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 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);
64 "%s: unable to resolve type %X addresses.\n",
65 dev->name, ntohs(veth->h_vlan_encapsulated_proto));
67 memcpy(veth->h_source, dev->dev_addr, ETH_ALEN);
74 static inline struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
76 if (VLAN_DEV_INFO(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
77 if (skb_shared(skb) || skb_cloned(skb)) {
78 struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
83 /* Lifted from Gleb's VLAN code... */
84 memmove(skb->data - ETH_HLEN,
85 skb->data - VLAN_ETH_HLEN, 12);
86 skb->mac_header += VLAN_HLEN;
94 * Determine the packet's protocol ID. The rule here is that we
95 * assume 802.3 if the type field is short enough to be a length.
96 * This is normal practice and works for any 'now in use' protocol.
98 * Also, at this point we assume that we ARE dealing exclusively with
99 * VLAN packets, or packets that should be made into VLAN packets based
100 * on a default VLAN ID.
102 * NOTE: Should be similar to ethernet/eth.c.
104 * SANITY NOTE: This method is called when a packet is moving up the stack
105 * towards userland. To get here, it would have already passed
106 * through the ethernet/eth.c eth_type_trans() method.
107 * SANITY NOTE 2: We are referencing to the VLAN_HDR frields, which MAY be
108 * stored UNALIGNED in the memory. RISC systems don't like
109 * such cases very much...
110 * SANITY NOTE 2a: According to Dave Miller & Alexey, it will always be aligned,
111 * so there doesn't need to be any of the unaligned stuff. It has
112 * been commented out now... --Ben
115 int vlan_skb_recv(struct sk_buff *skb, struct net_device *dev,
116 struct packet_type* ptype, struct net_device *orig_dev)
118 unsigned char *rawp = NULL;
119 struct vlan_hdr *vhdr;
121 struct net_device_stats *stats;
122 unsigned short vlan_TCI;
125 if (dev->nd_net != &init_net) {
130 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
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);
146 printk(VLAN_DBG "%s: skb: %p vlan_id: %hx\n",
147 __FUNCTION__, skb, vid);
150 /* Ok, we will find the correct VLAN device, strip the header,
151 * and then go on as usual.
154 /* We have 12 bits of vlan ID.
156 * We must not drop allow preempt until we hold a
157 * reference to the device (netif_rx does that) or we
162 skb->dev = __find_vlan_dev(dev, vid);
167 printk(VLAN_DBG "%s: ERROR: No net_device for VID: %i on dev: %s [%i]\n",
168 __FUNCTION__, (unsigned int)(vid), dev->name, dev->ifindex);
174 skb->dev->last_rx = jiffies;
176 /* Bump the rx counters for the VLAN device. */
177 stats = vlan_dev_get_stats(skb->dev);
179 stats->rx_bytes += skb->len;
181 /* Take off the VLAN header (4 bytes currently) */
182 skb_pull_rcsum(skb, VLAN_HLEN);
184 /* Ok, lets check to make sure the device (dev) we
185 * came in on is what this VLAN is attached to.
188 if (dev != VLAN_DEV_INFO(skb->dev)->real_dev) {
192 printk(VLAN_DBG "%s: dropping skb: %p because came in on wrong device, dev: %s real_dev: %s, skb_dev: %s\n",
193 __FUNCTION__, skb, dev->name,
194 VLAN_DEV_INFO(skb->dev)->real_dev->name,
203 * Deal with ingress priority mapping.
205 skb->priority = vlan_get_ingress_priority(skb->dev, ntohs(vhdr->h_vlan_TCI));
208 printk(VLAN_DBG "%s: priority: %lu for TCI: %hu (hbo)\n",
209 __FUNCTION__, (unsigned long)(skb->priority),
210 ntohs(vhdr->h_vlan_TCI));
213 /* The ethernet driver already did the pkt_type calculations
216 switch (skb->pkt_type) {
217 case PACKET_BROADCAST: /* Yeah, stats collect these together.. */
218 // stats->broadcast ++; // no such counter :-(
221 case PACKET_MULTICAST:
225 case PACKET_OTHERHOST:
226 /* Our lower layer thinks this is not local, let's make sure.
227 * This allows the VLAN to have a different MAC than the underlying
228 * device, and still route correctly.
230 if (!compare_ether_addr(eth_hdr(skb)->h_dest, skb->dev->dev_addr)) {
231 /* It is for our (changed) MAC-address! */
232 skb->pkt_type = PACKET_HOST;
239 /* Was a VLAN packet, grab the encapsulated protocol, which the layer
240 * three protocols care about.
