Loading...
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
122#define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
123
124static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
125 const unsigned char *addr)
126{
127 struct macvlan_dev *vlan;
128 u32 idx = macvlan_eth_hash(addr);
129
130 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) {
131 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
132 return vlan;
133 }
134 return NULL;
135}
136
137static struct macvlan_source_entry *macvlan_hash_lookup_source(
138 const struct macvlan_dev *vlan,
139 const unsigned char *addr)
140{
141 struct macvlan_source_entry *entry;
142 u32 idx = macvlan_eth_hash(addr);
143 struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
144
145 hlist_for_each_entry_rcu(entry, h, hlist) {
146 if (ether_addr_equal_64bits(entry->addr, addr) &&
147 entry->vlan == vlan)
148 return entry;
149 }
150 return NULL;
151}
152
153static int macvlan_hash_add_source(struct macvlan_dev *vlan,
154 const unsigned char *addr)
155{
156 struct macvlan_port *port = vlan->port;
157 struct macvlan_source_entry *entry;
158 struct hlist_head *h;
159
160 entry = macvlan_hash_lookup_source(vlan, addr);
161 if (entry)
162 return 0;
163
164 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
165 if (!entry)
166 return -ENOMEM;
167
168 ether_addr_copy(entry->addr, addr);
169 entry->vlan = vlan;
170 h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
171 hlist_add_head_rcu(&entry->hlist, h);
172 vlan->macaddr_count++;
173
174 return 0;
175}
176
177static void macvlan_hash_add(struct macvlan_dev *vlan)
178{
179 struct macvlan_port *port = vlan->port;
180 const unsigned char *addr = vlan->dev->dev_addr;
181 u32 idx = macvlan_eth_hash(addr);
182
183 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
184}
185
186static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
187{
188 hlist_del_rcu(&entry->hlist);
189 kfree_rcu(entry, rcu);
190}
191
192static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
193{
194 hlist_del_rcu(&vlan->hlist);
195 if (sync)
196 synchronize_rcu();
197}
198
199static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
200 const unsigned char *addr)
201{
202 macvlan_hash_del(vlan, true);
203 /* Now that we are unhashed it is safe to change the device
204 * address without confusing packet delivery.
205 */
206 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
207 macvlan_hash_add(vlan);
208}
209
210static bool macvlan_addr_busy(const struct macvlan_port *port,
211 const unsigned char *addr)
212{
213 /* Test to see if the specified address is
214 * currently in use by the underlying device or
215 * another macvlan.
216 */
217 if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
218 ether_addr_equal_64bits(port->dev->dev_addr, addr))
219 return true;
220
221 if (macvlan_hash_lookup(port, addr))
222 return true;
223
224 return false;
225}
226
227
228static int macvlan_broadcast_one(struct sk_buff *skb,
229 const struct macvlan_dev *vlan,
230 const struct ethhdr *eth, bool local)
231{
232 struct net_device *dev = vlan->dev;
233
234 if (local)
235 return __dev_forward_skb(dev, skb);
236
237 skb->dev = dev;
238 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
239 skb->pkt_type = PACKET_BROADCAST;
240 else
241 skb->pkt_type = PACKET_MULTICAST;
242
243 return 0;
244}
245
246static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
247{
248 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
249}
250
251
252static unsigned int mc_hash(const struct macvlan_dev *vlan,
253 const unsigned char *addr)
254{
255 u32 val = __get_unaligned_cpu32(addr + 2);
256
257 val ^= macvlan_hash_mix(vlan);
258 return hash_32(val, MACVLAN_MC_FILTER_BITS);
259}
260
261static void macvlan_broadcast(struct sk_buff *skb,
262 const struct macvlan_port *port,
263 struct net_device *src,
264 enum macvlan_mode mode)
265{
266 const struct ethhdr *eth = eth_hdr(skb);
267 const struct macvlan_dev *vlan;
268 struct sk_buff *nskb;
269 unsigned int i;
270 int err;
271 unsigned int hash;
272
273 if (skb->protocol == htons(ETH_P_PAUSE))
274 return;
275
276 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
277 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
278 if (vlan->dev == src || !(vlan->mode & mode))
279 continue;
280
281 hash = mc_hash(vlan, eth->h_dest);
282 if (!test_bit(hash, vlan->mc_filter))
283 continue;
284
285 err = NET_RX_DROP;
286 nskb = skb_clone(skb, GFP_ATOMIC);
287 if (likely(nskb))
288 err = macvlan_broadcast_one(
289 nskb, vlan, eth,
290 mode == MACVLAN_MODE_BRIDGE) ?:
291 netif_rx_ni(nskb);
292 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
293 err == NET_RX_SUCCESS, true);
294 }
295 }
296}
297
298static void macvlan_process_broadcast(struct work_struct *w)
299{
300 struct macvlan_port *port = container_of(w, struct macvlan_port,
301 bc_work);
302 struct sk_buff *skb;
303 struct sk_buff_head list;
304
305 __skb_queue_head_init(&list);
306
307 spin_lock_bh(&port->bc_queue.lock);
308 skb_queue_splice_tail_init(&port->bc_queue, &list);
309 spin_unlock_bh(&port->bc_queue.lock);
310
311 while ((skb = __skb_dequeue(&list))) {
312 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
313
314 rcu_read_lock();
315
316 if (!src)
317 /* frame comes from an external address */
318 macvlan_broadcast(skb, port, NULL,
319 MACVLAN_MODE_PRIVATE |
320 MACVLAN_MODE_VEPA |
321 MACVLAN_MODE_PASSTHRU|
322 MACVLAN_MODE_BRIDGE);
323 else if (src->mode == MACVLAN_MODE_VEPA)
324 /* flood to everyone except source */
325 macvlan_broadcast(skb, port, src->dev,
326 MACVLAN_MODE_VEPA |
327 MACVLAN_MODE_BRIDGE);
328 else
329 /*
330 * flood only to VEPA ports, bridge ports
331 * already saw the frame on the way out.
