dm snapshot: implement iterate devices
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / drivers / net / tun.c
blob42b6c6319bc29824fe4e2061ee882759cfeb05c1
1 /*
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 $
19 * Changes:
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.
30 * Added ethtool API.
31 * Minor cleanups
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>
57 #include <linux/if.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>
67 #include <net/sock.h>
69 #include <asm/system.h>
70 #include <asm/uaccess.h>
72 /* Uncomment to enable debugging */
73 /* #define TUN_DEBUG 1 */
75 #ifdef TUN_DEBUG
76 static int debug;
78 #define DBG if(tun->debug)printk
79 #define DBG1 if(debug==2)printk
80 #else
81 #define DBG( a... )
82 #define DBG1( a... )
83 #endif
85 #define FLT_EXACT_COUNT 8
86 struct tap_filter {
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];
92 struct tun_file {
93 atomic_t count;
94 struct tun_struct *tun;
95 struct net *net;
98 struct tun_sock;
100 struct tun_struct {
101 struct tun_file *tfile;
102 unsigned int flags;
103 uid_t owner;
104 gid_t group;
106 struct sk_buff_head readq;
108 struct net_device *dev;
109 struct fasync_struct *fasync;
111 struct tap_filter txflt;
112 struct sock *sk;
113 struct socket socket;
115 #ifdef TUN_DEBUG
116 int debug;
117 #endif
120 struct tun_sock {
121 struct sock sk;
122 struct tun_struct *tun;
125 static inline struct tun_sock *tun_sk(struct sock *sk)
127 return container_of(sk, struct tun_sock, sk);
130 static int tun_attach(struct tun_struct *tun, struct file *file)
132 struct tun_file *tfile = file->private_data;
133 const struct cred *cred = current_cred();
134 int err;
136 ASSERT_RTNL();
138 /* Check permissions */
139 if (((tun->owner != -1 && cred->euid != tun->owner) ||
140 (tun->group != -1 && !in_egroup_p(tun->group))) &&
141 !capable(CAP_NET_ADMIN))
142 return -EPERM;
144 netif_tx_lock_bh(tun->dev);
146 err = -EINVAL;
147 if (tfile->tun)
148 goto out;
150 err = -EBUSY;
151 if (tun->tfile)
152 goto out;
154 err = 0;
155 tfile->tun = tun;
156 tun->tfile = tfile;
157 dev_hold(tun->dev);
158 sock_hold(tun->sk);
159 atomic_inc(&tfile->count);
161 out:
162 netif_tx_unlock_bh(tun->dev);
163 return err;
166 static void __tun_detach(struct tun_struct *tun)
168 /* Detach from net device */
169 netif_tx_lock_bh(tun->dev);
170 tun->tfile = NULL;
171 netif_tx_unlock_bh(tun->dev);
173 /* Drop read queue */
174 skb_queue_purge(&tun->readq);
176 /* Drop the extra count on the net device */
177 dev_put(tun->dev);
180 static void tun_detach(struct tun_struct *tun)
182 rtnl_lock();
183 __tun_detach(tun);
184 rtnl_unlock();
187 static struct tun_struct *__tun_get(struct tun_file *tfile)
189 struct tun_struct *tun = NULL;
191 if (atomic_inc_not_zero(&tfile->count))
192 tun = tfile->tun;
194 return tun;
197 static struct tun_struct *tun_get(struct file *file)
199 return __tun_get(file->private_data);
202 static void tun_put(struct tun_struct *tun)
204 struct tun_file *tfile = tun->tfile;
206 if (atomic_dec_and_test(&tfile->count))
207 tun_detach(tfile->tun);
210 /* TAP filterting */
211 static void addr_hash_set(u32 *mask, const u8 *addr)
213 int n = ether_crc(ETH_ALEN, addr) >> 26;
214 mask[n >> 5] |= (1 << (n & 31));
217 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
219 int n = ether_crc(ETH_ALEN, addr) >> 26;
220 return mask[n >> 5] & (1 << (n & 31));
223 static int update_filter(struct tap_filter *filter, void __user *arg)
225 struct { u8 u[ETH_ALEN]; } *addr;
226 struct tun_filter uf;
227 int err, alen, n, nexact;
229 if (copy_from_user(&uf, arg, sizeof(uf)))
230 return -EFAULT;
232 if (!uf.count) {
233 /* Disabled */
234 filter->count = 0;
235 return 0;
238 alen = ETH_ALEN * uf.count;
239 addr = kmalloc(alen, GFP_KERNEL);
240 if (!addr)
241 return -ENOMEM;
243 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
244 err = -EFAULT;
245 goto done;
248 /* The filter is updated without holding any locks. Which is
249 * perfectly safe. We disable it first and in the worst
250 * case we'll accept a few undesired packets. */
251 filter->count = 0;
252 wmb();
254 /* Use first set of addresses as an exact filter */
255 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
256 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
258 nexact = n;
260 /* Remaining multicast addresses are hashed,
261 * unicast will leave the filter disabled. */
262 memset(filter->mask, 0, sizeof(filter->mask));
263 for (; n < uf.count; n++) {
264 if (!is_multicast_ether_addr(addr[n].u)) {
265 err = 0; /* no filter */
266 goto done;
268 addr_hash_set(filter->mask, addr[n].u);
271 /* For ALLMULTI just set the mask to all ones.
