250 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1253 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1255 struct tun_struct
*tun
= netdev_priv(dev
);
1257 tun
->flags
&= ~TUN_NOCHECKSUM
;
1259 tun
->flags
|= TUN_NOCHECKSUM
;
1263 static const struct ethtool_ops tun_ethtool_ops
= {
1264 .get_settings
= tun_get_settings
,
1265 .get_drvinfo
= tun_get_drvinfo
,
1266 .get_msglevel
= tun_get_msglevel
,
1267 .set_msglevel
= tun_set_msglevel
,
1268 .get_link
= tun_get_link
,
1269 .get_rx_csum
= tun_get_rx_csum
,
1270 .set_rx_csum
= tun_set_rx_csum
1273 static int tun_init_net(struct net
*net
)
1278 static void tun_exit_net(struct net
*net
)
1280 struct net_device
*dev
, *next
;
1283 for_each_netdev_safe(net
, dev
, next
) {
1284 if (dev
->ethtool_ops
!= &tun_ethtool_ops
)
1286 DBG(KERN_INFO
"%s cleaned up\n", dev
->name
);
1287 unregister_netdevice(dev
);
1292 static struct pernet_operations tun_net_ops
= {
1293 .init
= tun_init_net
,
1294 .exit
= tun_exit_net
,
1297 static int __init
tun_init(void)
1301 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1302 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1304 ret
= register_pernet_device(&tun_net_ops
);
1306 printk(KERN_ERR
"tun: Can't register pernet ops\n");
1310 ret
= misc_register(&tun_miscdev
);
1312 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1318 unregister_pernet_device(&tun_net_ops
);
1323 static void tun_cleanup(void)
1325 misc_deregister(&tun_miscdev
);
1326 unregister_pernet_device(&tun_net_ops
);
1329 module_init(tun_init
);
1330 module_exit(tun_cleanup
);
1331 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1332 MODULE_AUTHOR(DRV_COPYRIGHT
);
1333 MODULE_LICENSE("GPL");
1334 MODULE_ALIAS_MISCDEV(TUN_MINOR
);