2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
21 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
24 * Mark Smith <markzzzsmith@yahoo.com.au>
25 * Use random_ether_addr() for tap MAC address.
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
37 #define DRV_NAME "tun"
38 #define DRV_VERSION "1.6"
39 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
40 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
42 #include <linux/module.h>
43 #include <linux/errno.h>
44 #include <linux/kernel.h>
45 #include <linux/major.h>
46 #include <linux/slab.h>
47 #include <linux/poll.h>
48 #include <linux/fcntl.h>
49 #include <linux/init.h>
50 #include <linux/skbuff.h>
51 #include <linux/netdevice.h>
52 #include <linux/etherdevice.h>
53 #include <linux/miscdevice.h>
54 #include <linux/ethtool.h>
55 #include <linux/rtnetlink.h>
56 #include <linux/compat.h>
58 #include <linux/if_arp.h>
59 #include <linux/if_ether.h>
60 #include <linux/if_tun.h>
61 #include <linux/crc32.h>
62 #include <linux/nsproxy.h>
63 #include <linux/virtio_net.h>
64 #include <linux/rcupdate.h>
65 #include <net/net_namespace.h>
66 #include <net/netns/generic.h>
67 #include <net/rtnetlink.h>
70 #include <asm/system.h>
71 #include <asm/uaccess.h>
73 /* Uncomment to enable debugging */
74 /* #define TUN_DEBUG 1 */
79 #define DBG if(tun->debug)printk
80 #define DBG1 if(debug==2)printk
86 #define FLT_EXACT_COUNT 8
88 unsigned int count
; /* Number of addrs. Zero means disabled */
89 u32 mask
[2]; /* Mask of the hashed addrs */
90 unsigned char addr
[FLT_EXACT_COUNT
][ETH_ALEN
];
95 struct tun_struct
*tun
;
102 struct tun_file
*tfile
;
107 struct net_device
*dev
;
108 struct fasync_struct
*fasync
;
110 struct tap_filter txflt
;
111 struct socket socket
;
120 struct tun_struct
*tun
;
123 static inline struct tun_sock
*tun_sk(struct sock
*sk
)
125 return container_of(sk
, struct tun_sock
, sk
);
128 static int tun_attach(struct tun_struct
*tun
, struct file
*file
)
130 struct tun_file
*tfile
= file
->private_data
;
135 netif_tx_lock_bh(tun
->dev
);
148 tun
->socket
.file
= file
;
150 sock_hold(tun
->socket
.sk
);
151 atomic_inc(&tfile
->count
);
154 netif_tx_unlock_bh(tun
->dev
);
158 static void __tun_detach(struct tun_struct
*tun
)
160 /* Detach from net device */
161 netif_tx_lock_bh(tun
->dev
);
163 tun
->socket
.file
= NULL
;
164 netif_tx_unlock_bh(tun
->dev
);
166 /* Drop read queue */
167 skb_queue_purge(&tun
->socket
.sk
->sk_receive_queue
);
169 /* Drop the extra count on the net device */
173 static void tun_detach(struct tun_struct
*tun
)
180 static struct tun_struct
*__tun_get(struct tun_file
*tfile
)
182 struct tun_struct
*tun
= NULL
;
184 if (atomic_inc_not_zero(&tfile
->count
))
190 static struct tun_struct
*tun_get(struct file
*file
)
192 return __tun_get(file
->private_data
);
195 static void tun_put(struct tun_struct
*tun
)
197 struct tun_file
*tfile
= tun
->tfile
;
199 if (atomic_dec_and_test(&tfile
->count
))
200 tun_detach(tfile
->tun
);
204 static void addr_hash_set(u32
*mask
, const u8
*addr
)
206 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
207 mask
[n
>> 5] |= (1 << (n
& 31));
210 static unsigned int addr_hash_test(const u32
*mask
, const u8
*addr
)
212 int n
= ether_crc(ETH_ALEN
, addr
) >> 26;
213 return mask
[n
>> 5] & (1 << (n
& 31));
216 static int update_filter(struct tap_filter
*filter
, void __user
*arg
)
218 struct { u8 u
[ETH_ALEN
]; } *addr
;
219 struct tun_filter uf
;
220 int err
, alen
, n
, nexact
;
222 if (copy_from_user(&uf
, arg
, sizeof(uf
)))
231 alen
= ETH_ALEN
* uf
.count
;
232 addr
= kmalloc(alen
, GFP_KERNEL
);
236 if (copy_from_user(addr
, arg
+ sizeof(uf
), alen
)) {
241 /* The filter is updated without holding any locks. Which is
242 * perfectly safe. We disable it first and in the worst
243 * case we'll accept a few undesired packets. */
247 /* Use first set of addresses as an exact filter */
248 for (n
= 0; n
< uf
.count
&& n
< FLT_EXACT_COUNT
; n
++)
249 memcpy(filter
->addr
[n
], addr
[n
].u
, ETH_ALEN
);
253 /* Remaining multicast addresses are hashed,
254 * unicast will leave the filter disabled. */
255 memset(filter
->mask
, 0, sizeof(filter
->mask
));
256 for (; n
< uf
.count
; n
++) {
257 if (!is_multicast_ether_addr(addr
[n
].u
)) {
258 err
= 0; /* no filter */
261 addr_hash_set(filter
->mask
, addr
[n
].u
);
264 /* For ALLMULTI just set the mask to all ones.
