x86, apic: Fix spurious error interrupts triggering on all non-boot APs
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / drivers / net / tun.c
blob909b73db4a991e6d3d6646339c148a18b089dd10
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/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>
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 net_device *dev;
107 struct fasync_struct *fasync;
109 struct tap_filter txflt;
110 struct socket socket;
112 #ifdef TUN_DEBUG
113 int debug;
114 #endif
117 struct tun_sock {
118 struct sock sk;
119 struct tun_struct *tun;
122 static inline struct tun_sock *tun_sk(struct sock *sk)
124 return container_of(sk, struct tun_sock, sk);
127 static int tun_attach(struct tun_struct *tun, struct file *file)
129 struct tun_file *tfile = file->private_data;
130 int err;
132 ASSERT_RTNL();
134 netif_tx_lock_bh(tun->dev);
136 err = -EINVAL;
137 if (tfile->tun)
138 goto out;
140 err = -EBUSY;
141 if (tun->tfile)
142 goto out;
144 err = 0;
145 tfile->tun = tun;
146 tun->tfile = tfile;
147 dev_hold(tun->dev);
148 sock_hold(tun->socket.sk);
149 atomic_inc(&tfile->count);
151 out:
152 netif_tx_unlock_bh(tun->dev);
153 return err;
156 static void __tun_detach(struct tun_struct *tun)
158 /* Detach from net device */
159 netif_tx_lock_bh(tun->dev);
160 tun->tfile = NULL;
161 netif_tx_unlock_bh(tun->dev);
163 /* Drop read queue */
164 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
166 /* Drop the extra count on the net device */
167 dev_put(tun->dev);
170 static void tun_detach(struct tun_struct *tun)
172 rtnl_lock();
173 __tun_detach(tun);
174 rtnl_unlock();
177 static struct tun_struct *__tun_get(struct tun_file *tfile)
179 struct tun_struct *tun = NULL;
181 if (atomic_inc_not_zero(&tfile->count))
182 tun = tfile->tun;
184 return tun;
187 static struct tun_struct *tun_get(struct file *file)
189 return __tun_get(file->private_data);
192 static void tun_put(struct tun_struct *tun)
194 struct tun_file *tfile = tun->tfile;
196 if (atomic_dec_and_test(&tfile->count))
197 tun_detach(tfile->tun);
200 /* TAP filterting */
201 static void addr_hash_set(u32 *mask, const u8 *addr)
203 int n = ether_crc(ETH_ALEN, addr) >> 26;
204 mask[n >> 5] |= (1 << (n & 31));
207 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
209 int n = ether_crc(ETH_ALEN, addr) >> 26;
210 return mask[n >> 5] & (1 << (n & 31));
213 static int update_filter(struct tap_filter *filter, void __user *arg)
215 struct { u8 u[ETH_ALEN]; } *addr;
216 struct tun_filter uf;
217 int err, alen, n, nexact;
219 if (copy_from_user(&uf, arg, sizeof(uf)))
220 return -EFAULT;
222 if (!uf.count) {
223 /* Disabled */
224 filter->count = 0;
225 return 0;
228 alen = ETH_ALEN * uf.count;
229 addr = kmalloc(alen, GFP_KERNEL);
230 if (!addr)
231 return -ENOMEM;
233 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
234 err = -EFAULT;
235 goto done;
238 /* The filter is updated without holding any locks. Which is
239 * perfectly safe. We disable it first and in the worst
240 * case we'll accept a few undesired packets. */
241 filter->count = 0;
242 wmb();
244 /* Use first set of addresses as an exact filter */
245 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
246 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
248 nexact = n;
250 /* Remaining multicast addresses are hashed,
251 * unicast will leave the filter disabled. */
252 memset(filter->mask, 0, sizeof(filter->mask));
253 for (; n < uf.count; n++) {
254 if (!is_multicast_ether_addr(addr[n].u)) {
255 err = 0; /* no filter */
256 goto done;
258 addr_hash_set(filter->mask, addr[n].u);
261 /* For ALLMULTI just set the mask to all ones.
262 * This overrides the mask populated above. */
263 if ((uf.flags & TUN_FLT_ALLMULTI))
264 memset(filter->mask, ~0, sizeof(filter->mask));
266 /* Now enable the filter */
267 wmb();
268 filter->count = nexact;
270 /* Return the number of exact filters */
271 err = nexact;
273 done:
274 kfree(addr);
275 return err;
278 /* Returns: 0 - drop, !=0 - accept */
279 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
281 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
282 * at this point. */
283 struct ethhdr *eh = (struct ethhdr *) skb->data;
284 int i;
286 /* Exact match */
287 for (i = 0; i < filter->count; i++)
288 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
289 return 1;
291 /* Inexact match (multicast only) */
292 if (is_multicast_ether_addr(eh->h_dest))
293 return addr_hash_test(filter->mask, eh->h_dest);
295 return 0;
299 * Checks whether the packet is accepted or not.
