tun: reuse struct sock fields
[linux-2.6/x86.git] / drivers / net / tun.c
blob589a44acdc76ec7c77949beed85c713f612f6ce7
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 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 const struct cred *cred = current_cred();
131 int err;
133 ASSERT_RTNL();
135 /* Check permissions */
136 if (((tun->owner != -1 && cred->euid != tun->owner) ||
137 (tun->group != -1 && !in_egroup_p(tun->group))) &&
138 !capable(CAP_NET_ADMIN))
139 return -EPERM;
141 netif_tx_lock_bh(tun->dev);
143 err = -EINVAL;
144 if (tfile->tun)
145 goto out;
147 err = -EBUSY;
148 if (tun->tfile)
149 goto out;
151 err = 0;
152 tfile->tun = tun;
153 tun->tfile = tfile;
154 dev_hold(tun->dev);
155 sock_hold(tun->socket.sk);
156 atomic_inc(&tfile->count);
158 out:
159 netif_tx_unlock_bh(tun->dev);
160 return err;
163 static void __tun_detach(struct tun_struct *tun)
165 /* Detach from net device */
166 netif_tx_lock_bh(tun->dev);
167 tun->tfile = NULL;
168 netif_tx_unlock_bh(tun->dev);
170 /* Drop read queue */
171 skb_queue_purge(&tun->socket.sk->sk_receive_queue);
173 /* Drop the extra count on the net device */
174 dev_put(tun->dev);
177 static void tun_detach(struct tun_struct *tun)
179 rtnl_lock();
180 __tun_detach(tun);
181 rtnl_unlock();
184 static struct tun_struct *__tun_get(struct tun_file *tfile)
186 struct tun_struct *tun = NULL;
188 if (atomic_inc_not_zero(&tfile->count))
189 tun = tfile->tun;
191 return tun;
194 static struct tun_struct *tun_get(struct file *file)
196 return __tun_get(file->private_data);
199 static void tun_put(struct tun_struct *tun)
201 struct tun_file *tfile = tun->tfile;
203 if (atomic_dec_and_test(&tfile->count))
204 tun_detach(tfile->tun);
207 /* TAP filterting */
208 static void addr_hash_set(u32 *mask, const u8 *addr)
210 int n = ether_crc(ETH_ALEN, addr) >> 26;
211 mask[n >> 5] |= (1 << (n & 31));
214 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
216 int n = ether_crc(ETH_ALEN, addr) >> 26;
217 return mask[n >> 5] & (1 << (n & 31));
220 static int update_filter(struct tap_filter *filter, void __user *arg)
222 struct { u8 u[ETH_ALEN]; } *addr;
223 struct tun_filter uf;
224 int err, alen, n, nexact;
226 if (copy_from_user(&uf, arg, sizeof(uf)))
227 return -EFAULT;
229 if (!uf.count) {
230 /* Disabled */
231 filter->count = 0;
232 return 0;
235 alen = ETH_ALEN * uf.count;
236 addr = kmalloc(alen, GFP_KERNEL);
237 if (!addr)
238 return -ENOMEM;
240 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
241 err = -EFAULT;
242 goto done;
245 /* The filter is updated without holding any locks. Which is
246 * perfectly safe. We disable it first and in the worst
247 * case we'll accept a few undesired packets. */
248 filter->count = 0;
249 wmb();
251 /* Use first set of addresses as an exact filter */
252 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
253 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
255 nexact = n;
257 /* Remaining multicast addresses are hashed,
258 * unicast will leave the filter disabled. */
259 memset(filter->mask, 0, sizeof(filter->mask));
260 for (; n < uf.count; n++) {
261 if (!is_multicast_ether_addr(addr[n].u)) {
262 err = 0; /* no filter */
263 goto done;
265 addr_hash_set(filter->mask, addr[n].u);
268 /* For ALLMULTI just set the mask to all ones.