332 */
333 macvlan_broadcast(skb, port, src->dev,
334 MACVLAN_MODE_VEPA);
335
336 rcu_read_unlock();
337
338 if (src)
339 dev_put(src->dev);
340 kfree_skb(skb);
341 }
342}
343
344static void macvlan_broadcast_enqueue(struct macvlan_port *port,
345 const struct macvlan_dev *src,
346 struct sk_buff *skb)
347{
348 struct sk_buff *nskb;
349 int err = -ENOMEM;
350
351 nskb = skb_clone(skb, GFP_ATOMIC);
352 if (!nskb)
353 goto err;
354
355 MACVLAN_SKB_CB(nskb)->src = src;
356
357 spin_lock(&port->bc_queue.lock);
358 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
359 if (src)
360 dev_hold(src->dev);
361 __skb_queue_tail(&port->bc_queue, nskb);
362 err = 0;
363 }
364 spin_unlock(&port->bc_queue.lock);
365
366 if (err)
367 goto free_nskb;
368
369 schedule_work(&port->bc_work);
370 return;
371
372free_nskb:
373 kfree_skb(nskb);
374err:
375 atomic_long_inc(&skb->dev->rx_dropped);
376}
377
378static void macvlan_flush_sources(struct macvlan_port *port,
379 struct macvlan_dev *vlan)
380{
381 int i;
382
383 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
384 struct hlist_node *h, *n;
385
386 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
387 struct macvlan_source_entry *entry;
388
389 entry = hlist_entry(h, struct macvlan_source_entry,
390 hlist);
391 if (entry->vlan == vlan)
392 macvlan_hash_del_source(entry);
393 }
394 }
395 vlan->macaddr_count = 0;
396}
397
398static void macvlan_forward_source_one(struct sk_buff *skb,
399 struct macvlan_dev *vlan)
400{
401 struct sk_buff *nskb;
402 struct net_device *dev;
403 int len;
404 int ret;
405
406 dev = vlan->dev;
407 if (unlikely(!(dev->flags & IFF_UP)))
408 return;
409
410 nskb = skb_clone(skb, GFP_ATOMIC);
411 if (!nskb)
412 return;
413
414 len = nskb->len + ETH_HLEN;
415 nskb->dev = dev;
416
417 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
418 nskb->pkt_type = PACKET_HOST;
419
420 ret = netif_rx(nskb);
421 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
422}
423
424static void macvlan_forward_source(struct sk_buff *skb,
425 struct macvlan_port *port,
426 const unsigned char *addr)
427{
428 struct macvlan_source_entry *entry;
429 u32 idx = macvlan_eth_hash(addr);
430 struct hlist_head *h = &port->vlan_source_hash[idx];
431
432 hlist_for_each_entry_rcu(entry, h, hlist) {
433 if (ether_addr_equal_64bits(entry->addr, addr))
434 macvlan_forward_source_one(skb, entry->vlan);
435 }
436}
437
438/* called under rcu_read_lock() from netif_receive_skb */
439static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
440{
441 struct macvlan_port *port;
442 struct sk_buff *skb = *pskb;
443 const struct ethhdr *eth = eth_hdr(skb);
444 const struct macvlan_dev *vlan;
445 const struct macvlan_dev *src;
446 struct net_device *dev;
447 unsigned int len = 0;
448 int ret;
449 rx_handler_result_t handle_res;
450
451 port = macvlan_port_get_rcu(skb->dev);
452 if (is_multicast_ether_addr(eth->h_dest)) {
453 unsigned int hash;
454
455 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
456 if (!skb)
457 return RX_HANDLER_CONSUMED;
458 *pskb = skb;
459 eth = eth_hdr(skb);
460 macvlan_forward_source(skb, port, eth->h_source);
461 src = macvlan_hash_lookup(port, eth->h_source);
462 if (src && src->mode != MACVLAN_MODE_VEPA &&
463 src->mode != MACVLAN_MODE_BRIDGE) {
464 /* forward to original port. */
465 vlan = src;
466 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
467 netif_rx(skb);
468 handle_res = RX_HANDLER_CONSUMED;
469 goto out;
470 }
471
472 hash = mc_hash(NULL, eth->h_dest);
473 if (test_bit(hash, port->mc_filter))
474 macvlan_broadcast_enqueue(port, src, skb);
475
476 return RX_HANDLER_PASS;
477 }
478
479 macvlan_forward_source(skb, port, eth->h_source);
480 if (macvlan_passthru(port))
481 vlan = list_first_or_null_rcu(&port->vlans,
482 struct macvlan_dev, list);
483 else
484 vlan = macvlan_hash_lookup(port, eth->h_dest);
485 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
486 return RX_HANDLER_PASS;
487
488 dev = vlan->dev;
489 if (unlikely(!(dev->flags & IFF_UP))) {
490 kfree_skb(skb);
491 return RX_HANDLER_CONSUMED;
492 }
493 len = skb->len + ETH_HLEN;
494 skb = skb_share_check(skb, GFP_ATOMIC);
495 if (!skb) {
496 ret = NET_RX_DROP;
497 handle_res = RX_HANDLER_CONSUMED;
498 goto out;
499 }
500
501 *pskb = skb;
502 skb->dev = dev;
503 skb->pkt_type = PACKET_HOST;
504
505 ret = NET_RX_SUCCESS;
506 handle_res = RX_HANDLER_ANOTHER;
507out:
508 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
509 return handle_res;
510}
511
512static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
513{
514 const struct macvlan_dev *vlan = netdev_priv(dev);
515 const struct macvlan_port *port = vlan->port;
516 const struct macvlan_dev *dest;
517
518 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
519 const struct ethhdr *eth = (void *)skb->data;
520
521 /* send to other bridge ports directly */
522 if (is_multicast_ether_addr(eth->h_dest)) {
523 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
524 goto xmit_world;
525 }
526
527 dest = macvlan_hash_lookup(port, eth->h_dest);
528 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
529 /* send to lowerdev first for its network taps */
