2 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of
7 * the License, or (at your option) any later version.
9 * The code this is based on carried the following copyright notice:
11 * (C) Copyright 2001-2006
12 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
13 * Re-worked by Ben Greear <greearb@candelatech.com>
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/init.h>
20 #include <linux/errno.h>
21 #include <linux/slab.h>
22 #include <linux/string.h>
23 #include <linux/rculist.h>
24 #include <linux/notifier.h>
25 #include <linux/netdevice.h>
26 #include <linux/etherdevice.h>
27 #include <linux/ethtool.h>
28 #include <linux/if_arp.h>
29 #include <linux/if_vlan.h>
30 #include <linux/if_link.h>
31 #include <linux/if_macvlan.h>
32 #include <linux/hash.h>
33 #include <linux/workqueue.h>
34 #include <net/rtnetlink.h>
36 #include <linux/netpoll.h>
38 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE)
41 struct net_device
*dev
;
42 struct hlist_head vlan_hash
[MACVLAN_HASH_SIZE
];
43 struct list_head vlans
;
45 struct sk_buff_head bc_queue
;
46 struct work_struct bc_work
;
51 struct macvlan_skb_cb
{
52 const struct macvlan_dev
*src
;
55 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
57 static void macvlan_port_destroy(struct net_device
*dev
);
59 static struct macvlan_port
*macvlan_port_get_rcu(const struct net_device
*dev
)
61 return rcu_dereference(dev
->rx_handler_data
);
64 static struct macvlan_port
*macvlan_port_get_rtnl(const struct net_device
*dev
)
66 return rtnl_dereference(dev
->rx_handler_data
);
69 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
71 static struct macvlan_dev
*macvlan_hash_lookup(const struct macvlan_port
*port
,
72 const unsigned char *addr
)
74 struct macvlan_dev
*vlan
;
76 hlist_for_each_entry_rcu(vlan
, &port
->vlan_hash
[addr
[5]], hlist
) {
77 if (ether_addr_equal_64bits(vlan
->dev
->dev_addr
, addr
))
83 static void macvlan_hash_add(struct macvlan_dev
*vlan
)
85 struct macvlan_port
*port
= vlan
->port
;
86 const unsigned char *addr
= vlan
->dev
->dev_addr
;
88 hlist_add_head_rcu(&vlan
->hlist
, &port
->vlan_hash
[addr
[5]]);
91 static void macvlan_hash_del(struct macvlan_dev
*vlan
, bool sync
)
93 hlist_del_rcu(&vlan
->hlist
);
98 static void macvlan_hash_change_addr(struct macvlan_dev
*vlan
,
99 const unsigned char *addr
)
101 macvlan_hash_del(vlan
, true);
102 /* Now that we are unhashed it is safe to change the device
103 * address without confusing packet delivery.
105 memcpy(vlan
->dev
->dev_addr
, addr
, ETH_ALEN
);
106 macvlan_hash_add(vlan
);
109 static int macvlan_addr_busy(const struct macvlan_port
*port
,
110 const unsigned char *addr
)
112 /* Test to see if the specified multicast address is
113 * currently in use by the underlying device or
116 if (ether_addr_equal_64bits(port
->dev
->dev_addr
, addr
))
119 if (macvlan_hash_lookup(port
, addr
))
126 static int macvlan_broadcast_one(struct sk_buff
*skb
,
127 const struct macvlan_dev
*vlan
,
128 const struct ethhdr
*eth
, bool local
)
130 struct net_device
*dev
= vlan
->dev
;
133 return __dev_forward_skb(dev
, skb
);
136 if (ether_addr_equal_64bits(eth
->h_dest
, dev
->broadcast
))
137 skb
->pkt_type
= PACKET_BROADCAST
;
139 skb
->pkt_type
= PACKET_MULTICAST
;
144 static u32
macvlan_hash_mix(const struct macvlan_dev
*vlan
)
146 return (u32
)(((unsigned long)vlan
) >> L1_CACHE_SHIFT
);
150 static unsigned int mc_hash(const struct macvlan_dev
*vlan
,
151 const unsigned char *addr
)
153 u32 val
= __get_unaligned_cpu32(addr
+ 2);
155 val
^= macvlan_hash_mix(vlan
);
156 return hash_32(val
, MACVLAN_MC_FILTER_BITS
);
159 static void macvlan_broadcast(struct sk_buff
*skb
,
160 const struct macvlan_port
*port
,
161 struct net_device
*src
,
162 enum macvlan_mode mode
)
164 const struct ethhdr
*eth
= eth_hdr(skb
);
165 const struct macvlan_dev
*vlan
;
166 struct sk_buff
*nskb
;
171 if (skb
->protocol
== htons(ETH_P_PAUSE
))
174 for (i
= 0; i
< MACVLAN_HASH_SIZE
; i
++) {
175 hlist_for_each_entry_rcu(vlan
, &port
->vlan_hash
[i
], hlist
) {
176 if (vlan
->dev
== src
|| !(vlan
->mode
& mode
))
179 hash
= mc_hash(vlan
, eth
->h_dest
);
180 if (!test_bit(hash
, vlan
->mc_filter
))
184 nskb
= skb_clone(skb
, GFP_ATOMIC
);
186 err
= macvlan_broadcast_one(
188 mode
== MACVLAN_MODE_BRIDGE
) ?:
190 macvlan_count_rx(vlan
, skb
->len
+ ETH_HLEN
,
191 err
== NET_RX_SUCCESS
, 1);
196 static void macvlan_process_broadcast(struct work_struct
*w
)
198 struct macvlan_port
*port
= container_of(w
, struct macvlan_port
,
201 struct sk_buff_head list
;
203 skb_queue_head_init(&list
);
205 spin_lock_bh(&port
->bc_queue
.lock
);
206 skb_queue_splice_tail_init(&port
->bc_queue
, &list
);
207 spin_unlock_bh(&port
->bc_queue
.lock
);
209 while ((skb
= __skb_dequeue(&list
))) {
210 const struct macvlan_dev
*src
= MACVLAN_SKB_CB(skb
)->src
;
215 /* frame comes from an external address */
216 macvlan_broadcast(skb
, port
, NULL
,
217 MACVLAN_MODE_PRIVATE
|
219 MACVLAN_MODE_PASSTHRU
|
220 MACVLAN_MODE_BRIDGE
);
221 else if (src
->mode
== MACVLAN_MODE_VEPA
)
222 /* flood to everyone except source */
223 macvlan_broadcast(skb
, port
, src
->dev
,
225 MACVLAN_MODE_BRIDGE
);
228 * flood only to VEPA ports, bridge ports
229 * already saw the frame on the way out.
