?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sound: oxygen: handle cards with missing EEPROM
[linux-2.6/mini2440.git] / drivers / net / tun.c
blob828929212c18e7a8e750fc3b157395f13e023989
1 /*
2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
4 *
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.
9 *
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.
14 *
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
16 */
17
18 /*
19 * Changes:
20 *
21 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
23 *
24 * Mark Smith <markzzzsmith@yahoo.com.au>
25 * Use random_ether_addr() for tap MAC address.
26 *
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
32 *
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
35 */
36
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>"
41
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>
68
69 #include <asm/system.h>
70 #include <asm/uaccess.h>
71
72 /* Uncomment to enable debugging */
73 /* #define TUN_DEBUG 1 */
74
75 #ifdef TUN_DEBUG
76 static int debug;
77
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
84
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];
90 };
91
92 struct tun_file {
93 atomic_t count;
94 struct tun_struct *tun;
95 struct net *net;
96 };
97
98 struct tun_sock;
99
100 struct tun_struct {
101 struct tun_file *tfile;
102 unsigned int flags;
103 uid_t owner;
104 gid_t group;
105
106 struct sk_buff_head readq;
107
108 struct net_device *dev;
109 struct fasync_struct *fasync;
110
111 struct tap_filter txflt;
112 struct sock *sk;
113 struct socket socket;
114
115 #ifdef TUN_DEBUG
116 int debug;
117 #endif
118 };
119
120 struct tun_sock {
121 struct sock sk;
122 struct tun_struct *tun;
123 };
124
125 static inline struct tun_sock *tun_sk(struct sock *sk)
126 {
127 return container_of(sk, struct tun_sock, sk);
128 }
129
130 static int tun_attach(struct tun_struct *tun, struct file *file)
131 {
132 struct tun_file *tfile = file->private_data;
133 const struct cred *cred = current_cred();
134 int err;
135
136 ASSERT_RTNL();
137
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;
143
144 netif_tx_lock_bh(tun->dev);
145
146 err = -EINVAL;
147 if (tfile->tun)
148 goto out;
149
150 err = -EBUSY;
151 if (tun->tfile)
152 goto out;
153
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);
160
161 out:
162 netif_tx_unlock_bh(tun->dev);
163 return err;
164 }
165
166 static void __tun_detach(struct tun_struct *tun)
167 {
168 /* Detach from net device */
169 netif_tx_lock_bh(tun->dev);
170 tun->tfile = NULL;
171 netif_tx_unlock_bh(tun->dev);
172
173 /* Drop read queue */
174 skb_queue_purge(&tun->readq);
175
176 /* Drop the extra count on the net device */
177 dev_put(tun->dev);
178 }
179
180 static void tun_detach(struct tun_struct *tun)
181 {
182 rtnl_lock();
183 __tun_detach(tun);
184 rtnl_unlock();
185 }
186
187 static struct tun_struct *__tun_get(struct tun_file *tfile)
188 {
189 struct tun_struct *tun = NULL;
190
191 if (atomic_inc_not_zero(&tfile->count))
192 tun = tfile->tun;
193
194 return tun;
195 }
196
197 static struct tun_struct *tun_get(struct file *file)
198 {
199 return __tun_get(file->private_data);
200 }
201
202 static void tun_put(struct tun_struct *tun)
203 {
204 struct tun_file *tfile = tun->tfile;
205
206 if (atomic_dec_and_test(&tfile->count))
207 tun_detach(tfile->tun);
208 }
209
210 /* TAP filterting */
211 static void addr_hash_set(u32 *mask, const u8 *addr)
212 {
213 int n = ether_crc(ETH_ALEN, addr) >> 26;
214 mask[n >> 5] |= (1 << (n & 31));
215 }
216
217 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
218 {
219 int n = ether_crc(ETH_ALEN, addr) >> 26;
220 return mask[n >> 5] & (1 << (n & 31));
221 }
222
223 static int update_filter(struct tap_filter *filter, void __user *arg)
224 {
225 struct { u8 u[ETH_ALEN]; } *addr;
226 struct tun_filter uf;
227 int err, alen, n, nexact;
228
229 if (copy_from_user(&uf, arg, sizeof(uf)))
230 return -EFAULT;
231
232 if (!uf.count) {
233 /* Disabled */
234 filter->count = 0;
235 return 0;
236 }
237
238 alen = ETH_ALEN * uf.count;
239 addr = kmalloc(alen, GFP_KERNEL);
240 if (!addr)
241 return -ENOMEM;
242
243 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
244 err = -EFAULT;
245 goto done;
246 }
247
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();
253
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);
257
258 nexact = n;
259
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;
267 }
268 addr_hash_set(filter->mask, addr[n].u);
269 }
270
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));
275
276 /* Now enable the filter */
277 wmb();
278 filter->count = nexact;
279
280 /* Return the number of exact filters */
281 err = nexact;
282
283 done:
284 kfree(addr);
285 return err;
286 }
287
288 /* Returns: 0 - drop, !=0 - accept */
289 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
290 {
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;
295
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;
300
301 /* Inexact match (multicast only) */
302 if (is_multicast_ether_addr(eh->h_dest))
303 return addr_hash_test(filter->mask, eh->h_dest);
304
305 return 0;
306 }
307
308 /*
309 * Checks whether the packet is accepted or not.
