3 * Ethernet-type device handling.
5 * Authors: Ben Greear <greearb@candelatech.com>
6 * Please send support related email to: vlan@scry.wanfear.com
7 * VLAN Home Page: http://www.candelatech.com/~greear/vlan.html
9 * Fixes: Mar 22 2001: Martin Bokaemper <mbokaemper@unispherenetworks.com>
10 * - reset skb->pkt_type on incoming packets when MAC was changed
11 * - see that changed MAC is saddr for outgoing packets
12 * Oct 20, 2001: Ard van Breeman:
13 * - Fix MC-list, finally.
14 * - Flush MC-list on VLAN destroy.
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
23 #include <linux/module.h>
26 #include <linux/init.h>
27 #include <asm/uaccess.h> /* for copy_from_user */
28 #include <linux/skbuff.h>
29 #include <linux/netdevice.h>
30 #include <linux/etherdevice.h>
31 #include <net/datalink.h>
32 #include <net/p8022.h>
37 #include <linux/if_vlan.h>
41 * Rebuild the Ethernet MAC header. This is called after an ARP
42 * (or in future other address resolution) has completed on this
43 * sk_buff. We now let ARP fill in the other fields.
45 * This routine CANNOT use cached dst->neigh!
46 * Really, it is used only when dst->neigh is wrong.
48 * TODO: This needs a checkup, I'm ignorant here. --BLG
50 static int vlan_dev_rebuild_header(struct sk_buff *skb)
52 struct net_device *dev = skb->dev;
53 struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
55 switch (veth->h_vlan_encapsulated_proto) {
57 case __constant_htons(ETH_P_IP):
59 /* TODO: Confirm this will work with VLAN headers... */
60 return arp_find(veth->h_dest, skb);
63 pr_debug("%s: unable to resolve type %X addresses.\n",
64 dev->name, ntohs(veth->h_vlan_encapsulated_proto));
66 memcpy(veth->h_source, dev->dev_addr, ETH_ALEN);
73 static inline struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
75 if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
76 if (skb_shared(skb) || skb_cloned(skb)) {
77 struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
82 /* Lifted from Gleb's VLAN code... */
83 memmove(skb->data - ETH_HLEN,
84 skb->data - VLAN_ETH_HLEN, 12);
85 skb->mac_header += VLAN_HLEN;
93 * Determine the packet's protocol ID. The rule here is that we
94 * assume 802.3 if the type field is short enough to be a length.
95 * This is normal practice and works for any 'now in use' protocol.
97 * Also, at this point we assume that we ARE dealing exclusively with
98 * VLAN packets, or packets that should be made into VLAN packets based
99 * on a default VLAN ID.
101 * NOTE: Should be similar to ethernet/eth.c.
103 * SANITY NOTE: This method is called when a packet is moving up the stack
104 * towards userland. To get here, it would have already passed
105 * through the ethernet/eth.c eth_type_trans() method.
106 * SANITY NOTE 2: We are referencing to the VLAN_HDR frields, which MAY be
107 * stored UNALIGNED in the memory. RISC systems don't like
108 * such cases very much...
109 * SANITY NOTE 2a: According to Dave Miller & Alexey, it will always be
110 * aligned, so there doesn't need to be any of the unaligned
111 * stuff. It has been commented out now... --Ben
114 int vlan_skb_recv(struct sk_buff *skb, struct net_device *dev,
115 struct packet_type *ptype, struct net_device *orig_dev)
117 unsigned char *rawp = NULL;
118 struct vlan_hdr *vhdr;
120 struct net_device_stats *stats;
121 unsigned short vlan_TCI;
124 if (dev->nd_net != &init_net) {
129 skb = skb_share_check(skb, GFP_ATOMIC);
133 if (unlikely(!pskb_may_pull(skb, VLAN_HLEN))) {
138 vhdr = (struct vlan_hdr *)(skb->data);
140 /* vlan_TCI = ntohs(get_unaligned(&vhdr->h_vlan_TCI)); */
141 vlan_TCI = ntohs(vhdr->h_vlan_TCI);
143 vid = (vlan_TCI & VLAN_VID_MASK);
145 /* Ok, we will find the correct VLAN device, strip the header,
146 * and then go on as usual.
149 /* We have 12 bits of vlan ID.
