3 * Ethernet-type device handling.
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
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;
92 static inline void vlan_set_encap_proto(struct sk_buff *skb,
93 struct vlan_hdr *vhdr)
99 * Was a VLAN packet, grab the encapsulated protocol, which the layer
100 * three protocols care about.
103 proto = vhdr->h_vlan_encapsulated_proto;
104 if (ntohs(proto) >= 1536) {
105 skb->protocol = proto;
110 if (*(unsigned short *)rawp == 0xFFFF)
112 * This is a magic hack to spot IPX packets. Older Novell
113 * breaks the protocol design and runs IPX over 802.3 without
114 * an 802.2 LLC layer. We look for FFFF which isn't a used
115 * 802.2 SSAP/DSAP. This won't work for fault tolerant netware
116 * but does for the rest.
118 skb->protocol = htons(ETH_P_802_3);
123 skb->protocol = htons(ETH_P_802_2);
127 * Determine the packet's protocol ID. The rule here is that we
128 * assume 802.3 if the type field is short enough to be a length.
129 * This is normal practice and works for any 'now in use' protocol.
131 * Also, at this point we assume that we ARE dealing exclusively with
132 * VLAN packets, or packets that should be made into VLAN packets based
133 * on a default VLAN ID.
135 * NOTE: Should be similar to ethernet/eth.c.
137 * SANITY NOTE: This method is called when a packet is moving up the stack
138 * towards userland. To get here, it would have already passed
139 * through the ethernet/eth.c eth_type_trans() method.
140 * SANITY NOTE 2: We are referencing to the VLAN_HDR frields, which MAY be
141 * stored UNALIGNED in the memory. RISC systems don't like
142 * such cases very much...
143 * SANITY NOTE 2a: According to Dave Miller & Alexey, it will always be
144 * aligned, so there doesn't need to be any of the unaligned
145 * stuff. It has been commented out now... --Ben
148 int vlan_skb_recv(struct sk_buff *skb, struct net_device *dev,
149 struct packet_type *ptype, struct net_device *orig_dev)
151 struct vlan_hdr *vhdr;
153 struct net_device_stats *stats;
154 unsigned short vlan_TCI;
156 skb = skb_share_check(skb, GFP_ATOMIC);
160 if (unlikely(!pskb_may_pull(skb, VLAN_HLEN)))
163 vhdr = (struct vlan_hdr *)skb->data;
164 vlan_TCI = ntohs(vhdr->h_vlan_TCI);
165 vid = (vlan_TCI & VLAN_VID_MASK);
168 skb->dev = __find_vlan_dev(dev, vid);
170 pr_debug("%s: ERROR: No net_device for VID: %u on dev: %s\n",
171 __func__, (unsigned int)vid, dev->name);
175 skb->dev->last_rx = jiffies;
177 stats = &skb->dev->stats;
179 stats->rx_bytes += skb->len;
181 skb_pull_rcsum(skb, VLAN_HLEN);
183 skb->priority = vlan_get_ingress_priority(skb->dev,
184 ntohs(vhdr->h_vlan_TCI));
186 pr_debug("%s: priority: %u for TCI: %hu\n",
187 __func__, skb->priority, ntohs(vhdr->h_vlan_TCI));
189 switch (skb->pkt_type) {
190 case PACKET_BROADCAST: /* Yeah, stats collect these together.. */
191 /* stats->broadcast ++; // no such counter :-( */
194 case PACKET_MULTICAST:
198 case PACKET_OTHERHOST:
199 /* Our lower layer thinks this is not local, let's make sure.
200 * This allows the VLAN to have a different MAC than the
201 * underlying device, and still route correctly.