530 dev_forward_skb(vlan->lowerdev, skb);
531
532 return NET_XMIT_SUCCESS;
533 }
534 }
535
536xmit_world:
537 skb->dev = vlan->lowerdev;
538 return dev_queue_xmit(skb);
539}
540
541static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
542{
543#ifdef CONFIG_NET_POLL_CONTROLLER
544 if (vlan->netpoll)
545 netpoll_send_skb(vlan->netpoll, skb);
546#else
547 BUG();
548#endif
549 return NETDEV_TX_OK;
550}
551
552static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
553 struct net_device *dev)
554{
555 unsigned int len = skb->len;
556 int ret;
557 struct macvlan_dev *vlan = netdev_priv(dev);
558
559 if (unlikely(netpoll_tx_running(dev)))
560 return macvlan_netpoll_send_skb(vlan, skb);
561
562 if (vlan->fwd_priv) {
563 skb->dev = vlan->lowerdev;
564 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
565 } else {
566 ret = macvlan_queue_xmit(skb, dev);
567 }
568
569 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
570 struct vlan_pcpu_stats *pcpu_stats;
571
572 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
573 u64_stats_update_begin(&pcpu_stats->syncp);
574 pcpu_stats->tx_packets++;
575 pcpu_stats->tx_bytes += len;
576 u64_stats_update_end(&pcpu_stats->syncp);
577 } else {
578 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
579 }
580 return ret;
581}
582
583static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
584 unsigned short type, const void *daddr,
585 const void *saddr, unsigned len)
586{
587 const struct macvlan_dev *vlan = netdev_priv(dev);
588 struct net_device *lowerdev = vlan->lowerdev;
589
590 return dev_hard_header(skb, lowerdev, type, daddr,
591 saddr ? : dev->dev_addr, len);
592}
593
594static const struct header_ops macvlan_hard_header_ops = {
595 .create = macvlan_hard_header,
596 .parse = eth_header_parse,
597 .cache = eth_header_cache,
598 .cache_update = eth_header_cache_update,
599};
600
601static int macvlan_open(struct net_device *dev)
602{
603 struct macvlan_dev *vlan = netdev_priv(dev);
604 struct net_device *lowerdev = vlan->lowerdev;
605 int err;
606
607 if (macvlan_passthru(vlan->port)) {
608 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
609 err = dev_set_promiscuity(lowerdev, 1);
610 if (err < 0)
611 goto out;
612 }
613 goto hash_add;
614 }
615
616 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) {
617 vlan->fwd_priv =
618 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
619
620 /* If we get a NULL pointer back, or if we get an error
621 * then we should just fall through to the non accelerated path
622 */
623 if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
624 vlan->fwd_priv = NULL;
625 } else
626 return 0;
627 }
628
629 err = -EBUSY;
630 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
631 goto out;
632
633 err = dev_uc_add(lowerdev, dev->dev_addr);
634 if (err < 0)
635 goto out;
636 if (dev->flags & IFF_ALLMULTI) {
637 err = dev_set_allmulti(lowerdev, 1);
638 if (err < 0)
639 goto del_unicast;
640 }
641
642 if (dev->flags & IFF_PROMISC) {
643 err = dev_set_promiscuity(lowerdev, 1);
644 if (err < 0)
645 goto clear_multi;
646 }
647
648hash_add:
649 macvlan_hash_add(vlan);
650 return 0;
651
652clear_multi:
653 if (dev->flags & IFF_ALLMULTI)
654 dev_set_allmulti(lowerdev, -1);
655del_unicast:
656 dev_uc_del(lowerdev, dev->dev_addr);
657out:
658 if (vlan->fwd_priv) {
659 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
660 vlan->fwd_priv);
661 vlan->fwd_priv = NULL;
662 }
663 return err;
664}
665
666static int macvlan_stop(struct net_device *dev)
667{
668 struct macvlan_dev *vlan = netdev_priv(dev);
669 struct net_device *lowerdev = vlan->lowerdev;
670
671 if (vlan->fwd_priv) {
672 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
673 vlan->fwd_priv);
674 vlan->fwd_priv = NULL;
675 return 0;
676 }
677
678 dev_uc_unsync(lowerdev, dev);
679 dev_mc_unsync(lowerdev, dev);
680
681 if (macvlan_passthru(vlan->port)) {
682 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
683 dev_set_promiscuity(lowerdev, -1);
684 goto hash_del;
685 }
686
687 if (dev->flags & IFF_ALLMULTI)
688 dev_set_allmulti(lowerdev, -1);
689
690 if (dev->flags & IFF_PROMISC)
691 dev_set_promiscuity(lowerdev, -1);
692
693 dev_uc_del(lowerdev, dev->dev_addr);
694
695hash_del:
696 macvlan_hash_del(vlan, !dev->dismantle);
697 return 0;
698}
699
700static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
701{
702 struct macvlan_dev *vlan = netdev_priv(dev);
703 struct net_device *lowerdev = vlan->lowerdev;
704 struct macvlan_port *port = vlan->port;
705 int err;
706
707 if (!(dev->flags & IFF_UP)) {
708 /* Just copy in the new address */
709 ether_addr_copy(dev->dev_addr, addr);
710 } else {
711 /* Rehash and update the device filters */
712 if (macvlan_addr_busy(vlan->port, addr))
713 return -EBUSY;
714
715 if (!macvlan_passthru(port)) {
716 err = dev_uc_add(lowerdev, addr);
717 if (err)
718 return err;
719
720 dev_uc_del(lowerdev, dev->dev_addr);
721 }
722
723 macvlan_hash_change_addr(vlan, addr);
724 }
725 if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
726 /* Since addr_change isn't set, we are here due to lower
727 * device change. Save the lower-dev address so we can
728 * restore it later.
