1/*
2 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
3 *
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.
8 *
9 * The code this is based on carried the following copyright notice:
10 * ---
11 * (C) Copyright 2001-2006
12 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
13 * Re-worked by Ben Greear <greearb@candelatech.com>
14 * ---
15 */
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>
35#include <net/xfrm.h>
36#include <linux/netpoll.h>
37
38#define MACVLAN_HASH_BITS 8
39#define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS)
40#define MACVLAN_BC_QUEUE_LEN 1000
41
42#define MACVLAN_F_PASSTHRU 1
43#define MACVLAN_F_ADDRCHANGE 2
44
45struct macvlan_port {
46 struct net_device *dev;
47 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
48 struct list_head vlans;
49 struct sk_buff_head bc_queue;
50 struct work_struct bc_work;
51 u32 flags;
52 int count;
53 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE];
54 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
55 unsigned char perm_addr[ETH_ALEN];
56};
57
58struct macvlan_source_entry {
59 struct hlist_node hlist;
60 struct macvlan_dev *vlan;
61 unsigned char addr[6+2] __aligned(sizeof(u16));
62 struct rcu_head rcu;
63};
64
65struct macvlan_skb_cb {
66 const struct macvlan_dev *src;
67};
68
69#define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
70
71static void macvlan_port_destroy(struct net_device *dev);
72
73static inline bool macvlan_passthru(const struct macvlan_port *port)
74{
75 return port->flags & MACVLAN_F_PASSTHRU;
76}
77
78static inline void macvlan_set_passthru(struct macvlan_port *port)
79{
80 port->flags |= MACVLAN_F_PASSTHRU;
81}
82
83static inline bool macvlan_addr_change(const struct macvlan_port *port)
84{
85 return port->flags & MACVLAN_F_ADDRCHANGE;
86}
87
88static inline void macvlan_set_addr_change(struct macvlan_port *port)
89{
90 port->flags |= MACVLAN_F_ADDRCHANGE;
91}
92
93static inline void macvlan_clear_addr_change(struct macvlan_port *port)
94{
95 port->flags &= ~MACVLAN_F_ADDRCHANGE;
96}
97
98/* Hash Ethernet address */
99static u32 macvlan_eth_hash(const unsigned char *addr)
100{
101 u64 value = get_unaligned((u64 *)addr);
102
103 /* only want 6 bytes */
104#ifdef __BIG_ENDIAN
105 value >>= 16;
106#else
107 value <<= 16;
108#endif
109 return hash_64(value, MACVLAN_HASH_BITS);
110}
111
112static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
113{
114 return rcu_dereference(dev->rx_handler_data);
115}
116
117static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
118{
119 return rtnl_dereference(dev->rx_handler_data);
120}
121
122static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
123 const unsigned char *addr)
124{
125 struct macvlan_dev *vlan;
126 u32 idx = macvlan_eth_hash(addr);
127
128 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) {
129 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
130 return vlan;
131 }
132 return NULL;
133}
134
135static struct macvlan_source_entry *macvlan_hash_lookup_source(
136 const struct macvlan_dev *vlan,
137 const unsigned char *addr)
138{
139 struct macvlan_source_entry *entry;
140 u32 idx = macvlan_eth_hash(addr);
141 struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
142
143 hlist_for_each_entry_rcu(entry, h, hlist) {
144 if (ether_addr_equal_64bits(entry->addr, addr) &&
145 entry->vlan == vlan)
146 return entry;
147 }
148 return NULL;
149}
150
151static int macvlan_hash_add_source(struct macvlan_dev *vlan,
152 const unsigned char *addr)
153{
154 struct macvlan_port *port = vlan->port;
155 struct macvlan_source_entry *entry;
156 struct hlist_head *h;
157
158 entry = macvlan_hash_lookup_source(vlan, addr);
159 if (entry)
160 return 0;
161
162 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
163 if (!entry)
164 return -ENOMEM;
165
166 ether_addr_copy(entry->addr, addr);
167 entry->vlan = vlan;
168 h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
169 hlist_add_head_rcu(&entry->hlist, h);
170 vlan->macaddr_count++;
171
172 return 0;
173}
174
175static void macvlan_hash_add(struct macvlan_dev *vlan)
176{
177 struct macvlan_port *port = vlan->port;
178 const unsigned char *addr = vlan->dev->dev_addr;
179 u32 idx = macvlan_eth_hash(addr);
180
181 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
182}
183
184static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
185{
186 hlist_del_rcu(&entry->hlist);
187 kfree_rcu(entry, rcu);
188}
189
190static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
191{
192 hlist_del_rcu(&vlan->hlist);
193 if (sync)
194 synchronize_rcu();
195}
196
197static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
198 const unsigned char *addr)
199{
200 macvlan_hash_del(vlan, true);
201 /* Now that we are unhashed it is safe to change the device
202 * address without confusing packet delivery.
203 */
204 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
205 macvlan_hash_add(vlan);
206}
207
208static bool macvlan_addr_busy(const struct macvlan_port *port,
209 const unsigned char *addr)
210{
211 /* Test to see if the specified address is
212 * currently in use by the underlying device or
213 * another macvlan.