300 * Returns: 0 - drop, !=0 - accept
302 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
304 if (!filter->count)
305 return 1;
307 return run_filter(filter, skb);
310 /* Network device part of the driver */
312 static const struct ethtool_ops tun_ethtool_ops;
314 /* Net device detach from fd. */
315 static void tun_net_uninit(struct net_device *dev)
317 struct tun_struct *tun = netdev_priv(dev);
318 struct tun_file *tfile = tun->tfile;
320 /* Inform the methods they need to stop using the dev.
322 if (tfile) {
323 wake_up_all(&tun->socket.wait);
324 if (atomic_dec_and_test(&tfile->count))
325 __tun_detach(tun);
329 static void tun_free_netdev(struct net_device *dev)
331 struct tun_struct *tun = netdev_priv(dev);
333 sock_put(tun->socket.sk);
336 /* Net device open. */
337 static int tun_net_open(struct net_device *dev)
339 netif_start_queue(dev);
340 return 0;
343 /* Net device close. */
344 static int tun_net_close(struct net_device *dev)
346 netif_stop_queue(dev);
347 return 0;
350 /* Net device start xmit */
351 static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
353 struct tun_struct *tun = netdev_priv(dev);
355 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
357 /* Drop packet if interface is not attached */
358 if (!tun->tfile)
359 goto drop;
361 /* Drop if the filter does not like it.
362 * This is a noop if the filter is disabled.
363 * Filter can be enabled only for the TAP devices. */
364 if (!check_filter(&tun->txflt, skb))
365 goto drop;
367 if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
368 if (!(tun->flags & TUN_ONE_QUEUE)) {
369 /* Normal queueing mode. */
370 /* Packet scheduler handles dropping of further packets. */
371 netif_stop_queue(dev);
373 /* We won't see all dropped packets individually, so overrun
374 * error is more appropriate. */
375 dev->stats.tx_fifo_errors++;
376 } else {
377 /* Single queue mode.
378 * Driver handles dropping of all packets itself. */
379 goto drop;
383 /* Orphan the skb - required as we might hang on to it
384 * for indefinite time. */
385 skb_orphan(skb);
387 /* Enqueue packet */
388 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
389 dev->trans_start = jiffies;
391 /* Notify and wake up reader process */
392 if (tun->flags & TUN_FASYNC)
393 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
394 wake_up_interruptible(&tun->socket.wait);
395 return NETDEV_TX_OK;
397 drop:
398 dev->stats.tx_dropped++;
399 kfree_skb(skb);
400 return NETDEV_TX_OK;
403 static void tun_net_mclist(struct net_device *dev)
406 * This callback is supposed to deal with mc filter in
407 * _rx_ path and has nothing to do with the _tx_ path.
408 * In rx path we always accept everything userspace gives us.
410 return;
413 #define MIN_MTU 68
414 #define MAX_MTU 65535
416 static int
417 tun_net_change_mtu(struct net_device *dev, int new_mtu)
419 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
420 return -EINVAL;
421 dev->mtu = new_mtu;
422 return 0;
425 static const struct net_device_ops tun_netdev_ops = {
426 .ndo_uninit = tun_net_uninit,
427 .ndo_open = tun_net_open,
428 .ndo_stop = tun_net_close,
429 .ndo_start_xmit = tun_net_xmit,
430 .ndo_change_mtu = tun_net_change_mtu,
433 static const struct net_device_ops tap_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,
439 .ndo_set_multicast_list = tun_net_mclist,
440 .ndo_set_mac_address = eth_mac_addr,
441 .ndo_validate_addr = eth_validate_addr,
444 /* Initialize net device. */
445 static void tun_net_init(struct net_device *dev)
447 struct tun_struct *tun = netdev_priv(dev);
449 switch (tun->flags & TUN_TYPE_MASK) {
450 case TUN_TUN_DEV:
451 dev->netdev_ops = &tun_netdev_ops;
453 /* Point-to-Point TUN Device */
454 dev->hard_header_len = 0;
455 dev->addr_len = 0;
456 dev->mtu = 1500;
458 /* Zero header length */
459 dev->type = ARPHRD_NONE;
460 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
461 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
462 break;
464 case TUN_TAP_DEV:
465 dev->netdev_ops = &tap_netdev_ops;
466 /* Ethernet TAP Device */
467 ether_setup(dev);
469 random_ether_addr(dev->dev_addr);
471 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
472 break;
476 /* Character device part */