272 * This overrides the mask populated above. */
273 if ((uf.flags & TUN_FLT_ALLMULTI))
274 memset(filter->mask, ~0, sizeof(filter->mask));
276 /* Now enable the filter */
277 wmb();
278 filter->count = nexact;
280 /* Return the number of exact filters */
281 err = nexact;
283 done:
284 kfree(addr);
285 return err;
288 /* Returns: 0 - drop, !=0 - accept */
289 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
291 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
292 * at this point. */
293 struct ethhdr *eh = (struct ethhdr *) skb->data;
294 int i;
296 /* Exact match */
297 for (i = 0; i < filter->count; i++)
298 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
299 return 1;
301 /* Inexact match (multicast only) */
302 if (is_multicast_ether_addr(eh->h_dest))
303 return addr_hash_test(filter->mask, eh->h_dest);
305 return 0;
309 * Checks whether the packet is accepted or not.
310 * Returns: 0 - drop, !=0 - accept
312 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
314 if (!filter->count)
315 return 1;
317 return run_filter(filter, skb);
320 /* Network device part of the driver */
322 static const struct ethtool_ops tun_ethtool_ops;
324 /* Net device detach from fd. */
325 static void tun_net_uninit(struct net_device *dev)
327 struct tun_struct *tun = netdev_priv(dev);
328 struct tun_file *tfile = tun->tfile;
330 /* Inform the methods they need to stop using the dev.
332 if (tfile) {
333 wake_up_all(&tun->socket.wait);
334 if (atomic_dec_and_test(&tfile->count))
335 __tun_detach(tun);
339 static void tun_free_netdev(struct net_device *dev)
341 struct tun_struct *tun = netdev_priv(dev);
343 sock_put(tun->sk);
346 /* Net device open. */
347 static int tun_net_open(struct net_device *dev)
349 netif_start_queue(dev);
350 return 0;
353 /* Net device close. */
354 static int tun_net_close(struct net_device *dev)
356 netif_stop_queue(dev);
357 return 0;
360 /* Net device start xmit */
361 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
363 struct tun_struct *tun = netdev_priv(dev);
365 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
367 /* Drop packet if interface is not attached */
368 if (!tun->tfile)
369 goto drop;
371 /* Drop if the filter does not like it.
372 * This is a noop if the filter is disabled.
373 * Filter can be enabled only for the TAP devices. */
374 if (!check_filter(&tun->txflt, skb))
375 goto drop;
377 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
378 if (!(tun->flags & TUN_ONE_QUEUE)) {
379 /* Normal queueing mode. */
380 /* Packet scheduler handles dropping of further packets. */
381 netif_stop_queue(dev);
383 /* We won't see all dropped packets individually, so overrun
384 * error is more appropriate. */
385 dev->stats.tx_fifo_errors++;
386 } else {
387 /* Single queue mode.
388 * Driver handles dropping of all packets itself. */
389 goto drop;
393 /* Enqueue packet */
394 skb_queue_tail(&tun->readq, skb);
395 dev->trans_start = jiffies;
397 /* Notify and wake up reader process */
398 if (tun->flags & TUN_FASYNC)
399 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
400 wake_up_interruptible(&tun->socket.wait);
401 return 0;
403 drop:
404 dev->stats.tx_dropped++;
405 kfree_skb(skb);
406 return 0;
409 static void tun_net_mclist(struct net_device *dev)
412 * This callback is supposed to deal with mc filter in
413 * _rx_ path and has nothing to do with the _tx_ path.
414 * In rx path we always accept everything userspace gives us.