214 */
215 if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
216 ether_addr_equal_64bits(port->dev->dev_addr, addr))
217 return true;
218
219 if (macvlan_hash_lookup(port, addr))
220 return true;
221
222 return false;
223}
224
225
226static int macvlan_broadcast_one(struct sk_buff *skb,
227 const struct macvlan_dev *vlan,
228 const struct ethhdr *eth, bool local)
229{
230 struct net_device *dev = vlan->dev;
231
232 if (local)
233 return __dev_forward_skb(dev, skb);
234
235 skb->dev = dev;
236 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
237 skb->pkt_type = PACKET_BROADCAST;
238 else
239 skb->pkt_type = PACKET_MULTICAST;
240
241 return 0;
242}
243
244static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
245{
246 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
247}
248
249
250static unsigned int mc_hash(const struct macvlan_dev *vlan,
251 const unsigned char *addr)
252{
253 u32 val = __get_unaligned_cpu32(addr + 2);
254
255 val ^= macvlan_hash_mix(vlan);
256 return hash_32(val, MACVLAN_MC_FILTER_BITS);
257}
258
259static void macvlan_broadcast(struct sk_buff *skb,
260 const struct macvlan_port *port,
261 struct net_device *src,
262 enum macvlan_mode mode)
263{
264 const struct ethhdr *eth = eth_hdr(skb);
265 const struct macvlan_dev *vlan;
266 struct sk_buff *nskb;
267 unsigned int i;
268 int err;
269 unsigned int hash;
270
271 if (skb->protocol == htons(ETH_P_PAUSE))
272 return;
273
274 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
275 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
276 if (vlan->dev == src || !(vlan->mode & mode))
277 continue;
278
279 hash = mc_hash(vlan, eth->h_dest);
280 if (!test_bit(hash, vlan->mc_filter))
281 continue;
282
283 err = NET_RX_DROP;
284 nskb = skb_clone(skb, GFP_ATOMIC);
285 if (likely(nskb))
286 err = macvlan_broadcast_one(
287 nskb, vlan, eth,
288 mode == MACVLAN_MODE_BRIDGE) ?:
289 netif_rx_ni(nskb);
290 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
291 err == NET_RX_SUCCESS, true);
292 }
293 }
294}
295
296static void macvlan_process_broadcast(struct work_struct *w)
297{
298 struct macvlan_port *port = container_of(w, struct macvlan_port,
299 bc_work);
300 struct sk_buff *skb;
301 struct sk_buff_head list;
302
303 __skb_queue_head_init(&list);
304
305 spin_lock_bh(&port->bc_queue.lock);
306 skb_queue_splice_tail_init(&port->bc_queue, &list);
307 spin_unlock_bh(&port->bc_queue.lock);
308
309 while ((skb = __skb_dequeue(&list))) {
310 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
311
312 rcu_read_lock();
313
314 if (!src)
315 /* frame comes from an external address */
316 macvlan_broadcast(skb, port, NULL,
317 MACVLAN_MODE_PRIVATE |
318 MACVLAN_MODE_VEPA |
319 MACVLAN_MODE_PASSTHRU|
320 MACVLAN_MODE_BRIDGE);
321 else if (src->mode == MACVLAN_MODE_VEPA)
322 /* flood to everyone except source */
323 macvlan_broadcast(skb, port, src->dev,
324 MACVLAN_MODE_VEPA |
325 MACVLAN_MODE_BRIDGE);
326 else
327 /*
328 * flood only to VEPA ports, bridge ports
329 * already saw the frame on the way out.
330 */
331 macvlan_broadcast(skb, port, src->dev,
332 MACVLAN_MODE_VEPA);
333
334 rcu_read_unlock();
335
336 if (src)
337 dev_put(src->dev);
338 consume_skb(skb);
339 }
340}
341
342static void macvlan_broadcast_enqueue(struct macvlan_port *port,
343 const struct macvlan_dev *src,
344 struct sk_buff *skb)
345{
346 struct sk_buff *nskb;
347 int err = -ENOMEM;
348
349 nskb = skb_clone(skb, GFP_ATOMIC);
350 if (!nskb)
351 goto err;
352
353 MACVLAN_SKB_CB(nskb)->src = src;
354
355 spin_lock(&port->bc_queue.lock);
356 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
357 if (src)
358 dev_hold(src->dev);
359 __skb_queue_tail(&port->bc_queue, nskb);
360 err = 0;
361 }
362 spin_unlock(&port->bc_queue.lock);
363
364 if (err)
365 goto free_nskb;
366
367 schedule_work(&port->bc_work);
368 return;
369
370free_nskb:
371 kfree_skb(nskb);
372err:
373 atomic_long_inc(&skb->dev->rx_dropped);
374}
375
376static void macvlan_flush_sources(struct macvlan_port *port,
377 struct macvlan_dev *vlan)
378{
379 int i;
380
381 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
382 struct hlist_node *h, *n;
383
384 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
385 struct macvlan_source_entry *entry;
386
387 entry = hlist_entry(h, struct macvlan_source_entry,
388 hlist);
389 if (entry->vlan == vlan)
390 macvlan_hash_del_source(entry);
391 }
392 }
393 vlan->macaddr_count = 0;
394}
395
396static void macvlan_forward_source_one(struct sk_buff *skb,
397 struct macvlan_dev *vlan)
398{
399 struct sk_buff *nskb;
400 struct net_device *dev;
401 int len;
402 int ret;
403
404 dev = vlan->dev;
405 if (unlikely(!(dev->flags & IFF_UP)))
406 return;
407
408 nskb = skb_clone(skb, GFP_ATOMIC);
409 if (!nskb)
410 return;
411
412 len = nskb->len + ETH_HLEN;
413 nskb->dev = dev;
414
415 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
416 nskb->pkt_type = PACKET_HOST;
417
418 ret = netif_rx(nskb);
419 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
420}
421
422static void macvlan_forward_source(struct sk_buff *skb,
423 struct macvlan_port *port,
424 const unsigned char *addr)
425{
426 struct macvlan_source_entry *entry;
427 u32 idx = macvlan_eth_hash(addr);
428 struct hlist_head *h = &port->vlan_source_hash[idx];
429
430 hlist_for_each_entry_rcu(entry, h, hlist) {
431 if (ether_addr_equal_64bits(entry->addr, addr))
432 macvlan_forward_source_one(skb, entry->vlan);
433 }
434}
435
436/* called under rcu_read_lock() from netif_receive_skb */
437static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
438{
439 struct macvlan_port *port;
440 struct sk_buff *skb = *pskb;
441 const struct ethhdr *eth = eth_hdr(skb);
442 const struct macvlan_dev *vlan;
443 const struct macvlan_dev *src;
444 struct net_device *dev;
445 unsigned int len = 0;
446 int ret;
447 rx_handler_result_t handle_res;
448
449 port = macvlan_port_get_rcu(skb->dev);
450 if (is_multicast_ether_addr(eth->h_dest)) {
451 unsigned int hash;
452
453 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
454 if (!skb)
455 return RX_HANDLER_CONSUMED;
456 *pskb = skb;
457 eth = eth_hdr(skb);
458 macvlan_forward_source(skb, port, eth->h_source);
459 src = macvlan_hash_lookup(port, eth->h_source);
460 if (src && src->mode != MACVLAN_MODE_VEPA &&
461 src->mode != MACVLAN_MODE_BRIDGE) {
462 /* forward to original port. */
463 vlan = src;
464 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
465 netif_rx(skb);
466 handle_res = RX_HANDLER_CONSUMED;
467 goto out;
468 }
469
470 hash = mc_hash(NULL, eth->h_dest);
471 if (test_bit(hash, port->mc_filter))
472 macvlan_broadcast_enqueue(port, src, skb);
473
474 return RX_HANDLER_PASS;
475 }
476