269 * This overrides the mask populated above. */
270 if ((uf.flags & TUN_FLT_ALLMULTI))
271 memset(filter->mask, ~0, sizeof(filter->mask));
273 /* Now enable the filter */
274 wmb();
275 filter->count = nexact;
277 /* Return the number of exact filters */
278 err = nexact;
280 done:
281 kfree(addr);
282 return err;
285 /* Returns: 0 - drop, !=0 - accept */
286 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
288 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
289 * at this point. */
290 struct ethhdr *eh = (struct ethhdr *) skb->data;
291 int i;
293 /* Exact match */
294 for (i = 0; i < filter->count; i++)
295 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
296 return 1;
298 /* Inexact match (multicast only) */
299 if (is_multicast_ether_addr(eh->h_dest))
300 return addr_hash_test(filter->mask, eh->h_dest);
302 return 0;
306 * Checks whether the packet is accepted or not.
307 * Returns: 0 - drop, !=0 - accept
309 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
311 if (!filter->count)
312 return 1;
314 return run_filter(filter, skb);
317 /* Network device part of the driver */
319 static const struct ethtool_ops tun_ethtool_ops;
321 /* Net device detach from fd. */
322 static void tun_net_uninit(struct net_device *dev)
324 struct tun_struct *tun = netdev_priv(dev);
325 struct tun_file *tfile = tun->tfile;
327 /* Inform the methods they need to stop using the dev.
329 if (tfile) {
330 wake_up_all(&tun->socket.wait);
331 if (atomic_dec_and_test(&tfile->count))
332 __tun_detach(tun);
336 static void tun_free_netdev(struct net_device *dev)
338 struct tun_struct *tun = netdev_priv(dev);
340 sock_put(tun->socket.sk);
343 /* Net device open. */
344 static int tun_net_open(struct net_device *dev)
346 netif_start_queue(dev);
347 return 0;
350 /* Net device close. */
351 static int tun_net_close(struct net_device *dev)
353 netif_stop_queue(dev);
354 return 0;
357 /* Net device start xmit */
358 static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
360 struct tun_struct *tun = netdev_priv(dev);
362 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
364 /* Drop packet if interface is not attached */
365 if (!tun->tfile)
366 goto drop;
368 /* Drop if the filter does not like it.
369 * This is a noop if the filter is disabled.
370 * Filter can be enabled only for the TAP devices. */
371 if (!check_filter(&tun->txflt, skb))
372 goto drop;
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++;
383 } else {
384 /* Single queue mode.
385 * Driver handles dropping of all packets itself. */
386 goto drop;
390 /* Enqueue packet */
391 skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
392 dev->trans_start = jiffies;
394 /* Notify and wake up reader process */
395 if (tun->flags & TUN_FASYNC)
396 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
397 wake_up_interruptible(&tun->socket.wait);
398 return NETDEV_TX_OK;
400 drop:
401 dev->stats.tx_dropped++;
402 kfree_skb(skb);
403 return NETDEV_TX_OK;
406 static void tun_net_mclist(struct net_device *dev)
409 * This callback is supposed to deal with mc filter in
410 * _rx_ path and has nothing to do with the _tx_ path.
411 * In rx path we always accept everything userspace gives us.
