vlan.c revision 48752e1b1802231ef2a076f34d861918b7d571c3
1/*
2 * INET		802.1Q VLAN
3 *		Ethernet-type device handling.
4 *
5 * Authors:	Ben Greear <greearb@candelatech.com>
6 *              Please send support related email to: netdev@vger.kernel.org
7 *              VLAN Home Page: http://www.candelatech.com/~greear/vlan.html
8 *
9 * Fixes:
10 *              Fix for packet capture - Nick Eggleston <nick@dccinc.com>;
11 *		Add HW acceleration hooks - David S. Miller <davem@redhat.com>;
12 *		Correct all the locking - David S. Miller <davem@redhat.com>;
13 *		Use hash table for VLAN groups - David S. Miller <davem@redhat.com>
14 *
15 *		This program is free software; you can redistribute it and/or
16 *		modify it under the terms of the GNU General Public License
17 *		as published by the Free Software Foundation; either version
18 *		2 of the License, or (at your option) any later version.
19 */
20
21#include <linux/capability.h>
22#include <linux/module.h>
23#include <linux/netdevice.h>
24#include <linux/skbuff.h>
25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/rculist.h>
28#include <net/p8022.h>
29#include <net/arp.h>
30#include <linux/rtnetlink.h>
31#include <linux/notifier.h>
32#include <net/rtnetlink.h>
33#include <net/net_namespace.h>
34#include <net/netns/generic.h>
35#include <asm/uaccess.h>
36
37#include <linux/if_vlan.h>
38#include "vlan.h"
39#include "vlanproc.h"
40
41#define DRV_VERSION "1.8"
42
43/* Global VLAN variables */
44
45int vlan_net_id __read_mostly;
46
47const char vlan_fullname[] = "802.1Q VLAN Support";
48const char vlan_version[] = DRV_VERSION;
49static const char vlan_copyright[] = "Ben Greear <greearb@candelatech.com>";
50static const char vlan_buggyright[] = "David S. Miller <davem@redhat.com>";
51
52/* End of global variables definitions. */
53
54static void vlan_group_free(struct vlan_group *grp)
55{
56	int i;
57
58	for (i = 0; i < VLAN_GROUP_ARRAY_SPLIT_PARTS; i++)
59		kfree(grp->vlan_devices_arrays[i]);
60	kfree(grp);
61}
62
63static struct vlan_group *vlan_group_alloc(struct net_device *real_dev)
64{
65	struct vlan_group *grp;
66
67	grp = kzalloc(sizeof(struct vlan_group), GFP_KERNEL);
68	if (!grp)
69		return NULL;
70
71	grp->real_dev = real_dev;
72	return grp;
73}
74
75static int vlan_group_prealloc_vid(struct vlan_group *vg, u16 vlan_id)
76{
77	struct net_device **array;
78	unsigned int size;
79
80	ASSERT_RTNL();
81
82	array = vg->vlan_devices_arrays[vlan_id / VLAN_GROUP_ARRAY_PART_LEN];
83	if (array != NULL)
84		return 0;
85
86	size = sizeof(struct net_device *) * VLAN_GROUP_ARRAY_PART_LEN;
87	array = kzalloc(size, GFP_KERNEL);
88	if (array == NULL)
89		return -ENOBUFS;
90
91	vg->vlan_devices_arrays[vlan_id / VLAN_GROUP_ARRAY_PART_LEN] = array;
92	return 0;
93}
94
95static void vlan_rcu_free(struct rcu_head *rcu)
96{
97	vlan_group_free(container_of(rcu, struct vlan_group, rcu));
98}
99
100void unregister_vlan_dev(struct net_device *dev, struct list_head *head)
101{
102	struct vlan_dev_info *vlan = vlan_dev_info(dev);
103	struct net_device *real_dev = vlan->real_dev;
104	const struct net_device_ops *ops = real_dev->netdev_ops;
105	struct vlan_group *grp;
106	u16 vlan_id = vlan->vlan_id;
107
108	ASSERT_RTNL();
109
110	grp = rtnl_dereference(real_dev->vlgrp);
111	BUG_ON(!grp);
112
113	/* Take it out of our own structures, but be sure to interlock with
114	 * HW accelerating devices or SW vlan input packet processing if
115	 * VLAN is not 0 (leave it there for 802.1p).
