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();
127 /* Check permissions */
128 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
129 (tun
->group
!= -1 && cred
->egid
!= tun
->group
)) &&
130 !capable(CAP_NET_ADMIN
))
139 static void __tun_detach(struct tun_struct
*tun
)
141 struct tun_file
*tfile
= tun
->tfile
;
143 /* Detach from net device */
147 /* Drop read queue */
148 skb_queue_purge(&tun
->readq
);
151 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
156 static struct tun_struct
*tun_get(struct file
*file
)
158 return __tun_get(file
->private_data
);
161 static void tun_put(struct tun_struct
*tun
)
167 static void addr_hash_set(u32
*mask
, const u8
*addr
)
169 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
170 mask
[n
>> 5] |= (1 << (n
& 31));
173 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
175 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
176 return mask
[n
>> 5] & (1 << (n
& 31));
179 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
181 struct { u8 u
[ETH_ALEN
]; } *addr
;
182 struct tun_filter uf
;
183 int err
, alen
, n
, nexact
;
185 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
194 alen
= ETH_ALEN
* uf
.count
;
195 addr
= kmalloc(alen
, GFP_KERNEL
);
199 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
204 /* The filter is updated without holding any locks. Which is
205 * perfectly safe. We disable it first and in the worst
206 * case we'll accept a few undesired packets. */
210 /* Use first set of addresses as an exact filter */
211 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
212 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
216 /* The rest is hashed */
217 memset(filter
->mask
, 0, sizeof(filter
->mask
));
218 for (; n
< uf
.count
; n
++)
219 addr_hash_set(filter
->mask
, addr
[n
].u
);
221 /* For ALLMULTI just set the mask to all ones.
222 * This overrides the mask populated above. */
223 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
224 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
226 /* Now enable the filter */
228 filter
->count
= nexact
;
230 /* Return the number of exact filters */
238 /* Returns: 0 - drop, !=0 - accept */
239 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
241 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
243 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
247 for (i
= 0; i
< filter
->count
; i
++)
248 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
251 /* Inexact match (multicast only) */
252 if (is_multicast_ether_addr(eh
->h_dest
))
253 return addr_hash_test(filter
->mask
, eh
->h_dest
);
259 * Checks whether the packet is accepted or not.
260 * Returns: 0 - drop, !=0 - accept
262 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
267 return run_filter(filter
, skb
);
270 /* Network device part of the driver */
272 static const struct ethtool_ops tun_ethtool_ops
;
274 /* Net device open. */
275 static int tun_net_open(struct net_device
*dev
)
277 netif_start_queue(dev
);
281 /* Net device close. */
282 static int tun_net_close(struct net_device
*dev
)
284 netif_stop_queue(dev
);
288 /* Net device start xmit */
289 static int tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
291 struct tun_struct
*tun
= netdev_priv(dev
);
293 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
295 /* Drop packet if interface is not attached */
299 /* Drop if the filter does not like it.
300 * This is a noop if the filter is disabled.
301 * Filter can be enabled only for the TAP devices. */
302 if (!check_filter(&tun
->txflt
, skb
))
305 if (skb_queue_len(&tun
->readq
) >= dev
->tx_queue_len
) {
306 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
307 /* Normal queueing mode. */
308 /* Packet scheduler handles dropping of further packets. */
309 netif_stop_queue(dev
);
311 /* We won't see all dropped packets individually, so overrun
312 * error is more appropriate. */
313 dev
->stats
.tx_fifo_errors
++;
315 /* Single queue mode.
316 * Driver handles dropping of all packets itself. */
322 skb_queue_tail(&tun
->readq
, skb
);
323 dev
->trans_start
= jiffies
;
325 /* Notify and wake up reader process */
326 if (tun
->flags
& TUN_FASYNC
)
327 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
328 wake_up_interruptible(&tun
->read_wait
);
332 dev
->stats
.tx_dropped
++;
337 static void tun_net_mclist(struct net_device
*dev
)
340 * This callback is supposed to deal with mc filter in
341 * _rx_ path and has nothing to do with the _tx_ path.