231 macvlan_broadcast(skb
, port
, src
->dev
,
240 static void macvlan_broadcast_enqueue(struct macvlan_port
*port
,
243 struct sk_buff
*nskb
;
246 nskb
= skb_clone(skb
, GFP_ATOMIC
);
250 spin_lock(&port
->bc_queue
.lock
);
251 if (skb_queue_len(&port
->bc_queue
) < skb
->dev
->tx_queue_len
) {
252 __skb_queue_tail(&port
->bc_queue
, nskb
);
255 spin_unlock(&port
->bc_queue
.lock
);
260 schedule_work(&port
->bc_work
);
266 atomic_long_inc(&skb
->dev
->rx_dropped
);
269 /* called under rcu_read_lock() from netif_receive_skb */
270 static rx_handler_result_t
macvlan_handle_frame(struct sk_buff
**pskb
)
272 struct macvlan_port
*port
;
273 struct sk_buff
*skb
= *pskb
;
274 const struct ethhdr
*eth
= eth_hdr(skb
);
275 const struct macvlan_dev
*vlan
;
276 const struct macvlan_dev
*src
;
277 struct net_device
*dev
;
278 unsigned int len
= 0;
279 int ret
= NET_RX_DROP
;
281 port
= macvlan_port_get_rcu(skb
->dev
);
282 if (is_multicast_ether_addr(eth
->h_dest
)) {
283 skb
= ip_check_defrag(skb
, IP_DEFRAG_MACVLAN
);
285 return RX_HANDLER_CONSUMED
;
287 src
= macvlan_hash_lookup(port
, eth
->h_source
);
288 if (src
&& src
->mode
!= MACVLAN_MODE_VEPA
&&
289 src
->mode
!= MACVLAN_MODE_BRIDGE
) {
290 /* forward to original port. */
292 ret
= macvlan_broadcast_one(skb
, vlan
, eth
, 0) ?:
297 MACVLAN_SKB_CB(skb
)->src
= src
;
298 macvlan_broadcast_enqueue(port
, skb
);
300 return RX_HANDLER_PASS
;
304 vlan
= list_first_or_null_rcu(&port
->vlans
,
305 struct macvlan_dev
, list
);
307 vlan
= macvlan_hash_lookup(port
, eth
->h_dest
);
309 return RX_HANDLER_PASS
;
312 if (unlikely(!(dev
->flags
& IFF_UP
))) {
314 return RX_HANDLER_CONSUMED
;
316 len
= skb
->len
+ ETH_HLEN
;
317 skb
= skb_share_check(skb
, GFP_ATOMIC
);
322 skb
->pkt_type
= PACKET_HOST
;
327 macvlan_count_rx(vlan
, len
, ret
== NET_RX_SUCCESS
, 0);
328 return RX_HANDLER_CONSUMED
;
331 static int macvlan_queue_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
333 const struct macvlan_dev
*vlan
= netdev_priv(dev
);
334 const struct macvlan_port
*port
= vlan
->port
;
335 const struct macvlan_dev
*dest
;
337 if (vlan
->mode
== MACVLAN_MODE_BRIDGE
) {
338 const struct ethhdr
*eth
= (void *)skb
->data
;
340 /* send to other bridge ports directly */
341 if (is_multicast_ether_addr(eth
->h_dest
)) {
342 macvlan_broadcast(skb
, port
, dev
, MACVLAN_MODE_BRIDGE
);
346 dest
= macvlan_hash_lookup(port
, eth
->h_dest
);
347 if (dest
&& dest
->mode
== MACVLAN_MODE_BRIDGE
) {
348 /* send to lowerdev first for its network taps */
349 dev_forward_skb(vlan
->lowerdev
, skb
);
351 return NET_XMIT_SUCCESS
;
356 skb
->dev
= vlan
->lowerdev
;
357 return dev_queue_xmit(skb
);
360 static inline netdev_tx_t
macvlan_netpoll_send_skb(struct macvlan_dev
*vlan
, struct sk_buff
*skb
)
362 #ifdef CONFIG_NET_POLL_CONTROLLER
364 netpoll_send_skb(vlan
->netpoll
, skb
);
371 static netdev_tx_t
macvlan_start_xmit(struct sk_buff
*skb
,
372 struct net_device
*dev
)
374 unsigned int len
= skb
->len
;
376 struct macvlan_dev
*vlan
= netdev_priv(dev
);
378 if (unlikely(netpoll_tx_running(dev
)))
379 return macvlan_netpoll_send_skb(vlan
, skb
);
381 if (vlan
->fwd_priv
) {
382 skb
->dev
= vlan
->lowerdev
;
383 ret
= dev_queue_xmit_accel(skb
, vlan
->fwd_priv
);
385 ret
= macvlan_queue_xmit(skb
, dev
);
388 if (likely(ret
== NET_XMIT_SUCCESS
|| ret
== NET_XMIT_CN
)) {
389 struct vlan_pcpu_stats
*pcpu_stats
;
391 pcpu_stats
= this_cpu_ptr(vlan
->pcpu_stats
);
392 u64_stats_update_begin(&pcpu_stats
->syncp
);
393 pcpu_stats
->tx_packets
++;
394 pcpu_stats
->tx_bytes
+= len
;
395 u64_stats_update_end(&pcpu_stats
->syncp
);
397 this_cpu_inc(vlan
->pcpu_stats
->tx_dropped
);
402 static int macvlan_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
403 unsigned short type
, const void *daddr
,
404 const void *saddr
, unsigned len
)
406 const struct macvlan_dev
*vlan
= netdev_priv(dev
);
407 struct net_device
*lowerdev
= vlan
->lowerdev
;
409 return dev_hard_header(skb
, lowerdev
, type
, daddr
,
410 saddr
? : dev
->dev_addr
, len
);
413 static const struct header_ops macvlan_hard_header_ops
= {
414 .create
= macvlan_hard_header
,
415 .rebuild
= eth_rebuild_header
,
416 .parse
= eth_header_parse
,
417 .cache
= eth_header_cache
,
418 .cache_update
= eth_header_cache_update
,
421 static struct rtnl_link_ops macvlan_link_ops
