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 if (dev->nd_net != &init_net)
159 skb = skb_share_check(skb, GFP_ATOMIC);
163 if (unlikely(!pskb_may_pull(skb, VLAN_HLEN)))
166 vhdr = (struct vlan_hdr *)skb->data;
167 vlan_TCI = ntohs(vhdr->h_vlan_TCI);
168 vid = (vlan_TCI & VLAN_VID_MASK);
171 skb->dev = __find_vlan_dev(dev, vid);
173 pr_debug("%s: ERROR: No net_device for VID: %u on dev: %s\n",
174 __FUNCTION__, (unsigned int)vid, dev->name);
178 skb->dev->last_rx = jiffies;
180 stats = &skb->dev->stats;
182 stats->rx_bytes += skb->len;
184 skb_pull_rcsum(skb, VLAN_HLEN);
186 skb->priority = vlan_get_ingress_priority(skb->dev,
187 ntohs(vhdr->h_vlan_TCI));
189 pr_debug("%s: priority: %u for TCI: %hu\n",
190 __FUNCTION__, skb->priority, ntohs(vhdr->h_vlan_TCI));
192 switch (skb->pkt_type) {
193 case PACKET_BROADCAST: /* Yeah, stats collect these together.. */
194 /* stats->broadcast ++; // no such counter :-( */
197 case PACKET_MULTICAST:
201 case PACKET_OTHERHOST:
202 /* Our lower layer thinks this is not local, let's make sure.
203 * This allows the VLAN to have a different MAC than the
204 * underlying device, and still route correctly.
206 if (!compare_ether_addr(eth_hdr(skb)->h_dest,
208 skb->pkt_type = PACKET_HOST;
214 vlan_set_encap_proto(skb, vhdr);
216 skb = vlan_check_reorder_header(skb);
224 return NET_RX_SUCCESS;
233 static inline unsigned short
234 vlan_dev_get_egress_qos_mask(struct net_device *dev, struct sk_buff *skb)
236 struct vlan_priority_tci_mapping *mp;
238 mp = vlan_dev_info(dev)->egress_priority_map[(skb->priority & 0xF)];
240 if (mp->priority == skb->priority) {
241 return mp->vlan_qos; /* This should already be shifted
242 * to mask correctly with the
251 * Create the VLAN header for an arbitrary protocol layer
253 * saddr=NULL means use device source address
254 * daddr=NULL means leave destination address (eg unresolved arp)
256 * This is called when the SKB is moving down the stack towards the
259 static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
261 const void *daddr, const void *saddr,
264 struct vlan_hdr *vhdr;
265 unsigned short veth_TCI = 0;
267 int build_vlan_header = 0;
268 struct net_device *vdev = dev;
270 pr_debug("%s: skb: %p type: %hx len: %u vlan_id: %hx, daddr: %p\n",
271 __FUNCTION__, skb, type, len, vlan_dev_info(dev)->vlan_id,
274 /* build vlan header only if re_order_header flag is NOT set. This
275 * fixes some programs that get confused when they see a VLAN device
276 * sending a frame that is VLAN encoded (the consensus is that the VLAN
277 * device should look completely like an Ethernet device when the
278 * REORDER_HEADER flag is set) The drawback to this is some extra
279 * header shuffling in the hard_start_xmit. Users can turn off this
280 * REORDER behaviour with the vconfig tool.
282 if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR))
283 build_vlan_header = 1;
285 if (build_vlan_header) {
286 vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
288 /* build the four bytes that make this a VLAN header. */
290 /* Now, construct the second two bytes. This field looks
292 * usr_priority: 3 bits (high bits)
294 * VLAN ID 12 bits (low bits)
297 veth_TCI = vlan_dev_info(dev)->vlan_id;
298 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
300 vhdr->h_vlan_TCI = htons(veth_TCI);
303 * Set the protocol type. For a packet of type ETH_P_802_3 we
304 * put the length in here instead. It is up to the 802.2
305 * layer to carry protocol information.