342 * In rx path we always accept everything userspace gives us.
348 #define MAX_MTU 65535
351 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
353 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
359 static const struct net_device_ops tun_netdev_ops
= {
360 .ndo_open
= tun_net_open
,
361 .ndo_stop
= tun_net_close
,
362 .ndo_start_xmit
= tun_net_xmit
,
363 .ndo_change_mtu
= tun_net_change_mtu
,
366 static const struct net_device_ops tap_netdev_ops
= {
367 .ndo_open
= tun_net_open
,
368 .ndo_stop
= tun_net_close
,
369 .ndo_start_xmit
= tun_net_xmit
,
370 .ndo_change_mtu
= tun_net_change_mtu
,
371 .ndo_set_multicast_list
= tun_net_mclist
,
372 .ndo_set_mac_address
= eth_mac_addr
,
373 .ndo_validate_addr
= eth_validate_addr
,
376 /* Initialize net device. */
377 static void tun_net_init(struct net_device
*dev
)
379 struct tun_struct
*tun
= netdev_priv(dev
);
381 switch (tun
->flags
& TUN_TYPE_MASK
) {
383 dev
->netdev_ops
= &tun_netdev_ops
;
385 /* Point-to-Point TUN Device */
386 dev
->hard_header_len
= 0;
390 /* Zero header length */
391 dev
->type
= ARPHRD_NONE
;
392 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
393 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
397 dev
->netdev_ops
= &tap_netdev_ops
;
398 /* Ethernet TAP Device */
401 random_ether_addr(dev
->dev_addr
);
403 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
408 /* Character device part */
411 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
413 struct tun_struct
*tun
= tun_get(file
);
414 unsigned int mask
= POLLOUT
| POLLWRNORM
;
419 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
421 poll_wait(file
, &tun
->read_wait
, wait
);
423 if (!skb_queue_empty(&tun
->readq
))
424 mask
|= POLLIN
| POLLRDNORM
;
430 /* prepad is the amount to reserve at front. len is length after that.
431 * linear is a hint as to how much to copy (usually headers). */
432 static struct sk_buff
*tun_alloc_skb(size_t prepad
, size_t len
, size_t linear
,
438 skb
= alloc_skb(prepad
+ len
, gfp
|__GFP_NOWARN
);
440 skb_reserve(skb
, prepad
);
445 /* Under a page? Don't bother with paged skb. */
446 if (prepad
+ len
< PAGE_SIZE
)
449 /* Start with a normal skb, and add pages. */
450 skb
= alloc_skb(prepad
+ linear
, gfp
);
454 skb_reserve(skb
, prepad
);
455 skb_put(skb
, linear
);
459 for (i
= 0; i
< MAX_SKB_FRAGS
; i
++) {
460 skb_frag_t
*f
= &skb_shinfo(skb
)->frags
[i
];
462 f
->page
= alloc_page(gfp
|__GFP_ZERO
);
469 skb
->data_len
+= PAGE_SIZE
;
470 skb
->len
+= PAGE_SIZE
;
471 skb
->truesize
+= PAGE_SIZE
;
472 skb_shinfo(skb
)->nr_frags
++;
474 if (len
< PAGE_SIZE
) {
481 /* Too large, or alloc fail? */
490 /* Get packet from user space buffer */
491 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
, struct iovec
*iv
, size_t count
)
493 struct tun_pi pi
= { 0, __constant_htons(ETH_P_IP
) };
495 size_t len
= count
, align
= 0;
496 struct virtio_net_hdr gso
= { 0 };
498 if (!(tun
->flags
& TUN_NO_PI
)) {
499 if ((len
-= sizeof(pi
)) > count
)
502 if(memcpy_fromiovec((void *)&pi
, iv
, sizeof(pi
)))
506 if (tun
->flags
& TUN_VNET_HDR
) {
507 if ((len
-= sizeof(gso
)) > count
)
510 if (memcpy_fromiovec((void *)&gso
, iv
, sizeof(gso
)))
513 if (gso
.hdr_len
> len
)
517 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
518 align
= NET_IP_ALIGN
;
519 if (unlikely(len
< ETH_HLEN
))
523 if (!(skb
= tun_alloc_skb(align
, len
, gso
.hdr_len
, GFP_KERNEL
))) {
524 tun
->dev
->stats
.rx_dropped
++;
528 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, len
)) {
529 tun
->dev
->stats
.rx_dropped
++;
534 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
535 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
537 tun
->dev
->stats
.rx_frame_errors
++;
541 } else if (tun
->flags
& TUN_NOCHECKSUM
)
542 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
544 switch (tun
->flags
& TUN_TYPE_MASK
) {
546 if (tun
->flags