729 */
730 ether_addr_copy(vlan->port->perm_addr,
731 lowerdev->dev_addr);
732 }
733 macvlan_clear_addr_change(port);
734 return 0;
735}
736
737static int macvlan_set_mac_address(struct net_device *dev, void *p)
738{
739 struct macvlan_dev *vlan = netdev_priv(dev);
740 struct sockaddr *addr = p;
741
742 if (!is_valid_ether_addr(addr->sa_data))
743 return -EADDRNOTAVAIL;
744
745 /* If the addresses are the same, this is a no-op */
746 if (ether_addr_equal(dev->dev_addr, addr->sa_data))
747 return 0;
748
749 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
750 macvlan_set_addr_change(vlan->port);
751 return dev_set_mac_address(vlan->lowerdev, addr);
752 }
753
754 return macvlan_sync_address(dev, addr->sa_data);
755}
756
757static void macvlan_change_rx_flags(struct net_device *dev, int change)
758{
759 struct macvlan_dev *vlan = netdev_priv(dev);
760 struct net_device *lowerdev = vlan->lowerdev;
761
762 if (dev->flags & IFF_UP) {
763 if (change & IFF_ALLMULTI)
764 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
765 if (change & IFF_PROMISC)
766 dev_set_promiscuity(lowerdev,
767 dev->flags & IFF_PROMISC ? 1 : -1);
768
769 }
770}
771
772static void macvlan_compute_filter(unsigned long *mc_filter,
773 struct net_device *dev,
774 struct macvlan_dev *vlan)
775{
776 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
777 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
778 } else {
779 struct netdev_hw_addr *ha;
780 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
781
782 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
783 netdev_for_each_mc_addr(ha, dev) {
784 __set_bit(mc_hash(vlan, ha->addr), filter);
785 }
786
787 __set_bit(mc_hash(vlan, dev->broadcast), filter);
788
789 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
790 }
791}
792
793static void macvlan_set_mac_lists(struct net_device *dev)
794{
795 struct macvlan_dev *vlan = netdev_priv(dev);
796
797 macvlan_compute_filter(vlan->mc_filter, dev, vlan);
798
799 dev_uc_sync(vlan->lowerdev, dev);
800 dev_mc_sync(vlan->lowerdev, dev);
801
802 /* This is slightly inaccurate as we're including the subscription
803 * list of vlan->lowerdev too.
804 *
805 * Bug alert: This only works if everyone has the same broadcast
806 * address as lowerdev. As soon as someone changes theirs this
807 * will break.
808 *
809 * However, this is already broken as when you change your broadcast
810 * address we don't get called.
811 *
812 * The solution is to maintain a list of broadcast addresses like
813 * we do for uc/mc, if you care.
814 */
815 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
816}
817
818static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
819{
820 struct macvlan_dev *vlan = netdev_priv(dev);
821
822 if (vlan->lowerdev->mtu < new_mtu)
823 return -EINVAL;
824 dev->mtu = new_mtu;
825 return 0;
826}
827
828/*
829 * macvlan network devices have devices nesting below it and are a special
830 * "super class" of normal network devices; split their locks off into a
831 * separate class since they always nest.