265 * This overrides the mask populated above. */
266 if ((uf
.flags
& TUN_FLT_ALLMULTI
))
267 memset(filter
->mask
, ~0, sizeof(filter
->mask
));
269 /* Now enable the filter */
271 filter
->count
= nexact
;
273 /* Return the number of exact filters */
281 /* Returns: 0 - drop, !=0 - accept */
282 static int run_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
284 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
286 struct ethhdr
*eh
= (struct ethhdr
*) skb
->data
;
290 for (i
= 0; i
< filter
->count
; i
++)
291 if (!compare_ether_addr(eh
->h_dest
, filter
->addr
[i
]))
294 /* Inexact match (multicast only) */
295 if (is_multicast_ether_addr(eh
->h_dest
))
296 return addr_hash_test(filter
->mask
, eh
->h_dest
);
302 * Checks whether the packet is accepted or not.
303 * Returns: 0 - drop, !=0 - accept
305 static int check_filter(struct tap_filter
*filter
, const struct sk_buff
*skb
)
310 return run_filter(filter
, skb
);
313 /* Network device part of the driver */
315 static const struct ethtool_ops tun_ethtool_ops
;
317 /* Net device detach from fd. */
318 static void tun_net_uninit(struct net_device
*dev
)
320 struct tun_struct
*tun
= netdev_priv(dev
);
321 struct tun_file
*tfile
= tun
->tfile
;
323 /* Inform the methods they need to stop using the dev.
326 wake_up_all(&tun
->socket
.wait
);
327 if (atomic_dec_and_test(&tfile
->count
))
332 static void tun_free_netdev(struct net_device
*dev
)
334 struct tun_struct
*tun
= netdev_priv(dev
);
336 sock_put(tun
->socket
.sk
);
339 /* Net device open. */
340 static int tun_net_open(struct net_device
*dev
)
342 netif_start_queue(dev
);
346 /* Net device close. */
347 static int tun_net_close(struct net_device
*dev
)
349 netif_stop_queue(dev
);
353 /* Net device start xmit */
354 static netdev_tx_t
tun_net_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
356 struct tun_struct
*tun
= netdev_priv(dev
);
358 DBG(KERN_INFO
"%s: tun_net_xmit %d\n", tun
->dev
->name
, skb
->len
);
360 /* Drop packet if interface is not attached */
364 /* Drop if the filter does not like it.
365 * This is a noop if the filter is disabled.
366 * Filter can be enabled only for the TAP devices. */
367 if (!check_filter(&tun
->txflt
, skb
))
370 if (tun
->socket
.sk
->sk_filter
&&
371 sk_filter(tun
->socket
.sk
, skb
))
374 if (skb_queue_len(&tun
->socket
.sk
->sk_receive_queue
) >= dev
->tx_queue_len
) {
375 if (!(tun
->flags
& TUN_ONE_QUEUE
)) {
376 /* Normal queueing mode. */
377 /* Packet scheduler handles dropping of further packets. */
378 netif_stop_queue(dev
);
380 /* We won't see all dropped packets individually, so overrun
381 * error is more appropriate. */
382 dev
->stats
.tx_fifo_errors
++;
384 /* Single queue mode.
385 * Driver handles dropping of all packets itself. */
390 /* Orphan the skb - required as we might hang on to it
391 * for indefinite time. */
395 skb_queue_tail(&tun
->socket
.sk
->sk_receive_queue
, skb
);
396 dev
->trans_start
= jiffies
;
398 /* Notify and wake up reader process */
399 if (tun
->flags
& TUN_FASYNC
)
400 kill_fasync(&tun
->fasync
, SIGIO
, POLL_IN
);
401 wake_up_interruptible_poll(&tun
->socket
.wait
, POLLIN
|
402 POLLRDNORM
| POLLRDBAND
);
406 dev
->stats
.tx_dropped
++;
411 static void tun_net_mclist(struct net_device
*dev
)
414 * This callback is supposed to deal with mc filter in
415 * _rx_ path and has nothing to do with the _tx_ path.
416 * In rx path we always accept everything userspace gives us.