151 * We must not drop allow preempt until we hold a
152 * reference to the device (netif_rx does that) or we
157 skb->dev = __find_vlan_dev(dev, vid);
160 pr_debug("%s: ERROR: No net_device for VID: %u on dev: %s\n",
161 __FUNCTION__, (unsigned int)vid, dev->name);
166 skb->dev->last_rx = jiffies;
168 /* Bump the rx counters for the VLAN device. */
169 stats = &skb->dev->stats;
171 stats->rx_bytes += skb->len;
173 /* Take off the VLAN header (4 bytes currently) */
174 skb_pull_rcsum(skb, VLAN_HLEN);
177 * Deal with ingress priority mapping.
179 skb->priority = vlan_get_ingress_priority(skb->dev,
180 ntohs(vhdr->h_vlan_TCI));
182 pr_debug("%s: priority: %u for TCI: %hu\n",
183 __FUNCTION__, skb->priority, ntohs(vhdr->h_vlan_TCI));
185 /* The ethernet driver already did the pkt_type calculations
188 switch (skb->pkt_type) {
189 case PACKET_BROADCAST: /* Yeah, stats collect these together.. */
190 /* stats->broadcast ++; // no such counter :-( */
193 case PACKET_MULTICAST:
197 case PACKET_OTHERHOST:
198 /* Our lower layer thinks this is not local, let's make sure.
199 * This allows the VLAN to have a different MAC than the
200 * underlying device, and still route correctly.
202 if (!compare_ether_addr(eth_hdr(skb)->h_dest,
204 /* It is for our (changed) MAC-address! */
205 skb->pkt_type = PACKET_HOST;
211 /* Was a VLAN packet, grab the encapsulated protocol, which the layer
212 * three protocols care about.
214 /* proto = get_unaligned(&vhdr->h_vlan_encapsulated_proto); */
215 proto = vhdr->h_vlan_encapsulated_proto;
217 skb->protocol = proto;
218 if (ntohs(proto) >= 1536) {
219 /* place it back on the queue to be handled by
220 * true layer 3 protocols.
223 /* See if we are configured to re-write the VLAN header
224 * to make it look like ethernet...
226 skb = vlan_check_reorder_header(skb);
228 /* Can be null if skb-clone fails when re-ordering */
232 /* TODO: Add a more specific counter here. */
242 * This is a magic hack to spot IPX packets. Older Novell breaks
243 * the protocol design and runs IPX over 802.3 without an 802.2 LLC
244 * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
245 * won't work for fault tolerant netware but does for the rest.
247 if (*(unsigned short *)rawp == 0xFFFF) {
248 skb->protocol = htons(ETH_P_802_3);
249 /* place it back on the queue to be handled by true layer 3
252 /* See if we are configured to re-write the VLAN header
253 * to make it look like ethernet...
255 skb = vlan_check_reorder_header(skb);
257 /* Can be null if skb-clone fails when re-ordering */
261 /* TODO: Add a more specific counter here. */
271 skb->protocol = htons(ETH_P_802_2);
272 /* place it back on the queue to be handled by upper layer protocols.
275 /* See if we are configured to re-write the VLAN header
276 * to make it look like ethernet...
278 skb = vlan_check_reorder_header(skb);
280 /* Can be null if skb-clone fails when re-ordering */
284 /* TODO: Add a more specific counter here. */
291 static inline unsigned short
292 vlan_dev_get_egress_qos_mask(struct net_device *dev, struct sk_buff *skb)
294 struct vlan_priority_tci_mapping *mp;
296 mp = vlan_dev_info(dev)->egress_priority_map[(skb->priority & 0xF)];
298 if (mp->priority == skb->priority) {
299 return mp->vlan_qos; /* This should already be shifted
300 * to mask correctly with the
309 * Create the VLAN header for an arbitrary protocol layer
311 * saddr=NULL means use device source address
312 * daddr=NULL means leave destination address (eg unresolved arp)
314 * This is called when the SKB is moving down the stack towards the
317 static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
319 const void *daddr, const void *saddr,
322 struct vlan_hdr *vhdr;
323 unsigned short veth_TCI = 0;
325 int build_vlan_header = 0;
326 struct net_device *vdev = dev;
328 pr_debug("%s: skb: %p type: %hx len: %u vlan_id: %hx, daddr: %p\n",
329 __FUNCTION__, skb, type, len, vlan_dev_info(dev)->vlan_id,
332 /* build vlan header only if re_order_header flag is NOT set. This
333 * fixes some programs that get confused when they see a VLAN device
334 * sending a frame that is VLAN encoded (the consensus is that the VLAN
335 * device should look completely like an Ethernet device when the
336 * REORDER_HEADER flag is set) The drawback to this is some extra
337 * header shuffling in the hard_start_xmit. Users can turn off this
338 * REORDER behaviour with the vconfig tool.