832 */
833static struct lock_class_key macvlan_netdev_addr_lock_key;
834
835#define ALWAYS_ON_OFFLOADS \
836 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
837 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
838
839#define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
840
841#define MACVLAN_FEATURES \
842 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
843 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
844 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
845 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
846
847#define MACVLAN_STATE_MASK \
848 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
849
850static int macvlan_get_nest_level(struct net_device *dev)
851{
852 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
853}
854
855static void macvlan_set_lockdep_class(struct net_device *dev)
856{
857 netdev_lockdep_set_classes(dev);
858 lockdep_set_class_and_subclass(&dev->addr_list_lock,
859 &macvlan_netdev_addr_lock_key,
860 macvlan_get_nest_level(dev));
861}
862
863static int macvlan_init(struct net_device *dev)
864{
865 struct macvlan_dev *vlan = netdev_priv(dev);
866 const struct net_device *lowerdev = vlan->lowerdev;
867 struct macvlan_port *port = vlan->port;
868
869 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
870 (lowerdev->state & MACVLAN_STATE_MASK);
871 dev->features = lowerdev->features & MACVLAN_FEATURES;
872 dev->features |= ALWAYS_ON_FEATURES;
873 dev->hw_features |= NETIF_F_LRO;
874 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
875 dev->vlan_features |= ALWAYS_ON_OFFLOADS;
876 dev->hw_enc_features |= dev->features;
877 dev->gso_max_size = lowerdev->gso_max_size;
878 dev->gso_max_segs = lowerdev->gso_max_segs;
879 dev->hard_header_len = lowerdev->hard_header_len;
880
881 macvlan_set_lockdep_class(dev);
882
883 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
884 if (!vlan->pcpu_stats)
885 return -ENOMEM;
886
887 port->count += 1;
888
889 return 0;
890}
891
892static void macvlan_uninit(struct net_device *dev)
893{
894 struct macvlan_dev *vlan = netdev_priv(dev);
895 struct macvlan_port *port = vlan->port;
896
897 free_percpu(vlan->pcpu_stats);
898
899 macvlan_flush_sources(port, vlan);
900 port->count -= 1;
901 if (!port->count)
902 macvlan_port_destroy(port->dev);
903}
904
905static void macvlan_dev_get_stats64(struct net_device *dev,
906 struct rtnl_link_stats64 *stats)
907{
908 struct macvlan_dev *vlan = netdev_priv(dev);
909
910 if (vlan->pcpu_stats) {
911 struct vlan_pcpu_stats *p;
912 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
913 u32 rx_errors = 0, tx_dropped = 0;
914 unsigned int start;
915 int i;
916
917 for_each_possible_cpu(i) {
918 p = per_cpu_ptr(vlan->pcpu_stats, i);
919 do {
920 start = u64_stats_fetch_begin_irq(&p->syncp);
921 rx_packets = p->rx_packets;
922 rx_bytes = p->rx_bytes;
923 rx_multicast = p->rx_multicast;
924 tx_packets = p->tx_packets;
925 tx_bytes = p->tx_bytes;
926 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
927
928 stats->rx_packets += rx_packets;
929 stats->rx_bytes += rx_bytes;
930 stats->multicast += rx_multicast;
931 stats->tx_packets += tx_packets;
932 stats->tx_bytes += tx_bytes;
933 /* rx_errors & tx_dropped are u32, updated
934 * without syncp protection.
935 */
936 rx_errors += p->rx_errors;
937 tx_dropped += p->tx_dropped;
938 }
939 stats->rx_errors = rx_errors;
940 stats->rx_dropped = rx_errors;
941 stats->tx_dropped = tx_dropped;
942 }
943}
944
945static int macvlan_vlan_rx_add_vid(struct net_device *dev,
946 __be16 proto, u16 vid)
947{
948 struct macvlan_dev *vlan = netdev_priv(dev);
949 struct net_device *lowerdev = vlan->lowerdev;
950
951 return vlan_vid_add(lowerdev, proto, vid);
952}
953
954static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
955 __be16 proto, u16 vid)
956{
957 struct macvlan_dev *vlan = netdev_priv(dev);
958 struct net_device *lowerdev = vlan->lowerdev;
959
960 vlan_vid_del(lowerdev, proto, vid);
961 return 0;
962}
963
964static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
965 struct net_device *dev,
966 const unsigned char *addr, u16 vid,
967 u16 flags)
968{
969 struct macvlan_dev *vlan = netdev_priv(dev);
970 int err = -EINVAL;
971
972 /* Support unicast filter only on passthru devices.
973 * Multicast filter should be allowed on all devices.
974 */
975 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
976 return -EOPNOTSUPP;
977
978 if (flags & NLM_F_REPLACE)
979 return -EOPNOTSUPP;
980
981 if (is_unicast_ether_addr(addr))
982 err = dev_uc_add_excl(dev, addr);
983 else if (is_multicast_ether_addr(addr))
984 err = dev_mc_add_excl(dev, addr);
985
986 return err;
987}
988
989static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
990 struct net_device *dev,
991 const unsigned char *addr, u16 vid)
992{
993 struct macvlan_dev *vlan = netdev_priv(dev);
994 int err = -EINVAL;
995
996 /* Support unicast filter only on passthru devices.
997 * Multicast filter should be allowed on all devices.