477 macvlan_forward_source(skb, port, eth->h_source);
478 if (macvlan_passthru(port))
479 vlan = list_first_or_null_rcu(&port->vlans,
480 struct macvlan_dev, list);
481 else
482 vlan = macvlan_hash_lookup(port, eth->h_dest);
483 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
484 return RX_HANDLER_PASS;
485
486 dev = vlan->dev;
487 if (unlikely(!(dev->flags & IFF_UP))) {
488 kfree_skb(skb);
489 return RX_HANDLER_CONSUMED;
490 }
491 len = skb->len + ETH_HLEN;
492 skb = skb_share_check(skb, GFP_ATOMIC);
493 if (!skb) {
494 ret = NET_RX_DROP;
495 handle_res = RX_HANDLER_CONSUMED;
496 goto out;
497 }
498
499 *pskb = skb;
500 skb->dev = dev;
501 skb->pkt_type = PACKET_HOST;
502
503 ret = NET_RX_SUCCESS;
504 handle_res = RX_HANDLER_ANOTHER;
505out:
506 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
507 return handle_res;
508}
509
510static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
511{
512 const struct macvlan_dev *vlan = netdev_priv(dev);
513 const struct macvlan_port *port = vlan->port;
514 const struct macvlan_dev *dest;
515
516 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
517 const struct ethhdr *eth = (void *)skb->data;
518
519 /* send to other bridge ports directly */
520 if (is_multicast_ether_addr(eth->h_dest)) {
521 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
522 goto xmit_world;
523 }
524
525 dest = macvlan_hash_lookup(port, eth->h_dest);
526 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
527 /* send to lowerdev first for its network taps */
528 dev_forward_skb(vlan->lowerdev, skb);
529
530 return NET_XMIT_SUCCESS;
531 }
532 }
533xmit_world:
534 skb->dev = vlan->lowerdev;
535 return dev_queue_xmit_accel(skb,
536 netdev_get_sb_channel(dev) ? dev : NULL);
537}
538
539static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
540{
541#ifdef CONFIG_NET_POLL_CONTROLLER
542 if (vlan->netpoll)
543 netpoll_send_skb(vlan->netpoll, skb);
544#else
545 BUG();
546#endif
547 return NETDEV_TX_OK;
548}
549
550static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
551 struct net_device *dev)
552{
553 struct macvlan_dev *vlan = netdev_priv(dev);
554 unsigned int len = skb->len;
555 int ret;
556
557 if (unlikely(netpoll_tx_running(dev)))
558 return macvlan_netpoll_send_skb(vlan, skb);
559
560 ret = macvlan_queue_xmit(skb, dev);
561
562 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
563 struct vlan_pcpu_stats *pcpu_stats;
564
565 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
566 u64_stats_update_begin(&pcpu_stats->syncp);
567 pcpu_stats->tx_packets++;
568 pcpu_stats->tx_bytes += len;
569 u64_stats_update_end(&pcpu_stats->syncp);
570 } else {
571 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
572 }
573 return ret;
574}
575
576static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
577 unsigned short type, const void *daddr,
578 const void *saddr, unsigned len)
579{
580 const struct macvlan_dev *vlan = netdev_priv(dev);
581 struct net_device *lowerdev = vlan->lowerdev;
582
583 return dev_hard_header(skb, lowerdev, type, daddr,
584 saddr ? : dev->dev_addr, len);
585}
586
587static const struct header_ops macvlan_hard_header_ops = {
588 .create = macvlan_hard_header,
589 .parse = eth_header_parse,
590 .cache = eth_header_cache,
591 .cache_update = eth_header_cache_update,
592};
593
594static int macvlan_open(struct net_device *dev)
595{
596 struct macvlan_dev *vlan = netdev_priv(dev);
597 struct net_device *lowerdev = vlan->lowerdev;
598 int err;
599
600 if (macvlan_passthru(vlan->port)) {
601 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
602 err = dev_set_promiscuity(lowerdev, 1);
603 if (err < 0)
604 goto out;
605 }
606 goto hash_add;
607 }
608
609 err = -EADDRINUSE;
610 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
611 goto out;
612
613 /* Attempt to populate accel_priv which is used to offload the L2
614 * forwarding requests for unicast packets.
615 */
616 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
617 vlan->accel_priv =
618 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
619
620 /* If earlier attempt to offload failed, or accel_priv is not
621 * populated we must add the unicast address to the lower device.
622 */
623 if (IS_ERR_OR_NULL(vlan->accel_priv)) {
624 vlan->accel_priv = NULL;
625 err = dev_uc_add(lowerdev, dev->dev_addr);
626 if (err < 0)
627 goto out;
628 }
629
630 if (dev->flags & IFF_ALLMULTI) {
631 err = dev_set_allmulti(lowerdev, 1);
632 if (err < 0)
633 goto del_unicast;
634 }
635
636 if (dev->flags & IFF_PROMISC) {
637 err = dev_set_promiscuity(lowerdev, 1);
638 if (err < 0)
639 goto clear_multi;
640 }
641
642hash_add:
643 macvlan_hash_add(vlan);
644 return 0;
645
646clear_multi:
647 if (dev->flags & IFF_ALLMULTI)
648 dev_set_allmulti(lowerdev, -1);
649del_unicast:
650 if (vlan->accel_priv) {
651 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
652 vlan->accel_priv);
653 vlan->accel_priv = NULL;
654 } else {
655 dev_uc_del(lowerdev, dev->dev_addr);
656 }
657out:
658 return err;
659}
660
661static int macvlan_stop(struct net_device *dev)
662{
663 struct macvlan_dev *vlan = netdev_priv(dev);
664 struct net_device *lowerdev = vlan->lowerdev;
665
666 if (vlan->accel_priv) {
667 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
668 vlan->accel_priv);
669 vlan->accel_priv = NULL;
670 }
671
672 dev_uc_unsync(lowerdev, dev);
673 dev_mc_unsync(lowerdev, dev);
674
675 if (macvlan_passthru(vlan->port)) {
676 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
677 dev_set_promiscuity(lowerdev, -1);
678 goto hash_del;
679 }
680
681 if (dev->flags & IFF_ALLMULTI)
682 dev_set_allmulti(lowerdev, -1);
683
684 if (dev->flags & IFF_PROMISC)
685 dev_set_promiscuity(lowerdev, -1);
686
687 dev_uc_del(lowerdev, dev->dev_addr);
688
689hash_del:
690 macvlan_hash_del(vlan, !dev->dismantle);
691 return 0;
692}
693
694static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
695{
696 struct macvlan_dev *vlan = netdev_priv(dev);
697 struct net_device *lowerdev = vlan->lowerdev;
698 struct macvlan_port *port = vlan->port;
699 int err;
700
701 if (!(dev->flags & IFF_UP)) {
702 /* Just copy in the new address */
703 ether_addr_copy(dev->dev_addr, addr);
704 } else {
705 /* Rehash and update the device filters */
706 if (macvlan_addr_busy(vlan->port, addr))
707 return -EADDRINUSE;
708
709 if (!macvlan_passthru(port)) {
710 err = dev_uc_add(lowerdev, addr);
711 if (err)
712 return err;
713
714 dev_uc_del(lowerdev, dev->dev_addr);
715 }
716
717 macvlan_hash_change_addr(vlan, addr);
718 }
719 if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
720 /* Since addr_change isn't set, we are here due to lower
721 * device change. Save the lower-dev address so we can
722 * restore it later.