416 return;
419 #define MIN_MTU 68
420 #define MAX_MTU 65535
422 static int
423 tun_net_change_mtu(struct net_device *dev, int new_mtu)
425 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
426 return -EINVAL;
427 dev->mtu = new_mtu;
428 return 0;
431 static const struct net_device_ops tun_netdev_ops = {
432 .ndo_uninit = tun_net_uninit,
433 .ndo_open = tun_net_open,
434 .ndo_stop = tun_net_close,
435 .ndo_start_xmit = tun_net_xmit,
436 .ndo_change_mtu = tun_net_change_mtu,
439 static const struct net_device_ops tap_netdev_ops = {
440 .ndo_uninit = tun_net_uninit,
441 .ndo_open = tun_net_open,
442 .ndo_stop = tun_net_close,
443 .ndo_start_xmit = tun_net_xmit,
444 .ndo_change_mtu = tun_net_change_mtu,
445 .ndo_set_multicast_list = tun_net_mclist,
446 .ndo_set_mac_address = eth_mac_addr,
447 .ndo_validate_addr = eth_validate_addr,
450 /* Initialize net device. */
451 static void tun_net_init(struct net_device *dev)
453 struct tun_struct *tun = netdev_priv(dev);
455 switch (tun->flags & TUN_TYPE_MASK) {
456 case TUN_TUN_DEV:
457 dev->netdev_ops = &tun_netdev_ops;
459 /* Point-to-Point TUN Device */
460 dev->hard_header_len = 0;
461 dev->addr_len = 0;
462 dev->mtu = 1500;
464 /* Zero header length */
465 dev->type = ARPHRD_NONE;
466 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
467 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
468 break;
470 case TUN_TAP_DEV:
471 dev->netdev_ops = &tap_netdev_ops;
472 /* Ethernet TAP Device */
473 ether_setup(dev);
475 random_ether_addr(dev->dev_addr);
477 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
478 break;
482 /* Character device part */
484 /* Poll */
485 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
487 struct tun_file *tfile = file->private_data;
488 struct tun_struct *tun = __tun_get(tfile);
489 struct sock *sk;
490 unsigned int mask = 0;
492 if (!tun)
493 return POLLERR;
495 sk = tun->sk;
497 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
499 poll_wait(file, &tun->socket.wait, wait);
501 if (!skb_queue_empty(&tun->readq))
502 mask |= POLLIN | POLLRDNORM;
504 if (sock_writeable(sk) ||
505 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
506 sock_writeable(sk)))
507 mask |= POLLOUT | POLLWRNORM;
509 if (tun->dev->reg_state != NETREG_REGISTERED)
510 mask = POLLERR;
512 tun_put(tun);
513 return mask;
516 /* prepad is the amount to reserve at front. len is length after that.
517 * linear is a hint as to how much to copy (usually headers). */
518 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
519 size_t prepad, size_t len,
520 size_t linear, int noblock)
522 struct sock *sk = tun->sk;
523 struct sk_buff *skb;
524 int err;
526 /* Under a page? Don't bother with paged skb. */
527 if (prepad + len < PAGE_SIZE || !linear)
528 linear = len;
530 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
531 &err);
532 if (!skb)
533 return ERR_PTR(err);
535 skb_reserve(skb, prepad);
536 skb_put(skb, linear);
537 skb->data_len = len - linear;
538 skb->len += len - linear;
540 return skb;
543 /* Get packet from user space buffer */
544 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
545 const struct iovec *iv, size_t count,
546 int noblock)
548 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
549 struct sk_buff *skb;
550 size_t len = count, align = 0;
551 struct virtio_net_hdr gso = { 0 };
552 int offset = 0;
554 if (!(tun->flags & TUN_NO_PI)) {
555 if ((len -= sizeof(pi)) > count)
556 return -EINVAL;
558 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
559 return -EFAULT;
560 offset += sizeof(pi);
563 if (tun->flags & TUN_VNET_HDR) {
564 if ((len -= sizeof(gso)) > count)
565 return -EINVAL;
567 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
568 return -EFAULT;
570 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
571 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
572 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
574 if (gso.hdr_len > len)
575 return -EINVAL;
576 offset += sizeof(gso);
579 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
580 align = NET_IP_ALIGN;
581 if (unlikely(len < ETH_HLEN ||
582 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
583 return -EINVAL;
586 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