413 return;
416 #define MIN_MTU 68
417 #define MAX_MTU 65535
419 static int
420 tun_net_change_mtu(struct net_device *dev, int new_mtu)
422 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
423 return -EINVAL;
424 dev->mtu = new_mtu;
425 return 0;
428 static const struct net_device_ops tun_netdev_ops = {
429 .ndo_uninit = tun_net_uninit,
430 .ndo_open = tun_net_open,
431 .ndo_stop = tun_net_close,
432 .ndo_start_xmit = tun_net_xmit,
433 .ndo_change_mtu = tun_net_change_mtu,
436 static const struct net_device_ops tap_netdev_ops = {
437 .ndo_uninit = tun_net_uninit,
438 .ndo_open = tun_net_open,
439 .ndo_stop = tun_net_close,
440 .ndo_start_xmit = tun_net_xmit,
441 .ndo_change_mtu = tun_net_change_mtu,
442 .ndo_set_multicast_list = tun_net_mclist,
443 .ndo_set_mac_address = eth_mac_addr,
444 .ndo_validate_addr = eth_validate_addr,
447 /* Initialize net device. */
448 static void tun_net_init(struct net_device *dev)
450 struct tun_struct *tun = netdev_priv(dev);
452 switch (tun->flags & TUN_TYPE_MASK) {
453 case TUN_TUN_DEV:
454 dev->netdev_ops = &tun_netdev_ops;
456 /* Point-to-Point TUN Device */
457 dev->hard_header_len = 0;
458 dev->addr_len = 0;
459 dev->mtu = 1500;
461 /* Zero header length */
462 dev->type = ARPHRD_NONE;
463 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
464 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
465 break;
467 case TUN_TAP_DEV:
468 dev->netdev_ops = &tap_netdev_ops;
469 /* Ethernet TAP Device */
470 ether_setup(dev);
472 random_ether_addr(dev->dev_addr);
474 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
475 break;
479 /* Character device part */
481 /* Poll */
482 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
484 struct tun_file *tfile = file->private_data;
485 struct tun_struct *tun = __tun_get(tfile);
486 struct sock *sk;
487 unsigned int mask = 0;
489 if (!tun)
490 return POLLERR;
492 sk = tun->socket.sk;
494 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
496 poll_wait(file, &tun->socket.wait, wait);
498 if (!skb_queue_empty(&sk->sk_receive_queue))
499 mask |= POLLIN | POLLRDNORM;
501 if (sock_writeable(sk) ||
502 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
503 sock_writeable(sk)))
504 mask |= POLLOUT | POLLWRNORM;
506 if (tun->dev->reg_state != NETREG_REGISTERED)
507 mask = POLLERR;
509 tun_put(tun);
510 return mask;
513 /* prepad is the amount to reserve at front. len is length after that.
514 * linear is a hint as to how much to copy (usually headers). */
515 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
516 size_t prepad, size_t len,
517 size_t linear, int noblock)
519 struct sock *sk = tun->socket.sk;
520 struct sk_buff *skb;
521 int err;
523 /* Under a page? Don't bother with paged skb. */
524 if (prepad + len < PAGE_SIZE || !linear)
525 linear = len;
527 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
528 &err);
529 if (!skb)
530 return ERR_PTR(err);
532 skb_reserve(skb, prepad);
533 skb_put(skb, linear);
534 skb->data_len = len - linear;
535 skb->len += len - linear;
537 return skb;
540 /* Get packet from user space buffer */
541 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
542 const struct iovec *iv, size_t count,
543 int noblock)
545 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
546 struct sk_buff *skb;
547 size_t len = count, align = 0;
548 struct virtio_net_hdr gso = { 0 };
549 int offset = 0;
551 if (!(tun->flags & TUN_NO_PI)) {
552 if ((len -= sizeof(pi)) > count)
553 return -EINVAL;
555 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
556 return -EFAULT;
557 offset += sizeof(pi);
560 if (tun->flags & TUN_VNET_HDR) {
561 if ((len -= sizeof(gso)) > count)
562 return -EINVAL;
564 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
565 return -EFAULT;
567 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
568 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
569 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
571 if (gso.hdr_len > len)
572 return -EINVAL;
573 offset += sizeof(gso);
576 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
577 align = NET_IP_ALIGN;
578 if (unlikely(len < ETH_HLEN ||
579 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
580 return -EINVAL;
583 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
584 if (IS_ERR(skb)) {
585 if (PTR_ERR(skb) != -EAGAIN)
586 tun->dev->stats.rx_dropped++;
587 return PTR_ERR(skb);
590 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
591 tun->dev->stats.rx_dropped++;
592 kfree_skb(skb);
593 return -EFAULT;
596 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