478 /* Poll */
479 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
481 struct tun_file *tfile = file->private_data;
482 struct tun_struct *tun = __tun_get(tfile);
483 struct sock *sk;
484 unsigned int mask = 0;
486 if (!tun)
487 return POLLERR;
489 sk = tun->socket.sk;
491 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
493 poll_wait(file, &tun->socket.wait, wait);
495 if (!skb_queue_empty(&sk->sk_receive_queue))
496 mask |= POLLIN | POLLRDNORM;
498 if (sock_writeable(sk) ||
499 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
500 sock_writeable(sk)))
501 mask |= POLLOUT | POLLWRNORM;
503 if (tun->dev->reg_state != NETREG_REGISTERED)
504 mask = POLLERR;
506 tun_put(tun);
507 return mask;
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->socket.sk;
517 struct sk_buff *skb;
518 int err;
520 /* Under a page? Don't bother with paged skb. */
521 if (prepad + len < PAGE_SIZE || !linear)
522 linear = len;
524 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
525 &err);
526 if (!skb)
527 return ERR_PTR(err);
529 skb_reserve(skb, prepad);
530 skb_put(skb, linear);
531 skb->data_len = len - linear;
532 skb->len += len - linear;
534 return skb;
537 /* Get packet from user space buffer */
538 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
539 const struct iovec *iv, size_t count,
540 int noblock)
542 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
543 struct sk_buff *skb;
544 size_t len = count, align = 0;
545 struct virtio_net_hdr gso = { 0 };
546 int offset = 0;
548 if (!(tun->flags & TUN_NO_PI)) {
549 if ((len -= sizeof(pi)) > count)
550 return -EINVAL;
552 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
553 return -EFAULT;
554 offset += sizeof(pi);
557 if (tun->flags & TUN_VNET_HDR) {
558 if ((len -= sizeof(gso)) > count)
559 return -EINVAL;
561 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
562 return -EFAULT;
564 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
565 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
566 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
568 if (gso.hdr_len > len)
569 return -EINVAL;
570 offset += sizeof(gso);
573 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
574 align = NET_IP_ALIGN;
575 if (unlikely(len < ETH_HLEN ||
576 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
577 return -EINVAL;
580 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
581 if (IS_ERR(skb)) {
582 if (PTR_ERR(skb) != -EAGAIN)
583 tun->dev->stats.rx_dropped++;
584 return PTR_ERR(skb);
587 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
588 tun->dev->stats.rx_dropped++;
589 kfree_skb(skb);
590 return -EFAULT;
593 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
594 if (!skb_partial_csum_set(skb, gso.csum_start,
595 gso.csum_offset)) {
596 tun->dev->stats.rx_frame_errors++;
597 kfree_skb(skb);
598 return -EINVAL;
600 } else if (tun->flags & TUN_NOCHECKSUM)
601 skb->ip_summed = CHECKSUM_UNNECESSARY;
603 switch (tun->flags & TUN_TYPE_MASK) {
604 case TUN_TUN_DEV:
605 if (tun->flags & TUN_NO_PI) {
606 switch (skb->data[0] & 0xf0) {
607 case 0x40:
608 pi.proto = htons(ETH_P_IP);
609 break;
610 case 0x60:
611 pi.proto = htons(ETH_P_IPV6);
612 break;
613 default:
614 tun->dev->stats.rx_dropped++;
615 kfree_skb(skb);
616 return -EINVAL;
620 skb_reset_mac_header(skb);
621 skb->protocol = pi.proto;
622 skb->dev = tun->dev;
623 break;
624 case TUN_TAP_DEV:
625 skb->protocol = eth_type_trans(skb, tun->dev);
626 break;
629 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
630 pr_debug("GSO!\n");
631 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
632 case VIRTIO_NET_HDR_GSO_TCPV4:
633 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
634 break;
635 case VIRTIO_NET_HDR_GSO_TCPV6:
636 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
637 break;
638 case VIRTIO_NET_HDR_GSO_UDP:
639 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
640 break;
641 default:
642 tun->dev->stats.rx_frame_errors++;
643 kfree_skb(skb);
644 return -EINVAL;
647 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
648 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
650 skb_shinfo(skb)->gso_size = gso.gso_size;
651 if (skb_shinfo(skb)->gso_size == 0) {