340 if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR))
341 build_vlan_header = 1;
343 if (build_vlan_header) {
344 vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
346 /* build the four bytes that make this a VLAN header. */
348 /* Now, construct the second two bytes. This field looks
350 * usr_priority: 3 bits (high bits)
352 * VLAN ID 12 bits (low bits)
355 veth_TCI = vlan_dev_info(dev)->vlan_id;
356 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
358 vhdr->h_vlan_TCI = htons(veth_TCI);
361 * Set the protocol type. For a packet of type ETH_P_802_3 we
362 * put the length in here instead. It is up to the 802.2
363 * layer to carry protocol information.
366 if (type != ETH_P_802_3)
367 vhdr->h_vlan_encapsulated_proto = htons(type);
369 vhdr->h_vlan_encapsulated_proto = htons(len);
371 skb->protocol = htons(ETH_P_8021Q);
372 skb_reset_network_header(skb);
375 /* Before delegating work to the lower layer, enter our MAC-address */
377 saddr = dev->dev_addr;
379 dev = vlan_dev_info(dev)->real_dev;
381 /* MPLS can send us skbuffs w/out enough space. This check will grow
382 * the skb if it doesn't have enough headroom. Not a beautiful solution,
383 * so I'll tick a counter so that users can know it's happening...
387 /* NOTE: This may still break if the underlying device is not the final
388 * device (and thus there are more headers to add...) It should work for
389 * good-ole-ethernet though.
391 if (skb_headroom(skb) < dev->hard_header_len) {
392 struct sk_buff *sk_tmp = skb;
393 skb = skb_realloc_headroom(sk_tmp, dev->hard_header_len);
396 struct net_device_stats *stats = &vdev->stats;
400 vlan_dev_info(vdev)->cnt_inc_headroom_on_tx++;
401 pr_debug("%s: %s: had to grow skb\n", __FUNCTION__, vdev->name);
404 if (build_vlan_header) {
405 /* Now make the underlying real hard header */
406 rc = dev_hard_header(skb, dev, ETH_P_8021Q, daddr, saddr,
413 /* If here, then we'll just make a normal looking ethernet
414 * frame, but, the hard_start_xmit method will insert the tag
415 * (it has to be able to do this for bridged and other skbs
416 * that don't come down the protocol stack in an orderly manner.
418 rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
423 static int vlan_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
425 struct net_device_stats *stats = &dev->stats;
426 struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
428 /* Handle non-VLAN frames if they are sent to us, for example by DHCP.
430 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
431 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
434 if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
435 vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR) {
436 int orig_headroom = skb_headroom(skb);
437 unsigned short veth_TCI;
439 /* This is not a VLAN frame...but we can fix that! */
440 vlan_dev_info(dev)->cnt_encap_on_xmit++;
442 pr_debug("%s: proto to encap: 0x%hx\n",
443 __FUNCTION__, htons(veth->h_vlan_proto));
444 /* Construct the second two bytes. This field looks something
446 * usr_priority: 3 bits (high bits)
448 * VLAN ID 12 bits (low bits)
450 veth_TCI = vlan_dev_info(dev)->vlan_id;
451 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
453 skb = __vlan_put_tag(skb, veth_TCI);
459 if (orig_headroom < VLAN_HLEN)
460 vlan_dev_info(dev)->cnt_inc_headroom_on_tx++;
463 pr_debug("%s: about to send skb: %p to dev: %s\n",
464 __FUNCTION__, skb, skb->dev->name);
465 pr_debug(" " MAC_FMT " " MAC_FMT " %4hx %4hx %4hx\n",
466 veth->h_dest[0], veth->h_dest[1], veth->h_dest[2],
467 veth->h_dest[3], veth->h_dest[4], veth->h_dest[5],
468 veth->h_source[0], veth->h_source[1], veth->h_source[2],
469 veth->h_source[3], veth->h_source[4], veth->h_source[5],
470 veth->h_vlan_proto, veth->h_vlan_TCI,
471 veth->h_vlan_encapsulated_proto);
473 stats->tx_packets++; /* for statics only */
474 stats->tx_bytes += skb->len;
476 skb->dev = vlan_dev_info(dev)->real_dev;
482 static int vlan_dev_hwaccel_hard_start_xmit(struct sk_buff *skb,
483 struct net_device *dev)
485 struct net_device_stats *stats = &dev->stats;
486 unsigned short veth_TCI;
488 /* Construct the second two bytes. This field looks something
490 * usr_priority: 3 bits (high bits)
492 * VLAN ID 12 bits (low bits)
494 veth_TCI = vlan_dev_info(dev)->vlan_id;
495 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
496 skb = __vlan_hwaccel_put_tag(skb, veth_TCI);
499 stats->tx_bytes += skb->len;
501 skb->dev = vlan_dev_info(dev)->real_dev;
507 static int vlan_dev_change_mtu(struct net_device *dev, int new_mtu)
509 /* TODO: gotta make sure the underlying layer can handle it,
510 * maybe an IFF_VLAN_CAPABLE flag for devices?