597 if (!skb_partial_csum_set(skb, gso.csum_start,
598 gso.csum_offset)) {
599 tun->dev->stats.rx_frame_errors++;
600 kfree_skb(skb);
601 return -EINVAL;
603 } else if (tun->flags & TUN_NOCHECKSUM)
604 skb->ip_summed = CHECKSUM_UNNECESSARY;
606 switch (tun->flags & TUN_TYPE_MASK) {
607 case TUN_TUN_DEV:
608 if (tun->flags & TUN_NO_PI) {
609 switch (skb->data[0] & 0xf0) {
610 case 0x40:
611 pi.proto = htons(ETH_P_IP);
612 break;
613 case 0x60:
614 pi.proto = htons(ETH_P_IPV6);
615 break;
616 default:
617 tun->dev->stats.rx_dropped++;
618 kfree_skb(skb);
619 return -EINVAL;
623 skb_reset_mac_header(skb);
624 skb->protocol = pi.proto;
625 skb->dev = tun->dev;
626 break;
627 case TUN_TAP_DEV:
628 skb->protocol = eth_type_trans(skb, tun->dev);
629 break;
632 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
633 pr_debug("GSO!\n");
634 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
635 case VIRTIO_NET_HDR_GSO_TCPV4:
636 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
637 break;
638 case VIRTIO_NET_HDR_GSO_TCPV6:
639 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
640 break;
641 case VIRTIO_NET_HDR_GSO_UDP:
642 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
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 if (sinfo->gso_type & SKB_GSO_UDP)
730 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
731 else
732 BUG();
733 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
734 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
735 } else
736 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
738 if (skb->ip_summed == CHECKSUM_PARTIAL) {
739 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
740 gso.csum_start = skb->csum_start - skb_headroom(skb);
741 gso.csum_offset = skb->csum_offset;
742 } /* else everything is zero */
744 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
745 sizeof(gso))))
746 return -EFAULT;
747 total += sizeof(gso);
750 len = min_t(int, skb->len, len);
752 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
753 total += len;
755 tun->dev->stats.tx_packets++;
756 tun->dev->stats.tx_bytes += len;
758 return total;
761 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
762 unsigned long count, loff_t pos)
764 struct file *file = iocb->ki_filp;
765 struct tun_file *tfile = file->private_data;
766 struct tun_struct *tun = __tun_get(tfile);
767 DECLARE_WAITQUEUE(wait, current);
768 struct sk_buff *skb;
769 ssize_t len, ret = 0;
771 if (!tun)
772 return -EBADFD;
774 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
776 len = iov_length(iv, count);
777 if (len < 0) {
778 ret = -EINVAL;
779 goto out;
782 add_wait_queue(&tun->socket.wait, &wait);
783 while (len) {
784 current->state = TASK_INTERRUPTIBLE;
786 /* Read frames from the queue */
787 if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
788 if (file->f_flags & O_NONBLOCK) {
789 ret = -EAGAIN;
790 break;
792 if (signal_pending(current)) {
793 ret = -ERESTARTSYS;
794 break;
796 if (tun->dev->reg_state != NETREG_REGISTERED) {
797 ret = -EIO;
798 break;
801 /* Nothing to read, let's sleep */
802 schedule();
803 continue;
805 netif_wake_queue(tun->dev);
807 ret = tun_put_user(tun, skb, iv, len);
808 kfree_skb(skb);
809 break;
812 current->state = TASK_RUNNING;
813 remove_wait_queue(&tun->socket.wait, &wait);
815 out:
816 tun_put(tun);
817 return ret;
820 static void tun_setup(struct net_device *dev)
822 struct tun_struct *tun = netdev_priv(dev);
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 container_of(sk, struct tun_sock, sk)->tun = tun;
991 tun_net_init(dev);
993 if (strchr(dev->name, '%')) {
994 err = dev_alloc_name(dev, dev->name);
995 if (err < 0)
996 goto err_free_sk;
999 err = register_netdevice(tun->dev);
1000 if (err < 0)
1001 goto err_free_sk;
1003 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1004 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1005 device_create_file(&tun->dev->dev, &dev_attr_group))
1006 printk(KERN_ERR "Failed to create tun sysfs files\n");
1008 sk->sk_destruct = tun_sock_destruct;
1010 err = tun_attach(tun, file);
1011 if (err < 0)
1012 goto failed;
1015 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
1017 if (ifr->ifr_flags & IFF_NO_PI)
1018 tun->flags |= TUN_NO_PI;
1019 else
1020 tun->flags &= ~TUN_NO_PI;
1022 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1023 tun->flags |= TUN_ONE_QUEUE;
1024 else
1025 tun->flags &= ~TUN_ONE_QUEUE;
1027 if (ifr->ifr_flags & IFF_VNET_HDR)
1028 tun->flags |= TUN_VNET_HDR;
1029 else
1030 tun->flags &= ~TUN_VNET_HDR;
1032 /* Make sure persistent devices do not get stuck in
1033 * xoff state.
