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>
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
;
96 wait_queue_head_t read_wait
;
102 struct tun_file
*tfile
;
107 struct sk_buff_head readq
;
109 struct net_device
*dev
;
110 struct fasync_struct
*fasync
;
112 struct tap_filter txflt
;
114 struct socket socket
;
123 struct tun_struct
*tun
;
126 static inline struct tun_sock
*tun_sk(struct sock
*sk
)
128 return container_of(sk
, struct tun_sock
, sk
);
131 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
133 struct tun_file
*tfile
= file
->private_data
;
134 const struct cred
*cred
= current_cred();
139 /* Check permissions */
140 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
141 (tun
->group
!= -1 && !in_egroup_p(tun
->group
))) &&
142 !capable(CAP_NET_ADMIN
))
145 netif_tx_lock_bh(tun
->dev
);
159 atomic_inc(&tfile
->count
);
162 netif_tx_unlock_bh(tun
->dev
);
166 static void __tun_detach(struct tun_struct
*tun
)
168 struct tun_file
*tfile
= tun
->tfile
;
170 /* Detach from net device */
171 netif_tx_lock_bh(tun
->dev
);
174 netif_tx_unlock_bh(tun
->dev
);
176 /* Drop read queue */
177 skb_queue_purge(&tun
->readq
);
179 /* Drop the extra count on the net device */
183 static void tun_detach(struct tun_struct
*tun
)
190 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
192 struct tun_struct
*tun
= NULL
;
194 if (atomic_inc_not_zero(&tfile
->count
))
200 static struct tun_struct
*tun_get(struct file
*file
)
202 return __tun_get(file
->private_data
);
205 static void tun_put(struct tun_struct
*tun
)
207 struct tun_file
*tfile
= tun
->tfile
;
209 if (atomic_dec_and_test(&tfile
->count
))
210 tun_detach(tfile
->tun
);
214 static void addr_hash_set(u32
*mask
, const u8
*addr
)
216 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
217 mask
[n
>> 5] |= (1 << (n
& 31));
220 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
222 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
223 return mask
[n
>> 5] & (1 << (n
& 31));
226 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
228 struct { u8 u
[ETH_ALEN
]; } *addr
;
229 struct tun_filter uf
;
230 int err
, alen
, n
, nexact
;
232 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
241 alen
= ETH_ALEN
* uf
.count
;
242 addr
= kmalloc(alen
, GFP_KERNEL
);
246 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
251 /* The filter is updated without holding any locks. Which is
252 * perfectly safe. We disable it first and in the worst
253 * case we'll accept a few undesired packets. */
257 /* Use first set of addresses as an exact filter */
258 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
259 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
263 /* Remaining multicast addresses are hashed,
264 * unicast will leave the filter disabled. */
265 memset(filter
->mask
, 0, sizeof(filter
->mask
));
266 for (; n
< uf
.count
; n
++) {
267 if (!is_multicast_ether_addr(addr
[n
].u
)) {
268 err
= 0; /* no filter */
271 addr_hash_set(filter
->mask
, addr
[n
].u
);
274 /* For ALLMULTI just set the mask to all ones.
275 * This overrides the mask populated above. */
276 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
277 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
279 /* Now enable the filter */
281 filter
->count
= nexact
;
283 /* Return the number of exact filters */
291 /* Returns: 0 - drop, !=0 - accept */
292 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
294 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
296 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
300 for (i
= 0; i
< filter
->count
; i
++)
301 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
304 /* Inexact match (multicast only) */
305 if (is_multicast_ether_addr(eh
->h_dest
))
306 return addr_hash_test(filter
->mask
, eh
->h_dest
);
312 * Checks whether the packet is accepted or not.
313 * Returns: 0 - drop, !=0 - accept
315 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
320 return run_filter(filter
, skb
);
323 /* Network device part of the driver */
325 static const struct ethtool_ops tun_ethtool_ops
;
327 /* Net device detach from fd. */
328 static void tun_net_uninit(struct net_device
*dev
)
330 struct tun_struct
*tun
= netdev_priv(dev
);
331 struct tun_file
*tfile
= tun
->tfile
;
333 /* Inform the methods they need to stop using the dev.