308 if (type != ETH_P_802_3)
309 vhdr->h_vlan_encapsulated_proto = htons(type);
311 vhdr->h_vlan_encapsulated_proto = htons(len);
313 skb->protocol = htons(ETH_P_8021Q);
314 skb_reset_network_header(skb);
317 /* Before delegating work to the lower layer, enter our MAC-address */
319 saddr = dev->dev_addr;
321 dev = vlan_dev_info(dev)->real_dev;
323 /* MPLS can send us skbuffs w/out enough space. This check will grow
324 * the skb if it doesn't have enough headroom. Not a beautiful solution,
325 * so I'll tick a counter so that users can know it's happening...
329 /* NOTE: This may still break if the underlying device is not the final
330 * device (and thus there are more headers to add...) It should work for
331 * good-ole-ethernet though.
333 if (skb_headroom(skb) < dev->hard_header_len) {
334 struct sk_buff *sk_tmp = skb;
335 skb = skb_realloc_headroom(sk_tmp, dev->hard_header_len);
338 struct net_device_stats *stats = &vdev->stats;
342 vlan_dev_info(vdev)->cnt_inc_headroom_on_tx++;
343 pr_debug("%s: %s: had to grow skb\n", __FUNCTION__, vdev->name);
346 if (build_vlan_header) {
347 /* Now make the underlying real hard header */
348 rc = dev_hard_header(skb, dev, ETH_P_8021Q, daddr, saddr,
355 /* If here, then we'll just make a normal looking ethernet
356 * frame, but, the hard_start_xmit method will insert the tag
357 * (it has to be able to do this for bridged and other skbs
358 * that don't come down the protocol stack in an orderly manner.
360 rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
365 static int vlan_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
367 struct net_device_stats *stats = &dev->stats;
368 struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
370 /* Handle non-VLAN frames if they are sent to us, for example by DHCP.
372 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
373 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
376 if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
377 vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR) {
378 int orig_headroom = skb_headroom(skb);
379 unsigned short veth_TCI;
381 /* This is not a VLAN frame...but we can fix that! */
382 vlan_dev_info(dev)->cnt_encap_on_xmit++;
384 pr_debug("%s: proto to encap: 0x%hx\n",
385 __FUNCTION__, htons(veth->h_vlan_proto));
386 /* Construct the second two bytes. This field looks something
388 * usr_priority: 3 bits (high bits)
390 * VLAN ID 12 bits (low bits)
392 veth_TCI = vlan_dev_info(dev)->vlan_id;
393 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
395 skb = __vlan_put_tag(skb, veth_TCI);
401 if (orig_headroom < VLAN_HLEN)
402 vlan_dev_info(dev)->cnt_inc_headroom_on_tx++;
405 pr_debug("%s: about to send skb: %p to dev: %s\n",
406 __FUNCTION__, skb, skb->dev->name);
407 pr_debug(" " MAC_FMT " " MAC_FMT " %4hx %4hx %4hx\n",
408 veth->h_dest[0], veth->h_dest[1], veth->h_dest[2],
409 veth->h_dest[3], veth->h_dest[4], veth->h_dest[5],
410 veth->h_source[0], veth->h_source[1], veth->h_source[2],
411 veth->h_source[3], veth->h_source[4], veth->h_source[5],
412 veth->h_vlan_proto, veth->h_vlan_TCI,
413 veth->h_vlan_encapsulated_proto);
415 stats->tx_packets++; /* for statics only */
416 stats->tx_bytes += skb->len;
418 skb->dev = vlan_dev_info(dev)->real_dev;
424 static int vlan_dev_hwaccel_hard_start_xmit(struct sk_buff *skb,
425 struct net_device *dev)
427 struct net_device_stats *stats = &dev->stats;
428 unsigned short veth_TCI;
430 /* Construct the second two bytes. This field looks something
432 * usr_priority: 3 bits (high bits)
434 * VLAN ID 12 bits (low bits)
436 veth_TCI = vlan_dev_info(dev)->vlan_id;
437 veth_TCI |= vlan_dev_get_egress_qos_mask(dev, skb);
438 skb = __vlan_hwaccel_put_tag(skb, veth_TCI);
441 stats->tx_bytes += skb->len;
443 skb->dev = vlan_dev_info(dev)->real_dev;
449 static int vlan_dev_change_mtu(struct net_device *dev, int new_mtu)
451 /* TODO: gotta make sure the underlying layer can handle it,
452 * maybe an IFF_VLAN_CAPABLE flag for devices?