723 */
724 ether_addr_copy(vlan->port->perm_addr,
725 lowerdev->dev_addr);
726 }
727 macvlan_clear_addr_change(port);
728 return 0;
729}
730
731static int macvlan_set_mac_address(struct net_device *dev, void *p)
732{
733 struct macvlan_dev *vlan = netdev_priv(dev);
734 struct sockaddr *addr = p;
735
736 if (!is_valid_ether_addr(addr->sa_data))
737 return -EADDRNOTAVAIL;
738
739 /* If the addresses are the same, this is a no-op */
740 if (ether_addr_equal(dev->dev_addr, addr->sa_data))
741 return 0;
742
743 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
744 macvlan_set_addr_change(vlan->port);
745 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
746 }
747
748 if (macvlan_addr_busy(vlan->port, addr->sa_data))
749 return -EADDRINUSE;
750
751 return macvlan_sync_address(dev, addr->sa_data);
752}
753
754static void macvlan_change_rx_flags(struct net_device *dev, int change)
755{
756 struct macvlan_dev *vlan = netdev_priv(dev);
757 struct net_device *lowerdev = vlan->lowerdev;
758
759 if (dev->flags & IFF_UP) {
760 if (change & IFF_ALLMULTI)
761 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
762 if (change & IFF_PROMISC)
763 dev_set_promiscuity(lowerdev,
764 dev->flags & IFF_PROMISC ? 1 : -1);
765
766 }
767}
768
769static void macvlan_compute_filter(unsigned long *mc_filter,
770 struct net_device *dev,
771 struct macvlan_dev *vlan)
772{
773 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
774 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
775 } else {
776 struct netdev_hw_addr *ha;
777 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
778
779 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
780 netdev_for_each_mc_addr(ha, dev) {
781 __set_bit(mc_hash(vlan, ha->addr), filter);
782 }
783
784 __set_bit(mc_hash(vlan, dev->broadcast), filter);
785
786 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
787 }
788}
789
790static void macvlan_set_mac_lists(struct net_device *dev)
791{
792 struct macvlan_dev *vlan = netdev_priv(dev);
793
794 macvlan_compute_filter(vlan->mc_filter, dev, vlan);
795
796 dev_uc_sync(vlan->lowerdev, dev);
797 dev_mc_sync(vlan->lowerdev, dev);
798
799 /* This is slightly inaccurate as we're including the subscription
800 * list of vlan->lowerdev too.
801 *
802 * Bug alert: This only works if everyone has the same broadcast
803 * address as lowerdev. As soon as someone changes theirs this
804 * will break.
805 *
806 * However, this is already broken as when you change your broadcast
807 * address we don't get called.
808 *
809 * The solution is to maintain a list of broadcast addresses like
810 * we do for uc/mc, if you care.
811 */
812 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
813}
814
815static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
816{
817 struct macvlan_dev *vlan = netdev_priv(dev);
818
819 if (vlan->lowerdev->mtu < new_mtu)
820 return -EINVAL;
821 dev->mtu = new_mtu;
822 return 0;
823}
824
825/*
826 * macvlan network devices have devices nesting below it and are a special
827 * "super class" of normal network devices; split their locks off into a
828 * separate class since they always nest.
829 */
830static struct lock_class_key macvlan_netdev_addr_lock_key;
831
832#define ALWAYS_ON_OFFLOADS \
833 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
834 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
835
836#define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
837
838#define MACVLAN_FEATURES \
839 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
840 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
841 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
842 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
843
844#define MACVLAN_STATE_MASK \
845 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
846
847static int macvlan_get_nest_level(struct net_device *dev)
848{
849 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
850}
851
852static void macvlan_set_lockdep_class(struct net_device *dev)
853{
854 netdev_lockdep_set_classes(dev);
855 lockdep_set_class_and_subclass(&dev->addr_list_lock,
856 &macvlan_netdev_addr_lock_key,
857 macvlan_get_nest_level(dev));
858}
859
860static int macvlan_init(struct net_device *dev)
861{
862 struct macvlan_dev *vlan = netdev_priv(dev);
863 const struct net_device *lowerdev = vlan->lowerdev;
864 struct macvlan_port *port = vlan->port;
865
866 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
867 (lowerdev->state & MACVLAN_STATE_MASK);
868 dev->features = lowerdev->features & MACVLAN_FEATURES;
869 dev->features |= ALWAYS_ON_FEATURES;
870 dev->hw_features |= NETIF_F_LRO;
871 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
872 dev->vlan_features |= ALWAYS_ON_OFFLOADS;
873 dev->hw_enc_features |= dev->features;
874 dev->gso_max_size = lowerdev->gso_max_size;
875 dev->gso_max_segs = lowerdev->gso_max_segs;
876 dev->hard_header_len = lowerdev->hard_header_len;
877
878 macvlan_set_lockdep_class(dev);
879
880 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
881 if (!vlan->pcpu_stats)
882 return -ENOMEM;
883
884 port->count += 1;
885
886 return 0;
887}
888
889static void macvlan_uninit(struct net_device *dev)
890{
891 struct macvlan_dev *vlan = netdev_priv(dev);
892 struct macvlan_port *port = vlan->port;
893
894 free_percpu(vlan->pcpu_stats);
895
896 macvlan_flush_sources(port, vlan);
897 port->count -= 1;
898 if (!port->count)
899 macvlan_port_destroy(port->dev);
900}
901
902static void macvlan_dev_get_stats64(struct net_device *dev,
903 struct rtnl_link_stats64 *stats)
904{
905 struct macvlan_dev *vlan = netdev_priv(dev);
906
907 if (vlan->pcpu_stats) {
908 struct vlan_pcpu_stats *p;
909 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
910 u32 rx_errors = 0, tx_dropped = 0;
911 unsigned int start;
912 int i;
913
914 for_each_possible_cpu(i) {
915 p = per_cpu_ptr(vlan->pcpu_stats, i);
916 do {
917 start = u64_stats_fetch_begin_irq(&p->syncp);
918 rx_packets = p->rx_packets;
919 rx_bytes = p->rx_bytes;
920 rx_multicast = p->rx_multicast;
921 tx_packets = p->tx_packets;
922 tx_bytes = p->tx_bytes;
923 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
924
925 stats->rx_packets += rx_packets;
926 stats->rx_bytes += rx_bytes;
927 stats->multicast += rx_multicast;
928 stats->tx_packets += tx_packets;
929 stats->tx_bytes += tx_bytes;
930 /* rx_errors & tx_dropped are u32, updated
931 * without syncp protection.