310 * Returns: 0 - drop, !=0 - accept
311 */
312 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
313 {
314 if (!filter->count)
315 return 1;
316
317 return run_filter(filter, skb);
318 }
319
320 /* Network device part of the driver */
321
322 static const struct ethtool_ops tun_ethtool_ops;
323
324 /* Net device detach from fd. */
325 static void tun_net_uninit(struct net_device *dev)
326 {
327 struct tun_struct *tun = netdev_priv(dev);
328 struct tun_file *tfile = tun->tfile;
329
330 /* Inform the methods they need to stop using the dev.
331 */
332 if (tfile) {
333 wake_up_all(&tun->socket.wait);
334 if (atomic_dec_and_test(&tfile->count))
335 __tun_detach(tun);
336 }
337 }
338
339 static void tun_free_netdev(struct net_device *dev)
340 {
341 struct tun_struct *tun = netdev_priv(dev);
342
343 sock_put(tun->sk);
344 }
345
346 /* Net device open. */
347 static int tun_net_open(struct net_device *dev)
348 {
349 netif_start_queue(dev);
350 return 0;
351 }
352
353 /* Net device close. */
354 static int tun_net_close(struct net_device *dev)
355 {
356 netif_stop_queue(dev);
357 return 0;
358 }
359
360 /* Net device start xmit */
361 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
362 {
363 struct tun_struct *tun = netdev_priv(dev);
364
365 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
366
367 /* Drop packet if interface is not attached */
368 if (!tun->tfile)
369 goto drop;
370
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;
376
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);
382
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;
390 }
391 }
392
393 /* Enqueue packet */
394 skb_queue_tail(&tun->readq, skb);
395 dev->trans_start = jiffies;
396
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;
402
403 drop:
404 dev->stats.tx_dropped++;
405 kfree_skb(skb);
406 return 0;
407 }
408
409 static void tun_net_mclist(struct net_device *dev)
410 {
411 /*
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.
415 */
416 return;
417 }
418
419 #define MIN_MTU 68
420 #define MAX_MTU 65535
421
422 static int
423 tun_net_change_mtu(struct net_device *dev, int new_mtu)
424 {
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;
429 }
430
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,
437 };
438
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,
448 };
449
450 /* Initialize net device. */
451 static void tun_net_init(struct net_device *dev)
452 {
453 struct tun_struct *tun = netdev_priv(dev);
454
455 switch (tun->flags & TUN_TYPE_MASK) {
456 case TUN_TUN_DEV:
457 dev->netdev_ops = &tun_netdev_ops;
458
459 /* Point-to-Point TUN Device */
460 dev->hard_header_len = 0;
461 dev->addr_len = 0;
462 dev->mtu = 1500;
463
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;
469
470 case TUN_TAP_DEV:
471 dev->netdev_ops = &tap_netdev_ops;
472 /* Ethernet TAP Device */
473 ether_setup(dev);
474
475 random_ether_addr(dev->dev_addr);
476
477 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
478 break;
479 }
480 }
481
482 /* Character device part */
483
484 /* Poll */
485 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
486 {
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;
491
492 if (!tun)
493 return POLLERR;
494
495 sk = tun->sk;
496
497 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
498
499 poll_wait(file, &tun->socket.wait, wait);
500
501 if (!skb_queue_empty(&tun->readq))
502 mask |= POLLIN | POLLRDNORM;
503
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;
508
509 if (tun->dev->reg_state != NETREG_REGISTERED)
510 mask = POLLERR;
511
512 tun_put(tun);
513 return mask;
514 }
515
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)
521 {
522 struct sock *sk = tun->sk;
523 struct sk_buff *skb;
524 int err;
525