587 if (IS_ERR(skb)) {
588 if (PTR_ERR(skb) != -EAGAIN)
589 tun->dev->stats.rx_dropped++;
590 return PTR_ERR(skb);
593 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
594 tun->dev->stats.rx_dropped++;
595 kfree_skb(skb);
596 return -EFAULT;
599 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
600 if (!skb_partial_csum_set(skb, gso.csum_start,
601 gso.csum_offset)) {
602 tun->dev->stats.rx_frame_errors++;
603 kfree_skb(skb);
604 return -EINVAL;
606 } else if (tun->flags & TUN_NOCHECKSUM)
607 skb->ip_summed = CHECKSUM_UNNECESSARY;
609 switch (tun->flags & TUN_TYPE_MASK) {
610 case TUN_TUN_DEV:
611 if (tun->flags & TUN_NO_PI) {
612 switch (skb->data[0] & 0xf0) {
613 case 0x40:
614 pi.proto = htons(ETH_P_IP);
615 break;
616 case 0x60:
617 pi.proto = htons(ETH_P_IPV6);
618 break;
619 default:
620 tun->dev->stats.rx_dropped++;
621 kfree_skb(skb);
622 return -EINVAL;
626 skb_reset_mac_header(skb);
627 skb->protocol = pi.proto;
628 skb->dev = tun->dev;
629 break;
630 case TUN_TAP_DEV:
631 skb->protocol = eth_type_trans(skb, tun->dev);
632 break;
635 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
636 pr_debug("GSO!\n");
637 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
638 case VIRTIO_NET_HDR_GSO_TCPV4:
639 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
640 break;
641 case VIRTIO_NET_HDR_GSO_TCPV6:
642 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
643 break;
644 default:
645 tun->dev->stats.rx_frame_errors++;
646 kfree_skb(skb);
647 return -EINVAL;
650 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
651 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
653 skb_shinfo(skb)->gso_size = gso.gso_size;
654 if (skb_shinfo(skb)->gso_size == 0) {
655 tun->dev->stats.rx_frame_errors++;
656 kfree_skb(skb);
657 return -EINVAL;
660 /* Header must be checked, and gso_segs computed. */
661 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
662 skb_shinfo(skb)->gso_segs = 0;
665 netif_rx_ni(skb);
667 tun->dev->stats.rx_packets++;
668 tun->dev->stats.rx_bytes += len;
670 return count;
673 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
674 unsigned long count, loff_t pos)
676 struct file *file = iocb->ki_filp;
677 struct tun_struct *tun = tun_get(file);
678 ssize_t result;
680 if (!tun)
681 return -EBADFD;
683 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
685 result = tun_get_user(tun, iv, iov_length(iv, count),
686 file->f_flags & O_NONBLOCK);
688 tun_put(tun);
689 return result;
692 /* Put packet to the user space buffer */
693 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
694 struct sk_buff *skb,
695 const struct iovec *iv, int len)
697 struct tun_pi pi = { 0, skb->protocol };
698 ssize_t total = 0;
700 if (!(tun->flags & TUN_NO_PI)) {
701 if ((len -= sizeof(pi)) < 0)
702 return -EINVAL;
704 if (len < skb->len) {
705 /* Packet will be striped */
706 pi.flags |= TUN_PKT_STRIP;
709 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
710 return -EFAULT;
711 total += sizeof(pi);
714 if (tun->flags & TUN_VNET_HDR) {
715 struct virtio_net_hdr gso = { 0 }; /* no info leak */
716 if ((len -= sizeof(gso)) < 0)
717 return -EINVAL;
719 if (skb_is_gso(skb)) {
720 struct skb_shared_info *sinfo = skb_shinfo(skb);
722 /* This is a hint as to how much should be linear. */
723 gso.hdr_len = skb_headlen(skb);
724 gso.gso_size = sinfo->gso_size;
725 if (sinfo->gso_type & SKB_GSO_TCPV4)
726 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
727 else if (sinfo->gso_type & SKB_GSO_TCPV6)
728 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
729 else
730 BUG();
731 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
732 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
733 } else
734 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
736 if (skb->ip_summed == CHECKSUM_PARTIAL) {
737 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
738 gso.csum_start = skb->csum_start - skb_headroom(skb);
739 gso.csum_offset = skb->csum_offset;
740 } /* else everything is zero */
742 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
743 sizeof(gso))))
744 return -EFAULT;
745 total += sizeof(gso);
748 len = min_t(int, skb->len, len);
750 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