;
423 static int macvlan_open(struct net_device
*dev
)
425 struct macvlan_dev
*vlan
= netdev_priv(dev
);
426 struct net_device
*lowerdev
= vlan
->lowerdev
;
429 if (vlan
->port
->passthru
) {
430 if (!(vlan
->flags
& MACVLAN_FLAG_NOPROMISC
)) {
431 err
= dev_set_promiscuity(lowerdev
, 1);
438 if (lowerdev
->features
& NETIF_F_HW_L2FW_DOFFLOAD
&&
439 dev
->rtnl_link_ops
== &macvlan_link_ops
) {
441 lowerdev
->netdev_ops
->ndo_dfwd_add_station(lowerdev
, dev
);
443 /* If we get a NULL pointer back, or if we get an error
444 * then we should just fall through to the non accelerated path
446 if (IS_ERR_OR_NULL(vlan
->fwd_priv
)) {
447 vlan
->fwd_priv
= NULL
;
453 if (macvlan_addr_busy(vlan
->port
, dev
->dev_addr
))
456 err
= dev_uc_add(lowerdev
, dev
->dev_addr
);
459 if (dev
->flags
& IFF_ALLMULTI
) {
460 err
= dev_set_allmulti(lowerdev
, 1);
466 macvlan_hash_add(vlan
);
470 dev_uc_del(lowerdev
, dev
->dev_addr
);
472 if (vlan
->fwd_priv
) {
473 lowerdev
->netdev_ops
->ndo_dfwd_del_station(lowerdev
,
475 vlan
->fwd_priv
= NULL
;
480 static int macvlan_stop(struct net_device
*dev
)
482 struct macvlan_dev
*vlan
= netdev_priv(dev
);
483 struct net_device
*lowerdev
= vlan
->lowerdev
;
485 if (vlan
->fwd_priv
) {
486 lowerdev
->netdev_ops
->ndo_dfwd_del_station(lowerdev
,
488 vlan
->fwd_priv
= NULL
;
492 dev_uc_unsync(lowerdev
, dev
);
493 dev_mc_unsync(lowerdev
, dev
);
495 if (vlan
->port
->passthru
) {
496 if (!(vlan
->flags
& MACVLAN_FLAG_NOPROMISC
))
497 dev_set_promiscuity(lowerdev
, -1);
501 if (dev
->flags
& IFF_ALLMULTI
)
502 dev_set_allmulti(lowerdev
, -1);
504 dev_uc_del(lowerdev
, dev
->dev_addr
);
507 macvlan_hash_del(vlan
, !dev
->dismantle
);
511 static int macvlan_sync_address(struct net_device
*dev
, unsigned char *addr
)
513 struct macvlan_dev
*vlan
= netdev_priv(dev
);
514 struct net_device
*lowerdev
= vlan
->lowerdev
;
517 if (!(dev
->flags
& IFF_UP
)) {
518 /* Just copy in the new address */
519 ether_addr_copy(dev
->dev_addr
, addr
);
521 /* Rehash and update the device filters */
522 if (macvlan_addr_busy(vlan
->port
, addr
))
525 if (!vlan
->port
->passthru
) {
526 err
= dev_uc_add(lowerdev
, addr
);
530 dev_uc_del(lowerdev
, dev
->dev_addr
);
533 macvlan_hash_change_addr(vlan
, addr
);
538 static int macvlan_set_mac_address(struct net_device
*dev
, void *p
)
540 struct macvlan_dev
*vlan
= netdev_priv(dev
);
541 struct sockaddr
*addr
= p
;
543 if (!is_valid_ether_addr(addr
->sa_data
))
544 return -EADDRNOTAVAIL
;
546 if (vlan
->mode
== MACVLAN_MODE_PASSTHRU
) {
547 dev_set_mac_address(vlan
->lowerdev
, addr
);
551 return macvlan_sync_address(dev
, addr
->sa_data
);
554 static void macvlan_change_rx_flags(struct net_device
*dev
, int change
)
556 struct macvlan_dev
*vlan
= netdev_priv(dev
);
557 struct net_device
*lowerdev
= vlan
->lowerdev
;
559 if (dev
->flags
& IFF_UP
) {
560 if (change
& IFF_ALLMULTI
)
561 dev_set_allmulti(lowerdev
, dev
->flags
& IFF_ALLMULTI
? 1 : -1);
565 static void macvlan_set_mac_lists(struct net_device
*dev
)
567 struct macvlan_dev
*vlan
= netdev_priv(dev
);
569 if (dev
->flags
& (IFF_PROMISC
| IFF_ALLMULTI
)) {
570 bitmap_fill(vlan
->mc_filter
, MACVLAN_MC_FILTER_SZ
);
572 struct netdev_hw_addr
*ha
;
573 DECLARE_BITMAP(filter
, MACVLAN_MC_FILTER_SZ
);
575 bitmap_zero(filter
, MACVLAN_MC_FILTER_SZ
);
576 netdev_for_each_mc_addr(ha
, dev
) {
577 __set_bit(mc_hash(vlan
, ha
->addr
), filter
);
580 __set_bit(mc_hash(vlan
, dev
->broadcast
), filter
);
582 bitmap_copy(vlan
->mc_filter
, filter
, MACVLAN_MC_FILTER_SZ
);
584 dev_uc_sync(vlan
->lowerdev
, dev
);
585 dev_mc_sync(vlan
->lowerdev
, dev
);
588 static int macvlan_change_mtu(struct net_device
*dev
, int new_mtu
)
590 struct macvlan_dev
*vlan
= netdev_priv(dev
);
592 if (new_mtu
< 68 || vlan
->lowerdev
->mtu
< new_mtu
)
599 * macvlan network devices have devices nesting below it and are a special
600 * "super class" of normal network devices; split their locks off into a
601 * separate class since they always nest.
603 static struct lock_class_key macvlan_netdev_xmit_lock_key
;
604 static struct lock_class_key macvlan_netdev_addr_lock_key
;
606 #define ALWAYS_ON_FEATURES \
607 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX)
609 #define MACVLAN_FEATURES \
610 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