336 wake_up_all(&tfile
->read_wait
);
337 if (atomic_dec_and_test(&tfile
->count
))
342 /* Net device open. */
343 static int tun_net_open(struct net_device
*dev
)
345 netif_start_queue(dev
);
349 /* Net device close. */
350 static int tun_net_close(struct net_device
*dev
)
352 netif_stop_queue(dev
);
356 /* Net device start xmit */
357 static int tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
359 struct tun_struct
*tun
= netdev_priv(dev
);
361 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
363 /* Drop packet if interface is not attached */
367 /* Drop if the filter does not like it.
368 * This is a noop if the filter is disabled.
369 * Filter can be enabled only for the TAP devices. */
370 if (!check_filter(&tun
->txflt
, skb
))
373 if (skb_queue_len(&tun
->readq
) >= dev
->tx_queue_len
) {
374 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
375 /* Normal queueing mode. */
376 /* Packet scheduler handles dropping of further packets. */
377 netif_stop_queue(dev
);
379 /* We won't see all dropped packets individually, so overrun
380 * error is more appropriate. */
381 dev
->stats
.tx_fifo_errors
++;
383 /* Single queue mode.
384 * Driver handles dropping of all packets itself. */
390 skb_queue_tail(&tun
->readq
, skb
);
391 dev
->trans_start
= jiffies
;
393 /* Notify and wake up reader process */
394 if (tun
->flags
& TUN_FASYNC
)
395 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
396 wake_up_interruptible(&tun
->tfile
->read_wait
);
400 dev
->stats
.tx_dropped
++;
405 static void tun_net_mclist(struct net_device
*dev
)
408 * This callback is supposed to deal with mc filter in
409 * _rx_ path and has nothing to do with the _tx_ path.
410 * In rx path we always accept everything userspace gives us.
416 #define MAX_MTU 65535
419 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
421 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
427 static const struct net_device_ops tun_netdev_ops
= {
428 .ndo_uninit
= tun_net_uninit
,
429 .ndo_open
= tun_net_open
,
430 .ndo_stop
= tun_net_close
,
431 .ndo_start_xmit
= tun_net_xmit
,
432 .ndo_change_mtu
= tun_net_change_mtu
,
435 static const struct net_device_ops tap_netdev_ops
= {
436 .ndo_uninit
= tun_net_uninit
,
437 .ndo_open
= tun_net_open
,
438 .ndo_stop
= tun_net_close
,
439 .ndo_start_xmit
= tun_net_xmit
,
440 .ndo_change_mtu
= tun_net_change_mtu
,
441 .ndo_set_multicast_list
= tun_net_mclist
,
442 .ndo_set_mac_address
= eth_mac_addr
,
443 .ndo_validate_addr
= eth_validate_addr
,
446 /* Initialize net device. */
447 static void tun_net_init(struct net_device
*dev
)
449 struct tun_struct
*tun
= netdev_priv(dev
);
451 switch (tun
->flags
& TUN_TYPE_MASK
) {
453 dev
->netdev_ops
= &tun_netdev_ops
;
455 /* Point-to-Point TUN Device */
456 dev
->hard_header_len
= 0;
460 /* Zero header length */
461 dev
->type
= ARPHRD_NONE
;
462 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
463 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
467 dev
->netdev_ops
= &tap_netdev_ops
;
468 /* Ethernet TAP Device */
471 random_ether_addr(dev
->dev_addr
);
473 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
478 /* Character device part */
481 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
483 struct tun_file
*tfile
= file
->private_data
;
484 struct tun_struct
*tun
= __tun_get(tfile
);
485 struct sock
*sk
= tun
->sk
;
486 unsigned int mask
= 0;
491 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
493 poll_wait(file
, &tfile
->read_wait
, wait
);
495 if (!skb_queue_empty(&tun
->readq
))
496 mask
|= POLLIN
| POLLRDNORM
;
498 if (sock_writeable(sk
) ||
499 (!test_and_set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
) &&
501 mask
|= POLLOUT
| POLLWRNORM
;
503 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
)
510 /* prepad is the amount to reserve at front. len is length after that.