422 #define MAX_MTU 65535
425 tun_net_change_mtu(struct net_device
*dev
, int new_mtu
)
427 if (new_mtu
< MIN_MTU
|| new_mtu
+ dev
->hard_header_len
> MAX_MTU
)
433 static const struct net_device_ops tun_netdev_ops
= {
434 .ndo_uninit
= tun_net_uninit
,
435 .ndo_open
= tun_net_open
,
436 .ndo_stop
= tun_net_close
,
437 .ndo_start_xmit
= tun_net_xmit
,
438 .ndo_change_mtu
= tun_net_change_mtu
,
441 static const struct net_device_ops tap_netdev_ops
= {
442 .ndo_uninit
= tun_net_uninit
,
443 .ndo_open
= tun_net_open
,
444 .ndo_stop
= tun_net_close
,
445 .ndo_start_xmit
= tun_net_xmit
,
446 .ndo_change_mtu
= tun_net_change_mtu
,
447 .ndo_set_multicast_list
= tun_net_mclist
,
448 .ndo_set_mac_address
= eth_mac_addr
,
449 .ndo_validate_addr
= eth_validate_addr
,
452 /* Initialize net device. */
453 static void tun_net_init(struct net_device
*dev
)
455 struct tun_struct
*tun
= netdev_priv(dev
);
457 switch (tun
->flags
& TUN_TYPE_MASK
) {
459 dev
->netdev_ops
= &tun_netdev_ops
;
461 /* Point-to-Point TUN Device */
462 dev
->hard_header_len
= 0;
466 /* Zero header length */
467 dev
->type
= ARPHRD_NONE
;
468 dev
->flags
= IFF_POINTOPOINT
| IFF_NOARP
| IFF_MULTICAST
;
469 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
473 dev
->netdev_ops
= &tap_netdev_ops
;
474 /* Ethernet TAP Device */
477 random_ether_addr(dev
->dev_addr
);
479 dev
->tx_queue_len
= TUN_READQ_SIZE
; /* We prefer our own queue length */
484 /* Character device part */
487 static unsigned int tun_chr_poll(struct file
*file
, poll_table
* wait
)
489 struct tun_file
*tfile
= file
->private_data
;
490 struct tun_struct
*tun
= __tun_get(tfile
);
492 unsigned int mask
= 0;
499 DBG(KERN_INFO
"%s: tun_chr_poll\n", tun
->dev
->name
);
501 poll_wait(file
, &tun
->socket
.wait
, wait
);
503 if (!skb_queue_empty(&sk
->sk_receive_queue
))
504 mask
|= POLLIN
| POLLRDNORM
;
506 if (sock_writeable(sk
) ||
507 (!test_and_set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
) &&
509 mask
|= POLLOUT
| POLLWRNORM
;
511 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
)
518 /* prepad is the amount to reserve at front. len is length after that.
519 * linear is a hint as to how much to copy (usually headers). */
520 static inline struct sk_buff
*tun_alloc_skb(struct tun_struct
*tun
,
521 size_t prepad
, size_t len
,
522 size_t linear
, int noblock
)
524 struct sock
*sk
= tun
->socket
.sk
;
528 /* Under a page? Don't bother with paged skb. */
529 if (prepad
+ len
< PAGE_SIZE
|| !linear
)
532 skb
= sock_alloc_send_pskb(sk
, prepad
+ linear
, len
- linear
, noblock
,
537 skb_reserve(skb
, prepad
);
538 skb_put(skb
, linear
);
539 skb
->data_len
= len
- linear
;
540 skb
->len
+= len
- linear
;
545 /* Get packet from user space buffer */
546 static __inline__ ssize_t
tun_get_user(struct tun_struct
*tun
,
547 const struct iovec
*iv
, size_t count
,
550 struct tun_pi pi
= { 0, cpu_to_be16(ETH_P_IP
) };
552 size_t len
= count
, align
= 0;
553 struct virtio_net_hdr gso
= { 0 };
556 if (!(tun
->flags
& TUN_NO_PI
)) {
557 if ((len
-= sizeof(pi
)) > count
)
560 if (memcpy_fromiovecend((void *)&pi
, iv
, 0, sizeof(pi
)))
562 offset
+= sizeof(pi
);
565 if (tun
->flags
& TUN_VNET_HDR
) {
566 if ((len
-= sizeof(gso
)) > count
)
569 if (memcpy_fromiovecend((void *)&gso
, iv
, offset
, sizeof(gso
)))
572 if ((gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) &&
573 gso
.csum_start
+ gso
.csum_offset
+ 2 > gso
.hdr_len
)
574 gso
.hdr_len
= gso
.csum_start
+ gso
.csum_offset
+ 2;
576 if (gso
.hdr_len
> len
)
578 offset
+= sizeof(gso
);
581 if ((tun
->flags
& TUN_TYPE_MASK
) == TUN_TAP_DEV
) {
582 align
= NET_IP_ALIGN
;
583 if (unlikely(len
< ETH_HLEN
||
584 (gso
.hdr_len
&& gso
.hdr_len
< ETH_HLEN
)))
588 skb
= tun_alloc_skb(tun
, align
, len
, gso
.hdr_len
, noblock
);
590 if (PTR_ERR(skb
) != -EAGAIN
)
591 tun
->dev
->stats
.rx_dropped
++;
595 if (skb_copy_datagram_from_iovec(skb
, 0, iv
, offset
, len
)) {
596 tun
->dev
->stats
.rx_dropped
++;
601 if (gso
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
602 if (!skb_partial_csum_set(skb
, gso
.csum_start
,
604 tun
->dev
->stats
.rx_frame_errors
++;
608 } else if (tun
->flags
& TUN_NOCHECKSUM
)
609 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
611 switch (tun
->flags
& TUN_TYPE_MASK
) {
613 if (tun
->flags
& TUN_NO_PI
) {
614 switch (skb
->data
[0] & 0xf0) {
616 pi
.proto
= htons(ETH_P_IP
);
619 pi
.proto
= htons(ETH_P_IPV6
);
622 tun
->dev
->stats
.rx_dropped
++;
628 skb_reset_mac_header(skb
);
629 skb
->protocol
= pi
.proto
;
633 skb
->protocol
= eth_type_trans(skb
, tun
->dev