652 tun->dev->stats.rx_frame_errors++;
653 kfree_skb(skb);
654 return -EINVAL;
657 /* Header must be checked, and gso_segs computed. */
658 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
659 skb_shinfo(skb)->gso_segs = 0;
662 netif_rx_ni(skb);
664 tun->dev->stats.rx_packets++;
665 tun->dev->stats.rx_bytes += len;
667 return count;
670 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
671 unsigned long count, loff_t pos)
673 struct file *file = iocb->ki_filp;
674 struct tun_struct *tun = tun_get(file);
675 ssize_t result;
677 if (!tun)
678 return -EBADFD;
680 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
682 result = tun_get_user(tun, iv, iov_length(iv, count),
683 file->f_flags & O_NONBLOCK);
685 tun_put(tun);
686 return result;
689 /* Put packet to the user space buffer */
690 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
691 struct sk_buff *skb,
692 const struct iovec *iv, int len)
694 struct tun_pi pi = { 0, skb->protocol };
695 ssize_t total = 0;
697 if (!(tun->flags & TUN_NO_PI)) {
698 if ((len -= sizeof(pi)) < 0)
699 return -EINVAL;
701 if (len < skb->len) {
702 /* Packet will be striped */
703 pi.flags |= TUN_PKT_STRIP;
706 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
707 return -EFAULT;
708 total += sizeof(pi);
711 if (tun->flags & TUN_VNET_HDR) {
712 struct virtio_net_hdr gso = { 0 }; /* no info leak */
713 if ((len -= sizeof(gso)) < 0)
714 return -EINVAL;
716 if (skb_is_gso(skb)) {
717 struct skb_shared_info *sinfo = skb_shinfo(skb);
719 /* This is a hint as to how much should be linear. */
720 gso.hdr_len = skb_headlen(skb);
721 gso.gso_size = sinfo->gso_size;
722 if (sinfo->gso_type & SKB_GSO_TCPV4)
723 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
724 else if (sinfo->gso_type & SKB_GSO_TCPV6)
725 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
726 else if (sinfo->gso_type & SKB_GSO_UDP)
727 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
728 else
729 BUG();
730 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
731 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
732 } else
733 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
735 if (skb->ip_summed == CHECKSUM_PARTIAL) {
736 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
737 gso.csum_start = skb->csum_start - skb_headroom(skb);
738 gso.csum_offset = skb->csum_offset;
739 } /* else everything is zero */
741 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
742 sizeof(gso))))
743 return -EFAULT;
744 total += sizeof(gso);
747 len = min_t(int, skb->len, len);
749 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
750 total += len;
752 tun->dev->stats.tx_packets++;
753 tun->dev->stats.tx_bytes += len;
755 return total;
758 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
759 unsigned long count, loff_t pos)
761 struct file *file = iocb->ki_filp;
762 struct tun_file *tfile = file->private_data;
763 struct tun_struct *tun = __tun_get(tfile);
764 DECLARE_WAITQUEUE(wait, current);
765 struct sk_buff *skb;
766 ssize_t len, ret = 0;
768 if (!tun)
769 return -EBADFD;
771 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
773 len = iov_length(iv, count);
774 if (len < 0) {
775 ret = -EINVAL;
776 goto out;
779 add_wait_queue(&tun->socket.wait, &wait);
780 while (len) {
781 current->state = TASK_INTERRUPTIBLE;
783 /* Read frames from the queue */
784 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
785 if (file->f_flags & O_NONBLOCK) {
786 ret = -EAGAIN;
787 break;
789 if (signal_pending(current)) {
790 ret = -ERESTARTSYS;
791 break;
793 if (tun->dev->reg_state != NETREG_REGISTERED) {
794 ret = -EIO;
795 break;
798 /* Nothing to read, let's sleep */
799 schedule();
800 continue;
802 netif_wake_queue(tun->dev);
804 ret = tun_put_user(tun, skb, iv, len);
805 kfree_skb(skb);
806 break;
809 current->state = TASK_RUNNING;
810 remove_wait_queue(&tun->socket.wait, &wait);
812 out:
813 tun_put(tun);
814 return ret;
817 static void tun_setup(struct net_device *dev)
819 struct tun_struct *tun = netdev_priv(dev);
821 tun->owner = -1;
822 tun->group = -1;
824 dev->ethtool_ops = &tun_ethtool_ops;
825 dev->destructor = tun_free_netdev;
828 /* Trivial set of netlink ops to allow deleting tun or tap
829 * device with netlink.