& TUN_NO_PI
) {
547 switch (skb
->data
[0] & 0xf0) {
549 pi
.proto
= htons(ETH_P_IP
);
552 pi
.proto
= htons(ETH_P_IPV6
);
555 tun
->dev
->stats
.rx_dropped
++;
561 skb_reset_mac_header(skb
);
562 skb
->protocol
= pi
.proto
;
566 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
570 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
572 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
573 case VIRTIO_NET_HDR_GSO_TCPV4
:
574 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
576 case VIRTIO_NET_HDR_GSO_TCPV6
:
577 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
580 tun
->dev
->stats
.rx_frame_errors
++;
585 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
586 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
588 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
589 if (skb_shinfo(skb
)->gso_size
== 0) {
590 tun
->dev
->stats
.rx_frame_errors
++;
595 /* Header must be checked, and gso_segs computed. */
596 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
597 skb_shinfo(skb
)->gso_segs
= 0;
602 tun
->dev
->stats
.rx_packets
++;
603 tun
->dev
->stats
.rx_bytes
+= len
;
608 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
609 unsigned long count
, loff_t pos
)
611 struct tun_struct
*tun
= tun_get(iocb
->ki_filp
);
617 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
619 result
= tun_get_user(tun
, (struct iovec
*) iv
, iov_length(iv
, count
));
625 /* Put packet to the user space buffer */
626 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
628 struct iovec
*iv
, int len
)
630 struct tun_pi pi
= { 0, skb
->protocol
};
633 if (!(tun
->flags
& TUN_NO_PI
)) {
634 if ((len
-= sizeof(pi
)) < 0)
637 if (len
< skb
->len
) {
638 /* Packet will be striped */
639 pi
.flags
|= TUN_PKT_STRIP
;
642 if (memcpy_toiovec(iv
, (void *) &pi
, sizeof(pi
)))
647 if (tun
->flags
& TUN_VNET_HDR
) {
648 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
649 if ((len
-= sizeof(gso
)) < 0)
652 if (skb_is_gso(skb
)) {
653 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
655 /* This is a hint as to how much should be linear. */
656 gso
.hdr_len
= skb_headlen(skb
);
657 gso
.gso_size
= sinfo
->gso_size
;
658 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
659 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
660 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
661 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
664 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
665 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
667 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
669 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
670 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
671 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
672 gso
.csum_offset
= skb
->csum_offset
;
673 } /* else everything is zero */
675 if (unlikely(memcpy_toiovec(iv
, (void *)&gso
, sizeof(gso
))))
677 total
+= sizeof(gso
);
680 len
= min_t(int, skb
->len
, len
);
682 skb_copy_datagram_iovec(skb
, 0, iv
, len
);
685 tun
->dev
->stats
.tx_packets
++;
686 tun
->dev
->stats
.tx_bytes
+= len
;
691 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
692 unsigned long count
, loff_t pos
)
694 struct file
*file
= iocb
->ki_filp
;
695 struct tun_struct
*tun
= tun_get(file
);
696 DECLARE_WAITQUEUE(wait
, current
);
698 ssize_t len
, ret
= 0;
703 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
705 len
= iov_length(iv
, count
);
711 add_wait_queue(&tun
->read_wait
, &wait
);
713 current
->state
= TASK_INTERRUPTIBLE
;
715 /* Read frames from the queue */
716 if (!(skb
=skb_dequeue(&tun
->readq
))) {
717 if (file
->f_flags
& O_NONBLOCK
) {
721 if (signal_pending(current
)) {
726 /* Nothing to read, let's sleep */
730 netif_wake_queue(tun
->dev
);
732 ret
= tun_put_user(tun
, skb
, (struct iovec
*) iv
, len
);
737 current
->state
= TASK_RUNNING
;
738 remove_wait_queue(&tun