);
637 if (gso
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
639 switch (gso
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
640 case VIRTIO_NET_HDR_GSO_TCPV4
:
641 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
643 case VIRTIO_NET_HDR_GSO_TCPV6
:
644 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
646 case VIRTIO_NET_HDR_GSO_UDP
:
647 skb_shinfo(skb
)->gso_type
= SKB_GSO_UDP
;
650 tun
->dev
->stats
.rx_frame_errors
++;
655 if (gso
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
656 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
658 skb_shinfo(skb
)->gso_size
= gso
.gso_size
;
659 if (skb_shinfo(skb
)->gso_size
== 0) {
660 tun
->dev
->stats
.rx_frame_errors
++;
665 /* Header must be checked, and gso_segs computed. */
666 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
667 skb_shinfo(skb
)->gso_segs
= 0;
672 tun
->dev
->stats
.rx_packets
++;
673 tun
->dev
->stats
.rx_bytes
+= len
;
678 static ssize_t
tun_chr_aio_write(struct kiocb
*iocb
, const struct iovec
*iv
,
679 unsigned long count
, loff_t pos
)
681 struct file
*file
= iocb
->ki_filp
;
682 struct tun_struct
*tun
= tun_get(file
);
688 DBG(KERN_INFO
"%s: tun_chr_write %ld\n", tun
->dev
->name
, count
);
690 result
= tun_get_user(tun
, iv
, iov_length(iv
, count
),
691 file
->f_flags
& O_NONBLOCK
);
697 /* Put packet to the user space buffer */
698 static __inline__ ssize_t
tun_put_user(struct tun_struct
*tun
,
700 const struct iovec
*iv
, int len
)
702 struct tun_pi pi
= { 0, skb
->protocol
};
705 if (!(tun
->flags
& TUN_NO_PI
)) {
706 if ((len
-= sizeof(pi
)) < 0)
709 if (len
< skb
->len
) {
710 /* Packet will be striped */
711 pi
.flags
|= TUN_PKT_STRIP
;
714 if (memcpy_toiovecend(iv
, (void *) &pi
, 0, sizeof(pi
)))
719 if (tun
->flags
& TUN_VNET_HDR
) {
720 struct virtio_net_hdr gso
= { 0 }; /* no info leak */
721 if ((len
-= sizeof(gso
)) < 0)
724 if (skb_is_gso(skb
)) {
725 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
727 /* This is a hint as to how much should be linear. */
728 gso
.hdr_len
= skb_headlen(skb
);
729 gso
.gso_size
= sinfo
->gso_size
;
730 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
731 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
732 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
733 gso
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
734 else if (sinfo
->gso_type
& SKB_GSO_UDP
)
735 gso
.gso_type
= VIRTIO_NET_HDR_GSO_UDP
;
738 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
739 gso
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
741 gso
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
743 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
744 gso
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
745 gso
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
746 gso
.csum_offset
= skb
->csum_offset
;
747 } /* else everything is zero */
749 if (unlikely(memcpy_toiovecend(iv
, (void *)&gso
, total
,
752 total
+= sizeof(gso
);
755 len
= min_t(int, skb
->len
, len
);
757 skb_copy_datagram_const_iovec(skb
, 0, iv
, total
, len
);
760 tun
->dev
->stats
.tx_packets
++;
761 tun
->dev
->stats
.tx_bytes
+= len
;
766 static ssize_t
tun_do_read(struct tun_struct
*tun
,
767 struct kiocb
*iocb
, const struct iovec
*iv
,
768 ssize_t len
, int noblock
)
770 DECLARE_WAITQUEUE(wait
, current
);
774 DBG(KERN_INFO
"%s: tun_chr_read\n", tun
->dev
->name
);
776 add_wait_queue(&tun
->socket
.wait
, &wait
);
778 current
->state
= TASK_INTERRUPTIBLE
;
780 /* Read frames from the queue */
781 if (!(skb
=skb_dequeue(&tun
->socket
.sk
->sk_receive_queue
))) {
786 if (signal_pending(current
)) {
790 if (tun
->dev
->reg_state
!= NETREG_REGISTERED
) {
795 /* Nothing to read, let's sleep */
799 netif_wake_queue(tun
->dev
);
801 ret
= tun_put_user(tun
, skb
, iv
, len
);
806 current
->state
= TASK_RUNNING
;
807 remove_wait_queue(&tun
->socket
.wait
, &wait
);
812 static ssize_t
tun_chr_aio_read(struct kiocb
*iocb
, const struct iovec
*iv
,
813 unsigned long count
, loff_t pos
)
815 struct file
*file
= iocb
->ki_filp
;
816 struct tun_file
*tfile
= file
->private_data
;
817 struct tun_struct
*tun
= __tun_get(tfile
);
822 len
= iov_length(iv
, count
);
828 ret
= tun_do_read(tun
, iocb
, iv
, len
, file
->f_flags
& O_NONBLOCK
);
829 ret
= min_t(ssize_t
, ret
, len
);
835 static void tun_setup(struct net_device
*dev
)
837 struct tun_struct
*tun
= netdev_priv(dev
);
842 dev
->ethtool_ops
= &tun_ethtool_ops
;
843 dev
->destructor
= tun_free_netdev
;
846 /* Trivial set of netlink ops to allow deleting tun or tap
847 * device with netlink.