611 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
612 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
613 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
615 #define MACVLAN_STATE_MASK \
616 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
618 static int macvlan_get_nest_level(struct net_device
*dev
)
620 return ((struct macvlan_dev
*)netdev_priv(dev
))->nest_level
;
623 static void macvlan_set_lockdep_class_one(struct net_device
*dev
,
624 struct netdev_queue
*txq
,
627 lockdep_set_class(&txq
->_xmit_lock
,
628 &macvlan_netdev_xmit_lock_key
);
631 static void macvlan_set_lockdep_class(struct net_device
*dev
)
633 lockdep_set_class_and_subclass(&dev
->addr_list_lock
,
634 &macvlan_netdev_addr_lock_key
,
635 macvlan_get_nest_level(dev
));
636 netdev_for_each_tx_queue(dev
, macvlan_set_lockdep_class_one
, NULL
);
639 static int macvlan_init(struct net_device
*dev
)
641 struct macvlan_dev
*vlan
= netdev_priv(dev
);
642 const struct net_device
*lowerdev
= vlan
->lowerdev
;
644 dev
->state
= (dev
->state
& ~MACVLAN_STATE_MASK
) |
645 (lowerdev
->state
& MACVLAN_STATE_MASK
);
646 dev
->features
= lowerdev
->features
& MACVLAN_FEATURES
;
647 dev
->features
|= ALWAYS_ON_FEATURES
;
648 dev
->vlan_features
= lowerdev
->vlan_features
& MACVLAN_FEATURES
;
649 dev
->gso_max_size
= lowerdev
->gso_max_size
;
650 dev
->iflink
= lowerdev
->ifindex
;
651 dev
->hard_header_len
= lowerdev
->hard_header_len
;
653 macvlan_set_lockdep_class(dev
);
655 vlan
->pcpu_stats
= netdev_alloc_pcpu_stats(struct vlan_pcpu_stats
);
656 if (!vlan
->pcpu_stats
)
662 static void macvlan_uninit(struct net_device
*dev
)
664 struct macvlan_dev
*vlan
= netdev_priv(dev
);
665 struct macvlan_port
*port
= vlan
->port
;
667 free_percpu(vlan
->pcpu_stats
);
671 macvlan_port_destroy(port
->dev
);
674 static struct rtnl_link_stats64
*macvlan_dev_get_stats64(struct net_device
*dev
,
675 struct rtnl_link_stats64
*stats
)
677 struct macvlan_dev
*vlan
= netdev_priv(dev
);
679 if (vlan
->pcpu_stats
) {
680 struct vlan_pcpu_stats
*p
;
681 u64 rx_packets
, rx_bytes
, rx_multicast
, tx_packets
, tx_bytes
;
682 u32 rx_errors
= 0, tx_dropped
= 0;
686 for_each_possible_cpu(i
) {
687 p
= per_cpu_ptr(vlan
->pcpu_stats
, i
);
689 start
= u64_stats_fetch_begin_irq(&p
->syncp
);
690 rx_packets
= p
->rx_packets
;
691 rx_bytes
= p
->rx_bytes
;
692 rx_multicast
= p
->rx_multicast
;
693 tx_packets
= p
->tx_packets
;
694 tx_bytes
= p
->tx_bytes
;
695 } while (u64_stats_fetch_retry_irq(&p
->syncp
, start
));
697 stats
->rx_packets
+= rx_packets
;
698 stats
->rx_bytes
+= rx_bytes
;
699 stats
->multicast
+= rx_multicast
;
700 stats
->tx_packets
+= tx_packets
;
701 stats
->tx_bytes
+= tx_bytes
;
702 /* rx_errors & tx_dropped are u32, updated
703 * without syncp protection.
705 rx_errors
+= p
->rx_errors
;
706 tx_dropped
+= p
->tx_dropped
;
708 stats
->rx_errors
= rx_errors
;
709 stats
->rx_dropped
= rx_errors
;
710 stats
->tx_dropped
= tx_dropped
;
715 static int macvlan_vlan_rx_add_vid(struct net_device
*dev
,
716 __be16 proto
, u16 vid
)
718 struct macvlan_dev
*vlan
= netdev_priv(dev
);
719 struct net_device
*lowerdev
= vlan
->lowerdev
;
721 return vlan_vid_add(lowerdev
, proto
, vid
);
724 static int macvlan_vlan_rx_kill_vid(struct net_device
*dev
,
725 __be16 proto
, u16 vid
)
727 struct macvlan_dev
*vlan
= netdev_priv(dev
);
728 struct net_device
*lowerdev
= vlan
->lowerdev
;
730 vlan_vid_del(lowerdev
, proto
, vid
);
734 static int macvlan_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
735 struct net_device
*dev
,
736 const unsigned char *addr
,
739 struct macvlan_dev
*vlan
= netdev_priv(dev
);
742 /* Support unicast filter only on passthru devices.
743 * Multicast filter should be allowed on all devices.
745 if (!vlan
->port
->passthru
&& is_unicast_ether_addr(addr
))
748 if (flags
& NLM_F_REPLACE
)
751 if (is_unicast_ether_addr(addr
))
752 err
= dev_uc_add_excl(dev
, addr
);
753 else if (is_multicast_ether_addr(addr
))
754 err
= dev_mc_add_excl(dev
, addr
);
759 static int macvlan_fdb_del(struct ndmsg
*ndm
, struct nlattr
*tb
[],
760 struct net_device
*dev
,
761 const unsigned char *addr
)
763 struct macvlan_dev
*vlan
= netdev_priv(dev
);
766 /* Support unicast filter only on passthru devices.
767 * Multicast filter should be allowed on all devices.