831 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
833 return -EINVAL;
836 static struct rtnl_link_ops tun_link_ops __read_mostly = {
837 .kind = DRV_NAME,
838 .priv_size = sizeof(struct tun_struct),
839 .setup = tun_setup,
840 .validate = tun_validate,
843 static void tun_sock_write_space(struct sock *sk)
845 struct tun_struct *tun;
847 if (!sock_writeable(sk))
848 return;
850 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
851 return;
853 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
854 wake_up_interruptible_sync(sk->sk_sleep);
856 tun = tun_sk(sk)->tun;
857 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
860 static void tun_sock_destruct(struct sock *sk)
862 free_netdev(tun_sk(sk)->tun->dev);
865 static struct proto tun_proto = {
866 .name = "tun",
867 .owner = THIS_MODULE,
868 .obj_size = sizeof(struct tun_sock),
871 static int tun_flags(struct tun_struct *tun)
873 int flags = 0;
875 if (tun->flags & TUN_TUN_DEV)
876 flags |= IFF_TUN;
877 else
878 flags |= IFF_TAP;
880 if (tun->flags & TUN_NO_PI)
881 flags |= IFF_NO_PI;
883 if (tun->flags & TUN_ONE_QUEUE)
884 flags |= IFF_ONE_QUEUE;
886 if (tun->flags & TUN_VNET_HDR)
887 flags |= IFF_VNET_HDR;
889 return flags;
892 static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
893 char *buf)
895 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
896 return sprintf(buf, "0x%x\n", tun_flags(tun));
899 static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
900 char *buf)
902 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
903 return sprintf(buf, "%d\n", tun->owner);
906 static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
907 char *buf)
909 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
910 return sprintf(buf, "%d\n", tun->group);
913 static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
914 static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
915 static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
917 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
919 struct sock *sk;
920 struct tun_struct *tun;
921 struct net_device *dev;
922 int err;
924 dev = __dev_get_by_name(net, ifr->ifr_name);
925 if (dev) {
926 const struct cred *cred = current_cred();
928 if (ifr->ifr_flags & IFF_TUN_EXCL)
929 return -EBUSY;
930 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
931 tun = netdev_priv(dev);
932 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
933 tun = netdev_priv(dev);
934 else
935 return -EINVAL;
937 if (((tun->owner != -1 && cred->euid != tun->owner) ||
938 (tun->group != -1 && !in_egroup_p(tun->group))) &&
939 !capable(CAP_NET_ADMIN))
940 return -EPERM;
941 err = security_tun_dev_attach(tun->socket.sk);
942 if (err < 0)
943 return err;
945 err = tun_attach(tun, file);
946 if (err < 0)
947 return err;
949 else {
950 char *name;
951 unsigned long flags = 0;
953 if (!capable(CAP_NET_ADMIN))
954 return -EPERM;
955 err = security_tun_dev_create();
956 if (err < 0)
957 return err;
959 /* Set dev type */
960 if (ifr->ifr_flags & IFF_TUN) {
961 /* TUN device */
962 flags |= TUN_TUN_DEV;
963 name = "tun%d";
964 } else if (ifr->ifr_flags & IFF_TAP) {
965 /* TAP device */
966 flags |= TUN_TAP_DEV;
967 name = "tap%d";
968 } else
969 return -EINVAL;
971 if (*ifr->ifr_name)
972 name = ifr->ifr_name;
974 dev = alloc_netdev(sizeof(struct tun_struct), name,
975 tun_setup);
976 if (!dev)
977 return -ENOMEM;
979 dev_net_set(dev, net);
980 dev->rtnl_link_ops = &tun_link_ops;
982 tun = netdev_priv(dev);
983 tun->dev = dev;
984 tun->flags = flags;
985 tun->txflt.count = 0;
987 err = -ENOMEM;
988 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
989 if (!sk)
990 goto err_free_dev;
992 init_waitqueue_head(&tun->socket.wait);
993 sock_init_data(&tun->socket, sk);
994 sk->sk_write_space = tun_sock_write_space;
995 sk->sk_sndbuf = INT_MAX;
997 tun_sk(sk)->tun = tun;
999 security_tun_dev_post_create(sk);
1001 tun_net_init(dev);
1003 if (strchr(dev->name, '%')) {
1004 err = dev_alloc_name(dev, dev->name);
1005 if (err < 0)
1006 goto err_free_sk;
1009 err = register_netdevice(tun->dev);
1010 if (err < 0)
1011 goto err_free_sk;
1013 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1014 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1015 device_create_file(&tun->dev->dev, &dev_attr_group))
1016 printk(KERN_ERR "Failed to create tun sysfs files\n");
1018 sk->sk_destruct = tun_sock_destruct;
1020 err = tun_attach(tun, file);
1021 if (err < 0)
1022 goto failed;
1025 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1027 if (ifr->ifr_flags & IFF_NO_PI)
1028 tun->flags |= TUN_NO_PI;
1029 else
1030 tun->flags &= ~TUN_NO_PI;
1032 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1033 tun->flags |= TUN_ONE_QUEUE;
1034 else
1035 tun->flags &= ~TUN_ONE_QUEUE;
1037 if (ifr->ifr_flags & IFF_VNET_HDR)
1038 tun->flags |= TUN_VNET_HDR;
1039 else
1040 tun->flags &= ~TUN_VNET_HDR;
1042 /* Make sure persistent devices do not get stuck in
1043 * xoff state.