849 static int tun_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
854 static struct rtnl_link_ops tun_link_ops __read_mostly
= {
856 .priv_size
= sizeof(struct tun_struct
),
858 .validate
= tun_validate
,
861 static void tun_sock_write_space(struct sock
*sk
)
863 struct tun_struct
*tun
;
865 if (!sock_writeable(sk
))
868 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
))
871 if (sk
->sk_sleep
&& waitqueue_active(sk
->sk_sleep
))
872 wake_up_interruptible_sync_poll(sk
->sk_sleep
, POLLOUT
|
873 POLLWRNORM
| POLLWRBAND
);
875 tun
= tun_sk(sk
)->tun
;
876 kill_fasync(&tun
->fasync
, SIGIO
, POLL_OUT
);
879 static void tun_sock_destruct(struct sock
*sk
)
881 free_netdev(tun_sk(sk
)->tun
->dev
);
884 static int tun_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
885 struct msghdr
*m
, size_t total_len
)
887 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
888 return tun_get_user(tun
, m
->msg_iov
, total_len
,
889 m
->msg_flags
& MSG_DONTWAIT
);
892 static int tun_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
893 struct msghdr
*m
, size_t total_len
,
896 struct tun_struct
*tun
= container_of(sock
, struct tun_struct
, socket
);
898 if (flags
& ~(MSG_DONTWAIT
|MSG_TRUNC
))
900 ret
= tun_do_read(tun
, iocb
, m
->msg_iov
, total_len
,
901 flags
& MSG_DONTWAIT
);
902 if (ret
> total_len
) {
903 m
->msg_flags
|= MSG_TRUNC
;
904 ret
= flags
& MSG_TRUNC
? ret
: total_len
;
909 /* Ops structure to mimic raw sockets with tun */
910 static const struct proto_ops tun_socket_ops
= {
911 .sendmsg
= tun_sendmsg
,
912 .recvmsg
= tun_recvmsg
,
915 static struct proto tun_proto
= {
917 .owner
= THIS_MODULE
,
918 .obj_size
= sizeof(struct tun_sock
),
921 static int tun_flags(struct tun_struct
*tun
)
925 if (tun
->flags
& TUN_TUN_DEV
)
930 if (tun
->flags
& TUN_NO_PI
)
933 if (tun
->flags
& TUN_ONE_QUEUE
)
934 flags
|= IFF_ONE_QUEUE
;
936 if (tun
->flags
& TUN_VNET_HDR
)
937 flags
|= IFF_VNET_HDR
;
942 static ssize_t
tun_show_flags(struct device
*dev
, struct device_attribute
*attr
,
945 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
946 return sprintf(buf
, "0x%x\n", tun_flags(tun
));
949 static ssize_t
tun_show_owner(struct device
*dev
, struct device_attribute
*attr
,
952 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
953 return sprintf(buf
, "%d\n", tun
->owner
);
956 static ssize_t
tun_show_group(struct device
*dev
, struct device_attribute
*attr
,
959 struct tun_struct
*tun
= netdev_priv(to_net_dev(dev
));
960 return sprintf(buf
, "%d\n", tun
->group
);
963 static DEVICE_ATTR(tun_flags
, 0444, tun_show_flags
, NULL
);
964 static DEVICE_ATTR(owner
, 0444, tun_show_owner
, NULL
);
965 static DEVICE_ATTR(group
, 0444, tun_show_group
, NULL
);
967 static int tun_set_iff(struct net
*net
, struct file
*file
, struct ifreq
*ifr
)
970 struct tun_struct
*tun
;
971 struct net_device
*dev
;
974 dev
= __dev_get_by_name(net
, ifr
->ifr_name
);
976 const struct cred
*cred
= current_cred();
978 if (ifr
->ifr_flags
& IFF_TUN_EXCL
)
980 if ((ifr
->ifr_flags
& IFF_TUN
) && dev
->netdev_ops
== &tun_netdev_ops
)
981 tun
= netdev_priv(dev
);
982 else if ((ifr
->ifr_flags
& IFF_TAP
) && dev
->netdev_ops
== &tap_netdev_ops
)
983 tun
= netdev_priv(dev
);
987 if (((tun
->owner
!= -1 && cred
->euid
!= tun
->owner
) ||
988 (tun
->group
!= -1 && !in_egroup_p(tun
->group
))) &&
989 !capable(CAP_NET_ADMIN
))
991 err
= security_tun_dev_attach(tun
->socket
.sk
);
995 err
= tun_attach(tun
, file
);
1001 unsigned long flags
= 0;
1003 if (!capable(CAP_NET_ADMIN
))
1005 err
= security_tun_dev_create();