932 */
933 rx_errors += p->rx_errors;
934 tx_dropped += p->tx_dropped;
935 }
936 stats->rx_errors = rx_errors;
937 stats->rx_dropped = rx_errors;
938 stats->tx_dropped = tx_dropped;
939 }
940}
941
942static int macvlan_vlan_rx_add_vid(struct net_device *dev,
943 __be16 proto, u16 vid)
944{
945 struct macvlan_dev *vlan = netdev_priv(dev);
946 struct net_device *lowerdev = vlan->lowerdev;
947
948 return vlan_vid_add(lowerdev, proto, vid);
949}
950
951static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
952 __be16 proto, u16 vid)
953{
954 struct macvlan_dev *vlan = netdev_priv(dev);
955 struct net_device *lowerdev = vlan->lowerdev;
956
957 vlan_vid_del(lowerdev, proto, vid);
958 return 0;
959}
960
961static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
962 struct net_device *dev,
963 const unsigned char *addr, u16 vid,
964 u16 flags,
965 struct netlink_ext_ack *extack)
966{
967 struct macvlan_dev *vlan = netdev_priv(dev);
968 int err = -EINVAL;
969
970 /* Support unicast filter only on passthru devices.
971 * Multicast filter should be allowed on all devices.
972 */
973 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
974 return -EOPNOTSUPP;
975
976 if (flags & NLM_F_REPLACE)
977 return -EOPNOTSUPP;
978
979 if (is_unicast_ether_addr(addr))
980 err = dev_uc_add_excl(dev, addr);
981 else if (is_multicast_ether_addr(addr))
982 err = dev_mc_add_excl(dev, addr);
983
984 return err;
985}
986
987static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
988 struct net_device *dev,
989 const unsigned char *addr, u16 vid)
990{
991 struct macvlan_dev *vlan = netdev_priv(dev);
992 int err = -EINVAL;
993
994 /* Support unicast filter only on passthru devices.
995 * Multicast filter should be allowed on all devices.
996 */
997 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
998 return -EOPNOTSUPP;
999
1000 if (is_unicast_ether_addr(addr))
1001 err = dev_uc_del(dev, addr);
1002 else if (is_multicast_ether_addr(addr))
1003 err = dev_mc_del(dev, addr);
1004
1005 return err;
1006}
1007
1008static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1009 struct ethtool_drvinfo *drvinfo)
1010{
1011 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1012 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1013}
1014
1015static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1016 struct ethtool_link_ksettings *cmd)
1017{
1018 const struct macvlan_dev *vlan = netdev_priv(dev);
1019
1020 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1021}
1022
1023static netdev_features_t macvlan_fix_features(struct net_device *dev,
1024 netdev_features_t features)
1025{
1026 struct macvlan_dev *vlan = netdev_priv(dev);
1027 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1028 netdev_features_t mask;
1029
1030 features |= NETIF_F_ALL_FOR_ALL;
1031 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1032 mask = features;
1033
1034 lowerdev_features &= (features | ~NETIF_F_LRO);
1035 features = netdev_increment_features(lowerdev_features, features, mask);
1036 features |= ALWAYS_ON_FEATURES;
1037 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1038
1039 return features;
1040}
1041
1042#ifdef CONFIG_NET_POLL_CONTROLLER
1043static void macvlan_dev_poll_controller(struct net_device *dev)
1044{
1045 return;
1046}
1047
1048static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1049{
1050 struct macvlan_dev *vlan = netdev_priv(dev);
1051 struct net_device *real_dev = vlan->lowerdev;
1052 struct netpoll *netpoll;
1053 int err = 0;
1054
1055 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1056 err = -ENOMEM;
1057 if (!netpoll)
1058 goto out;
1059
1060 err = __netpoll_setup(netpoll, real_dev);
1061 if (err) {
1062 kfree(netpoll);
1063 goto out;
1064 }
1065
1066 vlan->netpoll = netpoll;
1067
1068out:
1069 return err;
1070}
1071
1072static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1073{
1074 struct macvlan_dev *vlan = netdev_priv(dev);
1075 struct netpoll *netpoll = vlan->netpoll;
1076
1077 if (!netpoll)
1078 return;
1079
1080 vlan->netpoll = NULL;
1081
1082 __netpoll_free(netpoll);
1083}
1084#endif /* CONFIG_NET_POLL_CONTROLLER */
1085
1086static int macvlan_dev_get_iflink(const struct net_device *dev)
1087{
1088 struct macvlan_dev *vlan = netdev_priv(dev);
1089
1090 return vlan->lowerdev->ifindex;
1091}
1092
1093static const struct ethtool_ops macvlan_ethtool_ops = {
1094 .get_link = ethtool_op_get_link,
1095 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1096 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1097};
1098
1099static const struct net_device_ops macvlan_netdev_ops = {
1100 .ndo_init = macvlan_init,
1101 .ndo_uninit = macvlan_uninit,
1102 .ndo_open = macvlan_open,
1103 .ndo_stop = macvlan_stop,
1104 .ndo_start_xmit = macvlan_start_xmit,
1105 .ndo_change_mtu = macvlan_change_mtu,
1106 .ndo_fix_features = macvlan_fix_features,
1107 .ndo_change_rx_flags = macvlan_change_rx_flags,
1108 .ndo_set_mac_address = macvlan_set_mac_address,
1109 .ndo_set_rx_mode = macvlan_set_mac_lists,
1110 .ndo_get_stats64 = macvlan_dev_get_stats64,
1111 .ndo_validate_addr = eth_validate_addr,
1112 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1113 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1114 .ndo_fdb_add = macvlan_fdb_add,
1115 .ndo_fdb_del = macvlan_fdb_del,
1116 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1117 .ndo_get_lock_subclass = macvlan_get_nest_level,
1118#ifdef CONFIG_NET_POLL_CONTROLLER
1119 .ndo_poll_controller = macvlan_dev_poll_controller,
1120 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1121 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1122#endif
1123 .ndo_get_iflink = macvlan_dev_get_iflink,
1124 .ndo_features_check = passthru_features_check,
1125 .ndo_change_proto_down = dev_change_proto_down_generic,
1126};
1127
1128void macvlan_common_setup(struct net_device *dev)
1129{
1130 ether_setup(dev);
1131
1132 dev->min_mtu = 0;
1133 dev->max_mtu = ETH_MAX_MTU;
1134 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1135 netif_keep_dst(dev);
1136 dev->priv_flags |= IFF_UNICAST_FLT;
1137 dev->netdev_ops = &macvlan_netdev_ops;
1138 dev->needs_free_netdev = true;