526 /* Under a page? Don't bother with paged skb. */
527 if (prepad + len < PAGE_SIZE || !linear)
528 linear = len;
529
530 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
531 &err);
532 if (!skb)
533 return ERR_PTR(err);
534
535 skb_reserve(skb, prepad);
536 skb_put(skb, linear);
537 skb->data_len = len - linear;
538 skb->len += len - linear;
539
540 return skb;
541 }
542
543 /* Get packet from user space buffer */
544 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
545 struct iovec *iv, size_t count,
546 int noblock)
547 {
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
553 if (!(tun->flags & TUN_NO_PI)) {
554 if ((len -= sizeof(pi)) > count)
555 return -EINVAL;
556
557 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
558 return -EFAULT;
559 }
560
561 if (tun->flags & TUN_VNET_HDR) {
562 if ((len -= sizeof(gso)) > count)
563 return -EINVAL;
564
565 if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso)))
566 return -EFAULT;
567
568 if (gso.hdr_len > len)
569 return -EINVAL;
570 }
571
572 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
573 align = NET_IP_ALIGN;
574 if (unlikely(len < ETH_HLEN ||
575 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
576 return -EINVAL;
577 }
578
579 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
580 if (IS_ERR(skb)) {
581 if (PTR_ERR(skb) != -EAGAIN)
582 tun->dev->stats.rx_dropped++;
583 return PTR_ERR(skb);
584 }
585
586 if (skb_copy_datagram_from_iovec(skb, 0, iv, len)) {
587 tun->dev->stats.rx_dropped++;
588 kfree_skb(skb);
589 return -EFAULT;
590 }
591
592 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
593 if (!skb_partial_csum_set(skb, gso.csum_start,
594 gso.csum_offset)) {
595 tun->dev->stats.rx_frame_errors++;
596 kfree_skb(skb);
597 return -EINVAL;
598 }
599 } else if (tun->flags & TUN_NOCHECKSUM)
600 skb->ip_summed = CHECKSUM_UNNECESSARY;
601
602 switch (tun->flags & TUN_TYPE_MASK) {
603 case TUN_TUN_DEV:
604 if (tun->flags & TUN_NO_PI) {
605 switch (skb->data[0] & 0xf0) {
606 case 0x40:
607 pi.proto = htons(ETH_P_IP);
608 break;
609 case 0x60:
610 pi.proto = htons(ETH_P_IPV6);
611 break;
612 default:
613 tun->dev->stats.rx_dropped++;
614 kfree_skb(skb);
615 return -EINVAL;
616 }
617 }
618
619 skb_reset_mac_header(skb);
620 skb->protocol = pi.proto;
621 skb->dev = tun->dev;
622 break;
623 case TUN_TAP_DEV:
624 skb->protocol = eth_type_trans(skb, tun->dev);
625 break;
626 };
627
628 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
629 pr_debug("GSO!\n");
630 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
631 case VIRTIO_NET_HDR_GSO_TCPV4:
632 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
633 break;
634 case VIRTIO_NET_HDR_GSO_TCPV6:
635 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
636 break;
637 default:
638 tun->dev->stats.rx_frame_errors++;
639 kfree_skb(skb);
640 return -EINVAL;
641 }
642
643 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
644 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
645
646 skb_shinfo(skb)->gso_size = gso.gso_size;
647 if (skb_shinfo(skb)->gso_size == 0) {
648 tun->dev->stats.rx_frame_errors++;
649 kfree_skb(skb);
650 return -EINVAL;
651 }
652
653 /* Header must be checked, and gso_segs computed. */
654 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
655 skb_shinfo(skb)->gso_segs = 0;
656 }
657
658 netif_rx_ni(skb);
659
660 tun->dev->stats.rx_packets++;
661 tun->dev->stats.rx_bytes += len;
662
663 return count;
664 }
665
666 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
667 unsigned long count, loff_t pos)
668 {
669 struct file *file = iocb->ki_filp;
670 struct tun_struct *tun = tun_get(file);
671 ssize_t result;
672
673 if (!tun)
674 return -EBADFD;
675
676 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