116	 */
117	if (vlan_id && (real_dev->features & NETIF_F_HW_VLAN_FILTER))
118		ops->ndo_vlan_rx_kill_vid(real_dev, vlan_id);
119
120	grp->nr_vlans--;
121
122	vlan_group_set_device(grp, vlan_id, NULL);
123	/* Because unregister_netdevice_queue() makes sure at least one rcu
124	 * grace period is respected before device freeing,
125	 * we dont need to call synchronize_net() here.
126	 */
127	unregister_netdevice_queue(dev, head);
128
129	/* If the group is now empty, kill off the group. */
130	if (grp->nr_vlans == 0) {
131		vlan_gvrp_uninit_applicant(real_dev);
132
133		rcu_assign_pointer(real_dev->vlgrp, NULL);
134		if (ops->ndo_vlan_rx_register)
135			ops->ndo_vlan_rx_register(real_dev, NULL);
136
137		/* Free the group, after all cpu's are done. */
138		call_rcu(&grp->rcu, vlan_rcu_free);
139	}
140
141	/* Get rid of the vlan's reference to real_dev */
142	dev_put(real_dev);
143}
144
145int vlan_check_real_dev(struct net_device *real_dev, u16 vlan_id)
146{
147	const char *name = real_dev->name;
148	const struct net_device_ops *ops = real_dev->netdev_ops;
149
150	if (real_dev->features & NETIF_F_VLAN_CHALLENGED) {
151		pr_info("8021q: VLANs not supported on %s\n", name);
152		return -EOPNOTSUPP;
153	}
154
155	if ((real_dev->features & NETIF_F_HW_VLAN_FILTER) &&
156	    (!ops->ndo_vlan_rx_add_vid || !ops->ndo_vlan_rx_kill_vid)) {
157		pr_info("8021q: Device %s has buggy VLAN hw accel\n", name);
158		return -EOPNOTSUPP;
159	}
160
161	if (vlan_find_dev(real_dev, vlan_id) != NULL)
162		return -EEXIST;
163
164	return 0;
165}
166
167int register_vlan_dev(struct net_device *dev)
168{
169	struct vlan_dev_info *vlan = vlan_dev_info(dev);
170	struct net_device *real_dev = vlan->real_dev;
171	const struct net_device_ops *ops = real_dev->netdev_ops;
172	u16 vlan_id = vlan->vlan_id;
173	struct vlan_group *grp, *ngrp = NULL;
174	int err;
175
176	grp = rtnl_dereference(real_dev->vlgrp);
177	if (!grp) {
178		ngrp = grp = vlan_group_alloc(real_dev);
179		if (!grp)
180			return -ENOBUFS;
181		err = vlan_gvrp_init_applicant(real_dev);
182		if (err < 0)
183			goto out_free_group;
184	}
185
186	err = vlan_group_prealloc_vid(grp, vlan_id);
187	if (err < 0)
188		goto out_uninit_applicant;
189
190	err = register_netdevice(dev);
191	if (err < 0)
192		goto out_uninit_applicant;
193
194	/* Account for reference in struct vlan_dev_info */
195	dev_hold(real_dev);
196
197	netif_stacked_transfer_operstate(real_dev, dev);
198	linkwatch_fire_event(dev); /* _MUST_ call rfc2863_policy() */
199
200	/* So, got the sucker initialized, now lets place
201	 * it into our local structure.
202	 */
203	vlan_group_set_device(grp, vlan_id, dev);
204	grp->nr_vlans++;
205
206	if (ngrp) {
207		if (ops->ndo_vlan_rx_register)
208			ops->ndo_vlan_rx_register(real_dev, ngrp);
209		rcu_assign_pointer(real_dev->vlgrp, ngrp);
210	}
211	if (real_dev->features & NETIF_F_HW_VLAN_FILTER)
212		ops->ndo_vlan_rx_add_vid(real_dev, vlan_id);
213
214	return 0;
215
216out_uninit_applicant:
217	if (ngrp)
218		vlan_gvrp_uninit_applicant(real_dev);
219out_free_group:
220	if (ngrp) {
221		/* Free the group, after all cpu's are done. */
222		call_rcu(&ngrp->rcu, vlan_rcu_free);
223	}
224	return err;
225}
226
227/*  Attach a VLAN device to a mac address (ie Ethernet Card).