751 total += len;
753 tun->dev->stats.tx_packets++;
754 tun->dev->stats.tx_bytes += len;
756 return total;
759 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
760 unsigned long count, loff_t pos)
762 struct file *file = iocb->ki_filp;
763 struct tun_file *tfile = file->private_data;
764 struct tun_struct *tun = __tun_get(tfile);
765 DECLARE_WAITQUEUE(wait, current);
766 struct sk_buff *skb;
767 ssize_t len, ret = 0;
769 if (!tun)
770 return -EBADFD;
772 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
774 len = iov_length(iv, count);
775 if (len < 0) {
776 ret = -EINVAL;
777 goto out;
780 add_wait_queue(&tun->socket.wait, &wait);
781 while (len) {
782 current->state = TASK_INTERRUPTIBLE;
784 /* Read frames from the queue */
785 if (!(skb=skb_dequeue(&tun->readq))) {
786 if (file->f_flags & O_NONBLOCK) {
787 ret = -EAGAIN;
788 break;
790 if (signal_pending(current)) {
791 ret = -ERESTARTSYS;
792 break;
794 if (tun->dev->reg_state != NETREG_REGISTERED) {
795 ret = -EIO;
796 break;
799 /* Nothing to read, let's sleep */
800 schedule();
801 continue;
803 netif_wake_queue(tun->dev);
805 ret = tun_put_user(tun, skb, iv, len);
806 kfree_skb(skb);
807 break;
810 current->state = TASK_RUNNING;
811 remove_wait_queue(&tun->socket.wait, &wait);
813 out:
814 tun_put(tun);
815 return ret;
818 static void tun_setup(struct net_device *dev)
820 struct tun_struct *tun = netdev_priv(dev);
822 skb_queue_head_init(&tun->readq);
824 tun->owner = -1;
825 tun->group = -1;
827 dev->ethtool_ops = &tun_ethtool_ops;
828 dev->destructor = tun_free_netdev;
831 /* Trivial set of netlink ops to allow deleting tun or tap
832 * device with netlink.
834 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
836 return -EINVAL;
839 static struct rtnl_link_ops tun_link_ops __read_mostly = {
840 .kind = DRV_NAME,
841 .priv_size = sizeof(struct tun_struct),
842 .setup = tun_setup,
843 .validate = tun_validate,
846 static void tun_sock_write_space(struct sock *sk)
848 struct tun_struct *tun;
850 if (!sock_writeable(sk))
851 return;
853 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
854 return;
856 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
857 wake_up_interruptible_sync(sk->sk_sleep);
859 tun = container_of(sk, struct tun_sock, sk)->tun;
860 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
863 static void tun_sock_destruct(struct sock *sk)
865 free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
868 static struct proto tun_proto = {
869 .name = "tun",
870 .owner = THIS_MODULE,
871 .obj_size = sizeof(struct tun_sock),
874 static int tun_flags(struct tun_struct *tun)
876 int flags = 0;
878 if (tun->flags & TUN_TUN_DEV)
879 flags |= IFF_TUN;
880 else
881 flags |= IFF_TAP;
883 if (tun->flags & TUN_NO_PI)
884 flags |= IFF_NO_PI;
886 if (tun->flags & TUN_ONE_QUEUE)
887 flags |= IFF_ONE_QUEUE;
889 if (tun->flags & TUN_VNET_HDR)
890 flags |= IFF_VNET_HDR;
892 return flags;
895 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
896 char *buf)
898 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
899 return sprintf(buf, "0x%x\n", tun_flags(tun));
902 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
903 char *buf)
905 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
906 return sprintf(buf, "%d\n", tun->owner);
909 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
910 char *buf)
912 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
913 return sprintf(buf, "%d\n", tun->group);
916 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
917 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
918 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
920 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
922 struct sock *sk;
923 struct tun_struct *tun;
924 struct net_device *dev;
925 int err;
927 dev = __dev_get_by_name(net, ifr->ifr_name);
928 if (dev) {
929 if (ifr->ifr_flags & IFF_TUN_EXCL)
930 return -EBUSY;
931 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
932 tun = netdev_priv(dev);
933 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
934 tun = netdev_priv(dev);
935 else
936 return -EINVAL;
938 err = tun_attach(tun, file);
939 if (err < 0)
940 return err;
942 else {