->read_wait
, &wait
);
745 static void tun_setup(struct net_device
*dev
)
747 struct tun_struct
*tun
= netdev_priv(dev
);
749 skb_queue_head_init(&tun
->readq
);
750 init_waitqueue_head(&tun
->read_wait
);
755 dev
->ethtool_ops
= &tun_ethtool_ops
;
756 dev
->destructor
= free_netdev
;
757 dev
->features
|= NETIF_F_NETNS_LOCAL
;
760 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
762 struct tun_struct
*tun
;
763 struct net_device
*dev
;
766 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
768 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
769 tun
= netdev_priv(dev
);
770 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
771 tun
= netdev_priv(dev
);
775 err
= tun_attach(tun
, file
);
781 unsigned long flags
= 0;
785 if (!capable(CAP_NET_ADMIN
))
789 if (ifr
->ifr_flags
& IFF_TUN
) {
791 flags
|= TUN_TUN_DEV
;
793 } else if (ifr
->ifr_flags
& IFF_TAP
) {
795 flags
|= TUN_TAP_DEV
;
801 name
= ifr
->ifr_name
;
803 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
808 dev_net_set(dev
, net
);
810 tun
= netdev_priv(dev
);
813 tun
->txflt
.count
= 0;
817 if (strchr(dev
->name
, '%')) {
818 err
= dev_alloc_name(dev
, dev
->name
);
823 err
= register_netdevice(tun
->dev
);
827 err
= tun_attach(tun
, file
);
832 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
834 if (ifr
->ifr_flags
& IFF_NO_PI
)
835 tun
->flags
|= TUN_NO_PI
;
837 tun
->flags
&= ~TUN_NO_PI
;
839 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
840 tun
->flags
|= TUN_ONE_QUEUE
;
842 tun
->flags
&= ~TUN_ONE_QUEUE
;
844 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
845 tun
->flags
|= TUN_VNET_HDR
;
847 tun
->flags
&= ~TUN_VNET_HDR
;
849 /* Make sure persistent devices do not get stuck in
852 if (netif_running(tun
->dev
))
853 netif_wake_queue(tun
->dev
);
855 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
864 static int tun_get_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
866 struct tun_struct
*tun
= tun_get(file
);
871 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
873 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
877 if (ifr
->ifr_flags
& TUN_TUN_DEV
)
878 ifr
->ifr_flags
|= IFF_TUN
;
880 ifr
->ifr_flags
|= IFF_TAP
;
882 if (tun
->flags
& TUN_NO_PI
)
883 ifr
->ifr_flags
|= IFF_NO_PI
;
885 if (tun
->flags
& TUN_ONE_QUEUE
)
886 ifr
->ifr_flags
|= IFF_ONE_QUEUE
;
888 if (tun
->flags
& TUN_VNET_HDR
)
889 ifr
->ifr_flags
|= IFF_VNET_HDR
;
895 /* This is like a cut-down ethtool ops, except done via tun fd so no
897 static int set_offload(struct net_device
*dev
, unsigned long arg
)
899 unsigned int old_features
, features
;
901 old_features
= dev
->features
;
902 /* Unset features, set them as we chew on the arg. */
903 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
904 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
));
906 if (arg
& TUN_F_CSUM
) {
907 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
910 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
911 if (arg
& TUN_F_TSO_ECN
) {
912 features
|= NETIF_F_TSO_ECN
;
913 arg
&= ~TUN_F_TSO_ECN
;
915 if (arg
& TUN_F_TSO4
)
916 features
|= NETIF_F_TSO
;
917 if (arg
& TUN_F_TSO6
)
918 features
|= NETIF_F_TSO6
;
919 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
923 /* This gives the user a way to test for new features in future by
924 * trying to set them. */
928 dev
->features
= features
;
929 if (old_features
!= dev
->features
)
930 netdev_features_change(dev
);
935 static int tun_chr_ioctl(struct inode
*inode
, struct file
*file
,
936 unsigned int cmd
, unsigned long arg
)
938 struct tun_file
*tfile
= file
->private_data
;
939 struct tun_struct
*tun
;
940 void __user
* argp
= (void __user
*)arg
;
944 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
945 if (copy_from_user(&ifr
, argp
, sizeof ifr
))
948 if (cmd
== TUNGETFEATURES
) {
949 /* Currently this just means: "what IFF flags are valid?".