998 */
999 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1000 return -EOPNOTSUPP;
1001
1002 if (is_unicast_ether_addr(addr))
1003 err = dev_uc_del(dev, addr);
1004 else if (is_multicast_ether_addr(addr))
1005 err = dev_mc_del(dev, addr);
1006
1007 return err;
1008}
1009
1010static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1011 struct ethtool_drvinfo *drvinfo)
1012{
1013 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1014 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1015}
1016
1017static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1018 struct ethtool_link_ksettings *cmd)
1019{
1020 const struct macvlan_dev *vlan = netdev_priv(dev);
1021
1022 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1023}
1024
1025static netdev_features_t macvlan_fix_features(struct net_device *dev,
1026 netdev_features_t features)
1027{
1028 struct macvlan_dev *vlan = netdev_priv(dev);
1029 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1030 netdev_features_t mask;
1031
1032 features |= NETIF_F_ALL_FOR_ALL;
1033 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1034 mask = features;
1035
1036 lowerdev_features &= (features | ~NETIF_F_LRO);
1037 features = netdev_increment_features(lowerdev_features, features, mask);
1038 features |= ALWAYS_ON_FEATURES;
1039 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1040
1041 return features;
1042}
1043
1044#ifdef CONFIG_NET_POLL_CONTROLLER
1045static void macvlan_dev_poll_controller(struct net_device *dev)
1046{
1047 return;
1048}
1049
1050static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1051{
1052 struct macvlan_dev *vlan = netdev_priv(dev);
1053 struct net_device *real_dev = vlan->lowerdev;
1054 struct netpoll *netpoll;
1055 int err = 0;
1056
1057 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1058 err = -ENOMEM;
1059 if (!netpoll)
1060 goto out;
1061
1062 err = __netpoll_setup(netpoll, real_dev);
1063 if (err) {
1064 kfree(netpoll);
1065 goto out;
1066 }
1067
1068 vlan->netpoll = netpoll;
1069
1070out:
1071 return err;
1072}
1073
1074static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1075{
1076 struct macvlan_dev *vlan = netdev_priv(dev);
1077 struct netpoll *netpoll = vlan->netpoll;
1078
1079 if (!netpoll)
1080 return;
1081
1082 vlan->netpoll = NULL;
1083
1084 __netpoll_free_async(netpoll);
1085}
1086#endif /* CONFIG_NET_POLL_CONTROLLER */
1087
1088static int macvlan_dev_get_iflink(const struct net_device *dev)
1089{
1090 struct macvlan_dev *vlan = netdev_priv(dev);
1091
1092 return vlan->lowerdev->ifindex;
1093}
1094
1095static const struct ethtool_ops macvlan_ethtool_ops = {
1096 .get_link = ethtool_op_get_link,
1097 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1098 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1099};
1100
1101static const struct net_device_ops macvlan_netdev_ops = {
1102 .ndo_init = macvlan_init,
1103 .ndo_uninit = macvlan_uninit,
1104 .ndo_open = macvlan_open,
1105 .ndo_stop = macvlan_stop,
1106 .ndo_start_xmit = macvlan_start_xmit,
1107 .ndo_change_mtu = macvlan_change_mtu,
1108 .ndo_fix_features = macvlan_fix_features,
1109 .ndo_change_rx_flags = macvlan_change_rx_flags,
1110 .ndo_set_mac_address = macvlan_set_mac_address,
1111 .ndo_set_rx_mode = macvlan_set_mac_lists,
1112 .ndo_get_stats64 = macvlan_dev_get_stats64,
1113 .ndo_validate_addr = eth_validate_addr,
1114 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1115 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1116 .ndo_fdb_add = macvlan_fdb_add,
1117 .ndo_fdb_del = macvlan_fdb_del,
1118 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1119 .ndo_get_lock_subclass = macvlan_get_nest_level,
1120#ifdef CONFIG_NET_POLL_CONTROLLER
1121 .ndo_poll_controller = macvlan_dev_poll_controller,
1122 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1123 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1124#endif
1125 .ndo_get_iflink = macvlan_dev_get_iflink,
1126 .ndo_features_check = passthru_features_check,
1127};
1128
1129void macvlan_common_setup(struct net_device *dev)
1130{
1131 ether_setup(dev);
1132
1133 dev->min_mtu = 0;
1134 dev->max_mtu = ETH_MAX_MTU;
1135 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1136 netif_keep_dst(dev);
1137 dev->priv_flags |= IFF_UNICAST_FLT;
1138 dev->netdev_ops = &macvlan_netdev_ops;
1139 dev->needs_free_netdev = true;
1140 dev->header_ops = &macvlan_hard_header_ops;
1141 dev->ethtool_ops = &macvlan_ethtool_ops;
1142}
1143EXPORT_SYMBOL_GPL(macvlan_common_setup);
1144
1145static void macvlan_setup(struct net_device *dev)
1146{
1147 macvlan_common_setup(dev);
1148 dev->priv_flags |= IFF_NO_QUEUE;
1149}
1150
1151static int macvlan_port_create(struct net_device *dev)
1152{
1153 struct macvlan_port *port;
1154 unsigned int i;
1155 int err;
1156
1157 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1158 return -EINVAL;
1159
1160 if (netdev_is_rx_handler_busy(dev))
1161 return -EBUSY;
1162
1163 port = kzalloc(sizeof(*port), GFP_KERNEL);
1164 if (port == NULL)
1165 return -ENOMEM;
1166
1167 port->dev = dev;
1168 ether_addr_copy(port->perm_addr, dev->dev_addr);
1169 INIT_LIST_HEAD(&port->vlans);
1170 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1171 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1172 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1173 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1174
1175 skb_queue_head_init(&port->bc_queue);
1176 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1177
1178 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1179 if (err)
1180 kfree(port);
1181 else
1182 dev->priv_flags |= IFF_MACVLAN_PORT;
1183 return err;
1184}
1185
1186static void macvlan_port_destroy(struct net_device *dev)
1187{
1188 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1189 struct sk_buff *skb;
1190
1191 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1192 netdev_rx_handler_unregister(dev);
1193
1194 /* After this point, no packet can schedule bc_work anymore,
1195 * but we need to cancel it and purge left skbs if any.
1196 */
1197 cancel_work_sync(&port->bc_work);
1198
1199 while ((skb = __skb_dequeue(&port->bc_queue))) {
1200 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1201
1202 if (src)
1203 dev_put(src->dev);
1204
1205 kfree_skb(skb);
1206 }
1207
1208 /* If the lower device address has been changed by passthru
1209 * macvlan, put it back.