228 *  Returns 0 if the device was created or a negative error code otherwise.
229 */
230static int register_vlan_device(struct net_device *real_dev, u16 vlan_id)
231{
232	struct net_device *new_dev;
233	struct net *net = dev_net(real_dev);
234	struct vlan_net *vn = net_generic(net, vlan_net_id);
235	char name[IFNAMSIZ];
236	int err;
237
238	if (vlan_id >= VLAN_VID_MASK)
239		return -ERANGE;
240
241	err = vlan_check_real_dev(real_dev, vlan_id);
242	if (err < 0)
243		return err;
244
245	/* Gotta set up the fields for the device. */
246	switch (vn->name_type) {
247	case VLAN_NAME_TYPE_RAW_PLUS_VID:
248		/* name will look like:	 eth1.0005 */
249		snprintf(name, IFNAMSIZ, "%s.%.4i", real_dev->name, vlan_id);
250		break;
251	case VLAN_NAME_TYPE_PLUS_VID_NO_PAD:
252		/* Put our vlan.VID in the name.
253		 * Name will look like:	 vlan5
254		 */
255		snprintf(name, IFNAMSIZ, "vlan%i", vlan_id);
256		break;
257	case VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD:
258		/* Put our vlan.VID in the name.
259		 * Name will look like:	 eth0.5
260		 */
261		snprintf(name, IFNAMSIZ, "%s.%i", real_dev->name, vlan_id);
262		break;
263	case VLAN_NAME_TYPE_PLUS_VID:
264		/* Put our vlan.VID in the name.
265		 * Name will look like:	 vlan0005
266		 */
267	default:
268		snprintf(name, IFNAMSIZ, "vlan%.4i", vlan_id);
269	}
270
271	new_dev = alloc_netdev(sizeof(struct vlan_dev_info), name, vlan_setup);
272
273	if (new_dev == NULL)
274		return -ENOBUFS;
275
276	dev_net_set(new_dev, net);
277	/* need 4 bytes for extra VLAN header info,
278	 * hope the underlying device can handle it.
279	 */
280	new_dev->mtu = real_dev->mtu;
281
282	vlan_dev_info(new_dev)->vlan_id = vlan_id;
283	vlan_dev_info(new_dev)->real_dev = real_dev;
284	vlan_dev_info(new_dev)->dent = NULL;
285	vlan_dev_info(new_dev)->flags = VLAN_FLAG_REORDER_HDR;
286
287	new_dev->rtnl_link_ops = &vlan_link_ops;
288	err = register_vlan_dev(new_dev);
289	if (err < 0)
290		goto out_free_newdev;
291
292	return 0;
293
294out_free_newdev:
295	free_netdev(new_dev);
296	return err;
297}
298
299static void vlan_sync_address(struct net_device *dev,
300			      struct net_device *vlandev)
301{
302	struct vlan_dev_info *vlan = vlan_dev_info(vlandev);
303
304	/* May be called without an actual change */
305	if (!compare_ether_addr(vlan->real_dev_addr, dev->dev_addr))
306		return;
307
308	/* vlan address was different from the old address and is equal to
309	 * the new address */
310	if (compare_ether_addr(vlandev->dev_addr, vlan->real_dev_addr) &&
311	    !compare_ether_addr(vlandev->dev_addr, dev->dev_addr))
312		dev_uc_del(dev, vlandev->dev_addr);
313
314	/* vlan address was equal to the old address and is different from
315	 * the new address */
316	if (!compare_ether_addr(vlandev->dev_addr, vlan->real_dev_addr) &&
317	    compare_ether_addr(vlandev->dev_addr, dev->dev_addr))
318		dev_uc_add(dev, vlandev->dev_addr);
319
320	memcpy(vlan->real_dev_addr, dev->dev_addr, ETH_ALEN);
321}
322
323static void vlan_transfer_features(struct net_device *dev,
324				   struct net_device *vlandev)
325{
326	vlandev->gso_max_size = dev->gso_max_size;
327
328	if (dev->features & NETIF_F_HW_VLAN_TX)
329		vlandev->hard_header_len = dev->hard_header_len;