950 * This is needed because we never checked for invalid flags on
952 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
954 (unsigned int __user
*)argp
);
957 tun
= __tun_get(tfile
);
958 if (cmd
== TUNSETIFF
&& !tun
) {
961 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
964 err
= tun_set_iff(tfile
->net
, file
, &ifr
);
970 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
979 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
984 ret
= tun_get_iff(current
->nsproxy
->net_ns
, file
, &ifr
);
988 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
993 /* Disable/Enable checksum */
995 tun
->flags
|= TUN_NOCHECKSUM
;
997 tun
->flags
&= ~TUN_NOCHECKSUM
;
999 DBG(KERN_INFO
"%s: checksum %s\n",
1000 tun
->dev
->name
, arg
? "disabled" : "enabled");
1004 /* Disable/Enable persist mode */
1006 tun
->flags
|= TUN_PERSIST
;
1008 tun
->flags
&= ~TUN_PERSIST
;
1010 DBG(KERN_INFO
"%s: persist %s\n",
1011 tun
->dev
->name
, arg
? "enabled" : "disabled");
1015 /* Set owner of the device */
1016 tun
->owner
= (uid_t
) arg
;
1018 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1022 /* Set group of the device */
1023 tun
->group
= (gid_t
) arg
;
1025 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1029 /* Only allow setting the type when the interface is down */
1031 if (tun
->dev
->flags
& IFF_UP
) {
1032 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1036 tun
->dev
->type
= (int) arg
;
1037 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1050 ret
= set_offload(tun
->dev
, arg
);
1054 case TUNSETTXFILTER
:
1055 /* Can be set only for TAPs */
1057 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1060 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1066 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1067 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1068 if (copy_to_user(argp
, &ifr
, sizeof ifr
))
1073 /* Set hw address */
1074 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1075 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1078 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1090 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1092 struct tun_struct
*tun
= tun_get(file
);
1098 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1101 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1105 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1108 tun
->flags
|= TUN_FASYNC
;
1110 tun
->flags
&= ~TUN_FASYNC
;
1118 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1120 struct tun_file
*tfile
;
1121 cycle_kernel_lock();
1122 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1124 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1128 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1129 file
->private_data
= tfile
;
1133 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1135 struct tun_file
*tfile
= file
->private_data
;
1136 struct tun_struct
*tun
= __tun_get(tfile
);
1140 DBG(KERN_INFO
"%s: tun_chr_close\n", tun
->dev
->name
);
1145 /* If desireable, unregister the netdevice. */
1146 if (!(tun
->flags
& TUN_PERSIST
))
1147 unregister_netdevice(tun
->dev
);
1152 put_net(tfile
->net
);
1158 static const struct file_operations tun_fops
= {
1159 .owner
= THIS_MODULE