1210 */
1211 if (macvlan_passthru(port) &&
1212 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1213 struct sockaddr sa;
1214
1215 sa.sa_family = port->dev->type;
1216 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1217 dev_set_mac_address(port->dev, &sa);
1218 }
1219
1220 kfree(port);
1221}
1222
1223static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1224 struct netlink_ext_ack *extack)
1225{
1226 if (tb[IFLA_ADDRESS]) {
1227 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1228 return -EINVAL;
1229 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1230 return -EADDRNOTAVAIL;
1231 }
1232
1233 if (!data)
1234 return 0;
1235
1236 if (data[IFLA_MACVLAN_FLAGS] &&
1237 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1238 return -EINVAL;
1239
1240 if (data[IFLA_MACVLAN_MODE]) {
1241 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1242 case MACVLAN_MODE_PRIVATE:
1243 case MACVLAN_MODE_VEPA:
1244 case MACVLAN_MODE_BRIDGE:
1245 case MACVLAN_MODE_PASSTHRU:
1246 case MACVLAN_MODE_SOURCE:
1247 break;
1248 default:
1249 return -EINVAL;
1250 }
1251 }
1252
1253 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1254 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1255 case MACVLAN_MACADDR_ADD:
1256 case MACVLAN_MACADDR_DEL:
1257 case MACVLAN_MACADDR_FLUSH:
1258 case MACVLAN_MACADDR_SET:
1259 break;
1260 default:
1261 return -EINVAL;
1262 }
1263 }
1264
1265 if (data[IFLA_MACVLAN_MACADDR]) {
1266 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1267 return -EINVAL;
1268
1269 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1270 return -EADDRNOTAVAIL;
1271 }
1272
1273 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1274 return -EINVAL;
1275
1276 return 0;
1277}
1278
1279/**
1280 * reconfigure list of remote source mac address
1281 * (only for macvlan devices in source mode)
1282 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1283 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1284 */
1285static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1286 struct nlattr *data[])
1287{
1288 char *addr = NULL;
1289 int ret, rem, len;
1290 struct nlattr *nla, *head;
1291 struct macvlan_source_entry *entry;
1292
1293 if (data[IFLA_MACVLAN_MACADDR])
1294 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1295
1296 if (mode == MACVLAN_MACADDR_ADD) {
1297 if (!addr)
1298 return -EINVAL;
1299
1300 return macvlan_hash_add_source(vlan, addr);
1301
1302 } else if (mode == MACVLAN_MACADDR_DEL) {
1303 if (!addr)
1304 return -EINVAL;
1305
1306 entry = macvlan_hash_lookup_source(vlan, addr);
1307 if (entry) {
1308 macvlan_hash_del_source(entry);
1309 vlan->macaddr_count--;
1310 }
1311 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1312 macvlan_flush_sources(vlan->port, vlan);
1313 } else if (mode == MACVLAN_MACADDR_SET) {
1314 macvlan_flush_sources(vlan->port, vlan);
1315
1316 if (addr) {
1317 ret = macvlan_hash_add_source(vlan, addr);
1318 if (ret)
1319 return ret;
1320 }
1321
1322 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1323 return 0;
1324
1325 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1326 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1327
1328 nla_for_each_attr(nla, head, len, rem) {
1329 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1330 nla_len(nla) != ETH_ALEN)
1331 continue;
1332
1333 addr = nla_data(nla);
1334 ret = macvlan_hash_add_source(vlan, addr);
1335 if (ret)
1336 return ret;
1337 }
1338 } else {
1339 return -EINVAL;
1340 }
1341
1342 return 0;
1343}
1344
1345int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1346 struct nlattr *tb[], struct nlattr *data[],
1347 struct netlink_ext_ack *extack)
1348{
1349 struct macvlan_dev *vlan = netdev_priv(dev);
1350 struct macvlan_port *port;
1351 struct net_device *lowerdev;
1352 int err;
1353 int macmode;
1354 bool create = false;
1355
1356 if (!tb[IFLA_LINK])
1357 return -EINVAL;
1358
1359 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1360 if (lowerdev == NULL)
1361 return -ENODEV;
1362
1363 /* When creating macvlans or macvtaps on top of other macvlans - use
1364 * the real device as the lowerdev.
1365 */
1366 if (netif_is_macvlan(lowerdev))
1367 lowerdev = macvlan_dev_real_dev(lowerdev);
1368
1369 if (!tb[IFLA_MTU])
1370 dev->mtu = lowerdev->mtu;
1371 else if (dev->mtu > lowerdev->mtu)
1372 return -EINVAL;
1373
1374 /* MTU range: 68 - lowerdev->max_mtu */
1375 dev->min_mtu = ETH_MIN_MTU;
1376 dev->max_mtu = lowerdev->max_mtu;
1377
1378 if (!tb[IFLA_ADDRESS])
1379 eth_hw_addr_random(dev);
1380
1381 if (!macvlan_port_exists(lowerdev)) {
1382 err = macvlan_port_create(lowerdev);
1383 if (err < 0)
1384 return err;
1385 create = true;
1386 }
1387 port = macvlan_port_get_rtnl(lowerdev);
1388
1389 /* Only 1 macvlan device can be created in passthru mode */
1390 if (macvlan_passthru(port)) {
1391 /* The macvlan port must be not created this time,
1392 * still goto destroy_macvlan_port for readability.