1035 if (netif_running(tun->dev))
1036 netif_wake_queue(tun->dev);
1038 strcpy(ifr->ifr_name, tun->dev->name);
1039 return 0;
1041 err_free_sk:
1042 sock_put(sk);
1043 err_free_dev:
1044 free_netdev(dev);
1045 failed:
1046 return err;
1049 static int tun_get_iff(struct net *net, struct tun_struct *tun,
1050 struct ifreq *ifr)
1052 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1054 strcpy(ifr->ifr_name, tun->dev->name);
1056 ifr->ifr_flags = tun_flags(tun);
1058 return 0;
1061 /* This is like a cut-down ethtool ops, except done via tun fd so no
1062 * privs required. */
1063 static int set_offload(struct net_device *dev, unsigned long arg)
1065 unsigned int old_features, features;
1067 old_features = dev->features;
1068 /* Unset features, set them as we chew on the arg. */
1069 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1070 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6
1071 |NETIF_F_UFO));
1073 if (arg & TUN_F_CSUM) {
1074 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1075 arg &= ~TUN_F_CSUM;
1077 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1078 if (arg & TUN_F_TSO_ECN) {
1079 features |= NETIF_F_TSO_ECN;
1080 arg &= ~TUN_F_TSO_ECN;
1082 if (arg & TUN_F_TSO4)
1083 features |= NETIF_F_TSO;
1084 if (arg & TUN_F_TSO6)
1085 features |= NETIF_F_TSO6;
1086 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1089 if (arg & TUN_F_UFO) {
1090 features |= NETIF_F_UFO;
1091 arg &= ~TUN_F_UFO;
1095 /* This gives the user a way to test for new features in future by
1096 * trying to set them. */
1097 if (arg)
1098 return -EINVAL;
1100 dev->features = features;
1101 if (old_features != dev->features)
1102 netdev_features_change(dev);
1104 return 0;
1107 static long tun_chr_ioctl(struct file *file, unsigned int cmd,
1108 unsigned long arg)
1110 struct tun_file *tfile = file->private_data;
1111 struct tun_struct *tun;
1112 void __user* argp = (void __user*)arg;
1113 struct ifreq ifr;
1114 int sndbuf;
1115 int ret;
1117 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1118 if (copy_from_user(&ifr, argp, sizeof ifr))
1119 return -EFAULT;
1121 if (cmd == TUNGETFEATURES) {
1122 /* Currently this just means: "what IFF flags are valid?".
1123 * This is needed because we never checked for invalid flags on
1124 * TUNSETIFF. */
1125 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1126 IFF_VNET_HDR,
1127 (unsigned int __user*)argp);
1130 rtnl_lock();
1132 tun = __tun_get(tfile);
1133 if (cmd == TUNSETIFF && !tun) {
1134 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1136 ret = tun_set_iff(tfile->net, file, &ifr);
1138 if (ret)
1139 goto unlock;
1141 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1142 ret = -EFAULT;
1143 goto unlock;
1146 ret = -EBADFD;
1147 if (!tun)
1148 goto unlock;
1150 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1152 ret = 0;
1153 switch (cmd) {
1154 case TUNGETIFF:
1155 ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
1156 if (ret)
1157 break;
1159 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1160 ret = -EFAULT;
1161 break;
1163 case TUNSETNOCSUM:
1164 /* Disable/Enable checksum */
1165 if (arg)
1166 tun->flags |= TUN_NOCHECKSUM;
1167 else
1168 tun->flags &= ~TUN_NOCHECKSUM;
1170 DBG(KERN_INFO "%s: checksum %s\n",
1171 tun->dev->name, arg ? "disabled" : "enabled");
1172 break;
1174 case TUNSETPERSIST:
1175 /* Disable/Enable persist mode */
1176 if (arg)
1177 tun->flags |= TUN_PERSIST;
1178 else
1179 tun->flags &= ~TUN_PERSIST;
1181 DBG(KERN_INFO "%s: persist %s\n",
1182 tun->dev->name, arg ? "enabled" : "disabled");