677
678 result = tun_get_user(tun, (struct iovec *)iv, iov_length(iv, count),
679 file->f_flags & O_NONBLOCK);
680
681 tun_put(tun);
682 return result;
683 }
684
685 /* Put packet to the user space buffer */
686 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
687 struct sk_buff *skb,
688 struct iovec *iv, int len)
689 {
690 struct tun_pi pi = { 0, skb->protocol };
691 ssize_t total = 0;
692
693 if (!(tun->flags & TUN_NO_PI)) {
694 if ((len -= sizeof(pi)) < 0)
695 return -EINVAL;
696
697 if (len < skb->len) {
698 /* Packet will be striped */
699 pi.flags |= TUN_PKT_STRIP;
700 }
701
702 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
703 return -EFAULT;
704 total += sizeof(pi);
705 }
706
707 if (tun->flags & TUN_VNET_HDR) {
708 struct virtio_net_hdr gso = { 0 }; /* no info leak */
709 if ((len -= sizeof(gso)) < 0)
710 return -EINVAL;
711
712 if (skb_is_gso(skb)) {
713 struct skb_shared_info *sinfo = skb_shinfo(skb);
714
715 /* This is a hint as to how much should be linear. */
716 gso.hdr_len = skb_headlen(skb);
717 gso.gso_size = sinfo->gso_size;
718 if (sinfo->gso_type & SKB_GSO_TCPV4)
719 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
720 else if (sinfo->gso_type & SKB_GSO_TCPV6)
721 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
722 else
723 BUG();
724 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
725 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
726 } else
727 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
728
729 if (skb->ip_summed == CHECKSUM_PARTIAL) {
730 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
731 gso.csum_start = skb->csum_start - skb_headroom(skb);
732 gso.csum_offset = skb->csum_offset;
733 } /* else everything is zero */
734
735 if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso))))
736 return -EFAULT;
737 total += sizeof(gso);
738 }
739
740 len = min_t(int, skb->len, len);
741
742 skb_copy_datagram_iovec(skb, 0, iv, len);
743 total += len;
744
745 tun->dev->stats.tx_packets++;
746 tun->dev->stats.tx_bytes += len;
747
748 return total;
749 }
750
751 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
752 unsigned long count, loff_t pos)
753 {
754 struct file *file = iocb->ki_filp;
755 struct tun_file *tfile = file->private_data;
756 struct tun_struct *tun = __tun_get(tfile);
757 DECLARE_WAITQUEUE(wait, current);
758 struct sk_buff *skb;
759 ssize_t len, ret = 0;
760
761 if (!tun)
762 return -EBADFD;
763
764 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
765
766 len = iov_length(iv, count);
767 if (len < 0) {
768 ret = -EINVAL;
769 goto out;
770 }
771
772 add_wait_queue(&tun->socket.wait, &wait);
773 while (len) {
774 current->state = TASK_INTERRUPTIBLE;
775
776 /* Read frames from the queue */
777 if (!(skb=skb_dequeue(&tun->readq))) {
778 if (file->f_flags & O_NONBLOCK) {
779 ret = -EAGAIN;
780 break;
781 }
782 if (signal_pending(current)) {
783 ret = -ERESTARTSYS;
784 break;
785 }
786 if (tun->dev->reg_state != NETREG_REGISTERED) {
787 ret = -EIO;
788 break;
789 }
790
791 /* Nothing to read, let's sleep */
792 schedule();
793 continue;
794 }
795 netif_wake_queue(tun->dev);
796
797 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
798 kfree_skb(skb);
799 break;
800 }
801
802 current->state = TASK_RUNNING;
803 remove_wait_queue(&tun->socket.wait, &wait);
804
805 out:
806 tun_put(tun);
807 return ret;
808 }
809
810 static void tun_setup(struct net_device *dev)
811 {
812 struct tun_struct *tun = netdev_priv(dev);
813
814 skb_queue_head_init(&tun->readq);
815
816 tun->owner = -1;
817 tun->group = -1;
818
819 dev->ethtool_ops = &tun_ethtool_ops;
820 dev->destructor = tun_free_netdev;
821 }
822
823 /* Trivial set of netlink ops to allow deleting tun or tap
824 * device with netlink.