769 if (!vlan
->port
->passthru
&& is_unicast_ether_addr(addr
))
772 if (is_unicast_ether_addr(addr
))
773 err
= dev_uc_del(dev
, addr
);
774 else if (is_multicast_ether_addr(addr
))
775 err
= dev_mc_del(dev
, addr
);
780 static void macvlan_ethtool_get_drvinfo(struct net_device
*dev
,
781 struct ethtool_drvinfo
*drvinfo
)
783 strlcpy(drvinfo
->driver
, "macvlan", sizeof(drvinfo
->driver
));
784 strlcpy(drvinfo
->version
, "0.1", sizeof(drvinfo
->version
));
787 static int macvlan_ethtool_get_settings(struct net_device
*dev
,
788 struct ethtool_cmd
*cmd
)
790 const struct macvlan_dev
*vlan
= netdev_priv(dev
);
792 return __ethtool_get_settings(vlan
->lowerdev
, cmd
);
795 static netdev_features_t
macvlan_fix_features(struct net_device
*dev
,
796 netdev_features_t features
)
798 struct macvlan_dev
*vlan
= netdev_priv(dev
);
799 netdev_features_t mask
;
801 features
|= NETIF_F_ALL_FOR_ALL
;
802 features
&= (vlan
->set_features
| ~MACVLAN_FEATURES
);
805 features
= netdev_increment_features(vlan
->lowerdev
->features
,
808 features
|= ALWAYS_ON_FEATURES
;
813 #ifdef CONFIG_NET_POLL_CONTROLLER
814 static void macvlan_dev_poll_controller(struct net_device
*dev
)
819 static int macvlan_dev_netpoll_setup(struct net_device
*dev
, struct netpoll_info
*npinfo
)
821 struct macvlan_dev
*vlan
= netdev_priv(dev
);
822 struct net_device
*real_dev
= vlan
->lowerdev
;
823 struct netpoll
*netpoll
;
826 netpoll
= kzalloc(sizeof(*netpoll
), GFP_KERNEL
);
831 err
= __netpoll_setup(netpoll
, real_dev
);
837 vlan
->netpoll
= netpoll
;
843 static void macvlan_dev_netpoll_cleanup(struct net_device
*dev
)
845 struct macvlan_dev
*vlan
= netdev_priv(dev
);
846 struct netpoll
*netpoll
= vlan
->netpoll
;
851 vlan
->netpoll
= NULL
;
853 __netpoll_free_async(netpoll
);
855 #endif /* CONFIG_NET_POLL_CONTROLLER */
857 static const struct ethtool_ops macvlan_ethtool_ops
= {
858 .get_link
= ethtool_op_get_link
,
859 .get_settings
= macvlan_ethtool_get_settings
,
860 .get_drvinfo
= macvlan_ethtool_get_drvinfo
,
863 static const struct net_device_ops macvlan_netdev_ops
= {
864 .ndo_init
= macvlan_init
,
865 .ndo_uninit
= macvlan_uninit
,
866 .ndo_open
= macvlan_open
,
867 .ndo_stop
= macvlan_stop
,
868 .ndo_start_xmit
= macvlan_start_xmit
,
869 .ndo_change_mtu
= macvlan_change_mtu
,
870 .ndo_fix_features
= macvlan_fix_features
,
871 .ndo_change_rx_flags
= macvlan_change_rx_flags
,
872 .ndo_set_mac_address
= macvlan_set_mac_address
,
873 .ndo_set_rx_mode
= macvlan_set_mac_lists
,
874 .ndo_get_stats64
= macvlan_dev_get_stats64
,
875 .ndo_validate_addr
= eth_validate_addr
,
876 .ndo_vlan_rx_add_vid
= macvlan_vlan_rx_add_vid
,
877 .ndo_vlan_rx_kill_vid
= macvlan_vlan_rx_kill_vid
,
878 .ndo_fdb_add
= macvlan_fdb_add
,
879 .ndo_fdb_del