943 char *name;
944 unsigned long flags = 0;
946 err = -EINVAL;
948 if (!capable(CAP_NET_ADMIN))
949 return -EPERM;
951 /* Set dev type */
952 if (ifr->ifr_flags & IFF_TUN) {
953 /* TUN device */
954 flags |= TUN_TUN_DEV;
955 name = "tun%d";
956 } else if (ifr->ifr_flags & IFF_TAP) {
957 /* TAP device */
958 flags |= TUN_TAP_DEV;
959 name = "tap%d";
960 } else
961 goto failed;
963 if (*ifr->ifr_name)
964 name = ifr->ifr_name;
966 dev = alloc_netdev(sizeof(struct tun_struct), name,
967 tun_setup);
968 if (!dev)
969 return -ENOMEM;
971 dev_net_set(dev, net);
972 dev->rtnl_link_ops = &tun_link_ops;
974 tun = netdev_priv(dev);
975 tun->dev = dev;
976 tun->flags = flags;
977 tun->txflt.count = 0;
979 err = -ENOMEM;
980 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
981 if (!sk)
982 goto err_free_dev;
984 init_waitqueue_head(&tun->socket.wait);
985 sock_init_data(&tun->socket, sk);
986 sk->sk_write_space = tun_sock_write_space;
987 sk->sk_sndbuf = INT_MAX;
989 tun->sk = sk;
990 container_of(sk, struct tun_sock, sk)->tun = tun;
992 tun_net_init(dev);
994 if (strchr(dev->name, '%')) {
995 err = dev_alloc_name(dev, dev->name);
996 if (err < 0)
997 goto err_free_sk;
1000 err = -EINVAL;
1001 err = register_netdevice(tun->dev);
1002 if (err < 0)
1003 goto err_free_sk;
1005 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1006 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1007 device_create_file(&tun->dev->dev, &dev_attr_group))
1008 printk(KERN_ERR "Failed to create tun sysfs files\n");
1010 sk->sk_destruct = tun_sock_destruct;
1012 err = tun_attach(tun, file);
1013 if (err < 0)
1014 goto failed;
1017 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1019 if (ifr->ifr_flags & IFF_NO_PI)
1020 tun->flags |= TUN_NO_PI;
1021 else
1022 tun->flags &= ~TUN_NO_PI;
1024 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1025 tun->flags |= TUN_ONE_QUEUE;
1026 else
1027 tun->flags &= ~TUN_ONE_QUEUE;
1029 if (ifr->ifr_flags & IFF_VNET_HDR)
1030 tun->flags |= TUN_VNET_HDR;
1031 else
1032 tun->flags &= ~TUN_VNET_HDR;
1034 /* Make sure persistent devices do not get stuck in
1035 * xoff state.
1037 if (netif_running(tun->dev))
1038 netif_wake_queue(tun->dev);
1040 strcpy(ifr->ifr_name, tun->dev->name);
1041 return 0;
1043 err_free_sk:
1044 sock_put(sk);
1045 err_free_dev:
1046 free_netdev(dev);
1047 failed:
1048 return err;
1051 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1052 struct ifreq *ifr)
1054 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1056 strcpy(ifr->ifr_name, tun->dev->name);
1058 ifr->ifr_flags = tun_flags(tun);
1060 return 0;
1063 /* This is like a cut-down ethtool ops, except done via tun fd so no
1064 * privs required. */
1065 static int set_offload(struct net_device *dev, unsigned long arg)
1067 unsigned int old_features, features;
1069 old_features = dev->features;
1070 /* Unset features, set them as we chew on the arg. */
1071 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1072 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
1074 if (arg & TUN_F_CSUM) {
1075 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1076 arg &= ~TUN_F_CSUM;
1078 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1079 if (arg & TUN_F_TSO_ECN) {
1080 features |= NETIF_F_TSO_ECN;
1081 arg &= ~TUN_F_TSO_ECN;
1083 if (arg & TUN_F_TSO4)
1084 features |= NETIF_F_TSO;
1085 if (arg & TUN_F_TSO6)
1086 features |= NETIF_F_TSO6;
1087 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1091 /* This gives the user a way to test for new features in future by
1092 * trying to set them. */
1093 if (arg)
1094 return -EINVAL;
1096 dev->features = features;
1097 if (old_features != dev->features)
1098 netdev_features_change(dev);
1100 return 0;
1103 static long tun_chr_ioctl(struct file *file, unsigned int cmd,
1104 unsigned long arg)
1106 struct tun_file *tfile = file->private_data;
1107 struct tun_struct *tun;
1108 void __user* argp = (void __user*)arg;
1109 struct ifreq ifr;
1110 int sndbuf;
1111 int ret;
1113 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1114 if (copy_from_user(&ifr, argp, sizeof ifr))
1115 return -EFAULT;
1117 if (cmd == TUNGETFEATURES) {
1118 /* Currently this just means: "what IFF flags are valid?".