512 if (vlan_dev_info(dev)->real_dev->mtu < new_mtu)
520 void vlan_dev_set_ingress_priority(const struct net_device *dev,
521 u32 skb_prio, short vlan_prio)
523 struct vlan_dev_info *vlan = vlan_dev_info(dev);
525 if (vlan->ingress_priority_map[vlan_prio & 0x7] && !skb_prio)
526 vlan->nr_ingress_mappings--;
527 else if (!vlan->ingress_priority_map[vlan_prio & 0x7] && skb_prio)
528 vlan->nr_ingress_mappings++;
530 vlan->ingress_priority_map[vlan_prio & 0x7] = skb_prio;
533 int vlan_dev_set_egress_priority(const struct net_device *dev,
534 u32 skb_prio, short vlan_prio)
536 struct vlan_dev_info *vlan = vlan_dev_info(dev);
537 struct vlan_priority_tci_mapping *mp = NULL;
538 struct vlan_priority_tci_mapping *np;
539 u32 vlan_qos = (vlan_prio << 13) & 0xE000;
541 /* See if a priority mapping exists.. */
542 mp = vlan->egress_priority_map[skb_prio & 0xF];
544 if (mp->priority == skb_prio) {
545 if (mp->vlan_qos && !vlan_qos)
546 vlan->nr_egress_mappings--;
547 else if (!mp->vlan_qos && vlan_qos)
548 vlan->nr_egress_mappings++;
549 mp->vlan_qos = vlan_qos;
555 /* Create a new mapping then. */
556 mp = vlan->egress_priority_map[skb_prio & 0xF];
557 np = kmalloc(sizeof(struct vlan_priority_tci_mapping), GFP_KERNEL);
562 np->priority = skb_prio;
563 np->vlan_qos = vlan_qos;
564 vlan->egress_priority_map[skb_prio & 0xF] = np;
566 vlan->nr_egress_mappings++;
570 /* Flags are defined in the vlan_flags enum in include/linux/if_vlan.h file. */
571 int vlan_dev_set_vlan_flag(const struct net_device *dev,
572 u32 flag, short flag_val)
574 /* verify flag is supported */
575 if (flag == VLAN_FLAG_REORDER_HDR) {
577 vlan_dev_info(dev)->flags |= VLAN_FLAG_REORDER_HDR;
579 vlan_dev_info(dev)->flags &= ~VLAN_FLAG_REORDER_HDR;
585 void vlan_dev_get_realdev_name(const struct net_device *dev, char *result)
587 strncpy(result, vlan_dev_info(dev)->real_dev->name, 23);
590 void vlan_dev_get_vid(const struct net_device *dev, unsigned short *result)
592 *result = vlan_dev_info(dev)->vlan_id;
595 static int vlan_dev_open(struct net_device *dev)
597 struct vlan_dev_info *vlan = vlan_dev_info(dev);
598 struct net_device *real_dev = vlan->real_dev;
601 if (!(real_dev->flags & IFF_UP))
604 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr)) {
605 err = dev_unicast_add(real_dev, dev->dev_addr, ETH_ALEN);
609 memcpy(vlan->real_dev_addr, real_dev->dev_addr, ETH_ALEN);
611 if (dev->flags & IFF_ALLMULTI)
612 dev_set_allmulti(real_dev, 1);
613 if (dev->flags & IFF_PROMISC)
614 dev_set_promiscuity(real_dev, 1);
619 static int vlan_dev_stop(struct net_device *dev)
621 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
623 dev_mc_unsync(real_dev, dev);
624 if (dev->flags & IFF_ALLMULTI)
625 dev_set_allmulti(real_dev, -1);
626 if (dev->flags & IFF_PROMISC)
627 dev_set_promiscuity(real_dev, -1);
629 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
630 dev_unicast_delete(real_dev, dev->dev_addr, dev->addr_len);
635 static int vlan_dev_set_mac_address(struct net_device *dev, void *p)
637 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
638 struct sockaddr *addr = p;
641 if (!is_valid_ether_addr(addr->sa_data))
642 return -EADDRNOTAVAIL;
644 if (!(dev->flags & IFF_UP))
647 if (compare_ether_addr(addr->sa_data, real_dev->dev_addr)) {