1183 break;
1185 case TUNSETOWNER:
1186 /* Set owner of the device */
1187 tun->owner = (uid_t) arg;
1189 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1190 break;
1192 case TUNSETGROUP:
1193 /* Set group of the device */
1194 tun->group= (gid_t) arg;
1196 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1197 break;
1199 case TUNSETLINK:
1200 /* Only allow setting the type when the interface is down */
1201 if (tun->dev->flags & IFF_UP) {
1202 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1203 tun->dev->name);
1204 ret = -EBUSY;
1205 } else {
1206 tun->dev->type = (int) arg;
1207 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1208 ret = 0;
1210 break;
1212 #ifdef TUN_DEBUG
1213 case TUNSETDEBUG:
1214 tun->debug = arg;
1215 break;
1216 #endif
1217 case TUNSETOFFLOAD:
1218 ret = set_offload(tun->dev, arg);
1219 break;
1221 case TUNSETTXFILTER:
1222 /* Can be set only for TAPs */
1223 ret = -EINVAL;
1224 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1225 break;
1226 ret = update_filter(&tun->txflt, (void __user *)arg);
1227 break;
1229 case SIOCGIFHWADDR:
1230 /* Get hw addres */
1231 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1232 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1233 if (copy_to_user(argp, &ifr, sizeof ifr))
1234 ret = -EFAULT;
1235 break;
1237 case SIOCSIFHWADDR:
1238 /* Set hw address */
1239 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1240 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1242 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1243 break;
1245 case TUNGETSNDBUF:
1246 sndbuf = tun->socket.sk->sk_sndbuf;
1247 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1248 ret = -EFAULT;
1249 break;
1251 case TUNSETSNDBUF:
1252 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1253 ret = -EFAULT;
1254 break;
1257 tun->socket.sk->sk_sndbuf = sndbuf;
1258 break;
1260 default:
1261 ret = -EINVAL;
1262 break;
1265 unlock:
1266 rtnl_unlock();
1267 if (tun)
1268 tun_put(tun);
1269 return ret;
1272 static int tun_chr_fasync(int fd, struct file *file, int on)
1274 struct tun_struct *tun = tun_get(file);
1275 int ret;
1277 if (!tun)
1278 return -EBADFD;
1280 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1282 lock_kernel();
1283 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1284 goto out;
1286 if (on) {
1287 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1288 if (ret)
1289 goto out;
1290 tun->flags |= TUN_FASYNC;
1291 } else
1292 tun->flags &= ~TUN_FASYNC;
1293 ret = 0;
1294 out:
1295 unlock_kernel();
1296 tun_put(tun);
1297 return ret;
1300 static int tun_chr_open(struct inode *inode, struct file * file)
1302 struct tun_file *tfile;
1303 cycle_kernel_lock();
1304 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1306 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1307 if (!tfile)
1308 return -ENOMEM;
1309 atomic_set(&tfile->count, 0);
1310 tfile->tun = NULL;
1311 tfile->net = get_net(current->nsproxy->net_ns);
1312 file->private_data = tfile;
1313 return 0;
1316 static int tun_chr_close(struct inode *inode, struct file *file)
1318 struct tun_file *tfile = file->private_data;
1319 struct tun_struct *tun;
1321 tun = __tun_get(tfile);
1322 if (tun) {
1323 struct net_device *dev = tun->dev;
1325 DBG(KERN_INFO "%s: tun_chr_close\n", dev->name);
1327 __tun_detach(tun);
1329 /* If desireable, unregister the netdevice. */
1330 if (!(tun->flags & TUN_PERSIST)) {
1331 rtnl_lock();
1332 if (dev->reg_state == NETREG_REGISTERED)
1333 unregister_netdevice(dev);