330	else
331		vlandev->hard_header_len = dev->hard_header_len + VLAN_HLEN;
332
333#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
334	vlandev->fcoe_ddp_xid = dev->fcoe_ddp_xid;
335#endif
336
337	netdev_update_features(vlandev);
338}
339
340static void __vlan_device_event(struct net_device *dev, unsigned long event)
341{
342	switch (event) {
343	case NETDEV_CHANGENAME:
344		vlan_proc_rem_dev(dev);
345		if (vlan_proc_add_dev(dev) < 0)
346			pr_warning("8021q: failed to change proc name for %s\n",
347					dev->name);
348		break;
349	case NETDEV_REGISTER:
350		if (vlan_proc_add_dev(dev) < 0)
351			pr_warning("8021q: failed to add proc entry for %s\n",
352					dev->name);
353		break;
354	case NETDEV_UNREGISTER:
355		vlan_proc_rem_dev(dev);
356		break;
357	}
358}
359
360static int vlan_device_event(struct notifier_block *unused, unsigned long event,
361			     void *ptr)
362{
363	struct net_device *dev = ptr;
364	struct vlan_group *grp;
365	int i, flgs;
366	struct net_device *vlandev;
367	struct vlan_dev_info *vlan;
368	LIST_HEAD(list);
369
370	if (is_vlan_dev(dev))
371		__vlan_device_event(dev, event);
372
373	if ((event == NETDEV_UP) &&
374	    (dev->features & NETIF_F_HW_VLAN_FILTER) &&
375	    dev->netdev_ops->ndo_vlan_rx_add_vid) {
376		pr_info("8021q: adding VLAN 0 to HW filter on device %s\n",
377			dev->name);
378		dev->netdev_ops->ndo_vlan_rx_add_vid(dev, 0);
379	}
380
381	grp = rtnl_dereference(dev->vlgrp);
382	if (!grp)
383		goto out;
384
385	/* It is OK that we do not hold the group lock right now,
386	 * as we run under the RTNL lock.
387	 */
388
389	switch (event) {
390	case NETDEV_CHANGE:
391		/* Propagate real device state to vlan devices */
392		for (i = 0; i < VLAN_N_VID; i++) {
393			vlandev = vlan_group_get_device(grp, i);
394			if (!vlandev)
395				continue;
396
397			netif_stacked_transfer_operstate(dev, vlandev);
398		}
399		break;
400
401	case NETDEV_CHANGEADDR:
402		/* Adjust unicast filters on underlying device */
403		for (i = 0; i < VLAN_N_VID; i++) {
404			vlandev = vlan_group_get_device(grp, i);
405			if (!vlandev)
406				continue;
407
408			flgs = vlandev->flags;
409			if (!(flgs & IFF_UP))
410				continue;
411
412			vlan_sync_address(dev, vlandev);
413		}
414		break;
415
416	case NETDEV_CHANGEMTU:
417		for (i = 0; i < VLAN_N_VID; i++) {
418			vlandev = vlan_group_get_device(grp, i);
419			if (!vlandev)
420				continue;
421
422			if (vlandev->mtu <= dev->mtu)
423				continue;
424
425			dev_set_mtu(vlandev, dev->mtu);
426		}
427		break;
428
429	case NETDEV_FEAT_CHANGE:
430		/* Propagate device features to underlying device */
431		for (i = 0; i < VLAN_N_VID; i++) {
432			vlandev = vlan_group_get_device(grp, i);
433			if (!vlandev)
434				continue;
435
436			vlan_transfer_features(dev, vlandev);
437		}
438
439		break;
440
441	case NETDEV_DOWN:
442		/* Put all VLANs for this dev in the down state too.  */
443		for (i = 0; i < VLAN_N_VID; i++) {
444			vlandev = vlan_group_get_device(grp, i);
445			if (!vlandev)
446				continue;
447
448			flgs = vlandev->flags;
449			if (!(flgs & IFF_UP))
450				continue;
451
452			vlan = vlan_dev_info(vlandev);
453			if (!(vlan->flags & VLAN_FLAG_LOOSE_BINDING))