1045 if (netif_running(tun->dev))
1046 netif_wake_queue(tun->dev);
1048 strcpy(ifr->ifr_name, tun->dev->name);
1049 return 0;
1051 err_free_sk:
1052 sock_put(sk);
1053 err_free_dev:
1054 free_netdev(dev);
1055 failed:
1056 return err;
1059 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1060 struct ifreq *ifr)
1062 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1064 strcpy(ifr->ifr_name, tun->dev->name);
1066 ifr->ifr_flags = tun_flags(tun);
1068 return 0;
1071 /* This is like a cut-down ethtool ops, except done via tun fd so no
1072 * privs required. */
1073 static int set_offload(struct net_device *dev, unsigned long arg)
1075 unsigned int old_features, features;
1077 old_features = dev->features;
1078 /* Unset features, set them as we chew on the arg. */
1079 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1080 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6
1081 |NETIF_F_UFO));
1083 if (arg & TUN_F_CSUM) {
1084 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1085 arg &= ~TUN_F_CSUM;
1087 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1088 if (arg & TUN_F_TSO_ECN) {
1089 features |= NETIF_F_TSO_ECN;
1090 arg &= ~TUN_F_TSO_ECN;
1092 if (arg & TUN_F_TSO4)
1093 features |= NETIF_F_TSO;
1094 if (arg & TUN_F_TSO6)
1095 features |= NETIF_F_TSO6;
1096 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1099 if (arg & TUN_F_UFO) {
1100 features |= NETIF_F_UFO;
1101 arg &= ~TUN_F_UFO;
1105 /* This gives the user a way to test for new features in future by
1106 * trying to set them. */
1107 if (arg)
1108 return -EINVAL;
1110 dev->features = features;
1111 if (old_features != dev->features)
1112 netdev_features_change(dev);
1114 return 0;
1117 static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1118 unsigned long arg, int ifreq_len)
1120 struct tun_file *tfile = file->private_data;
1121 struct tun_struct *tun;
1122 void __user* argp = (void __user*)arg;
1123 struct ifreq ifr;
1124 int sndbuf;
1125 int ret;
1127 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1128 if (copy_from_user(&ifr, argp, ifreq_len))
1129 return -EFAULT;
1131 if (cmd == TUNGETFEATURES) {
1132 /* Currently this just means: "what IFF flags are valid?".
1133 * This is needed because we never checked for invalid flags on
1134 * TUNSETIFF. */
1135 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1136 IFF_VNET_HDR,
1137 (unsigned int __user*)argp);
1140 rtnl_lock();
1142 tun = __tun_get(tfile);
1143 if (cmd == TUNSETIFF && !tun) {
1144 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1146 ret = tun_set_iff(tfile->net, file, &ifr);
1148 if (ret)
1149 goto unlock;
1151 if (copy_to_user(argp, &ifr, ifreq_len))
1152 ret = -EFAULT;
1153 goto unlock;
1156 ret = -EBADFD;
1157 if (!tun)
1158 goto unlock;
1160 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1162 ret = 0;
1163 switch (cmd) {
1164 case TUNGETIFF:
1165 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1166 if (ret)
1167 break;
1169 if (copy_to_user(argp, &ifr, ifreq_len))
1170 ret = -EFAULT;
1171 break;
1173 case TUNSETNOCSUM:
1174 /* Disable/Enable checksum */
1175 if (arg)
1176 tun->flags |= TUN_NOCHECKSUM;
1177 else
1178 tun->flags &= ~TUN_NOCHECKSUM;
1180 DBG(KERN_INFO "%s: checksum %s\n",
1181 tun->dev->name, arg ? "disabled" : "enabled");
1182 break;
1184 case TUNSETPERSIST:
1185 /* Disable/Enable persist mode */
1186 if (arg)
1187 tun->flags |= TUN_PERSIST;
1188 else
1189 tun->flags &= ~TUN_PERSIST;
1191 DBG(KERN_INFO "%s: persist %s\n",
1192 tun->dev->name, arg ? "enabled" : "disabled");
1193 break;
1195 case TUNSETOWNER:
1196 /* Set owner of the device */
1197 tun->owner = (uid_t) arg;
1199 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1200 break;
1202 case TUNSETGROUP:
1203 /* Set group of the device */
1204 tun->group= (gid_t) arg;
1206 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1207 break;