,
1160 .llseek
= no_llseek
,
1161 .read
= do_sync_read
,
1162 .aio_read
= tun_chr_aio_read
,
1163 .write
= do_sync_write
,
1164 .aio_write
= tun_chr_aio_write
,
1165 .poll
= tun_chr_poll
,
1166 .ioctl
= tun_chr_ioctl
,
1167 .open
= tun_chr_open
,
1168 .release
= tun_chr_close
,
1169 .fasync
= tun_chr_fasync
1172 static struct miscdevice tun_miscdev
= {
1178 /* ethtool interface */
1180 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1183 cmd
->advertising
= 0;
1184 cmd
->speed
= SPEED_10
;
1185 cmd
->duplex
= DUPLEX_FULL
;
1186 cmd
->port
= PORT_TP
;
1187 cmd
->phy_address
= 0;
1188 cmd
->transceiver
= XCVR_INTERNAL
;
1189 cmd
->autoneg
= AUTONEG_DISABLE
;
1195 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1197 struct tun_struct
*tun
= netdev_priv(dev
);
1199 strcpy(info
->driver
, DRV_NAME
);
1200 strcpy(info
->version
, DRV_VERSION
);
1201 strcpy(info
->fw_version
, "N/A");
1203 switch (tun
->flags
& TUN_TYPE_MASK
) {
1205 strcpy(info
->bus_info
, "tun");
1208 strcpy(info
->bus_info
, "tap");
1213 static u32
tun_get_msglevel(struct net_device
*dev
)
1216 struct tun_struct
*tun
= netdev_priv(dev
);
1223 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1226 struct tun_struct
*tun
= netdev_priv(dev
);
1231 static u32
tun_get_link(struct net_device
*dev
)
1233 struct tun_struct
*tun
= netdev_priv(dev
);
1234 return !!tun
->tfile
;
1237 static u32
tun_get_rx_csum(struct net_device
*dev
)
1239 struct tun_struct
*tun
= netdev_priv(dev
);
1240 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1243 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1245 struct tun_struct
*tun
= netdev_priv(dev
);
1247 tun
->flags
&= ~TUN_NOCHECKSUM
;
1249 tun
->flags
|= TUN_NOCHECKSUM
;
1253 static const struct ethtool_ops tun_ethtool_ops
= {
1254 .get_settings
= tun_get_settings
,
1255 .get_drvinfo
= tun_get_drvinfo
,
1256 .get_msglevel
= tun_get_msglevel
,
1257 .set_msglevel
= tun_set_msglevel
,
1258 .get_link
= tun_get_link
,
1259 .get_rx_csum
= tun_get_rx_csum
,
1260 .set_rx_csum
= tun_set_rx_csum
1263 static int tun_init_net(struct net
*net
)
1268 static void tun_exit_net(struct net
*net
)
1270 struct net_device
*dev
, *next
;
1273 for_each_netdev_safe(net
, dev
, next
) {
1274 if (dev
->ethtool_ops
!= &tun_ethtool_ops
)
1276 DBG(KERN_INFO
"%s cleaned up\n", dev
->name
);
1277 unregister_netdevice(dev
);
1282 static struct pernet_operations tun_net_ops
= {
1283 .init
= tun_init_net
,
1284 .exit
= tun_exit_net
,
1287 static int __init
tun_init(void)
1291 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1292 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1294 ret
= register_pernet_device(&tun_net_ops
);
1296 printk(KERN_ERR
"tun: Can't register pernet ops\n");
1300 ret
= misc_register(&tun_miscdev
);
1302 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1308 unregister_pernet_device(&tun_net_ops
);
1313 static void tun_cleanup(void)
1315 misc_deregister(&tun_miscdev
);
1316 unregister_pernet_device(&tun_net_ops
);
1319 module_init(tun_init
);
1320 module_exit(tun_cleanup
);
1321 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1322 MODULE_AUTHOR(DRV_COPYRIGHT
);
1323 MODULE_LICENSE("GPL");
1324 MODULE_ALIAS_MISCDEV(TUN_MINOR
);