1139 dev->header_ops = &macvlan_hard_header_ops;
1140 dev->ethtool_ops = &macvlan_ethtool_ops;
1141}
1142EXPORT_SYMBOL_GPL(macvlan_common_setup);
1143
1144static void macvlan_setup(struct net_device *dev)
1145{
1146 macvlan_common_setup(dev);
1147 dev->priv_flags |= IFF_NO_QUEUE;
1148}
1149
1150static int macvlan_port_create(struct net_device *dev)
1151{
1152 struct macvlan_port *port;
1153 unsigned int i;
1154 int err;
1155
1156 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1157 return -EINVAL;
1158
1159 if (netdev_is_rx_handler_busy(dev))
1160 return -EBUSY;
1161
1162 port = kzalloc(sizeof(*port), GFP_KERNEL);
1163 if (port == NULL)
1164 return -ENOMEM;
1165
1166 port->dev = dev;
1167 ether_addr_copy(port->perm_addr, dev->dev_addr);
1168 INIT_LIST_HEAD(&port->vlans);
1169 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1170 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1171 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1172 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1173
1174 skb_queue_head_init(&port->bc_queue);
1175 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1176
1177 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1178 if (err)
1179 kfree(port);
1180 else
1181 dev->priv_flags |= IFF_MACVLAN_PORT;
1182 return err;
1183}
1184
1185static void macvlan_port_destroy(struct net_device *dev)
1186{
1187 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1188 struct sk_buff *skb;
1189
1190 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1191 netdev_rx_handler_unregister(dev);
1192
1193 /* After this point, no packet can schedule bc_work anymore,
1194 * but we need to cancel it and purge left skbs if any.
1195 */
1196 cancel_work_sync(&port->bc_work);
1197
1198 while ((skb = __skb_dequeue(&port->bc_queue))) {
1199 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1200
1201 if (src)
1202 dev_put(src->dev);
1203
1204 kfree_skb(skb);
1205 }
1206
1207 /* If the lower device address has been changed by passthru
1208 * macvlan, put it back.
1209 */
1210 if (macvlan_passthru(port) &&
1211 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1212 struct sockaddr sa;
1213
1214 sa.sa_family = port->dev->type;
1215 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1216 dev_set_mac_address(port->dev, &sa, NULL);
1217 }
1218
1219 kfree(port);
1220}
1221
1222static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1223 struct netlink_ext_ack *extack)
1224{
1225 if (tb[IFLA_ADDRESS]) {
1226 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1227 return -EINVAL;
1228 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1229 return -EADDRNOTAVAIL;
1230 }
1231
1232 if (!data)
1233 return 0;
1234
1235 if (data[IFLA_MACVLAN_FLAGS] &&
1236 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1237 return -EINVAL;
1238
1239 if (data[IFLA_MACVLAN_MODE]) {
1240 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1241 case MACVLAN_MODE_PRIVATE:
1242 case MACVLAN_MODE_VEPA:
1243 case MACVLAN_MODE_BRIDGE:
1244 case MACVLAN_MODE_PASSTHRU:
1245 case MACVLAN_MODE_SOURCE:
1246 break;
1247 default:
1248 return -EINVAL;
1249 }
1250 }
1251
1252 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1253 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1254 case MACVLAN_MACADDR_ADD:
1255 case MACVLAN_MACADDR_DEL:
1256 case MACVLAN_MACADDR_FLUSH:
1257 case MACVLAN_MACADDR_SET:
1258 break;
1259 default:
1260 return -EINVAL;
1261 }
1262 }
1263
1264 if (data[IFLA_MACVLAN_MACADDR]) {
1265 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1266 return -EINVAL;
1267
1268 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1269 return -EADDRNOTAVAIL;
1270 }
1271
1272 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1273 return -EINVAL;
1274
1275 return 0;
1276}
1277
1278/**
1279 * reconfigure list of remote source mac address
1280 * (only for macvlan devices in source mode)
1281 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1282 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1283 */
1284static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1285 struct nlattr *data[])
1286{
1287 char *addr = NULL;
1288 int ret, rem, len;
1289 struct nlattr *nla, *head;
1290 struct macvlan_source_entry *entry;
1291
1292 if (data[IFLA_MACVLAN_MACADDR])
1293 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1294
1295 if (mode == MACVLAN_MACADDR_ADD) {
1296 if (!addr)
1297 return -EINVAL;
1298
1299 return macvlan_hash_add_source(vlan, addr);
1300
1301 } else if (mode == MACVLAN_MACADDR_DEL) {
1302 if (!addr)
1303 return -EINVAL;
1304
1305 entry = macvlan_hash_lookup_source(vlan, addr);
1306 if (entry) {
1307 macvlan_hash_del_source(entry);
1308 vlan->macaddr_count--;
1309 }
1310 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1311 macvlan_flush_sources(vlan->port, vlan);
1312 } else if (mode == MACVLAN_MACADDR_SET) {
1313 macvlan_flush_sources(vlan->port, vlan);
1314
1315 if (addr) {
1316 ret = macvlan_hash_add_source(vlan, addr);
1317 if (ret)
1318 return ret;
1319 }
1320
1321 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1322 return 0;
1323
1324 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1325 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1326
1327 nla_for_each_attr(nla, head, len, rem) {
1328 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1329 nla_len(nla) != ETH_ALEN)
1330 continue;
1331
1332 addr = nla_data(nla);
1333 ret = macvlan_hash_add_source(vlan, addr);
1334 if (ret)
1335 return ret;
1336 }
1337 } else {
1338 return -EINVAL;
1339 }
1340
1341 return 0;
1342}
1343
1344int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1345 struct nlattr *tb[], struct nlattr *data[],
1346 struct netlink_ext_ack *extack)
1347{
1348 struct macvlan_dev *vlan = netdev_priv(dev);
1349 struct macvlan_port *port;
1350 struct net_device *lowerdev;
1351 int err;
1352 int macmode;
1353 bool create = false;
1354
1355 if (!tb[IFLA_LINK])
1356 return -EINVAL;
1357
1358 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1359 if (lowerdev == NULL)
1360 return -ENODEV;
1361
1362 /* When creating macvlans or macvtaps on top of other macvlans - use
1363 * the real device as the lowerdev.