1119 * This is needed because we never checked for invalid flags on
1120 * TUNSETIFF. */
1121 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1122 IFF_VNET_HDR,
1123 (unsigned int __user*)argp);
1126 rtnl_lock();
1128 tun = __tun_get(tfile);
1129 if (cmd == TUNSETIFF && !tun) {
1130 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1132 ret = tun_set_iff(tfile->net, file, &ifr);
1134 if (ret)
1135 goto unlock;
1137 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1138 ret = -EFAULT;
1139 goto unlock;
1142 ret = -EBADFD;
1143 if (!tun)
1144 goto unlock;
1146 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1148 ret = 0;
1149 switch (cmd) {
1150 case TUNGETIFF:
1151 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1152 if (ret)
1153 break;
1155 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1156 ret = -EFAULT;
1157 break;
1159 case TUNSETNOCSUM:
1160 /* Disable/Enable checksum */
1161 if (arg)
1162 tun->flags |= TUN_NOCHECKSUM;
1163 else
1164 tun->flags &= ~TUN_NOCHECKSUM;
1166 DBG(KERN_INFO "%s: checksum %s\n",
1167 tun->dev->name, arg ? "disabled" : "enabled");
1168 break;
1170 case TUNSETPERSIST:
1171 /* Disable/Enable persist mode */
1172 if (arg)
1173 tun->flags |= TUN_PERSIST;
1174 else
1175 tun->flags &= ~TUN_PERSIST;
1177 DBG(KERN_INFO "%s: persist %s\n",
1178 tun->dev->name, arg ? "enabled" : "disabled");
1179 break;
1181 case TUNSETOWNER:
1182 /* Set owner of the device */
1183 tun->owner = (uid_t) arg;
1185 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1186 break;
1188 case TUNSETGROUP:
1189 /* Set group of the device */
1190 tun->group= (gid_t) arg;
1192 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1193 break;
1195 case TUNSETLINK:
1196 /* Only allow setting the type when the interface is down */
1197 if (tun->dev->flags & IFF_UP) {
1198 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1199 tun->dev->name);
1200 ret = -EBUSY;
1201 } else {
1202 tun->dev->type = (int) arg;
1203 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1204 ret = 0;
1206 break;
1208 #ifdef TUN_DEBUG
1209 case TUNSETDEBUG:
1210 tun->debug = arg;
1211 break;
1212 #endif
1213 case TUNSETOFFLOAD:
1214 ret = set_offload(tun->dev, arg);
1215 break;
1217 case TUNSETTXFILTER:
1218 /* Can be set only for TAPs */
1219 ret = -EINVAL;
1220 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1221 break;
1222 ret = update_filter(&tun->txflt, (void __user *)arg);
1223 break;
1225 case SIOCGIFHWADDR:
1226 /* Get hw addres */
1227 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1228 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1229 if (copy_to_user(argp, &ifr, sizeof ifr))
1230 ret = -EFAULT;
1231 break;
1233 case SIOCSIFHWADDR:
1234 /* Set hw address */
1235 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1236 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1238 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1239 break;
1241 case TUNGETSNDBUF:
1242 sndbuf = tun->sk->sk_sndbuf;
1243 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1244 ret = -EFAULT;
1245 break;
1247 case TUNSETSNDBUF:
1248 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1249 ret = -EFAULT;
1250 break;
1253 tun->sk->sk_sndbuf = sndbuf;
1254 break;
1256 default:
1257 ret = -EINVAL;
1258 break;
1261 unlock:
1262 rtnl_unlock();
1263 if (tun)
1264 tun_put(tun);
1265 return ret;
1268 static int tun_chr_fasync(int fd, struct file *file, int on)
1270 struct tun_struct *tun = tun_get(file);
1271 int ret;
1273 if (!tun)
1274 return -EBADFD;
1276 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1278 lock_kernel();
1279 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1280 goto out;
1282 if (on) {
1283 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1284 if (ret)
1285 goto out;
1286 tun->flags |= TUN_FASYNC;
1287 } else
1288 tun->flags &= ~TUN_FASYNC;
1289 ret = 0;
1290 out:
1291 unlock_kernel();
1292 tun_put(tun);
1293 return ret;
1296 static int tun_chr_open(struct inode *inode, struct file * file)
1298 struct tun_file *tfile;
1299 cycle_kernel_lock();
1300 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1302 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1303 if (!tfile)
1304 return -ENOMEM;