511 * linear is a hint as to how much to copy (usually headers). */
512 static inline struct sk_buff
*tun_alloc_skb(struct tun_struct
*tun
,
513 size_t prepad
, size_t len
,
514 size_t linear
, int noblock
)
516 struct sock
*sk
= tun
->sk
;
520 /* Under a page? Don't bother with paged skb. */
521 if (prepad
+ len
< PAGE_SIZE
|| !linear
)
524 skb
= sock_alloc_send_pskb(sk
, prepad
+ linear
, len
- linear
, noblock
,
529 skb_reserve(skb
, prepad
);
530 skb_put(skb
, linear
);
531 skb
->data_len
= len
- linear
;
532 skb
->len
+= len
- linear
;
537 /* Get packet from user space buffer */
538 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
,
539 struct iovec
*iv
, size_t count
,
542 struct tun_pi pi
= { 0, cpu_to_be16(ETH_P_IP
) };
544 size_t len
= count
, align
= 0;
545 struct virtio_net_hdr gso
= { 0 };
547 if (!(tun
->flags
& TUN_NO_PI
)) {
548 if ((len
-= sizeof(pi
)) > count
)
551 if(memcpy_fromiovec((void *)&pi
, iv
, sizeof(pi
)))
555 if (tun
->flags
& TUN_VNET_HDR
) {
556 if ((len
-= sizeof(gso
)) > count
)
559 if (memcpy_fromiovec((void *)&gso
, iv
, sizeof(gso
)))
562 if (gso
.hdr_len
> len
)
566 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
567 align
= NET_IP_ALIGN
;
568 if (unlikely(len
< ETH_HLEN
||
569 (gso
.hdr_len
&& gso
.hdr_len
< ETH_HLEN
)))
573 skb
= tun_alloc_skb(tun
, align
, len
, gso
.hdr_len
, noblock
);
575 if (PTR_ERR(skb
) != -EAGAIN
)
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 file
*file
= iocb
->ki_filp
;
664 struct tun_struct
*tun
= tun_get(file
);
670 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
672 result
= tun_get_user(tun
, (struct iovec
*)iv
, iov_length(iv
, count
),
673 file
->f_flags
& O_NONBLOCK
);
679 /* Put packet to the user space buffer */
680 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
682 struct iovec
*iv
, int len
)
684 struct tun_pi pi
= { 0, skb
->protocol
};
687 if (!(tun
->flags
& TUN_NO_PI
)) {
688 if ((len
-= sizeof(pi
)) < 0)
691 if (len
< skb
->len
) {
692 /* Packet will be striped */
693 pi
.flags
|= TUN_PKT_STRIP
;
696 if (memcpy_toiovec(iv
, (void *) &pi
, sizeof(pi
)))
701 if (tun
->flags
& TUN_VNET_HDR
) {
702 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
703 if ((len
-= sizeof(gso
)) < 0)
706 if (skb_is_gso(skb
)) {
707 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
709 /* This is a hint as to how much should be linear. */
710 gso
.hdr_len
= skb_headlen(skb
);
711 gso
.gso_size
= sinfo
->gso_size
;
712 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
713 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
714 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
715 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
718 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
719 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
721 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
723 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
724 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
725 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
726 gso
.csum_offset
= skb
->csum_offset
;
727 } /* else everything is zero */
729 if (unlikely(memcpy_toiovec(iv
, (void *)&gso
, sizeof(gso
))))
731 total
+= sizeof(gso
);
734 len
= min_t(int, skb
->len
, len
);
736 skb_copy_datagram_iovec(skb
, 0, iv
, len
);
739 tun
->dev
->stats
.tx_packets
++;
740 tun
->dev
->stats
.tx_bytes
+= len
;
745 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
746 unsigned long count
, loff_t pos
)
748 struct file
*file
= iocb
->ki_filp
;
749 struct tun_file
*tfile
= file
->private_data
;
750 struct tun_struct
*tun
= __tun_get(tfile
);
751 DECLARE_WAITQUEUE(wait
, current
);
753 ssize_t len
, ret
= 0;
758 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
760 len
= iov_length(iv
, count
);
766 add_wait_queue(&tfile
->read_wait
, &wait
);
768 current
->state
= TASK_INTERRUPTIBLE
;
770 /* Read frames from the queue */
771 if (!(skb
=skb_dequeue(&tun
->readq
))) {
772 if (file
->f_flags
& O_NONBLOCK
) {
776 if (signal_pending(current
)) {
780 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
) {
785 /* Nothing to read, let's sleep */
789 netif_wake_queue(tun
->dev
);
791 ret
= tun_put_user(tun
, skb
, (struct iovec
*) iv
, len
);
796 current
->state
= TASK_RUNNING
;
797 remove_wait_queue(&tfile
->read_wait
, &wait
);
804 static void tun_setup(struct net_device
*dev
)
806 struct tun_struct
*tun
= netdev_priv(dev
);
808 skb_queue_head_init(&tun
->readq
);
813 dev
->ethtool_ops
= &tun_ethtool_ops
;
814 dev
->destructor
= free_netdev
;
817 /* Trivial set of netlink ops to allow deleting tun or tap
818 * device with netlink.