454 if (vlan_dev_info(dev)->real_dev->mtu < new_mtu)
462 void vlan_dev_set_ingress_priority(const struct net_device *dev,
463 u32 skb_prio, short vlan_prio)
465 struct vlan_dev_info *vlan = vlan_dev_info(dev);
467 if (vlan->ingress_priority_map[vlan_prio & 0x7] && !skb_prio)
468 vlan->nr_ingress_mappings--;
469 else if (!vlan->ingress_priority_map[vlan_prio & 0x7] && skb_prio)
470 vlan->nr_ingress_mappings++;
472 vlan->ingress_priority_map[vlan_prio & 0x7] = skb_prio;
475 int vlan_dev_set_egress_priority(const struct net_device *dev,
476 u32 skb_prio, short vlan_prio)
478 struct vlan_dev_info *vlan = vlan_dev_info(dev);
479 struct vlan_priority_tci_mapping *mp = NULL;
480 struct vlan_priority_tci_mapping *np;
481 u32 vlan_qos = (vlan_prio << 13) & 0xE000;
483 /* See if a priority mapping exists.. */
484 mp = vlan->egress_priority_map[skb_prio & 0xF];
486 if (mp->priority == skb_prio) {
487 if (mp->vlan_qos && !vlan_qos)
488 vlan->nr_egress_mappings--;
489 else if (!mp->vlan_qos && vlan_qos)
490 vlan->nr_egress_mappings++;
491 mp->vlan_qos = vlan_qos;
497 /* Create a new mapping then. */
498 mp = vlan->egress_priority_map[skb_prio & 0xF];
499 np = kmalloc(sizeof(struct vlan_priority_tci_mapping), GFP_KERNEL);
504 np->priority = skb_prio;
505 np->vlan_qos = vlan_qos;
506 vlan->egress_priority_map[skb_prio & 0xF] = np;
508 vlan->nr_egress_mappings++;
512 /* Flags are defined in the vlan_flags enum in include/linux/if_vlan.h file. */
513 int vlan_dev_set_vlan_flag(const struct net_device *dev,
514 u32 flag, short flag_val)
516 /* verify flag is supported */
517 if (flag == VLAN_FLAG_REORDER_HDR) {
519 vlan_dev_info(dev)->flags |= VLAN_FLAG_REORDER_HDR;
521 vlan_dev_info(dev)->flags &= ~VLAN_FLAG_REORDER_HDR;
527 void vlan_dev_get_realdev_name(const struct net_device *dev, char *result)
529 strncpy(result, vlan_dev_info(dev)->real_dev->name, 23);
532 void vlan_dev_get_vid(const struct net_device *dev, unsigned short *result)
534 *result = vlan_dev_info(dev)->vlan_id;
537 static int vlan_dev_open(struct net_device *dev)
539 struct vlan_dev_info *vlan = vlan_dev_info(dev);
540 struct net_device *real_dev = vlan->real_dev;
543 if (!(real_dev->flags & IFF_UP))
546 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr)) {
547 err = dev_unicast_add(real_dev, dev->dev_addr, ETH_ALEN);
551 memcpy(vlan->real_dev_addr, real_dev->dev_addr, ETH_ALEN);
553 if (dev->flags & IFF_ALLMULTI)
554 dev_set_allmulti(real_dev, 1);
555 if (dev->flags & IFF_PROMISC)
556 dev_set_promiscuity(real_dev, 1);
561 static int vlan_dev_stop(struct net_device *dev)
563 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
565 dev_mc_unsync(real_dev, dev);
566 dev_unicast_unsync(real_dev, dev);
567 if (dev->flags & IFF_ALLMULTI)
568 dev_set_allmulti(real_dev, -1);
569 if (dev->flags & IFF_PROMISC)
570 dev_set_promiscuity(real_dev, -1);
572 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
573 dev_unicast_delete(real_dev, dev->dev_addr, dev->addr_len);
578 static int vlan_dev_set_mac_address(struct net_device *dev, void *p)
580 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
581 struct sockaddr *addr = p;
584 if (!is_valid_ether_addr(addr->sa_data))
585 return -EADDRNOTAVAIL;
587 if (!(dev->flags & IFF_UP))
590 if (compare_ether_addr(addr->sa_data, real_dev->dev_addr)) {