648 err = dev_unicast_add(real_dev, addr->sa_data, ETH_ALEN);
653 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
654 dev_unicast_delete(real_dev, dev->dev_addr, ETH_ALEN);
657 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
661 static int vlan_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
663 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
665 int err = -EOPNOTSUPP;
667 strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
668 ifrr.ifr_ifru = ifr->ifr_ifru;
674 if (real_dev->do_ioctl && netif_device_present(real_dev))
675 err = real_dev->do_ioctl(real_dev, &ifrr, cmd);
680 ifr->ifr_ifru = ifrr.ifr_ifru;
685 static void vlan_dev_change_rx_flags(struct net_device *dev, int change)
687 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
689 if (change & IFF_ALLMULTI)
690 dev_set_allmulti(real_dev, dev->flags & IFF_ALLMULTI ? 1 : -1);
691 if (change & IFF_PROMISC)
692 dev_set_promiscuity(real_dev, dev->flags & IFF_PROMISC ? 1 : -1);
695 static void vlan_dev_set_multicast_list(struct net_device *vlan_dev)
697 dev_mc_sync(vlan_dev_info(vlan_dev)->real_dev, vlan_dev);
701 * vlan network devices have devices nesting below it, and are a special
702 * "super class" of normal network devices; split their locks off into a
703 * separate class since they always nest.
705 static struct lock_class_key vlan_netdev_xmit_lock_key;
707 static const struct header_ops vlan_header_ops = {
708 .create = vlan_dev_hard_header,
709 .rebuild = vlan_dev_rebuild_header,
710 .parse = eth_header_parse,
713 static int vlan_dev_init(struct net_device *dev)
715 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
718 /* IFF_BROADCAST|IFF_MULTICAST; ??? */
719 dev->flags = real_dev->flags & ~IFF_UP;
720 dev->iflink = real_dev->ifindex;
721 dev->state = (real_dev->state & ((1<<__LINK_STATE_NOCARRIER) |
722 (1<<__LINK_STATE_DORMANT))) |
723 (1<<__LINK_STATE_PRESENT);
725 /* ipv6 shared card related stuff */
726 dev->dev_id = real_dev->dev_id;
728 if (is_zero_ether_addr(dev->dev_addr))
729 memcpy(dev->dev_addr, real_dev->dev_addr, dev->addr_len);
730 if (is_zero_ether_addr(dev->broadcast))
731 memcpy(dev->broadcast, real_dev->broadcast, dev->addr_len);
733 if (real_dev->features & NETIF_F_HW_VLAN_TX) {
734 dev->header_ops = real_dev->header_ops;
735 dev->hard_header_len = real_dev->hard_header_len;
736 dev->hard_start_xmit = vlan_dev_hwaccel_hard_start_xmit;
738 dev->header_ops = &vlan_header_ops;
739 dev->hard_header_len = real_dev->hard_header_len + VLAN_HLEN;
740 dev->hard_start_xmit = vlan_dev_hard_start_xmit;
743 if (real_dev->priv_flags & IFF_802_1Q_VLAN)
746 lockdep_set_class_and_subclass(&dev->_xmit_lock,
747 &vlan_netdev_xmit_lock_key, subclass);
751 void vlan_setup(struct net_device *dev)
755 dev->priv_flags |= IFF_802_1Q_VLAN;
756 dev->tx_queue_len = 0;
758 dev->change_mtu = vlan_dev_change_mtu;
759 dev->init = vlan_dev_init;
760 dev->open = vlan_dev_open;
761 dev->stop = vlan_dev_stop;
762 dev->set_mac_address = vlan_dev_set_mac_address;
763 dev->set_multicast_list = vlan_dev_set_multicast_list;
764 dev->change_rx_flags = vlan_dev_change_rx_flags;
765 dev->do_ioctl = vlan_dev_ioctl;
766 dev->destructor = free_netdev;
768 memset(dev->broadcast, 0, ETH_ALEN);