242 /* proto = get_unaligned(&vhdr->h_vlan_encapsulated_proto); */
243 proto = vhdr->h_vlan_encapsulated_proto;
245 skb->protocol = proto;
246 if (ntohs(proto) >= 1536) {
247 /* place it back on the queue to be handled by
248 * true layer 3 protocols.
251 /* See if we are configured to re-write the VLAN header
252 * to make it look like ethernet...
254 skb = vlan_check_reorder_header(skb);
256 /* Can be null if skb-clone fails when re-ordering */
260 /* TODO: Add a more specific counter here. */
270 * This is a magic hack to spot IPX packets. Older Novell breaks
271 * the protocol design and runs IPX over 802.3 without an 802.2 LLC
272 * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
273 * won't work for fault tolerant netware but does for the rest.
275 if (*(unsigned short *)rawp == 0xFFFF) {
276 skb->protocol = htons(ETH_P_802_3);
277 /* place it back on the queue to be handled by true layer 3 protocols.
280 /* See if we are configured to re-write the VLAN header
281 * to make it look like ethernet...
283 skb = vlan_check_reorder_header(skb);
285 /* Can be null if skb-clone fails when re-ordering */
289 /* TODO: Add a more specific counter here. */
299 skb->protocol = htons(ETH_P_802_2);
300 /* place it back on the queue to be handled by upper layer protocols.
303 /* See if we are configured to re-write the VLAN header
304 * to make it look like ethernet...
306 skb = vlan_check_reorder_header(skb);
308 /* Can be null if skb-clone fails when re-ordering */
312 /* TODO: Add a more specific counter here. */
319 static inline unsigned short vlan_dev_get_egress_qos_mask(struct net_device* dev,
322 struct vlan_priority_tci_mapping *mp =
323 VLAN_DEV_INFO(dev)->egress_priority_map[(skb->priority & 0xF)];
326 if (mp->priority == skb->priority) {
327 return mp->vlan_qos; /* This should already be shifted to mask
328 * correctly with the VLAN's TCI
337 * Create the VLAN header for an arbitrary protocol layer
339 * saddr=NULL means use device source address
340 * daddr=NULL means leave destination address (eg unresolved arp)
342 * This is called when the SKB is moving down the stack towards the
345 int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
346 unsigned short type, void *daddr, void *saddr,
349 struct vlan_hdr *vhdr;
350 unsigned short veth_TCI = 0;
352 int build_vlan_header = 0;
353 struct net_device *vdev = dev; /* save this for the bottom of the method */
356 printk(VLAN_DBG "%s: skb: %p type: %hx len: %x vlan_id: %hx, daddr: %p\n",
357 __FUNCTION__, skb, type, len, VLAN_DEV_INFO(dev)->vlan_id, daddr);
360 /* build vlan header only if re_order_header flag is NOT set. This
361 * fixes some programs that get confused when they see a VLAN device
362 * sending a frame that is VLAN encoded (the consensus is that the VLAN
363 * device should look completely like an Ethernet device when the
364 * REORDER_HEADER flag is set) The drawback to this is some extra
365 * header shuffling in the hard_start_xmit. Users can turn off this
366 * REORDER behaviour with the vconfig tool.
368 if (!(VLAN_DEV_INFO(dev)->flags & VLAN_FLAG_REORDER_HDR))
369 build_vlan_header = 1;
371 if (build_vlan_header) {
372 vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
374 /* build the four bytes that make this a VLAN header. */
376 /* Now, construct the second two bytes. This field looks something
378 * usr_priority: 3 bits (high bits)
380 * VLAN ID 12 bits (low bits)
383 veth_TCI = VLAN_DEV_INFO(dev)->vlan_id;
384 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
386 vhdr->h_vlan_TCI = htons(veth_TCI);
389 * Set the protocol type.
390 * For a packet of type ETH_P_802_3 we put the length in here instead.
391 * It is up to the 802.2 layer to carry protocol information.
394 if (type != ETH_P_802_3) {
395 vhdr->h_vlan_encapsulated_proto = htons(type);
397 vhdr->h_vlan_encapsulated_proto = htons(len);
400 skb->protocol = htons(ETH_P_8021Q);
401 skb_reset_network_header(skb);
404 /* Before delegating work to the lower layer, enter our MAC-address */
406 saddr = dev->dev_addr;
408 dev = VLAN_DEV_INFO(dev)->real_dev;
410 /* MPLS can send us skbuffs w/out enough space. This check will grow the
411 * skb if it doesn't have enough headroom. Not a beautiful solution, so
412 * I'll tick a counter so that users can know it's happening... If they
416 /* NOTE: This may still break if the underlying device is not the final
417 * device (and thus there are more headers to add...) It should work for
418 * good-ole-ethernet though.