825 */
826 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
827 {
828 return -EINVAL;
829 }
830
831 static struct rtnl_link_ops tun_link_ops __read_mostly = {
832 .kind = DRV_NAME,
833 .priv_size = sizeof(struct tun_struct),
834 .setup = tun_setup,
835 .validate = tun_validate,
836 };
837
838 static void tun_sock_write_space(struct sock *sk)
839 {
840 struct tun_struct *tun;
841
842 if (!sock_writeable(sk))
843 return;
844
845 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
846 wake_up_interruptible_sync(sk->sk_sleep);
847
848 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
849 return;
850
851 tun = container_of(sk, struct tun_sock, sk)->tun;
852 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
853 }
854
855 static void tun_sock_destruct(struct sock *sk)
856 {
857 free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
858 }
859
860 static struct proto tun_proto = {
861 .name = "tun",
862 .owner = THIS_MODULE,
863 .obj_size = sizeof(struct tun_sock),
864 };
865
866 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
867 {
868 struct sock *sk;
869 struct tun_struct *tun;
870 struct net_device *dev;
871 int err;
872
873 dev = __dev_get_by_name(net, ifr->ifr_name);
874 if (dev) {
875 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
876 tun = netdev_priv(dev);
877 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
878 tun = netdev_priv(dev);
879 else
880 return -EINVAL;
881
882 err = tun_attach(tun, file);
883 if (err < 0)
884 return err;
885 }
886 else {
887 char *name;
888 unsigned long flags = 0;
889
890 err = -EINVAL;
891
892 if (!capable(CAP_NET_ADMIN))
893 return -EPERM;
894
895 /* Set dev type */
896 if (ifr->ifr_flags & IFF_TUN) {
897 /* TUN device */
898 flags |= TUN_TUN_DEV;
899 name = "tun%d";
900 } else if (ifr->ifr_flags & IFF_TAP) {
901 /* TAP device */
902 flags |= TUN_TAP_DEV;
903 name = "tap%d";
904 } else
905 goto failed;
906
907 if (*ifr->ifr_name)
908 name = ifr->ifr_name;
909
910 dev = alloc_netdev(sizeof(struct tun_struct), name,
911 tun_setup);
912 if (!dev)
913 return -ENOMEM;
914
915 dev_net_set(dev, net);
916 dev->rtnl_link_ops = &tun_link_ops;
917
918 tun = netdev_priv(dev);
919 tun->dev = dev;
920 tun->flags = flags;
921 tun->txflt.count = 0;
922
923 err = -ENOMEM;
924 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
925 if (!sk)
926 goto err_free_dev;
927
928 init_waitqueue_head(&tun->socket.wait);
929 sock_init_data(&tun->socket, sk);
930 sk->sk_write_space = tun_sock_write_space;
931 sk->sk_sndbuf = INT_MAX;
932
933 tun->sk = sk;
934 container_of(sk, struct tun_sock, sk)->tun = tun;
935
936 tun_net_init(dev);
937
938 if (strchr(dev->name, '%')) {
939 err = dev_alloc_name(dev, dev->name);
940 if (err < 0)
941 goto err_free_sk;
942 }
943
944 err = -EINVAL;
945 err = register_netdevice(tun->dev);
946 if (err < 0)
947 goto err_free_sk;
948
949 sk->sk_destruct = tun_sock_destruct;
950
951 err = tun_attach(tun, file);
952 if (err < 0)
953 goto failed;
954 }
955
956 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
957
958 if (ifr->ifr_flags & IFF_NO_PI)
959 tun->flags |= TUN_NO_PI;
960 else
961 tun->flags &= ~TUN_NO_PI;
962
963 if (ifr->ifr_flags & IFF_ONE_QUEUE)
964 tun->flags |= TUN_ONE_QUEUE;
965 else
966 tun->flags &= ~TUN_ONE_QUEUE;
967
968 if (ifr->ifr_flags & IFF_VNET_HDR)
969 tun->flags |= TUN_VNET_HDR;
970 else
971 tun->flags &= ~TUN_VNET_HDR;
972
973 /* Make sure persistent devices do not get stuck in
974 * xoff state.