820 static int tun_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
825 static struct rtnl_link_ops tun_link_ops __read_mostly
= {
827 .priv_size
= sizeof(struct tun_struct
),
829 .validate
= tun_validate
,
832 static void tun_sock_write_space(struct sock
*sk
)
834 struct tun_struct
*tun
;
836 if (!sock_writeable(sk
))
839 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
840 wake_up_interruptible_sync(sk
->sk_sleep
);
842 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
))
845 tun
= container_of(sk
, struct tun_sock
, sk
)->tun
;
846 kill_fasync(&tun
->fasync
, SIGIO
, POLL_OUT
);
849 static void tun_sock_destruct(struct sock
*sk
)
851 dev_put(container_of(sk
, struct tun_sock
, sk
)->tun
->dev
);
854 static struct proto tun_proto
= {
856 .owner
= THIS_MODULE
,
857 .obj_size
= sizeof(struct tun_sock
),
860 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
863 struct tun_struct
*tun
;
864 struct net_device
*dev
;
865 struct tun_file
*tfile
= file
->private_data
;
868 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
870 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
871 tun
= netdev_priv(dev
);
872 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
873 tun
= netdev_priv(dev
);
877 err
= tun_attach(tun
, file
);
883 unsigned long flags
= 0;
887 if (!capable(CAP_NET_ADMIN
))
891 if (ifr
->ifr_flags
& IFF_TUN
) {
893 flags
|= TUN_TUN_DEV
;
895 } else if (ifr
->ifr_flags
& IFF_TAP
) {
897 flags
|= TUN_TAP_DEV
;
903 name
= ifr
->ifr_name
;
905 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
910 dev_net_set(dev
, net
);
911 dev
->rtnl_link_ops
= &tun_link_ops
;
913 tun
= netdev_priv(dev
);
916 tun
->txflt
.count
= 0;
919 sk
= sk_alloc(net
, AF_UNSPEC
, GFP_KERNEL
, &tun_proto
);
923 /* This ref count is for tun->sk. */
925 sock_init_data(&tun
->socket
, sk
);
926 sk
->sk_write_space
= tun_sock_write_space
;
927 sk
->sk_destruct
= tun_sock_destruct
;
928 sk
->sk_sndbuf
= INT_MAX
;
929 sk
->sk_sleep
= &tfile
->read_wait
;
932 container_of(sk
, struct tun_sock
, sk
)->tun
= tun
;
936 if (strchr(dev
->name
, '%')) {
937 err
= dev_alloc_name(dev
, dev
->name
);
943 err
= register_netdevice(tun
->dev
);
947 err
= tun_attach(tun
, file
);
952 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
954 if (ifr
->ifr_flags
& IFF_NO_PI
)
955 tun
->flags
|= TUN_NO_PI
;
957 tun
->flags
&= ~TUN_NO_PI
;
959 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
960 tun
->flags
|= TUN_ONE_QUEUE
;
962 tun
->flags
&= ~TUN_ONE_QUEUE
;
964 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
965 tun
->flags
|= TUN_VNET_HDR
;
967 tun
->flags
&= ~TUN_VNET_HDR
;
969 /* Make sure persistent devices do not get stuck in
972 if (netif_running(tun
->dev
))
973 netif_wake_queue(tun
->dev
);
975 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
986 static int tun_get_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
988 struct tun_struct
*tun
= tun_get(file
);
993 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
995 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
999 if (ifr
->ifr_flags
& TUN_TUN_DEV
)
1000 ifr
->ifr_flags
|= IFF_TUN
;
1002 ifr
->ifr_flags
|= IFF_TAP
;
1004 if (tun
->flags
& TUN_NO_PI
)
1005 ifr
->ifr_flags
|= IFF_NO_PI
;
1007 if (tun
->flags
& TUN_ONE_QUEUE
)
1008 ifr
->ifr_flags
|= IFF_ONE_QUEUE
;
1010 if (tun
->flags
& TUN_VNET_HDR
)
1011 ifr
->ifr_flags
|= IFF_VNET_HDR
;
1017 /* This is like a cut-down ethtool ops, except done via tun fd so no
1018 * privs required. */
1019 static int set_offload(struct net_device
*dev
, unsigned long arg
)
1021 unsigned int old_features
, features
;
1023 old_features
= dev
->features
;
1024 /* Unset features, set them as we chew on the arg. */
1025 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