1010 if (ifr
->ifr_flags
& IFF_TUN
) {
1012 flags
|= TUN_TUN_DEV
;
1014 } else if (ifr
->ifr_flags
& IFF_TAP
) {
1016 flags
|= TUN_TAP_DEV
;
1022 name
= ifr
->ifr_name
;
1024 dev
= alloc_netdev(sizeof(struct tun_struct
), name
,
1029 dev_net_set(dev
, net
);
1030 dev
->rtnl_link_ops
= &tun_link_ops
;
1032 tun
= netdev_priv(dev
);
1035 tun
->txflt
.count
= 0;
1038 sk
= sk_alloc(net
, AF_UNSPEC
, GFP_KERNEL
, &tun_proto
);
1042 init_waitqueue_head(&tun
->socket
.wait
);
1043 tun
->socket
.ops
= &tun_socket_ops
;
1044 sock_init_data(&tun
->socket
, sk
);
1045 sk
->sk_write_space
= tun_sock_write_space
;
1046 sk
->sk_sndbuf
= INT_MAX
;
1048 tun_sk(sk
)->tun
= tun
;
1050 security_tun_dev_post_create(sk
);
1054 if (strchr(dev
->name
, '%')) {
1055 err
= dev_alloc_name(dev
, dev
->name
);
1060 err
= register_netdevice(tun
->dev
);
1064 if (device_create_file(&tun
->dev
->dev
, &dev_attr_tun_flags
) ||
1065 device_create_file(&tun
->dev
->dev
, &dev_attr_owner
) ||
1066 device_create_file(&tun
->dev
->dev
, &dev_attr_group
))
1067 printk(KERN_ERR
"Failed to create tun sysfs files\n");
1069 sk
->sk_destruct
= tun_sock_destruct
;
1071 err
= tun_attach(tun
, file
);
1076 DBG(KERN_INFO
"%s: tun_set_iff\n", tun
->dev
->name
);
1078 if (ifr
->ifr_flags
& IFF_NO_PI
)
1079 tun
->flags
|= TUN_NO_PI
;
1081 tun
->flags
&= ~TUN_NO_PI
;
1083 if (ifr
->ifr_flags
& IFF_ONE_QUEUE
)
1084 tun
->flags
|= TUN_ONE_QUEUE
;
1086 tun
->flags
&= ~TUN_ONE_QUEUE
;
1088 if (ifr
->ifr_flags
& IFF_VNET_HDR
)
1089 tun
->flags
|= TUN_VNET_HDR
;
1091 tun
->flags
&= ~TUN_VNET_HDR
;
1093 /* Make sure persistent devices do not get stuck in
1096 if (netif_running(tun
->dev
))
1097 netif_wake_queue(tun
->dev
);
1099 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1110 static int tun_get_iff(struct net
*net
, struct tun_struct
*tun
,
1113 DBG(KERN_INFO
"%s: tun_get_iff\n", tun
->dev
->name
);
1115 strcpy(ifr
->ifr_name
, tun
->dev
->name
);
1117 ifr
->ifr_flags
= tun_flags(tun
);
1122 /* This is like a cut-down ethtool ops, except done via tun fd so no
1123 * privs required. */
1124 static int set_offload(struct net_device
*dev
, unsigned long arg
)
1126 unsigned int old_features
, features
;
1128 old_features
= dev
->features
;
1129 /* Unset features, set them as we chew on the arg. */
1130 features
= (old_features
& ~(NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
1131 |NETIF_F_TSO_ECN
|NETIF_F_TSO
|NETIF_F_TSO6
1134 if (arg
& TUN_F_CSUM
) {
1135 features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
1138 if (arg
& (TUN_F_TSO4
|TUN_F_TSO6
)) {
1139 if (arg
& TUN_F_TSO_ECN
) {
1140 features
|= NETIF_F_TSO_ECN
;
1141 arg
&= ~TUN_F_TSO_ECN
;
1143 if (arg
& TUN_F_TSO4
)
1144 features
|= NETIF_F_TSO
;
1145 if (arg
& TUN_F_TSO6
)
1146 features
|= NETIF_F_TSO6
;
1147 arg
&= ~(TUN_F_TSO4
|TUN_F_TSO6
);
1150 if (arg
& TUN_F_UFO
) {
1151 features
|= NETIF_F_UFO
;
1156 /* This gives the user a way to test for new features in future by
1157 * trying to set them. */
1161 dev
->features
= features
;
1162 if (old_features
!= dev
->features
)
1163 netdev_features_change(dev
);
1168 static long __tun_chr_ioctl(struct file
*file
, unsigned int cmd
,
1169 unsigned long arg
, int ifreq_len
)
1171 struct tun_file
*tfile
= file
->private_data
;
1172 struct tun_struct
*tun
;
1173 void __user
* argp
= (void __user
*)arg
;
1174 struct sock_fprog fprog
;
1179 if (cmd
== TUNSETIFF
|| _IOC_TYPE(cmd
) == 0x89)
1180 if (copy_from_user(&ifr
, argp
, ifreq_len
))
1183 if (cmd
== TUNGETFEATURES
) {
1184 /* Currently this just means: "what IFF flags are valid?".