975 */
976 if (netif_running(tun->dev))
977 netif_wake_queue(tun->dev);
978
979 strcpy(ifr->ifr_name, tun->dev->name);
980 return 0;
981
982 err_free_sk:
983 sock_put(sk);
984 err_free_dev:
985 free_netdev(dev);
986 failed:
987 return err;
988 }
989
990 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
991 {
992 struct tun_struct *tun = tun_get(file);
993
994 if (!tun)
995 return -EBADFD;
996
997 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
998
999 strcpy(ifr->ifr_name, tun->dev->name);
1000
1001 ifr->ifr_flags = 0;
1002
1003 if (ifr->ifr_flags & TUN_TUN_DEV)
1004 ifr->ifr_flags |= IFF_TUN;
1005 else
1006 ifr->ifr_flags |= IFF_TAP;
1007
1008 if (tun->flags & TUN_NO_PI)
1009 ifr->ifr_flags |= IFF_NO_PI;
1010
1011 if (tun->flags & TUN_ONE_QUEUE)
1012 ifr->ifr_flags |= IFF_ONE_QUEUE;
1013
1014 if (tun->flags & TUN_VNET_HDR)
1015 ifr->ifr_flags |= IFF_VNET_HDR;
1016
1017 tun_put(tun);
1018 return 0;
1019 }
1020
1021 /* This is like a cut-down ethtool ops, except done via tun fd so no
1022 * privs required. */
1023 static int set_offload(struct net_device *dev, unsigned long arg)
1024 {
1025 unsigned int old_features, features;
1026
1027 old_features = dev->features;
1028 /* Unset features, set them as we chew on the arg. */
1029 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1030 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
1031
1032 if (arg & TUN_F_CSUM) {
1033 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1034 arg &= ~TUN_F_CSUM;
1035
1036 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1037 if (arg & TUN_F_TSO_ECN) {
1038 features |= NETIF_F_TSO_ECN;
1039 arg &= ~TUN_F_TSO_ECN;
1040 }
1041 if (arg & TUN_F_TSO4)
1042 features |= NETIF_F_TSO;
1043 if (arg & TUN_F_TSO6)
1044 features |= NETIF_F_TSO6;
1045 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1046 }
1047 }
1048
1049 /* This gives the user a way to test for new features in future by
1050 * trying to set them. */
1051 if (arg)
1052 return -EINVAL;
1053
1054 dev->features = features;
1055 if (old_features != dev->features)
1056 netdev_features_change(dev);
1057
1058 return 0;
1059 }
1060
1061 static int tun_chr_ioctl(struct inode *inode, struct file *file,
1062 unsigned int cmd, unsigned long arg)
1063 {
1064 struct tun_file *tfile = file->private_data;
1065 struct tun_struct *tun;
1066 void __user* argp = (void __user*)arg;
1067 struct ifreq ifr;
1068 int sndbuf;
1069 int ret;
1070
1071 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1072 if (copy_from_user(&ifr, argp, sizeof ifr))
1073 return -EFAULT;
1074
1075 if (cmd == TUNGETFEATURES) {
1076 /* Currently this just means: "what IFF flags are valid?".
1077 * This is needed because we never checked for invalid flags on
1078 * TUNSETIFF. */
1079 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1080 IFF_VNET_HDR,
1081 (unsigned int __user*)argp);
1082 }
1083
1084 tun = __tun_get(tfile);
1085 if (cmd == TUNSETIFF && !tun) {
1086 int err;
1087
1088 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1089
1090 rtnl_lock();
1091 err = tun_set_iff(tfile->net, file, &ifr);
1092 rtnl_unlock();
1093
1094 if (err)
1095 return err;
1096
1097 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1098 return -EFAULT;
1099 return 0;
1100 }
1101
1102
1103 if (!tun)
1104 return -EBADFD;
1105
1106 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1107
1108 ret = 0;
1109 switch (cmd) {
1110 case TUNGETIFF:
1111 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
1112 if (ret)
1113 break;
1114
1115 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1116 ret = -EFAULT;
1117 break;
1118
1119 case TUNSETNOCSUM:
1120 /* Disable/Enable checksum */
1121 if (arg)
1122 tun->flags |= TUN_NOCHECKSUM;
1123 else
1124 tun->flags &= ~TUN_NOCHECKSUM;
1125
1126 DBG(KERN_INFO "%s: checksum %s\n",
1127 tun->dev->name, arg ? "disabled" : "enabled");
1128 break;
1129
1130 case TUNSETPERSIST:
1131 /* Disable/Enable persist mode */
1132 if (arg)
1133 tun->flags |= TUN_PERSIST;
1134 else
1135 tun->flags &= ~TUN_PERSIST;
1136
1137 DBG(KERN_INFO "%s: persist %s\n",
1138 tun->dev->name, arg ? "enabled" : "disabled");
1139 break;
1140
1141 case TUNSETOWNER:
1142 /* Set owner of the device */
1143 tun->owner = (uid_t) arg;
1144
1145 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1146 break;
1147
1148 case TUNSETGROUP:
1149 /* Set group of the device */
1150 tun->group= (gid_t) arg;
1151
1152 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1153 break;
1154
1155 case TUNSETLINK:
1156 /* Only allow setting the type when the interface is down */
1157 rtnl_lock();
1158 if (tun->dev->flags & IFF_UP) {
1159 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1160 tun->dev->name);
1161 ret = -EBUSY;
1162 } else {
1163 tun->dev->type = (int) arg;
1164 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1165 ret = 0;
1166 }
1167 rtnl_unlock();
1168 break;
1169
1170 #ifdef TUN_DEBUG
1171 case TUNSETDEBUG:
1172 tun->debug = arg;
1173 break;
1174 #endif
1175 case TUNSETOFFLOAD:
1176 rtnl_lock();
1177 ret = set_offload(tun->dev, arg);
1178 rtnl_unlock();
1179 break;
1180
1181 case TUNSETTXFILTER:
1182 /* Can be set only for TAPs */
1183 ret = -EINVAL;
1184 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1185 break;
1186 rtnl_lock();
1187 ret = update_filter(&tun->txflt, (void __user *)arg);
1188 rtnl_unlock();
1189 break;
1190
1191 case SIOCGIFHWADDR:
1192 /* Get hw addres */
1193 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1194 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1195 if (copy_to_user(argp, &ifr, sizeof ifr))
1196 ret = -EFAULT;
1197 break;
1198
1199 case SIOCSIFHWADDR:
1200 /* Set hw address */
1201 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1202 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1203
1204 rtnl_lock();
1205 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1206 rtnl_unlock();
1207 break;
1208
1209 case TUNGETSNDBUF:
1210 sndbuf = tun->sk->sk_sndbuf;
1211 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1212 ret = -EFAULT;
1213 break;
1214
1215 case TUNSETSNDBUF:
1216 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1217 ret = -EFAULT;
1218 break;
1219 }
1220
1221 tun->sk->sk_sndbuf = sndbuf;
1222 break;
1223
1224 default:
1225 ret = -EINVAL;
1226 break;
1227 };
1228
1229 tun_put(tun);
1230 return ret;
1231 }
1232
1233 static int tun_chr_fasync(int fd, struct file *file, int on)
1234 {
1235 struct tun_struct *tun = tun_get(file);
1236 int ret;
1237
1238 if (!tun)
1239 return -EBADFD;
1240
1241 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1242
1243 lock_kernel();
1244 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1245 goto out;
1246
1247 if (on) {
1248 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1249 if (ret)
1250 goto out;
1251 tun->flags |= TUN_FASYNC;
1252 } else
1253 tun->flags &= ~TUN_FASYNC;
1254 ret = 0;
1255 out:
1256 unlock_kernel();
1257 tun_put(tun);
1258 return ret;
1259 }
1260
1261 static int tun_chr_open(struct inode *inode, struct file * file)
1262 {
1263 struct tun_file *tfile;
1264 cycle_kernel_lock();
1265 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1266
1267 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1268 if (!tfile)
1269 return -ENOMEM;
1270 atomic_set(&tfile->count, 0);
1271 tfile->tun = NULL;
1272 tfile->net = get_net(current->nsproxy->net_ns);
1273 file->private_data = tfile;
1274 return 0;
1275 }
1276
1277 static int tun_chr_close(struct inode *inode, struct file *file)
1278 {
1279 struct tun_file *tfile = file->private_data;
1280 struct tun_struct *tun;
1281
1282
1283 rtnl_lock();
1284 tun = __tun_get(tfile);
1285 if (tun) {
1286 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1287
1288 __tun_detach(tun);
1289
1290 /* If desireable, unregister the netdevice. */
1291 if (!(tun->flags & TUN_PERSIST))