= macvlan_fdb_del
,
880 .ndo_fdb_dump
= ndo_dflt_fdb_dump
,
881 .ndo_get_lock_subclass
= macvlan_get_nest_level
,
882 #ifdef CONFIG_NET_POLL_CONTROLLER
883 .ndo_poll_controller
= macvlan_dev_poll_controller
,
884 .ndo_netpoll_setup
= macvlan_dev_netpoll_setup
,
885 .ndo_netpoll_cleanup
= macvlan_dev_netpoll_cleanup
,
889 void macvlan_common_setup(struct net_device
*dev
)
893 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_TX_SKB_SHARING
);
894 dev
->priv_flags
|= IFF_UNICAST_FLT
;
895 dev
->netdev_ops
= &macvlan_netdev_ops
;
896 dev
->destructor
= free_netdev
;
897 dev
->header_ops
= &macvlan_hard_header_ops
;
898 dev
->ethtool_ops
= &macvlan_ethtool_ops
;
900 EXPORT_SYMBOL_GPL(macvlan_common_setup
);
902 static void macvlan_setup(struct net_device
*dev
)
904 macvlan_common_setup(dev
);
905 dev
->tx_queue_len
= 0;
908 static int macvlan_port_create(struct net_device
*dev
)
910 struct macvlan_port
*port
;
914 if (dev
->type
!= ARPHRD_ETHER
|| dev
->flags
& IFF_LOOPBACK
)
917 port
= kzalloc(sizeof(*port
), GFP_KERNEL
);
921 port
->passthru
= false;
923 INIT_LIST_HEAD(&port
->vlans
);
924 for (i
= 0; i
< MACVLAN_HASH_SIZE
; i
++)
925 INIT_HLIST_HEAD(&port
->vlan_hash
[i
]);
927 skb_queue_head_init(&port
->bc_queue
);
928 INIT_WORK(&port
->bc_work
, macvlan_process_broadcast
);
930 err
= netdev_rx_handler_register(dev
, macvlan_handle_frame
, port
);
934 dev
->priv_flags
|= IFF_MACVLAN_PORT
;
938 static void macvlan_port_destroy(struct net_device
*dev
)
940 struct macvlan_port
*port
= macvlan_port_get_rtnl(dev
);
942 cancel_work_sync(&port
->bc_work
);
943 dev
->priv_flags
&= ~IFF_MACVLAN_PORT
;
944 netdev_rx_handler_unregister(dev
);
945 kfree_rcu(port
, rcu
);
948 static int macvlan_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
950 if (tb
[IFLA_ADDRESS
]) {
951 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
)
953 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
])))
954 return -EADDRNOTAVAIL
;
957 if (data
&& data
[IFLA_MACVLAN_FLAGS
] &&
958 nla_get_u16(data
[IFLA_MACVLAN_FLAGS
]) & ~MACVLAN_FLAG_NOPROMISC
)
961 if (data
&& data
[IFLA_MACVLAN_MODE
]) {
962 switch (nla_get_u32(data
[IFLA_MACVLAN_MODE
])) {
963 case MACVLAN_MODE_PRIVATE
:
964 case MACVLAN_MODE_VEPA
:
965 case MACVLAN_MODE_BRIDGE
:
966 case MACVLAN_MODE_PASSTHRU
:
975 int macvlan_common_newlink(struct net
*src_net
, struct net_device
*dev
,
976 struct nlattr
*tb
[], struct nlattr
*data
[])
978 struct macvlan_dev
*vlan
= netdev_priv(dev
);
979 struct macvlan_port
*port
;
980 struct net_device
*lowerdev
;
986 lowerdev
= __dev_get_by_index(src_net
, nla_get_u32(tb
[IFLA_LINK
]));
987 if (lowerdev
== NULL
)
990 /* When creating macvlans or macvtaps on top of other macvlans - use
991 * the real device as the lowerdev.