1185 * This is needed because we never checked for invalid flags on
1187 return put_user(IFF_TUN
| IFF_TAP
| IFF_NO_PI
| IFF_ONE_QUEUE
|
1189 (unsigned int __user
*)argp
);
1194 tun
= __tun_get(tfile
);
1195 if (cmd
== TUNSETIFF
&& !tun
) {
1196 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
1198 ret
= tun_set_iff(tfile
->net
, file
, &ifr
);
1203 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1212 DBG(KERN_INFO
"%s: tun_chr_ioctl cmd %d\n", tun
->dev
->name
, cmd
);
1217 ret
= tun_get_iff(current
->nsproxy
->net_ns
, tun
, &ifr
);
1221 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1226 /* Disable/Enable checksum */
1228 tun
->flags
|= TUN_NOCHECKSUM
;
1230 tun
->flags
&= ~TUN_NOCHECKSUM
;
1232 DBG(KERN_INFO
"%s: checksum %s\n",
1233 tun
->dev
->name
, arg
? "disabled" : "enabled");
1237 /* Disable/Enable persist mode */
1239 tun
->flags
|= TUN_PERSIST
;
1241 tun
->flags
&= ~TUN_PERSIST
;
1243 DBG(KERN_INFO
"%s: persist %s\n",
1244 tun
->dev
->name
, arg
? "enabled" : "disabled");
1248 /* Set owner of the device */
1249 tun
->owner
= (uid_t
) arg
;
1251 DBG(KERN_INFO
"%s: owner set to %d\n", tun
->dev
->name
, tun
->owner
);
1255 /* Set group of the device */
1256 tun
->group
= (gid_t
) arg
;
1258 DBG(KERN_INFO
"%s: group set to %d\n", tun
->dev
->name
, tun
->group
);
1262 /* Only allow setting the type when the interface is down */
1263 if (tun
->dev
->flags
& IFF_UP
) {
1264 DBG(KERN_INFO
"%s: Linktype set failed because interface is up\n",
1268 tun
->dev
->type
= (int) arg
;
1269 DBG(KERN_INFO
"%s: linktype set to %d\n", tun
->dev
->name
, tun
->dev
->type
);
1280 ret
= set_offload(tun
->dev
, arg
);
1283 case TUNSETTXFILTER
:
1284 /* Can be set only for TAPs */
1286 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1288 ret
= update_filter(&tun
->txflt
, (void __user
*)arg
);
1293 memcpy(ifr
.ifr_hwaddr
.sa_data
, tun
->dev
->dev_addr
, ETH_ALEN
);
1294 ifr
.ifr_hwaddr
.sa_family
= tun
->dev
->type
;
1295 if (copy_to_user(argp
, &ifr
, ifreq_len
))
1300 /* Set hw address */
1301 DBG(KERN_DEBUG
"%s: set hw address: %pM\n",
1302 tun
->dev
->name
, ifr
.ifr_hwaddr
.sa_data
);
1304 ret
= dev_set_mac_address(tun
->dev
, &ifr
.ifr_hwaddr
);
1308 sndbuf
= tun
->socket
.sk
->sk_sndbuf
;
1309 if (copy_to_user(argp
, &sndbuf
, sizeof(sndbuf
)))
1314 if (copy_from_user(&sndbuf
, argp
, sizeof(sndbuf
))) {
1319 tun
->socket
.sk
->sk_sndbuf
= sndbuf
;
1322 case TUNATTACHFILTER
:
1323 /* Can be set only for TAPs */
1325 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1328 if (copy_from_user(&fprog
, argp
, sizeof(fprog
)))
1331 ret
= sk_attach_filter(&fprog
, tun
->socket
.sk
);
1334 case TUNDETACHFILTER
:
1335 /* Can be set only for TAPs */
1337 if ((tun
->flags
& TUN_TYPE_MASK
) != TUN_TAP_DEV
)
1339 ret
= sk_detach_filter(tun
->socket
.sk
);
1354 static long tun_chr_ioctl(struct file
*file
,
1355 unsigned int cmd
, unsigned long arg
)
1357 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof (struct ifreq
));
1360 #ifdef CONFIG_COMPAT
1361 static long tun_chr_compat_ioctl(struct file
*file
,
1362 unsigned int cmd
, unsigned long arg
)
1367 case TUNSETTXFILTER
:
1372 arg
= (unsigned long)compat_ptr(arg
);
1375 arg
= (compat_ulong_t
)arg
;
1380 * compat_ifreq is shorter than ifreq, so we must not access beyond
1381 * the end of that structure. All fields that are used in this
1382 * driver are compatible though, we don't need to convert the
1385 return __tun_chr_ioctl(file
, cmd
, arg
, sizeof(struct compat_ifreq
));
1387 #endif /* CONFIG_COMPAT */
1389 static int tun_chr_fasync(int fd
, struct file
*file
, int on
)
1391 struct tun_struct
*tun
= tun_get(file
);
1397 DBG(KERN_INFO
"%s: tun_chr_fasync %d\n", tun
->dev
->name
, on
);
1399 if ((ret
= fasync_helper(fd
, file
, on
, &tun
->fasync
)) < 0)
1403 ret
= __f_setown(file
, task_pid(current
), PIDTYPE_PID
, 0);
1406 tun
->flags
|= TUN_FASYNC
;
1408 tun
->flags
&= ~TUN_FASYNC
;
1415 static int tun_chr_open(struct inode
*inode
, struct file
* file
)
1417 struct tun_file
*tfile
;
1419 DBG1(KERN_INFO
"tunX: tun_chr_open\n");
1421 tfile
= kmalloc(sizeof(*tfile
), GFP_KERNEL
);
1424 atomic_set(&tfile
->count
, 0);
1426 tfile
->net
= get_net(current