1026 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
));
1028 if (arg
& TUN_F_CSUM
) {
1029 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
1032 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
1033 if (arg
& TUN_F_TSO_ECN
) {
1034 features
|= NETIF_F_TSO_ECN
;
1035 arg
&= ~TUN_F_TSO_ECN
;
1037 if (arg
& TUN_F_TSO4
)
1038 features
|= NETIF_F_TSO
;
1039 if (arg
& TUN_F_TSO6
)
1040 features
|= NETIF_F_TSO6
;
1041 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
1045 /* This gives the user a way to test for new features in future by
1046 * trying to set them. */
1050 dev
->features
= features
;
1051 if (old_features
!= dev
->features
)
1052 netdev_features_change(dev
);
1057 static int tun_chr_ioctl(struct inode
*inode
, struct file
*file
,
1058 unsigned int cmd
, unsigned long arg
)
1060 struct tun_file
*tfile
= file
->private_data
;
1061 struct tun_struct
*tun
;
1062 void __user
* argp
= (void __user
*)arg
;
1067 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
1068 if (copy_from_user(&ifr
, argp
, sizeof ifr
))
1071 if (cmd
== TUNGETFEATURES
) {
1072 /* Currently this just means: "what IFF flags are valid?".
1073 * This is needed because we never checked for invalid flags on
1075 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
1077 (unsigned int __user
*)argp
);
1080 tun
= __tun_get(tfile
);
1081 if (cmd
== TUNSETIFF
&& !tun
) {
1084 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
1087 err
= tun_set_iff(tfile
->net
, file
, &ifr
);
1093 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
1102 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
1107 ret
= tun_get_iff(current
->nsproxy
->net_ns
, file
, &ifr
);
1111 if (copy_to_user(argp
, &ifr
, sizeof(ifr
)))
1116 /* Disable/Enable checksum */
1118 tun
->flags
|= TUN_NOCHECKSUM
;
1120 tun
->flags
&= ~TUN_NOCHECKSUM
;
1122 DBG(KERN_INFO
"%s: checksum %s\n",
1123 tun
->dev
->name
, arg
? "disabled" : "enabled");
1127 /* Disable/Enable persist mode */
1129 tun
->flags
|= TUN_PERSIST
;
1131 tun
->flags
&= ~TUN_PERSIST
;
1133 DBG(KERN_INFO
"%s: persist %s\n",
1134 tun
->dev
->name
, arg
? "enabled" : "disabled");
1138 /* Set owner of the device */
1139 tun
->owner
= (uid_t
) arg
;
1141 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1145 /* Set group of the device */
1146 tun
->group
= (gid_t
) arg
;
1148 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1152 /* Only allow setting the type when the interface is down */
1154 if (tun
->dev
->flags
& IFF_UP
) {
1155 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1159 tun
->dev
->type
= (int) arg
;
1160 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1173 ret
= set_offload(tun
->dev
, arg
);
1177 case TUNSETTXFILTER
:
1178 /* Can be set only for TAPs */
1180 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1183 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1189 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1190 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1191 if (copy_to_user(argp
, &ifr
, sizeof ifr
))
1196 /* Set hw address */
1197 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1198 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1201 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1206 sndbuf
= tun
->sk
->sk_sndbuf
;
1207 if (copy_to_user(argp
, &sndbuf
, sizeof(sndbuf
)))
1212 if (copy_from_user(&sndbuf
, argp
, sizeof(sndbuf
))) {
1217 tun
->sk
->sk_sndbuf
= sndbuf
;
1229 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1231 struct tun_struct
*tun
= tun_get(file
);
1237 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1240 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1244 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1247 tun