993 if (netif_is_macvlan(lowerdev
))
994 lowerdev
= macvlan_dev_real_dev(lowerdev
);
997 dev
->mtu
= lowerdev
->mtu
;
998 else if (dev
->mtu
> lowerdev
->mtu
)
1001 if (!tb
[IFLA_ADDRESS
])
1002 eth_hw_addr_random(dev
);
1004 if (!macvlan_port_exists(lowerdev
)) {
1005 err
= macvlan_port_create(lowerdev
);
1009 port
= macvlan_port_get_rtnl(lowerdev
);
1011 /* Only 1 macvlan device can be created in passthru mode */
1015 vlan
->lowerdev
= lowerdev
;
1018 vlan
->set_features
= MACVLAN_FEATURES
;
1019 vlan
->nest_level
= dev_get_nest_level(lowerdev
, netif_is_macvlan
) + 1;
1021 vlan
->mode
= MACVLAN_MODE_VEPA
;
1022 if (data
&& data
[IFLA_MACVLAN_MODE
])
1023 vlan
->mode
= nla_get_u32(data
[IFLA_MACVLAN_MODE
]);
1025 if (data
&& data
[IFLA_MACVLAN_FLAGS
])
1026 vlan
->flags
= nla_get_u16(data
[IFLA_MACVLAN_FLAGS
]);
1028 if (vlan
->mode
== MACVLAN_MODE_PASSTHRU
) {
1031 port
->passthru
= true;
1032 eth_hw_addr_inherit(dev
, lowerdev
);
1036 err
= register_netdevice(dev
);
1040 dev
->priv_flags
|= IFF_MACVLAN
;
1041 err
= netdev_upper_dev_link(lowerdev
, dev
);
1043 goto unregister_netdev
;
1045 list_add_tail_rcu(&vlan
->list
, &port
->vlans
);
1046 netif_stacked_transfer_operstate(lowerdev
, dev
);
1051 unregister_netdevice(dev
);
1055 macvlan_port_destroy(lowerdev
);
1059 EXPORT_SYMBOL_GPL(macvlan_common_newlink
);
1061 static int macvlan_newlink(struct net
*src_net
, struct net_device
*dev
,
1062 struct nlattr
*tb
[], struct nlattr
*data
[])
1064 return macvlan_common_newlink(src_net
, dev
, tb
, data
);
1067 void macvlan_dellink(struct net_device
*dev
, struct list_head
*head
)
1069 struct macvlan_dev
*vlan
= netdev_priv(dev
);
1071 list_del_rcu(&vlan
->list
);
1072 unregister_netdevice_queue(dev
, head
);
1073 netdev_upper_dev_unlink(vlan
->lowerdev
, dev
);
1075 EXPORT_SYMBOL_GPL(macvlan_dellink
);
1077 static int macvlan_changelink(struct net_device
*dev
,
1078 struct nlattr
*tb
[], struct nlattr
*data
[])
1080 struct macvlan_dev
*vlan
= netdev_priv(dev
);
1081 enum macvlan_mode mode
;
1082 bool set_mode
= false;
1084 /* Validate mode, but don't set yet: setting flags may fail. */
1085 if (data
&& data
[IFLA_MACVLAN_MODE
]) {
1087 mode
= nla_get_u32(data
[IFLA_MACVLAN_MODE
]);
1088 /* Passthrough mode can't be set or cleared dynamically */
1089 if ((mode
== MACVLAN_MODE_PASSTHRU
) !=
1090 (vlan
->mode
== MACVLAN_MODE_PASSTHRU
))
1094 if (data
&& data
[IFLA_MACVLAN_FLAGS
]) {
1095 __u16 flags
= nla_get_u16(data
[IFLA_MACVLAN_FLAGS
]);
1096 bool promisc
= (flags
^ vlan
->flags
) & MACVLAN_FLAG_NOPROMISC
;
1097 if (vlan
->port
->passthru
&& promisc
) {
1100 if (flags
& MACVLAN_FLAG_NOPROMISC
)
1101 err
= dev_set_promiscuity(vlan
->lowerdev
, -1);
1103 err
= dev_set_promiscuity(vlan
->lowerdev
, 1);
1107 vlan
->flags
= flags
;
1114 static size_t macvlan_get_size(const struct net_device
*dev
)
1117 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1118 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1122 static int macvlan_fill_info(struct sk_buff
*skb
,
1123 const struct net_device
*dev
)
1125 struct macvlan_dev
*vlan
= netdev_priv(dev
);
1127 if (nla_put_u32(skb
, IFLA_MACVLAN_MODE
, vlan
->mode
))
1128 goto nla_put_failure
;
1129 if (nla_put_u16(skb