1292 unregister_netdevice(tun->dev);
1293
1294 }
1295 rtnl_unlock();
1296
1297 tun = tfile->tun;
1298 if (tun)
1299 sock_put(tun->sk);
1300
1301 put_net(tfile->net);
1302 kfree(tfile);
1303
1304 return 0;
1305 }
1306
1307 static const struct file_operations tun_fops = {
1308 .owner = THIS_MODULE,
1309 .llseek = no_llseek,
1310 .read = do_sync_read,
1311 .aio_read = tun_chr_aio_read,
1312 .write = do_sync_write,
1313 .aio_write = tun_chr_aio_write,
1314 .poll = tun_chr_poll,
1315 .ioctl = tun_chr_ioctl,
1316 .open = tun_chr_open,
1317 .release = tun_chr_close,
1318 .fasync = tun_chr_fasync
1319 };
1320
1321 static struct miscdevice tun_miscdev = {
1322 .minor = TUN_MINOR,
1323 .name = "tun",
1324 .fops = &tun_fops,
1325 };
1326
1327 /* ethtool interface */
1328
1329 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1330 {
1331 cmd->supported = 0;
1332 cmd->advertising = 0;
1333 cmd->speed = SPEED_10;
1334 cmd->duplex = DUPLEX_FULL;
1335 cmd->port = PORT_TP;
1336 cmd->phy_address = 0;
1337 cmd->transceiver = XCVR_INTERNAL;
1338 cmd->autoneg = AUTONEG_DISABLE;
1339 cmd->maxtxpkt = 0;
1340 cmd->maxrxpkt = 0;
1341 return 0;
1342 }
1343
1344 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1345 {
1346 struct tun_struct *tun = netdev_priv(dev);
1347
1348 strcpy(info->driver, DRV_NAME);
1349 strcpy(info->version, DRV_VERSION);
1350 strcpy(info->fw_version, "N/A");
1351
1352 switch (tun->flags & TUN_TYPE_MASK) {
1353 case TUN_TUN_DEV:
1354 strcpy(info->bus_info, "tun");
1355 break;
1356 case TUN_TAP_DEV:
1357 strcpy(info->bus_info, "tap");
1358 break;
1359 }
1360 }
1361
1362 static u32 tun_get_msglevel(struct net_device *dev)
1363 {
1364 #ifdef TUN_DEBUG
1365 struct tun_struct *tun = netdev_priv(dev);
1366 return tun->debug;
1367 #else
1368 return -EOPNOTSUPP;
1369 #endif
1370 }
1371
1372 static void tun_set_msglevel(struct net_device *dev, u32 value)
1373 {
1374 #ifdef TUN_DEBUG
1375 struct tun_struct *tun = netdev_priv(dev);
1376 tun->debug = value;
1377 #endif
1378 }
1379
1380 static u32 tun_get_link(struct net_device *dev)
1381 {
1382 struct tun_struct *tun = netdev_priv(dev);
1383 return !!tun->tfile;
1384 }
1385
1386 static u32 tun_get_rx_csum(struct net_device *dev)
1387 {
1388 struct tun_struct *tun = netdev_priv(dev);
1389 return (tun->flags & TUN_NOCHECKSUM) == 0;
1390 }
1391
1392 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1393 {
1394 struct tun_struct *tun = netdev_priv(dev);
1395 if (data)
1396 tun->flags &= ~TUN_NOCHECKSUM;
1397 else
1398 tun->flags |= TUN_NOCHECKSUM;
1399 return 0;
1400 }
1401
1402 static const struct ethtool_ops tun_ethtool_ops = {
1403 .get_settings = tun_get_settings,
1404 .get_drvinfo = tun_get_drvinfo,
1405 .get_msglevel = tun_get_msglevel,
1406 .set_msglevel = tun_set_msglevel,
1407 .get_link = tun_get_link,
1408 .get_rx_csum = tun_get_rx_csum,
1409 .set_rx_csum = tun_set_rx_csum
1410 };
1411
1412
1413 static int __init tun_init(void)
1414 {
1415 int ret = 0;
1416
1417 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1418 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1419
1420 ret = rtnl_link_register(&tun_link_ops);
1421 if (ret) {
1422 printk(KERN_ERR "tun: Can't register link_ops\n");
1423 goto err_linkops;
1424 }
1425
1426 ret = misc_register(&tun_miscdev);
1427 if (ret) {
1428 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1429 goto err_misc;
1430 }
1431 return 0;
1432 err_misc:
1433 rtnl_link_unregister(&tun_link_ops);
1434 err_linkops:
1435 return ret;
1436 }
1437
1438 static void tun_cleanup(void)
1439 {
1440 misc_deregister(&tun_miscdev);
1441 rtnl_link_unregister(&tun_link_ops);
1442 }
1443
1444 module_init(tun_init);
1445 module_exit(tun_cleanup);
1446 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1447 MODULE_AUTHOR(DRV_COPYRIGHT);
1448 MODULE_LICENSE("GPL");
1449 MODULE_ALIAS_MISCDEV(TUN_MINOR);
Support for the Chinese Samsung S3C2440 based development boards