203 if (!compare_ether_addr(eth_hdr(skb)->h_dest,
205 skb->pkt_type = PACKET_HOST;
211 vlan_set_encap_proto(skb, vhdr);
213 skb = vlan_check_reorder_header(skb);
221 return NET_RX_SUCCESS;
230 static inline unsigned short
231 vlan_dev_get_egress_qos_mask(struct net_device *dev, struct sk_buff *skb)
233 struct vlan_priority_tci_mapping *mp;
235 mp = vlan_dev_info(dev)->egress_priority_map[(skb->priority & 0xF)];
237 if (mp->priority == skb->priority) {
238 return mp->vlan_qos; /* This should already be shifted
239 * to mask correctly with the
248 * Create the VLAN header for an arbitrary protocol layer
250 * saddr=NULL means use device source address
251 * daddr=NULL means leave destination address (eg unresolved arp)
253 * This is called when the SKB is moving down the stack towards the
256 static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
258 const void *daddr, const void *saddr,
261 struct vlan_hdr *vhdr;
262 unsigned short veth_TCI = 0;
264 int build_vlan_header = 0;
265 struct net_device *vdev = dev;
267 pr_debug("%s: skb: %p type: %hx len: %u vlan_id: %hx, daddr: %p\n",
268 __func__, skb, type, len, vlan_dev_info(dev)->vlan_id,
271 /* build vlan header only if re_order_header flag is NOT set. This
272 * fixes some programs that get confused when they see a VLAN device
273 * sending a frame that is VLAN encoded (the consensus is that the VLAN
274 * device should look completely like an Ethernet device when the
275 * REORDER_HEADER flag is set) The drawback to this is some extra
276 * header shuffling in the hard_start_xmit. Users can turn off this
277 * REORDER behaviour with the vconfig tool.
279 if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR))
280 build_vlan_header = 1;
282 if (build_vlan_header) {
283 vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
285 /* build the four bytes that make this a VLAN header. */
287 /* Now, construct the second two bytes. This field looks
289 * usr_priority: 3 bits (high bits)
291 * VLAN ID 12 bits (low bits)
294 veth_TCI = vlan_dev_info(dev)->vlan_id;
295 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
297 vhdr->h_vlan_TCI = htons(veth_TCI);
300 * Set the protocol type. For a packet of type ETH_P_802_3 we
301 * put the length in here instead. It is up to the 802.2
302 * layer to carry protocol information.
305 if (type != ETH_P_802_3)
306 vhdr->h_vlan_encapsulated_proto = htons(type);
308 vhdr->h_vlan_encapsulated_proto = htons(len);
310 skb->protocol = htons(ETH_P_8021Q);
311 skb_reset_network_header(skb);
314 /* Before delegating work to the lower layer, enter our MAC-address */
316 saddr = dev->dev_addr;
318 dev = vlan_dev_info(dev)->real_dev;
320 /* MPLS can send us skbuffs w/out enough space. This check will grow
321 * the skb if it doesn't have enough headroom. Not a beautiful solution,
322 * so I'll tick a counter so that users can know it's happening...
326 /* NOTE: This may still break if the underlying device is not the final
327 * device (and thus there are more headers to add...) It should work for
328 * good-ole-ethernet though.
330 if (skb_headroom(skb) < dev->hard_header_len) {
331 struct sk_buff *sk_tmp = skb;
332 skb = skb_realloc_headroom(sk_tmp, dev->hard_header_len);
335 struct net_device_stats *stats = &vdev->stats;
339 vlan_dev_info(vdev)->cnt_inc_headroom_on_tx++;
340 pr_debug("%s: %s: had to grow skb\n", __func__, vdev->name);
343 if (build_vlan_header) {
344 /* Now make the underlying real hard header */
345 rc = dev_hard_header(skb, dev, ETH_P_8021Q, daddr, saddr,
352 /* If here, then we'll just make a normal looking ethernet
353 * frame, but, the hard_start_xmit method will insert the tag
354 * (it has to be able to do this for bridged and other skbs
355 * that don't come down the protocol stack in an orderly manner.