454				dev_change_flags(vlandev, flgs & ~IFF_UP);
455			netif_stacked_transfer_operstate(dev, vlandev);
456		}
457		break;
458
459	case NETDEV_UP:
460		/* Put all VLANs for this dev in the up state too.  */
461		for (i = 0; i < VLAN_N_VID; i++) {
462			vlandev = vlan_group_get_device(grp, i);
463			if (!vlandev)
464				continue;
465
466			flgs = vlandev->flags;
467			if (flgs & IFF_UP)
468				continue;
469
470			vlan = vlan_dev_info(vlandev);
471			if (!(vlan->flags & VLAN_FLAG_LOOSE_BINDING))
472				dev_change_flags(vlandev, flgs | IFF_UP);
473			netif_stacked_transfer_operstate(dev, vlandev);
474		}
475		break;
476
477	case NETDEV_UNREGISTER:
478		/* twiddle thumbs on netns device moves */
479		if (dev->reg_state != NETREG_UNREGISTERING)
480			break;
481
482		for (i = 0; i < VLAN_N_VID; i++) {
483			vlandev = vlan_group_get_device(grp, i);
484			if (!vlandev)
485				continue;
486
487			/* unregistration of last vlan destroys group, abort
488			 * afterwards */
489			if (grp->nr_vlans == 1)
490				i = VLAN_N_VID;
491
492			unregister_vlan_dev(vlandev, &list);
493		}
494		unregister_netdevice_many(&list);
495		break;
496
497	case NETDEV_PRE_TYPE_CHANGE:
498		/* Forbid underlaying device to change its type. */
499		return NOTIFY_BAD;
500
501	case NETDEV_NOTIFY_PEERS:
502	case NETDEV_BONDING_FAILOVER:
503		/* Propagate to vlan devices */
504		for (i = 0; i < VLAN_N_VID; i++) {
505			vlandev = vlan_group_get_device(grp, i);
506			if (!vlandev)
507				continue;
508
509			call_netdevice_notifiers(event, vlandev);
510		}
511		break;
512	}
513
514out:
515	return NOTIFY_DONE;
516}
517
518static struct notifier_block vlan_notifier_block __read_mostly = {
519	.notifier_call = vlan_device_event,
520};
521
522/*
523 *	VLAN IOCTL handler.
524 *	o execute requested action or pass command to the device driver
525 *   arg is really a struct vlan_ioctl_args __user *.
526 */
527static int vlan_ioctl_handler(struct net *net, void __user *arg)
528{
529	int err;
530	struct vlan_ioctl_args args;
531	struct net_device *dev = NULL;
532
533	if (copy_from_user(&args, arg, sizeof(struct vlan_ioctl_args)))
534		return -EFAULT;
535
536	/* Null terminate this sucker, just in case. */
537	args.device1[23] = 0;
538	args.u.device2[23] = 0;
539
540	rtnl_lock();
541
542	switch (args.cmd) {
543	case SET_VLAN_INGRESS_PRIORITY_CMD:
544	case SET_VLAN_EGRESS_PRIORITY_CMD:
545	case SET_VLAN_FLAG_CMD:
546	case ADD_VLAN_CMD:
547	case DEL_VLAN_CMD:
548	case GET_VLAN_REALDEV_NAME_CMD:
549	case GET_VLAN_VID_CMD:
550		err = -ENODEV;
551		dev = __dev_get_by_name(net, args.device1);
552		if (!dev)
553			goto out;
554
555		err = -EINVAL;
556		if (args.cmd != ADD_VLAN_CMD && !is_vlan_dev(dev))
557			goto out;
558	}
559
560	switch (args.cmd) {
561	case SET_VLAN_INGRESS_PRIORITY_CMD:
562		err = -EPERM;
563		if (!capable(CAP_NET_ADMIN))
564			break;
565		vlan_dev_set_ingress_priority(dev,
566					      args.u.skb_priority,
567					      args.vlan_qos);
568		err = 0;
569		break;
570
571	case SET_VLAN_EGRESS_PRIORITY_CMD:
572		err = -EPERM;
573		if (!capable(CAP_NET_ADMIN))
574			break;
575		err = vlan_dev_set_egress_priority(dev,
576						   args.u.skb_priority,
577						   args.vlan_qos);