1305 atomic_set(&tfile->count, 0);
1306 tfile->tun = NULL;
1307 tfile->net = get_net(current->nsproxy->net_ns);
1308 file->private_data = tfile;
1309 return 0;
1312 static int tun_chr_close(struct inode *inode, struct file *file)
1314 struct tun_file *tfile = file->private_data;
1315 struct tun_struct *tun;
1317 tun = __tun_get(tfile);
1318 if (tun) {
1319 struct net_device *dev = tun->dev;
1321 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
1323 __tun_detach(tun);
1325 /* If desireable, unregister the netdevice. */
1326 if (!(tun->flags & TUN_PERSIST)) {
1327 rtnl_lock();
1328 if (dev->reg_state == NETREG_REGISTERED)
1329 unregister_netdevice(dev);
1330 rtnl_unlock();
1334 tun = tfile->tun;
1335 if (tun)
1336 sock_put(tun->sk);
1338 put_net(tfile->net);
1339 kfree(tfile);
1341 return 0;
1344 static const struct file_operations tun_fops = {
1345 .owner = THIS_MODULE,
1346 .llseek = no_llseek,
1347 .read = do_sync_read,
1348 .aio_read = tun_chr_aio_read,
1349 .write = do_sync_write,
1350 .aio_write = tun_chr_aio_write,
1351 .poll = tun_chr_poll,
1352 .unlocked_ioctl = tun_chr_ioctl,
1353 .open = tun_chr_open,
1354 .release = tun_chr_close,
1355 .fasync = tun_chr_fasync
1358 static struct miscdevice tun_miscdev = {
1359 .minor = TUN_MINOR,
1360 .name = "tun",
1361 .devnode = "net/tun",
1362 .fops = &tun_fops,
1365 /* ethtool interface */
1367 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1369 cmd->supported = 0;
1370 cmd->advertising = 0;
1371 cmd->speed = SPEED_10;
1372 cmd->duplex = DUPLEX_FULL;
1373 cmd->port = PORT_TP;
1374 cmd->phy_address = 0;
1375 cmd->transceiver = XCVR_INTERNAL;
1376 cmd->autoneg = AUTONEG_DISABLE;
1377 cmd->maxtxpkt = 0;
1378 cmd->maxrxpkt = 0;
1379 return 0;
1382 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1384 struct tun_struct *tun = netdev_priv(dev);
1386 strcpy(info->driver, DRV_NAME);
1387 strcpy(info->version, DRV_VERSION);
1388 strcpy(info->fw_version, "N/A");
1390 switch (tun->flags & TUN_TYPE_MASK) {
1391 case TUN_TUN_DEV:
1392 strcpy(info->bus_info, "tun");
1393 break;
1394 case TUN_TAP_DEV:
1395 strcpy(info->bus_info, "tap");
1396 break;
1400 static u32 tun_get_msglevel(struct net_device *dev)
1402 #ifdef TUN_DEBUG
1403 struct tun_struct *tun = netdev_priv(dev);
1404 return tun->debug;
1405 #else
1406 return -EOPNOTSUPP;
1407 #endif
1410 static void tun_set_msglevel(struct net_device *dev, u32 value)
1412 #ifdef TUN_DEBUG
1413 struct tun_struct *tun = netdev_priv(dev);
1414 tun->debug = value;
1415 #endif
1418 static u32 tun_get_link(struct net_device *dev)
1420 struct tun_struct *tun = netdev_priv(dev);
1421 return !!tun->tfile;
1424 static u32 tun_get_rx_csum(struct net_device *dev)
1426 struct tun_struct *tun = netdev_priv(dev);
1427 return (tun->flags & TUN_NOCHECKSUM) == 0;
1430 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1432 struct tun_struct *tun = netdev_priv(dev);
1433 if (data)
1434 tun->flags &= ~TUN_NOCHECKSUM;
1435 else
1436 tun->flags |= TUN_NOCHECKSUM;
1437 return 0;
1440 static const struct ethtool_ops tun_ethtool_ops = {
1441 .get_settings = tun_get_settings,
1442 .get_drvinfo = tun_get_drvinfo,
1443 .get_msglevel = tun_get_msglevel,
1444 .set_msglevel = tun_set_msglevel,
1445 .get_link = tun_get_link,
1446 .get_rx_csum = tun_get_rx_csum,
1447 .set_rx_csum = tun_set_rx_csum
1451 static int __init tun_init(void)
1453 int ret = 0;
1455 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1456 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1458 ret = rtnl_link_register(&tun_link_ops);
1459 if (ret) {
1460 printk(KERN_ERR "tun: Can't register link_ops\n");
1461 goto err_linkops;
1464 ret = misc_register(&tun_miscdev);
1465 if (ret) {
1466 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1467 goto err_misc;
1469 return 0;
1470 err_misc:
1471 rtnl_link_unregister(&tun_link_ops);
1472 err_linkops:
1473 return ret;
1476 static void tun_cleanup(void)
1478 misc_deregister(&tun_miscdev);
1479 rtnl_link_unregister(&tun_link_ops);
1482 module_init(tun_init);
1483 module_exit(tun_cleanup);
1484 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1485 MODULE_AUTHOR(DRV_COPYRIGHT);
1486 MODULE_LICENSE("GPL");
1487 MODULE_ALIAS_MISCDEV(TUN_MINOR);