1364 */
1365 if (netif_is_macvlan(lowerdev))
1366 lowerdev = macvlan_dev_real_dev(lowerdev);
1367
1368 if (!tb[IFLA_MTU])
1369 dev->mtu = lowerdev->mtu;
1370 else if (dev->mtu > lowerdev->mtu)
1371 return -EINVAL;
1372
1373 /* MTU range: 68 - lowerdev->max_mtu */
1374 dev->min_mtu = ETH_MIN_MTU;
1375 dev->max_mtu = lowerdev->max_mtu;
1376
1377 if (!tb[IFLA_ADDRESS])
1378 eth_hw_addr_random(dev);
1379
1380 if (!netif_is_macvlan_port(lowerdev)) {
1381 err = macvlan_port_create(lowerdev);
1382 if (err < 0)
1383 return err;
1384 create = true;
1385 }
1386 port = macvlan_port_get_rtnl(lowerdev);
1387
1388 /* Only 1 macvlan device can be created in passthru mode */
1389 if (macvlan_passthru(port)) {
1390 /* The macvlan port must be not created this time,
1391 * still goto destroy_macvlan_port for readability.
1392 */
1393 err = -EINVAL;
1394 goto destroy_macvlan_port;
1395 }
1396
1397 vlan->lowerdev = lowerdev;
1398 vlan->dev = dev;
1399 vlan->port = port;
1400 vlan->set_features = MACVLAN_FEATURES;
1401 vlan->nest_level = dev_get_nest_level(lowerdev) + 1;
1402
1403 vlan->mode = MACVLAN_MODE_VEPA;
1404 if (data && data[IFLA_MACVLAN_MODE])
1405 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1406
1407 if (data && data[IFLA_MACVLAN_FLAGS])
1408 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1409
1410 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1411 if (port->count) {
1412 err = -EINVAL;
1413 goto destroy_macvlan_port;
1414 }
1415 macvlan_set_passthru(port);
1416 eth_hw_addr_inherit(dev, lowerdev);
1417 }
1418
1419 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1420 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1421 err = -EINVAL;
1422 goto destroy_macvlan_port;
1423 }
1424 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1425 err = macvlan_changelink_sources(vlan, macmode, data);
1426 if (err)
1427 goto destroy_macvlan_port;
1428 }
1429
1430 err = register_netdevice(dev);
1431 if (err < 0)
1432 goto destroy_macvlan_port;
1433
1434 dev->priv_flags |= IFF_MACVLAN;
1435 err = netdev_upper_dev_link(lowerdev, dev, extack);
1436 if (err)
1437 goto unregister_netdev;
1438
1439 list_add_tail_rcu(&vlan->list, &port->vlans);
1440 netif_stacked_transfer_operstate(lowerdev, dev);
1441 linkwatch_fire_event(dev);
1442
1443 return 0;
1444
1445unregister_netdev:
1446 /* macvlan_uninit would free the macvlan port */
1447 unregister_netdevice(dev);
1448 return err;
1449destroy_macvlan_port:
1450 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1451 * so we destroy the macvlan port only when it's valid.
1452 */
1453 if (create && macvlan_port_get_rtnl(lowerdev))
1454 macvlan_port_destroy(port->dev);
1455 return err;
1456}
1457EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1458
1459static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1460 struct nlattr *tb[], struct nlattr *data[],
1461 struct netlink_ext_ack *extack)
1462{
1463 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1464}
1465
1466void macvlan_dellink(struct net_device *dev, struct list_head *head)
1467{
1468 struct macvlan_dev *vlan = netdev_priv(dev);
1469
1470 if (vlan->mode == MACVLAN_MODE_SOURCE)
1471 macvlan_flush_sources(vlan->port, vlan);
1472 list_del_rcu(&vlan->list);
1473 unregister_netdevice_queue(dev, head);
1474 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1475}
1476EXPORT_SYMBOL_GPL(macvlan_dellink);
1477
1478static int macvlan_changelink(struct net_device *dev,
1479 struct nlattr *tb[], struct nlattr *data[],
1480 struct netlink_ext_ack *extack)
1481{
1482 struct macvlan_dev *vlan = netdev_priv(dev);
1483 enum macvlan_mode mode;
1484 bool set_mode = false;
1485 enum macvlan_macaddr_mode macmode;
1486 int ret;
1487
1488 /* Validate mode, but don't set yet: setting flags may fail. */
1489 if (data && data[IFLA_MACVLAN_MODE]) {
1490 set_mode = true;
1491 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1492 /* Passthrough mode can't be set or cleared dynamically */
1493 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1494 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1495 return -EINVAL;
1496 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1497 vlan->mode != mode)
1498 macvlan_flush_sources(vlan->port, vlan);
1499 }
1500
1501 if (data && data[IFLA_MACVLAN_FLAGS]) {
1502 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1503 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1504 if (macvlan_passthru(vlan->port) && promisc) {
1505 int err;
1506
1507 if (flags & MACVLAN_FLAG_NOPROMISC)
1508 err = dev_set_promiscuity(vlan->lowerdev, -1);
1509 else
1510 err = dev_set_promiscuity(vlan->lowerdev, 1);
1511 if (err < 0)
1512 return err;
1513 }
1514 vlan->flags = flags;
1515 }
1516 if (set_mode)
1517 vlan->mode = mode;
1518 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1519 if (vlan->mode != MACVLAN_MODE_SOURCE)
1520 return -EINVAL;
1521 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1522 ret = macvlan_changelink_sources(vlan, macmode, data);
1523 if (ret)
1524 return ret;
1525 }
1526 return 0;
1527}
1528
1529static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1530{
1531 if (vlan->macaddr_count == 0)
1532 return 0;
1533 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1534 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1535}
1536
1537static size_t macvlan_get_size(const struct net_device *dev)
1538{
1539 struct macvlan_dev *vlan = netdev_priv(dev);
1540
1541 return (0
1542 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1543 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1544 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1545 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1546 );