591 err = dev_unicast_add(real_dev, addr->sa_data, ETH_ALEN);
596 if (compare_ether_addr(dev->dev_addr, real_dev->dev_addr))
597 dev_unicast_delete(real_dev, dev->dev_addr, ETH_ALEN);
600 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
604 static int vlan_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
606 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
608 int err = -EOPNOTSUPP;
610 strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
611 ifrr.ifr_ifru = ifr->ifr_ifru;
617 if (real_dev->do_ioctl && netif_device_present(real_dev))
618 err = real_dev->do_ioctl(real_dev, &ifrr, cmd);
623 ifr->ifr_ifru = ifrr.ifr_ifru;
628 static void vlan_dev_change_rx_flags(struct net_device *dev, int change)
630 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
632 if (change & IFF_ALLMULTI)
633 dev_set_allmulti(real_dev, dev->flags & IFF_ALLMULTI ? 1 : -1);
634 if (change & IFF_PROMISC)
635 dev_set_promiscuity(real_dev, dev->flags & IFF_PROMISC ? 1 : -1);
638 static void vlan_dev_set_rx_mode(struct net_device *vlan_dev)
640 dev_mc_sync(vlan_dev_info(vlan_dev)->real_dev, vlan_dev);
641 dev_unicast_sync(vlan_dev_info(vlan_dev)->real_dev, vlan_dev);
645 * vlan network devices have devices nesting below it, and are a special
646 * "super class" of normal network devices; split their locks off into a
647 * separate class since they always nest.
649 static struct lock_class_key vlan_netdev_xmit_lock_key;
651 static const struct header_ops vlan_header_ops = {
652 .create = vlan_dev_hard_header,
653 .rebuild = vlan_dev_rebuild_header,
654 .parse = eth_header_parse,
657 static int vlan_dev_init(struct net_device *dev)
659 struct net_device *real_dev = vlan_dev_info(dev)->real_dev;
662 /* IFF_BROADCAST|IFF_MULTICAST; ??? */
663 dev->flags = real_dev->flags & ~IFF_UP;
664 dev->iflink = real_dev->ifindex;
665 dev->state = (real_dev->state & ((1<<__LINK_STATE_NOCARRIER) |
666 (1<<__LINK_STATE_DORMANT))) |
667 (1<<__LINK_STATE_PRESENT);
669 /* ipv6 shared card related stuff */
670 dev->dev_id = real_dev->dev_id;
672 if (is_zero_ether_addr(dev->dev_addr))
673 memcpy(dev->dev_addr, real_dev->dev_addr, dev->addr_len);
674 if (is_zero_ether_addr(dev->broadcast))
675 memcpy(dev->broadcast, real_dev->broadcast, dev->addr_len);
677 if (real_dev->features & NETIF_F_HW_VLAN_TX) {
678 dev->header_ops = real_dev->header_ops;
679 dev->hard_header_len = real_dev->hard_header_len;
680 dev->hard_start_xmit = vlan_dev_hwaccel_hard_start_xmit;
682 dev->header_ops = &vlan_header_ops;
683 dev->hard_header_len = real_dev->hard_header_len + VLAN_HLEN;
684 dev->hard_start_xmit = vlan_dev_hard_start_xmit;
687 if (real_dev->priv_flags & IFF_802_1Q_VLAN)
690 lockdep_set_class_and_subclass(&dev->_xmit_lock,
691 &vlan_netdev_xmit_lock_key, subclass);
695 void vlan_setup(struct net_device *dev)
699 dev->priv_flags |= IFF_802_1Q_VLAN;
700 dev->tx_queue_len = 0;
702 dev->change_mtu = vlan_dev_change_mtu;
703 dev->init = vlan_dev_init;
704 dev->open = vlan_dev_open;
705 dev->stop = vlan_dev_stop;
706 dev->set_mac_address = vlan_dev_set_mac_address;
707 dev->set_rx_mode = vlan_dev_set_rx_mode;
708 dev->set_multicast_list = vlan_dev_set_rx_mode;
709 dev->change_rx_flags = vlan_dev_change_rx_flags;
710 dev->do_ioctl = vlan_dev_ioctl;
711 dev->destructor = free_netdev;
713 memset(dev->broadcast, 0, ETH_ALEN);