1209 case TUNSETLINK:
1210 /* Only allow setting the type when the interface is down */
1211 if (tun->dev->flags & IFF_UP) {
1212 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1213 tun->dev->name);
1214 ret = -EBUSY;
1215 } else {
1216 tun->dev->type = (int) arg;
1217 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1218 ret = 0;
1220 break;
1222 #ifdef TUN_DEBUG
1223 case TUNSETDEBUG:
1224 tun->debug = arg;
1225 break;
1226 #endif
1227 case TUNSETOFFLOAD:
1228 ret = set_offload(tun->dev, arg);
1229 break;
1231 case TUNSETTXFILTER:
1232 /* Can be set only for TAPs */
1233 ret = -EINVAL;
1234 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1235 break;
1236 ret = update_filter(&tun->txflt, (void __user *)arg);
1237 break;
1239 case SIOCGIFHWADDR:
1240 /* Get hw addres */
1241 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1242 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1243 if (copy_to_user(argp, &ifr, ifreq_len))
1244 ret = -EFAULT;
1245 break;
1247 case SIOCSIFHWADDR:
1248 /* Set hw address */
1249 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1250 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1252 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1253 break;
1255 case TUNGETSNDBUF:
1256 sndbuf = tun->socket.sk->sk_sndbuf;
1257 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1258 ret = -EFAULT;
1259 break;
1261 case TUNSETSNDBUF:
1262 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1263 ret = -EFAULT;
1264 break;
1267 tun->socket.sk->sk_sndbuf = sndbuf;
1268 break;
1270 default:
1271 ret = -EINVAL;
1272 break;
1275 unlock:
1276 rtnl_unlock();
1277 if (tun)
1278 tun_put(tun);
1279 return ret;
1282 static long tun_chr_ioctl(struct file *file,
1283 unsigned int cmd, unsigned long arg)
1285 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
1288 #ifdef CONFIG_COMPAT
1289 static long tun_chr_compat_ioctl(struct file *file,
1290 unsigned int cmd, unsigned long arg)
1292 switch (cmd) {
1293 case TUNSETIFF:
1294 case TUNGETIFF:
1295 case TUNSETTXFILTER:
1296 case TUNGETSNDBUF:
1297 case TUNSETSNDBUF:
1298 case SIOCGIFHWADDR:
1299 case SIOCSIFHWADDR:
1300 arg = (unsigned long)compat_ptr(arg);
1301 break;
1302 default:
1303 arg = (compat_ulong_t)arg;
1304 break;
1308 * compat_ifreq is shorter than ifreq, so we must not access beyond
1309 * the end of that structure. All fields that are used in this
1310 * driver are compatible though, we don't need to convert the
1311 * contents.
1313 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
1315 #endif /* CONFIG_COMPAT */
1317 static int tun_chr_fasync(int fd, struct file *file, int on)
1319 struct tun_struct *tun = tun_get(file);
1320 int ret;
1322 if (!tun)
1323 return -EBADFD;
1325 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1327 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1328 goto out;
1330 if (on) {
1331 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1332 if (ret)
1333 goto out;
1334 tun->flags |= TUN_FASYNC;
1335 } else
1336 tun->flags &= ~TUN_FASYNC;
1337 ret = 0;
1338 out:
1339 tun_put(tun);
1340 return ret;
1343 static int tun_chr_open(struct inode *inode, struct file * file)
1345 struct tun_file *tfile;
1347 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1349 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1350 if (!tfile)
1351 return -ENOMEM;
1352 atomic_set(&tfile->count, 0);
1353 tfile->tun = NULL;
1354 tfile->net = get_net(current->nsproxy->net_ns);
1355 file->private_data = tfile;
1356 return 0;
1359 static int tun_chr_close(struct inode *inode, struct file *file)
1361 struct tun_file *tfile = file->private_data;
1362 struct tun_struct *tun;
1364 tun = __tun_get(tfile);
1365 if (tun) {
1366 struct net_device *dev = tun->dev;
1368 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