1547}
1548
1549static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1550 const struct macvlan_dev *vlan,
1551 const int i)
1552{
1553 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1554 struct macvlan_source_entry *entry;
1555
1556 hlist_for_each_entry_rcu(entry, h, hlist) {
1557 if (entry->vlan != vlan)
1558 continue;
1559 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1560 return 1;
1561 }
1562 return 0;
1563}
1564
1565static int macvlan_fill_info(struct sk_buff *skb,
1566 const struct net_device *dev)
1567{
1568 struct macvlan_dev *vlan = netdev_priv(dev);
1569 int i;
1570 struct nlattr *nest;
1571
1572 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1573 goto nla_put_failure;
1574 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1575 goto nla_put_failure;
1576 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1577 goto nla_put_failure;
1578 if (vlan->macaddr_count > 0) {
1579 nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA);
1580 if (nest == NULL)
1581 goto nla_put_failure;
1582
1583 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1584 if (macvlan_fill_info_macaddr(skb, vlan, i))
1585 goto nla_put_failure;
1586 }
1587 nla_nest_end(skb, nest);
1588 }
1589 return 0;
1590
1591nla_put_failure:
1592 return -EMSGSIZE;
1593}
1594
1595static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1596 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1597 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1598 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1599 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1600 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1601 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1602};
1603
1604int macvlan_link_register(struct rtnl_link_ops *ops)
1605{
1606 /* common fields */
1607 ops->validate = macvlan_validate;
1608 ops->maxtype = IFLA_MACVLAN_MAX;
1609 ops->policy = macvlan_policy;
1610 ops->changelink = macvlan_changelink;
1611 ops->get_size = macvlan_get_size;
1612 ops->fill_info = macvlan_fill_info;
1613
1614 return rtnl_link_register(ops);
1615};
1616EXPORT_SYMBOL_GPL(macvlan_link_register);
1617
1618static struct net *macvlan_get_link_net(const struct net_device *dev)
1619{
1620 return dev_net(macvlan_dev_real_dev(dev));
1621}
1622
1623static struct rtnl_link_ops macvlan_link_ops = {
1624 .kind = "macvlan",
1625 .setup = macvlan_setup,
1626 .newlink = macvlan_newlink,
1627 .dellink = macvlan_dellink,
1628 .get_link_net = macvlan_get_link_net,
1629 .priv_size = sizeof(struct macvlan_dev),
1630};
1631
1632static int macvlan_device_event(struct notifier_block *unused,
1633 unsigned long event, void *ptr)
1634{
1635 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1636 struct macvlan_dev *vlan, *next;
1637 struct macvlan_port *port;
1638 LIST_HEAD(list_kill);
1639
1640 if (!netif_is_macvlan_port(dev))
1641 return NOTIFY_DONE;
1642
1643 port = macvlan_port_get_rtnl(dev);
1644
1645 switch (event) {
1646 case NETDEV_UP:
1647 case NETDEV_DOWN:
1648 case NETDEV_CHANGE:
1649 list_for_each_entry(vlan, &port->vlans, list)
1650 netif_stacked_transfer_operstate(vlan->lowerdev,
1651 vlan->dev);
1652 break;
1653 case NETDEV_FEAT_CHANGE:
1654 list_for_each_entry(vlan, &port->vlans, list) {
1655 vlan->dev->gso_max_size = dev->gso_max_size;
1656 vlan->dev->gso_max_segs = dev->gso_max_segs;
1657 netdev_update_features(vlan->dev);
1658 }
1659 break;
1660 case NETDEV_CHANGEMTU:
1661 list_for_each_entry(vlan, &port->vlans, list) {
1662 if (vlan->dev->mtu <= dev->mtu)
1663 continue;
1664 dev_set_mtu(vlan->dev, dev->mtu);
1665 }
1666 break;
1667 case NETDEV_CHANGEADDR:
1668 if (!macvlan_passthru(port))
1669 return NOTIFY_DONE;
1670
1671 vlan = list_first_entry_or_null(&port->vlans,
1672 struct macvlan_dev,
1673 list);
1674
1675 if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1676 return NOTIFY_BAD;
1677
1678 break;
1679 case NETDEV_UNREGISTER:
1680 /* twiddle thumbs on netns device moves */
1681 if (dev->reg_state != NETREG_UNREGISTERING)
1682 break;
1683
1684 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1685 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1686 unregister_netdevice_many(&list_kill);
1687 break;
1688 case NETDEV_PRE_TYPE_CHANGE:
1689 /* Forbid underlaying device to change its type. */
1690 return NOTIFY_BAD;
1691
1692 case NETDEV_NOTIFY_PEERS:
1693 case NETDEV_BONDING_FAILOVER:
1694 case NETDEV_RESEND_IGMP:
1695 /* Propagate to all vlans */
1696 list_for_each_entry(vlan, &port->vlans, list)
1697 call_netdevice_notifiers(event, vlan->dev);
1698 }
1699 return NOTIFY_DONE;
1700}
1701
1702static struct notifier_block macvlan_notifier_block __read_mostly = {
1703 .notifier_call = macvlan_device_event,
1704};
1705
1706static int __init macvlan_init_module(void)
1707{
1708 int err;
1709
1710 register_netdevice_notifier(&macvlan_notifier_block);
1711
1712 err = macvlan_link_register(&macvlan_link_ops);
1713 if (err < 0)
1714 goto err1;
1715 return 0;
1716err1:
1717 unregister_netdevice_notifier(&macvlan_notifier_block);
1718 return err;
1719}
1720
1721static void __exit macvlan_cleanup_module(void)
1722{
1723 rtnl_link_unregister(&macvlan_link_ops);
1724 unregister_netdevice_notifier(&macvlan_notifier_block);
1725}
1726
1727module_init(macvlan_init_module);
1728module_exit(macvlan_cleanup_module);
1729
1730MODULE_LICENSE("GPL");
1731MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1732MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1733MODULE_ALIAS_RTNL_LINK("macvlan");
1734