420 if (skb_headroom(skb) < dev->hard_header_len) {
421 struct sk_buff *sk_tmp = skb;
422 skb = skb_realloc_headroom(sk_tmp, dev->hard_header_len);
425 struct net_device_stats *stats = vlan_dev_get_stats(vdev);
429 VLAN_DEV_INFO(vdev)->cnt_inc_headroom_on_tx++;
431 printk(VLAN_DBG "%s: %s: had to grow skb.\n", __FUNCTION__, vdev->name);
435 if (build_vlan_header) {
436 /* Now make the underlying real hard header */
437 rc = dev->hard_header(skb, dev, ETH_P_8021Q, daddr, saddr, len + VLAN_HLEN);
445 /* If here, then we'll just make a normal looking ethernet frame,
446 * but, the hard_start_xmit method will insert the tag (it has to
447 * be able to do this for bridged and other skbs that don't come
448 * down the protocol stack in an orderly manner.
450 rc = dev->hard_header(skb, dev, type, daddr, saddr, len);
456 int vlan_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
458 struct net_device_stats *stats = vlan_dev_get_stats(dev);
459 struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
461 /* Handle non-VLAN frames if they are sent to us, for example by DHCP.
463 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
464 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
467 if (veth->h_vlan_proto != htons(ETH_P_8021Q)) {
468 int orig_headroom = skb_headroom(skb);
469 unsigned short veth_TCI;
471 /* This is not a VLAN frame...but we can fix that! */
472 VLAN_DEV_INFO(dev)->cnt_encap_on_xmit++;
475 printk(VLAN_DBG "%s: proto to encap: 0x%hx (hbo)\n",
476 __FUNCTION__, htons(veth->h_vlan_proto));
478 /* Construct the second two bytes. This field looks something
480 * usr_priority: 3 bits (high bits)
482 * VLAN ID 12 bits (low bits)
484 veth_TCI = VLAN_DEV_INFO(dev)->vlan_id;
485 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
487 skb = __vlan_put_tag(skb, veth_TCI);
493 if (orig_headroom < VLAN_HLEN) {
494 VLAN_DEV_INFO(dev)->cnt_inc_headroom_on_tx++;
499 printk(VLAN_DBG "%s: about to send skb: %p to dev: %s\n",
500 __FUNCTION__, skb, skb->dev->name);
501 printk(VLAN_DBG " %2hx.%2hx.%2hx.%2xh.%2hx.%2hx %2hx.%2hx.%2hx.%2hx.%2hx.%2hx %4hx %4hx %4hx\n",
502 veth->h_dest[0], veth->h_dest[1], veth->h_dest[2], veth->h_dest[3], veth->h_dest[4], veth->h_dest[5],
503 veth->h_source[0], veth->h_source[1], veth->h_source[2], veth->h_source[3], veth->h_source[4], veth->h_source[5],
504 veth->h_vlan_proto, veth->h_vlan_TCI, veth->h_vlan_encapsulated_proto);
507 stats->tx_packets++; /* for statics only */
508 stats->tx_bytes += skb->len;
510 skb->dev = VLAN_DEV_INFO(dev)->real_dev;
516 int vlan_dev_hwaccel_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
518 struct net_device_stats *stats = vlan_dev_get_stats(dev);
519 unsigned short veth_TCI;
521 /* Construct the second two bytes. This field looks something
523 * usr_priority: 3 bits (high bits)
525 * VLAN ID 12 bits (low bits)
527 veth_TCI = VLAN_DEV_INFO(dev)->vlan_id;
528 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
529 skb = __vlan_hwaccel_put_tag(skb, veth_TCI);
532 stats->tx_bytes += skb->len;
534 skb->dev = VLAN_DEV_INFO(dev)->real_dev;
540 int vlan_dev_change_mtu(struct net_device *dev, int new_mtu)
542 /* TODO: gotta make sure the underlying layer can handle it,
543 * maybe an IFF_VLAN_CAPABLE flag for devices?