357 rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
362 static int vlan_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
364 struct net_device_stats *stats = &dev->stats;
365 struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
367 /* Handle non-VLAN frames if they are sent to us, for example by DHCP.
369 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
370 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
373 if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
374 vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR) {
375 int orig_headroom = skb_headroom(skb);
376 unsigned short veth_TCI;
378 /* This is not a VLAN frame...but we can fix that! */
379 vlan_dev_info(dev)->cnt_encap_on_xmit++;
381 pr_debug("%s: proto to encap: 0x%hx\n",
382 __func__, ntohs(veth->h_vlan_proto));
383 /* Construct the second two bytes. This field looks something
385 * usr_priority: 3 bits (high bits)
387 * VLAN ID 12 bits (low bits)
389 veth_TCI = vlan_dev_info(dev)->vlan_id;
390 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
392 skb = __vlan_put_tag(skb, veth_TCI);
398 if (orig_headroom < VLAN_HLEN)
399 vlan_dev_info(dev)->cnt_inc_headroom_on_tx++;
402 pr_debug("%s: about to send skb: %p to dev: %s\n",
403 __func__, skb, skb->dev->name);
404 pr_debug(" " MAC_FMT " " MAC_FMT " %4hx %4hx %4hx\n",
405 veth->h_dest[0], veth->h_dest[1], veth->h_dest[2],
406 veth->h_dest[3], veth->h_dest[4], veth->h_dest[5],
407 veth->h_source[0], veth->h_source[1], veth->h_source[2],
408 veth->h_source[3], veth->h_source[4], veth->h_source[5],
409 veth->h_vlan_proto, veth->h_vlan_TCI,
410 veth->h_vlan_encapsulated_proto);
412 stats->tx_packets++; /* for statics only */
413 stats->tx_bytes += skb->len;
415 skb->dev = vlan_dev_info(dev)->real_dev;
421 static int vlan_dev_hwaccel_hard_start_xmit(struct sk_buff *skb,
422 struct net_device *dev)
424 struct net_device_stats *stats = &dev->stats;
425 unsigned short veth_TCI;
427 /* Construct the second two bytes. This field looks something
429 * usr_priority: 3 bits (high bits)
431 * VLAN ID 12 bits (low bits)
433 veth_TCI = vlan_dev_info(dev)->vlan_id;
434 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
435 skb = __vlan_hwaccel_put_tag(skb, veth_TCI);
438 stats->tx_bytes += skb->len;
440 skb->dev = vlan_dev_info(dev)->real_dev;
446 static int vlan_dev_change_mtu(struct net_device *dev, int new_mtu)
448 /* TODO: gotta make sure the underlying layer can handle it,
449 * maybe an IFF_VLAN_CAPABLE flag for devices?
451 if (vlan_dev_info(dev)->real_dev->mtu < new_mtu)
459 void vlan_dev_set_ingress_priority(const struct net_device *dev,
460 u32 skb_prio, short vlan_prio)
462 struct vlan_dev_info *vlan = vlan_dev_info(dev);
464 if (vlan->ingress_priority_map[vlan_prio & 0x7] && !skb_prio)
465 vlan->nr_ingress_mappings--;
466 else if (!vlan->ingress_priority_map[vlan_prio & 0x7] && skb_prio)
467 vlan->nr_ingress_mappings++;
469 vlan->ingress_priority_map[vlan_prio & 0x7] = skb_prio;
472 int vlan_dev_set_egress_priority(const struct net_device *dev,
473 u32 skb_prio, short vlan_prio)
475 struct vlan_dev_info *vlan = vlan_dev_info(dev);
476 struct vlan_priority_tci_mapping *mp = NULL;
477 struct vlan_priority_tci_mapping *np;
478 u32 vlan_qos = (vlan_prio << 13) & 0xE000;
480 /* See if a priority mapping exists.. */
481 mp = vlan->egress_priority_map[skb_prio & 0xF];
483 if (mp->priority == skb_prio) {
484 if (mp->vlan_qos && !vlan_qos)
485 vlan->nr_egress_mappings--;
486 else if (!mp->vlan_qos && vlan_qos)
487 vlan->nr_egress_mappings++;
488 mp->vlan_qos = vlan_qos;