, IFLA_MACVLAN_FLAGS
, vlan
->flags
))
1130 goto nla_put_failure
;
1137 static const struct nla_policy macvlan_policy
[IFLA_MACVLAN_MAX
+ 1] = {
1138 [IFLA_MACVLAN_MODE
] = { .type
= NLA_U32
},
1139 [IFLA_MACVLAN_FLAGS
] = { .type
= NLA_U16
},
1142 int macvlan_link_register(struct rtnl_link_ops
*ops
)
1145 ops
->priv_size
= sizeof(struct macvlan_dev
);
1146 ops
->validate
= macvlan_validate
;
1147 ops
->maxtype
= IFLA_MACVLAN_MAX
;
1148 ops
->policy
= macvlan_policy
;
1149 ops
->changelink
= macvlan_changelink
;
1150 ops
->get_size
= macvlan_get_size
;
1151 ops
->fill_info
= macvlan_fill_info
;
1153 return rtnl_link_register(ops
);
1155 EXPORT_SYMBOL_GPL(macvlan_link_register
);
1157 static struct rtnl_link_ops macvlan_link_ops
= {
1159 .setup
= macvlan_setup
,
1160 .newlink
= macvlan_newlink
,
1161 .dellink
= macvlan_dellink
,
1164 static int macvlan_device_event(struct notifier_block
*unused
,
1165 unsigned long event
, void *ptr
)
1167 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1168 struct macvlan_dev
*vlan
, *next
;
1169 struct macvlan_port
*port
;
1170 LIST_HEAD(list_kill
);
1172 if (!macvlan_port_exists(dev
))
1175 port
= macvlan_port_get_rtnl(dev
);
1179 list_for_each_entry(vlan
, &port
->vlans
, list
)
1180 netif_stacked_transfer_operstate(vlan
->lowerdev
,
1183 case NETDEV_FEAT_CHANGE
:
1184 list_for_each_entry(vlan
, &port
->vlans
, list
) {
1185 vlan
->dev
->gso_max_size
= dev
->gso_max_size
;
1186 netdev_update_features(vlan
->dev
);
1189 case NETDEV_CHANGEMTU
:
1190 list_for_each_entry(vlan
, &port
->vlans
, list
) {
1191 if (vlan
->dev
->mtu
<= dev
->mtu
)
1193 dev_set_mtu(vlan
->dev
, dev
->mtu
);
1196 case NETDEV_CHANGEADDR
:
1197 if (!port
->passthru
)
1200 vlan
= list_first_entry_or_null(&port
->vlans
,
1204 if (macvlan_sync_address(vlan
->dev
, dev
->dev_addr
))
1208 case NETDEV_UNREGISTER
:
1209 /* twiddle thumbs on netns device moves */
1210 if (dev
->reg_state
!= NETREG_UNREGISTERING
)
1213 list_for_each_entry_safe(vlan
, next
, &port
->vlans
, list
)
1214 vlan
->dev
->rtnl_link_ops
->dellink(vlan
->dev
, &list_kill
);
1215 unregister_netdevice_many(&list_kill
);
1217 case NETDEV_PRE_TYPE_CHANGE
:
1218 /* Forbid underlaying device to change its type. */
1221 case NETDEV_NOTIFY_PEERS
:
1222 case NETDEV_BONDING_FAILOVER
:
1223 case NETDEV_RESEND_IGMP
:
1224 /* Propagate to all vlans */
1225 list_for_each_entry(vlan
, &port
->vlans
, list
)
1226 call_netdevice_notifiers(event
, vlan
->dev
);
1231 static struct notifier_block macvlan_notifier_block __read_mostly
= {
1232 .notifier_call
= macvlan_device_event
,
1235 static int __init
macvlan_init_module(void)
1239 register_netdevice_notifier(&macvlan_notifier_block
);
1241 err
= macvlan_link_register(&macvlan_link_ops
);
1246 unregister_netdevice_notifier(&macvlan_notifier_block
);
1250 static void __exit
macvlan_cleanup_module(void)
1252 rtnl_link_unregister(&macvlan_link_ops
);
1253 unregister_netdevice_notifier(&macvlan_notifier_block
);
1256 module_init(macvlan_init_module
);
1257 module_exit(macvlan_cleanup_module
);
1259 MODULE_LICENSE("GPL");
1260 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1261 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1262 MODULE_ALIAS_RTNL_LINK("macvlan");