578		break;
579
580	case SET_VLAN_FLAG_CMD:
581		err = -EPERM;
582		if (!capable(CAP_NET_ADMIN))
583			break;
584		err = vlan_dev_change_flags(dev,
585					    args.vlan_qos ? args.u.flag : 0,
586					    args.u.flag);
587		break;
588
589	case SET_VLAN_NAME_TYPE_CMD:
590		err = -EPERM;
591		if (!capable(CAP_NET_ADMIN))
592			break;
593		if ((args.u.name_type >= 0) &&
594		    (args.u.name_type < VLAN_NAME_TYPE_HIGHEST)) {
595			struct vlan_net *vn;
596
597			vn = net_generic(net, vlan_net_id);
598			vn->name_type = args.u.name_type;
599			err = 0;
600		} else {
601			err = -EINVAL;
602		}
603		break;
604
605	case ADD_VLAN_CMD:
606		err = -EPERM;
607		if (!capable(CAP_NET_ADMIN))
608			break;
609		err = register_vlan_device(dev, args.u.VID);
610		break;
611
612	case DEL_VLAN_CMD:
613		err = -EPERM;
614		if (!capable(CAP_NET_ADMIN))
615			break;
616		unregister_vlan_dev(dev, NULL);
617		err = 0;
618		break;
619
620	case GET_VLAN_REALDEV_NAME_CMD:
621		err = 0;
622		vlan_dev_get_realdev_name(dev, args.u.device2);
623		if (copy_to_user(arg, &args,
624				 sizeof(struct vlan_ioctl_args)))
625			err = -EFAULT;
626		break;
627
628	case GET_VLAN_VID_CMD:
629		err = 0;
630		args.u.VID = vlan_dev_vlan_id(dev);
631		if (copy_to_user(arg, &args,
632				 sizeof(struct vlan_ioctl_args)))
633		      err = -EFAULT;
634		break;
635
636	default:
637		err = -EOPNOTSUPP;
638		break;
639	}
640out:
641	rtnl_unlock();
642	return err;
643}
644
645static int __net_init vlan_init_net(struct net *net)
646{
647	struct vlan_net *vn = net_generic(net, vlan_net_id);
648	int err;
649
650	vn->name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD;
651
652	err = vlan_proc_init(net);
653
654	return err;
655}
656
657static void __net_exit vlan_exit_net(struct net *net)
658{
659	vlan_proc_cleanup(net);
660}
661
662static struct pernet_operations vlan_net_ops = {
663	.init = vlan_init_net,
664	.exit = vlan_exit_net,
665	.id   = &vlan_net_id,
666	.size = sizeof(struct vlan_net),
667};
668
669static int __init vlan_proto_init(void)
670{
671	int err;
672
673	pr_info("%s v%s %s\n", vlan_fullname, vlan_version, vlan_copyright);
674	pr_info("All bugs added by %s\n", vlan_buggyright);
675
676	err = register_pernet_subsys(&vlan_net_ops);
677	if (err < 0)
678		goto err0;
679
680	err = register_netdevice_notifier(&vlan_notifier_block);
681	if (err < 0)
682		goto err2;
683
684	err = vlan_gvrp_init();
685	if (err < 0)
686		goto err3;
687
688	err = vlan_netlink_init();
689	if (err < 0)
690		goto err4;
691
692	vlan_ioctl_set(vlan_ioctl_handler);
693	return 0;
694
695err4:
696	vlan_gvrp_uninit();
697err3:
698	unregister_netdevice_notifier(&vlan_notifier_block);
699err2:
700	unregister_pernet_subsys(&vlan_net_ops);
701err0:
702	return err;
703}
704
705static void __exit vlan_cleanup_module(void)
706{
707	vlan_ioctl_set(NULL);
708	vlan_netlink_fini();
709
710	unregister_netdevice_notifier(&vlan_notifier_block);
711
712	unregister_pernet_subsys(&vlan_net_ops);
713	rcu_barrier(); /* Wait for completion of call_rcu()'s */
714
715	vlan_gvrp_uninit();
716}
717
718module_init(vlan_proto_init);
719module_exit(vlan_cleanup_module);
720
721MODULE_LICENSE("GPL");
722MODULE_VERSION(DRV_VERSION);
723