1370 __tun_detach(tun);
1372 /* If desireable, unregister the netdevice. */
1373 if (!(tun->flags & TUN_PERSIST)) {
1374 rtnl_lock();
1375 if (dev->reg_state == NETREG_REGISTERED)
1376 unregister_netdevice(dev);
1377 rtnl_unlock();
1381 tun = tfile->tun;
1382 if (tun)
1383 sock_put(tun->socket.sk);
1385 put_net(tfile->net);
1386 kfree(tfile);
1388 return 0;
1391 static const struct file_operations tun_fops = {
1392 .owner = THIS_MODULE,
1393 .llseek = no_llseek,
1394 .read = do_sync_read,
1395 .aio_read = tun_chr_aio_read,
1396 .write = do_sync_write,
1397 .aio_write = tun_chr_aio_write,
1398 .poll = tun_chr_poll,
1399 .unlocked_ioctl = tun_chr_ioctl,
1400 #ifdef CONFIG_COMPAT
1401 .compat_ioctl = tun_chr_compat_ioctl,
1402 #endif
1403 .open = tun_chr_open,
1404 .release = tun_chr_close,
1405 .fasync = tun_chr_fasync
1408 static struct miscdevice tun_miscdev = {
1409 .minor = TUN_MINOR,
1410 .name = "tun",
1411 .nodename = "net/tun",
1412 .fops = &tun_fops,
1415 /* ethtool interface */
1417 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1419 cmd->supported = 0;
1420 cmd->advertising = 0;
1421 cmd->speed = SPEED_10;
1422 cmd->duplex = DUPLEX_FULL;
1423 cmd->port = PORT_TP;
1424 cmd->phy_address = 0;
1425 cmd->transceiver = XCVR_INTERNAL;
1426 cmd->autoneg = AUTONEG_DISABLE;
1427 cmd->maxtxpkt = 0;
1428 cmd->maxrxpkt = 0;
1429 return 0;
1432 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1434 struct tun_struct *tun = netdev_priv(dev);
1436 strcpy(info->driver, DRV_NAME);
1437 strcpy(info->version, DRV_VERSION);
1438 strcpy(info->fw_version, "N/A");
1440 switch (tun->flags & TUN_TYPE_MASK) {
1441 case TUN_TUN_DEV:
1442 strcpy(info->bus_info, "tun");
1443 break;
1444 case TUN_TAP_DEV:
1445 strcpy(info->bus_info, "tap");
1446 break;
1450 static u32 tun_get_msglevel(struct net_device *dev)
1452 #ifdef TUN_DEBUG
1453 struct tun_struct *tun = netdev_priv(dev);
1454 return tun->debug;
1455 #else
1456 return -EOPNOTSUPP;
1457 #endif
1460 static void tun_set_msglevel(struct net_device *dev, u32 value)
1462 #ifdef TUN_DEBUG
1463 struct tun_struct *tun = netdev_priv(dev);
1464 tun->debug = value;
1465 #endif
1468 static u32 tun_get_link(struct net_device *dev)
1470 struct tun_struct *tun = netdev_priv(dev);
1471 return !!tun->tfile;
1474 static u32 tun_get_rx_csum(struct net_device *dev)
1476 struct tun_struct *tun = netdev_priv(dev);
1477 return (tun->flags & TUN_NOCHECKSUM) == 0;
1480 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1482 struct tun_struct *tun = netdev_priv(dev);
1483 if (data)
1484 tun->flags &= ~TUN_NOCHECKSUM;
1485 else
1486 tun->flags |= TUN_NOCHECKSUM;
1487 return 0;
1490 static const struct ethtool_ops tun_ethtool_ops = {
1491 .get_settings = tun_get_settings,
1492 .get_drvinfo = tun_get_drvinfo,
1493 .get_msglevel = tun_get_msglevel,
1494 .set_msglevel = tun_set_msglevel,
1495 .get_link = tun_get_link,
1496 .get_rx_csum = tun_get_rx_csum,
1497 .set_rx_csum = tun_set_rx_csum
1501 static int __init tun_init(void)
1503 int ret = 0;
1505 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1506 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1508 ret = rtnl_link_register(&tun_link_ops);
1509 if (ret) {
1510 printk(KERN_ERR "tun: Can't register link_ops\n");
1511 goto err_linkops;
1514 ret = misc_register(&tun_miscdev);
1515 if (ret) {
1516 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1517 goto err_misc;
1519 return 0;
1520 err_misc:
1521 rtnl_link_unregister(&tun_link_ops);
1522 err_linkops:
1523 return ret;
1526 static void tun_cleanup(void)
1528 misc_deregister(&tun_miscdev);
1529 rtnl_link_unregister(&tun_link_ops);
1532 module_init(tun_init);
1533 module_exit(tun_cleanup);
1534 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1535 MODULE_AUTHOR(DRV_COPYRIGHT);
1536 MODULE_LICENSE("GPL");
1537 MODULE_ALIAS_MISCDEV(TUN_MINOR);