->nsproxy
->net_ns
);
1427 file
->private_data
= tfile
;
1431 static int tun_chr_close(struct inode
*inode
, struct file
*file
)
1433 struct tun_file
*tfile
= file
->private_data
;
1434 struct tun_struct
*tun
;
1436 tun
= __tun_get(tfile
);
1438 struct net_device
*dev
= tun
->dev
;
1440 DBG(KERN_INFO
"%s: tun_chr_close\n", dev
->name
);
1444 /* If desirable, unregister the netdevice. */
1445 if (!(tun
->flags
& TUN_PERSIST
)) {
1447 if (dev
->reg_state
== NETREG_REGISTERED
)
1448 unregister_netdevice(dev
);
1455 sock_put(tun
->socket
.sk
);
1457 put_net(tfile
->net
);
1463 static const struct file_operations tun_fops
= {
1464 .owner
= THIS_MODULE
,
1465 .llseek
= no_llseek
,
1466 .read
= do_sync_read
,
1467 .aio_read
= tun_chr_aio_read
,
1468 .write
= do_sync_write
,
1469 .aio_write
= tun_chr_aio_write
,
1470 .poll
= tun_chr_poll
,
1471 .unlocked_ioctl
= tun_chr_ioctl
,
1472 #ifdef CONFIG_COMPAT
1473 .compat_ioctl
= tun_chr_compat_ioctl
,
1475 .open
= tun_chr_open
,
1476 .release
= tun_chr_close
,
1477 .fasync
= tun_chr_fasync
1480 static struct miscdevice tun_miscdev
= {
1483 .nodename
= "net/tun",
1487 /* ethtool interface */
1489 static int tun_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
1492 cmd
->advertising
= 0;
1493 cmd
->speed
= SPEED_10
;
1494 cmd
->duplex
= DUPLEX_FULL
;
1495 cmd
->port
= PORT_TP
;
1496 cmd
->phy_address
= 0;
1497 cmd
->transceiver
= XCVR_INTERNAL
;
1498 cmd
->autoneg
= AUTONEG_DISABLE
;
1504 static void tun_get_drvinfo(struct net_device
*dev
, struct ethtool_drvinfo
*info
)
1506 struct tun_struct
*tun
= netdev_priv(dev
);
1508 strcpy(info
->driver
, DRV_NAME
);
1509 strcpy(info
->version
, DRV_VERSION
);
1510 strcpy(info
->fw_version
, "N/A");
1512 switch (tun
->flags
& TUN_TYPE_MASK
) {
1514 strcpy(info
->bus_info
, "tun");
1517 strcpy(info
->bus_info
, "tap");
1522 static u32
tun_get_msglevel(struct net_device
*dev
)
1525 struct tun_struct
*tun
= netdev_priv(dev
);
1532 static void tun_set_msglevel(struct net_device
*dev
, u32 value
)
1535 struct tun_struct
*tun
= netdev_priv(dev
);
1540 static u32
tun_get_link(struct net_device
*dev
)
1542 struct tun_struct
*tun
= netdev_priv(dev
);
1543 return !!tun
->tfile
;
1546 static u32
tun_get_rx_csum(struct net_device
*dev
)
1548 struct tun_struct
*tun
= netdev_priv(dev
);
1549 return (tun
->flags
& TUN_NOCHECKSUM
) == 0;
1552 static int tun_set_rx_csum(struct net_device
*dev
, u32 data
)
1554 struct tun_struct
*tun
= netdev_priv(dev
);
1556 tun
->flags
&= ~TUN_NOCHECKSUM
;
1558 tun
->flags
|= TUN_NOCHECKSUM
;
1562 static const struct ethtool_ops tun_ethtool_ops
= {
1563 .get_settings
= tun_get_settings
,
1564 .get_drvinfo
= tun_get_drvinfo
,
1565 .get_msglevel
= tun_get_msglevel
,
1566 .set_msglevel
= tun_set_msglevel
,
1567 .get_link
= tun_get_link
,
1568 .get_rx_csum
= tun_get_rx_csum
,
1569 .set_rx_csum
= tun_set_rx_csum
1573 static int __init
tun_init(void)
1577 printk(KERN_INFO
"tun: %s, %s\n", DRV_DESCRIPTION
, DRV_VERSION
);
1578 printk(KERN_INFO
"tun: %s\n", DRV_COPYRIGHT
);
1580 ret
= rtnl_link_register(&tun_link_ops
);
1582 printk(KERN_ERR
"tun: Can't register link_ops\n");
1586 ret
= misc_register(&tun_miscdev
);
1588 printk(KERN_ERR
"tun: Can't register misc device %d\n", TUN_MINOR
);
1593 rtnl_link_unregister(&tun_link_ops
);
1598 static void tun_cleanup(void)
1600 misc_deregister(&tun_miscdev
);
1601 rtnl_link_unregister(&tun_link_ops
);
1604 /* Get an underlying socket object from tun file. Returns error unless file is
1605 * attached to a device. The returned object works like a packet socket, it
1606 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
1607 * holding a reference to the file for as long as the socket is in use. */
1608 struct socket
*tun_get_socket(struct file
*file
)
1610 struct tun_struct
*tun
;
1611 if (file
->f_op
!= &tun_fops
)
1612 return ERR_PTR(-EINVAL
);
1613 tun
= tun_get(file
);
1615 return ERR_PTR(-EBADFD
);
1617 return &tun
->socket
;
1619 EXPORT_SYMBOL_GPL(tun_get_socket
);
1621 module_init(tun_init
);
1622 module_exit(tun_cleanup
);
1623 MODULE_DESCRIPTION(DRV_DESCRIPTION
);
1624 MODULE_AUTHOR(DRV_COPYRIGHT
);
1625 MODULE_LICENSE("GPL");
1626 MODULE_ALIAS_MISCDEV(TUN_MINOR
);