494 /* Create a new mapping then. */
495 mp = vlan->egress_priority_map[skb_prio & 0xF];
496 np = kmalloc(sizeof(struct vlan_priority_tci_mapping), GFP_KERNEL);
501 np->priority = skb_prio;
502 np->vlan_qos = vlan_qos;
503 vlan->egress_priority_map[skb_prio & 0xF] = np;
505 vlan->nr_egress_mappings++;
509 /* Flags are defined in the vlan_flags enum in include/linux/if_vlan.h file. */
510 int vlan_dev_set_vlan_flag(const struct net_device *dev,
511 u32 flag, short flag_val)
513 /* verify flag is supported */
514 if (flag == VLAN_FLAG_REORDER_HDR) {
516 vlan_dev_info(dev)->flags |= VLAN_FLAG_REORDER_HDR;
518 vlan_dev_info(dev)->flags &= ~VLAN_FLAG_REORDER_HDR;
524 void vlan_dev_get_realdev_name(const struct net_device *dev, char *result)
526 strncpy(result, vlan_dev_info(dev)->real_dev->name, 23);
529 void vlan_dev_get_vid(const struct net_device *dev, unsigned short *result)
531 *result = vlan_dev_info(dev)->vlan_id;
534 static int vlan_dev_open(struct net_device *dev)
536 struct vlan_dev_info *vlan = vlan_dev_info(dev);
537 struct net_device *real_dev = vlan->real_dev;
540 if (!(real_dev->flags & IFF_UP))
543 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr)) {
544 err = dev_unicast_add(real_dev, dev->dev_addr, ETH_ALEN);
548 memcpy(vlan->real_dev_addr, real_dev->dev_addr, ETH_ALEN);
550 if (dev->flags & IFF_ALLMULTI)
551 dev_set_allmulti(real_dev, 1);
552 if (dev->flags & IFF_PROMISC)
553 dev_set_promiscuity(real_dev, 1);
558 static int vlan_dev_stop(struct net_device *dev)
560 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
562 dev_mc_unsync(real_dev, dev);
563 dev_unicast_unsync(real_dev, dev);
564 if (dev->flags & IFF_ALLMULTI)
565 dev_set_allmulti(real_dev, -1);
566 if (dev->flags & IFF_PROMISC)
567 dev_set_promiscuity(real_dev, -1);
569 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
570 dev_unicast_delete(real_dev, dev->dev_addr, dev->addr_len);
575 static int vlan_dev_set_mac_address(struct net_device *dev, void *p)
577 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
578 struct sockaddr *addr = p;
581 if (!is_valid_ether_addr(addr->sa_data))
582 return -EADDRNOTAVAIL;
584 if (!(dev->flags & IFF_UP))
587 if (compare_ether_addr(addr->sa_data, real_dev->dev_addr)) {
588 err = dev_unicast_add(real_dev, addr->sa_data, ETH_ALEN);
593 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
594 dev_unicast_delete(real_dev, dev->dev_addr, ETH_ALEN);
597 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
601 static int vlan_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
603 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
605 int err = -EOPNOTSUPP;
607 strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
608 ifrr.ifr_ifru = ifr->ifr_ifru;
614 if (real_dev->do_ioctl && netif_device_present(real_dev))
615 err = real_dev->do_ioctl(real_dev, &ifrr, cmd);
620 ifr->ifr_ifru = ifrr.ifr_ifru;
625 static void vlan_dev_change_rx_flags(struct net_device *dev, int change)
627 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
629 if (change & IFF_ALLMULTI)
630 dev_set_allmulti(real_dev, dev->flags & IFF_ALLMULTI ? 1 : -1);
631 if (change & IFF_PROMISC)
632 dev_set_promiscuity(real_dev, dev->flags & IFF_PROMISC ? 1 : -1);
635 static void vlan_dev_set_rx_mode(struct net_device *vlan_dev)
637 dev_mc_sync(vlan_dev_info(vlan_dev)->real_dev, vlan_dev);
638 dev_unicast_sync(vlan_dev_info(vlan_dev)->real_dev, vlan_dev);
642 * vlan network devices have devices nesting below it, and are a special
643 * "super class" of normal network devices; split their locks off into a
644 * separate class since they always nest.