545 if (VLAN_DEV_INFO(dev)->real_dev->mtu < new_mtu)
553 void vlan_dev_set_ingress_priority(const struct net_device *dev,
554 u32 skb_prio, short vlan_prio)
556 struct vlan_dev_info *vlan = VLAN_DEV_INFO(dev);
558 if (vlan->ingress_priority_map[vlan_prio & 0x7] && !skb_prio)
559 vlan->nr_ingress_mappings--;
560 else if (!vlan->ingress_priority_map[vlan_prio & 0x7] && skb_prio)
561 vlan->nr_ingress_mappings++;
563 vlan->ingress_priority_map[vlan_prio & 0x7] = skb_prio;
566 int vlan_dev_set_egress_priority(const struct net_device *dev,
567 u32 skb_prio, short vlan_prio)
569 struct vlan_dev_info *vlan = VLAN_DEV_INFO(dev);
570 struct vlan_priority_tci_mapping *mp = NULL;
571 struct vlan_priority_tci_mapping *np;
572 u32 vlan_qos = (vlan_prio << 13) & 0xE000;
574 /* See if a priority mapping exists.. */
575 mp = vlan->egress_priority_map[skb_prio & 0xF];
577 if (mp->priority == skb_prio) {
578 if (mp->vlan_qos && !vlan_qos)
579 vlan->nr_egress_mappings--;
580 else if (!mp->vlan_qos && vlan_qos)
581 vlan->nr_egress_mappings++;
582 mp->vlan_qos = vlan_qos;
588 /* Create a new mapping then. */
589 mp = vlan->egress_priority_map[skb_prio & 0xF];
590 np = kmalloc(sizeof(struct vlan_priority_tci_mapping), GFP_KERNEL);
595 np->priority = skb_prio;
596 np->vlan_qos = vlan_qos;
597 vlan->egress_priority_map[skb_prio & 0xF] = np;
599 vlan->nr_egress_mappings++;
603 /* Flags are defined in the vlan_flags enum in include/linux/if_vlan.h file. */
604 int vlan_dev_set_vlan_flag(const struct net_device *dev,
605 u32 flag, short flag_val)
607 /* verify flag is supported */
608 if (flag == VLAN_FLAG_REORDER_HDR) {
610 VLAN_DEV_INFO(dev)->flags |= VLAN_FLAG_REORDER_HDR;
612 VLAN_DEV_INFO(dev)->flags &= ~VLAN_FLAG_REORDER_HDR;
616 printk(KERN_ERR "%s: flag %i is not valid.\n", __FUNCTION__, flag);
620 void vlan_dev_get_realdev_name(const struct net_device *dev, char *result)
622 strncpy(result, VLAN_DEV_INFO(dev)->real_dev->name, 23);
625 void vlan_dev_get_vid(const struct net_device *dev, unsigned short *result)
627 *result = VLAN_DEV_INFO(dev)->vlan_id;
630 int vlan_dev_open(struct net_device *dev)
632 struct vlan_dev_info *vlan = VLAN_DEV_INFO(dev);
633 struct net_device *real_dev = vlan->real_dev;
636 if (!(real_dev->flags & IFF_UP))
639 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr)) {
640 err = dev_unicast_add(real_dev, dev->dev_addr, ETH_ALEN);
644 memcpy(vlan->real_dev_addr, real_dev->dev_addr, ETH_ALEN);
646 if (dev->flags & IFF_ALLMULTI)
647 dev_set_allmulti(real_dev, 1);
648 if (dev->flags & IFF_PROMISC)
649 dev_set_promiscuity(real_dev, 1);
654 int vlan_dev_stop(struct net_device *dev)
656 struct net_device *real_dev = VLAN_DEV_INFO(dev)->real_dev;
658 dev_mc_unsync(real_dev, dev);
659 if (dev->flags & IFF_ALLMULTI)
660 dev_set_allmulti(real_dev, -1);
661 if (dev->flags & IFF_PROMISC)
662 dev_set_promiscuity(real_dev, -1);
664 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
665 dev_unicast_delete(real_dev, dev->dev_addr, dev->addr_len);
670 int vlan_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
672 struct net_device *real_dev = VLAN_DEV_INFO(dev)->real_dev;
674 int err = -EOPNOTSUPP;
676 strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
677 ifrr.ifr_ifru = ifr->ifr_ifru;
683 if (real_dev->do_ioctl && netif_device_present(real_dev))
684 err = real_dev->do_ioctl(real_dev, &ifrr, cmd);
689 ifr->ifr_ifru = ifrr.ifr_ifru;
694 void vlan_change_rx_flags(struct net_device *dev, int change)
696 struct net_device *real_dev = VLAN_DEV_INFO(dev)->real_dev;
698 if (change & IFF_ALLMULTI)
699 dev_set_allmulti(real_dev, dev->flags & IFF_ALLMULTI ? 1 : -1);
700 if (change & IFF_PROMISC)
701 dev_set_promiscuity(real_dev, dev->flags & IFF_PROMISC ? 1 : -1);
704 /** Taken from Gleb + Lennert's VLAN code, and modified... */
705 void vlan_dev_set_multicast_list(struct net_device *vlan_dev)
707 dev_mc_sync(VLAN_DEV_INFO(vlan_dev)->real_dev, vlan_dev);