->flags
|= TUN_FASYNC
;
1249 tun
->flags
&= ~TUN_FASYNC
;
1257 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1259 struct tun_file
*tfile
;
1260 cycle_kernel_lock();
1261 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1263 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1266 atomic_set(&tfile
->count
, 0);
1268 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1269 init_waitqueue_head(&tfile
->read_wait
);
1270 file
->private_data
= tfile
;
1274 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1276 struct tun_file
*tfile
= file
->private_data
;
1277 struct tun_struct
*tun
= __tun_get(tfile
);
1281 DBG(KERN_INFO
"%s: tun_chr_close\n", tun
->dev
->name
);
1286 /* If desireable, unregister the netdevice. */
1287 if (!(tun
->flags
& TUN_PERSIST
)) {
1289 unregister_netdevice(tun
->dev
);
1295 put_net(tfile
->net
);
1301 static const struct file_operations tun_fops
= {
1302 .owner
= THIS_MODULE
,
1303 .llseek
= no_llseek
,
1304 .read
= do_sync_read
,
1305 .aio_read
= tun_chr_aio_read
,
1306 .write
= do_sync_write
,
1307 .aio_write
= tun_chr_aio_write
,
1308 .poll
= tun_chr_poll
,
1309 .ioctl
= tun_chr_ioctl
,
1310 .open
= tun_chr_open
,
1311 .release
= tun_chr_close
,
1312 .fasync
= tun_chr_fasync
1315 static struct miscdevice tun_miscdev
= {
1321 /* ethtool interface */
1323 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1326 cmd
->advertising
= 0;
1327 cmd
->speed
= SPEED_10
;
1328 cmd
->duplex
= DUPLEX_FULL
;
1329 cmd
->port
= PORT_TP
;
1330 cmd
->phy_address
= 0;
1331 cmd
->transceiver
= XCVR_INTERNAL
;
1332 cmd
->autoneg
= AUTONEG_DISABLE
;
1338 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1340 struct tun_struct
*tun
= netdev_priv(dev
);
1342 strcpy(info
->driver
, DRV_NAME
);
1343 strcpy(info
->version
, DRV_VERSION
);
1344 strcpy(info
->fw_version
, "N/A");
1346 switch (tun
->flags
& TUN_TYPE_MASK
) {
1348 strcpy(info
->bus_info
, "tun");
1351 strcpy(info
->bus_info
, "tap");
1356 static u32
tun_get_msglevel(struct net_device
*dev
)
1359 struct tun_struct
*tun
= netdev_priv(dev
);
1366 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1369 struct tun_struct
*tun
= netdev_priv(dev
);
1374 static u32
tun_get_link(struct net_device
*dev
)
1376 struct tun_struct
*tun
= netdev_priv(dev
);
1377 return !!tun
->tfile
;
1380 static u32
tun_get_rx_csum(struct net_device
*dev
)
1382 struct tun_struct
*tun
= netdev_priv(dev
);
1383 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1386 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1388 struct tun_struct
*tun
= netdev_priv(dev
);
1390 tun
->flags
&= ~TUN_NOCHECKSUM
;
1392 tun
->flags
|= TUN_NOCHECKSUM
;
1396 static const struct ethtool_ops tun_ethtool_ops
= {
1397 .get_settings
= tun_get_settings
,
1398 .get_drvinfo
= tun_get_drvinfo
,
1399 .get_msglevel
= tun_get_msglevel
,
1400 .set_msglevel
= tun_set_msglevel
,
1401 .get_link
= tun_get_link
,
1402 .get_rx_csum
= tun_get_rx_csum
,
1403 .set_rx_csum
= tun_set_rx_csum
1407 static int __init
tun_init(void)
1411 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1412 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1414 ret
= rtnl_link_register(&tun_link_ops
);
1416 printk(KERN_ERR
"tun: Can't register link_ops\n");
1420 ret
= misc_register(&tun_miscdev
);
1422 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1427 rtnl_link_unregister(&tun_link_ops
);
1432 static void tun_cleanup(void)
1434 misc_deregister(&tun_miscdev
);
1435 rtnl_link_unregister(&tun_link_ops
);
1438 module_init(tun_init
);
1439 module_exit(tun_cleanup
);
1440 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1441 MODULE_AUTHOR(DRV_COPYRIGHT
);
1442 MODULE_LICENSE("GPL");
1443 MODULE_ALIAS_MISCDEV(TUN_MINOR
);