1334 rtnl_unlock();
1338 tun = tfile->tun;
1339 if (tun)
1340 sock_put(tun->socket.sk);
1342 put_net(tfile->net);
1343 kfree(tfile);
1345 return 0;
1348 static const struct file_operations tun_fops = {
1349 .owner = THIS_MODULE,
1350 .llseek = no_llseek,
1351 .read = do_sync_read,
1352 .aio_read = tun_chr_aio_read,
1353 .write = do_sync_write,
1354 .aio_write = tun_chr_aio_write,
1355 .poll = tun_chr_poll,
1356 .unlocked_ioctl = tun_chr_ioctl,
1357 .open = tun_chr_open,
1358 .release = tun_chr_close,
1359 .fasync = tun_chr_fasync
1362 static struct miscdevice tun_miscdev = {
1363 .minor = TUN_MINOR,
1364 .name = "tun",
1365 .devnode = "net/tun",
1366 .fops = &tun_fops,
1369 /* ethtool interface */
1371 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1373 cmd->supported = 0;
1374 cmd->advertising = 0;
1375 cmd->speed = SPEED_10;
1376 cmd->duplex = DUPLEX_FULL;
1377 cmd->port = PORT_TP;
1378 cmd->phy_address = 0;
1379 cmd->transceiver = XCVR_INTERNAL;
1380 cmd->autoneg = AUTONEG_DISABLE;
1381 cmd->maxtxpkt = 0;
1382 cmd->maxrxpkt = 0;
1383 return 0;
1386 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1388 struct tun_struct *tun = netdev_priv(dev);
1390 strcpy(info->driver, DRV_NAME);
1391 strcpy(info->version, DRV_VERSION);
1392 strcpy(info->fw_version, "N/A");
1394 switch (tun->flags & TUN_TYPE_MASK) {
1395 case TUN_TUN_DEV:
1396 strcpy(info->bus_info, "tun");
1397 break;
1398 case TUN_TAP_DEV:
1399 strcpy(info->bus_info, "tap");
1400 break;
1404 static u32 tun_get_msglevel(struct net_device *dev)
1406 #ifdef TUN_DEBUG
1407 struct tun_struct *tun = netdev_priv(dev);
1408 return tun->debug;
1409 #else
1410 return -EOPNOTSUPP;
1411 #endif
1414 static void tun_set_msglevel(struct net_device *dev, u32 value)
1416 #ifdef TUN_DEBUG
1417 struct tun_struct *tun = netdev_priv(dev);
1418 tun->debug = value;
1419 #endif
1422 static u32 tun_get_link(struct net_device *dev)
1424 struct tun_struct *tun = netdev_priv(dev);
1425 return !!tun->tfile;
1428 static u32 tun_get_rx_csum(struct net_device *dev)
1430 struct tun_struct *tun = netdev_priv(dev);
1431 return (tun->flags & TUN_NOCHECKSUM) == 0;
1434 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1436 struct tun_struct *tun = netdev_priv(dev);
1437 if (data)
1438 tun->flags &= ~TUN_NOCHECKSUM;
1439 else
1440 tun->flags |= TUN_NOCHECKSUM;
1441 return 0;
1444 static const struct ethtool_ops tun_ethtool_ops = {
1445 .get_settings = tun_get_settings,
1446 .get_drvinfo = tun_get_drvinfo,
1447 .get_msglevel = tun_get_msglevel,
1448 .set_msglevel = tun_set_msglevel,
1449 .get_link = tun_get_link,
1450 .get_rx_csum = tun_get_rx_csum,
1451 .set_rx_csum = tun_set_rx_csum
1455 static int __init tun_init(void)
1457 int ret = 0;
1459 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1460 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1462 ret = rtnl_link_register(&tun_link_ops);
1463 if (ret) {
1464 printk(KERN_ERR "tun: Can't register link_ops\n");
1465 goto err_linkops;
1468 ret = misc_register(&tun_miscdev);
1469 if (ret) {
1470 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1471 goto err_misc;
1473 return 0;
1474 err_misc:
1475 rtnl_link_unregister(&tun_link_ops);
1476 err_linkops:
1477 return ret;
1480 static void tun_cleanup(void)
1482 misc_deregister(&tun_miscdev);
1483 rtnl_link_unregister(&tun_link_ops);
1486 module_init(tun_init);
1487 module_exit(tun_cleanup);
1488 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1489 MODULE_AUTHOR(DRV_COPYRIGHT);
1490 MODULE_LICENSE("GPL");
1491 MODULE_ALIAS_MISCDEV(TUN_MINOR);