646 static struct lock_class_key vlan_netdev_xmit_lock_key;
648 static const struct header_ops vlan_header_ops = {
649 .create = vlan_dev_hard_header,
650 .rebuild = vlan_dev_rebuild_header,
651 .parse = eth_header_parse,
654 static int vlan_dev_init(struct net_device *dev)
656 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
659 /* IFF_BROADCAST|IFF_MULTICAST; ??? */
660 dev->flags = real_dev->flags & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI);
661 dev->iflink = real_dev->ifindex;
662 dev->state = (real_dev->state & ((1<<__LINK_STATE_NOCARRIER) |
663 (1<<__LINK_STATE_DORMANT))) |
664 (1<<__LINK_STATE_PRESENT);
666 dev->features |= real_dev->features & real_dev->vlan_features;
668 /* ipv6 shared card related stuff */
669 dev->dev_id = real_dev->dev_id;
671 if (is_zero_ether_addr(dev->dev_addr))
672 memcpy(dev->dev_addr, real_dev->dev_addr, dev->addr_len);
673 if (is_zero_ether_addr(dev->broadcast))
674 memcpy(dev->broadcast, real_dev->broadcast, dev->addr_len);
676 if (real_dev->features & NETIF_F_HW_VLAN_TX) {
677 dev->header_ops = real_dev->header_ops;
678 dev->hard_header_len = real_dev->hard_header_len;
679 dev->hard_start_xmit = vlan_dev_hwaccel_hard_start_xmit;
681 dev->header_ops = &vlan_header_ops;
682 dev->hard_header_len = real_dev->hard_header_len + VLAN_HLEN;
683 dev->hard_start_xmit = vlan_dev_hard_start_xmit;
686 if (real_dev->priv_flags & IFF_802_1Q_VLAN)
689 lockdep_set_class_and_subclass(&dev->_xmit_lock,
690 &vlan_netdev_xmit_lock_key, subclass);
694 static void vlan_dev_uninit(struct net_device *dev)
696 struct vlan_priority_tci_mapping *pm;
697 struct vlan_dev_info *vlan = vlan_dev_info(dev);
700 for (i = 0; i < ARRAY_SIZE(vlan->egress_priority_map); i++) {
701 while ((pm = vlan->egress_priority_map[i]) != NULL) {
702 vlan->egress_priority_map[i] = pm->next;
708 void vlan_setup(struct net_device *dev)
712 dev->priv_flags |= IFF_802_1Q_VLAN;
713 dev->tx_queue_len = 0;
715 dev->change_mtu = vlan_dev_change_mtu;
716 dev->init = vlan_dev_init;
717 dev->uninit = vlan_dev_uninit;
718 dev->open = vlan_dev_open;
719 dev->stop = vlan_dev_stop;
720 dev->set_mac_address = vlan_dev_set_mac_address;
721 dev->set_rx_mode = vlan_dev_set_rx_mode;
722 dev->set_multicast_list = vlan_dev_set_rx_mode;
723 dev->change_rx_flags = vlan_dev_change_rx_flags;
724 dev->do_ioctl = vlan_dev_ioctl;
725 dev->destructor = free_netdev;
727 memset(dev->broadcast, 0, ETH_ALEN);