1393 */
1394 err = -EINVAL;
1395 goto destroy_macvlan_port;
1396 }
1397
1398 vlan->lowerdev = lowerdev;
1399 vlan->dev = dev;
1400 vlan->port = port;
1401 vlan->set_features = MACVLAN_FEATURES;
1402 vlan->nest_level = dev_get_nest_level(lowerdev) + 1;
1403
1404 vlan->mode = MACVLAN_MODE_VEPA;
1405 if (data && data[IFLA_MACVLAN_MODE])
1406 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1407
1408 if (data && data[IFLA_MACVLAN_FLAGS])
1409 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1410
1411 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1412 if (port->count) {
1413 err = -EINVAL;
1414 goto destroy_macvlan_port;
1415 }
1416 macvlan_set_passthru(port);
1417 eth_hw_addr_inherit(dev, lowerdev);
1418 }
1419
1420 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1421 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1422 err = -EINVAL;
1423 goto destroy_macvlan_port;
1424 }
1425 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1426 err = macvlan_changelink_sources(vlan, macmode, data);
1427 if (err)
1428 goto destroy_macvlan_port;
1429 }
1430
1431 err = register_netdevice(dev);
1432 if (err < 0)
1433 goto destroy_macvlan_port;
1434
1435 dev->priv_flags |= IFF_MACVLAN;
1436 err = netdev_upper_dev_link(lowerdev, dev, extack);
1437 if (err)
1438 goto unregister_netdev;
1439
1440 list_add_tail_rcu(&vlan->list, &port->vlans);
1441 netif_stacked_transfer_operstate(lowerdev, dev);
1442 linkwatch_fire_event(dev);
1443
1444 return 0;
1445
1446unregister_netdev:
1447 /* macvlan_uninit would free the macvlan port */
1448 unregister_netdevice(dev);
1449 return err;
1450destroy_macvlan_port:
1451 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1452 * so we destroy the macvlan port only when it's valid.
1453 */
1454 if (create && macvlan_port_get_rtnl(lowerdev))
1455 macvlan_port_destroy(port->dev);
1456 return err;
1457}
1458EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1459
1460static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1461 struct nlattr *tb[], struct nlattr *data[],
1462 struct netlink_ext_ack *extack)
1463{
1464 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1465}
1466
1467void macvlan_dellink(struct net_device *dev, struct list_head *head)
1468{
1469 struct macvlan_dev *vlan = netdev_priv(dev);
1470
1471 if (vlan->mode == MACVLAN_MODE_SOURCE)
1472 macvlan_flush_sources(vlan->port, vlan);
1473 list_del_rcu(&vlan->list);
1474 unregister_netdevice_queue(dev, head);
1475 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1476}
1477EXPORT_SYMBOL_GPL(macvlan_dellink);
1478
1479static int macvlan_changelink(struct net_device *dev,
1480 struct nlattr *tb[], struct nlattr *data[],
1481 struct netlink_ext_ack *extack)
1482{
1483 struct macvlan_dev *vlan = netdev_priv(dev);
1484 enum macvlan_mode mode;
1485 bool set_mode = false;
1486 enum macvlan_macaddr_mode macmode;
1487 int ret;
1488
1489 /* Validate mode, but don't set yet: setting flags may fail. */
1490 if (data && data[IFLA_MACVLAN_MODE]) {
1491 set_mode = true;
1492 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1493 /* Passthrough mode can't be set or cleared dynamically */
1494 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1495 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1496 return -EINVAL;
1497 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1498 vlan->mode != mode)
1499 macvlan_flush_sources(vlan->port, vlan);
1500 }
1501
1502 if (data && data[IFLA_MACVLAN_FLAGS]) {
1503 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1504 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1505 if (macvlan_passthru(vlan->port) && promisc) {
1506 int err;
1507
1508 if (flags & MACVLAN_FLAG_NOPROMISC)
1509 err = dev_set_promiscuity(vlan->lowerdev, -1);
1510 else
1511 err = dev_set_promiscuity(vlan->lowerdev, 1);
1512 if (err < 0)
1513 return err;
1514 }
1515 vlan->flags = flags;
1516 }
1517 if (set_mode)
1518 vlan->mode = mode;
1519 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1520 if (vlan->mode != MACVLAN_MODE_SOURCE)
1521 return -EINVAL;
1522 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1523 ret = macvlan_changelink_sources(vlan, macmode, data);
1524 if (ret)
1525 return ret;
1526 }
1527 return 0;
1528}
1529
1530static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1531{
1532 if (vlan->macaddr_count == 0)
1533 return 0;
1534 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1535 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1536}
1537
1538static size_t macvlan_get_size(const struct net_device *dev)
1539{
1540 struct macvlan_dev *vlan = netdev_priv(dev);
1541
1542 return (0
1543 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1544 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1545 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1546 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1547 );
1548}
1549
1550static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1551 const struct macvlan_dev *vlan,
1552 const int i)
1553{
1554 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1555 struct macvlan_source_entry *entry;
1556
1557 hlist_for_each_entry_rcu(entry, h, hlist) {
1558 if (entry->vlan != vlan)
1559 continue;
1560 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1561 return 1;
1562 }
1563 return 0;
1564}
1565
1566static int macvlan_fill_info(struct sk_buff *skb,
1567 const struct net_device *dev)
1568{
1569 struct macvlan_dev *vlan = netdev_priv(dev);
1570 int i;
1571 struct nlattr *nest;
1572
1573 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1574 goto nla_put_failure;
1575 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1576 goto nla_put_failure;
1577 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1578 goto nla_put_failure;
1579 if (vlan->macaddr_count > 0) {
1580 nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA);
1581 if (nest == NULL)
1582 goto nla_put_failure;
1583
1584 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1585 if (macvlan_fill_info_macaddr(skb, vlan, i))
1586 goto nla_put_failure;
1587 }
1588 nla_nest_end(skb, nest);
1589 }
1590 return 0;
1591
1592nla_put_failure:
1593 return -EMSGSIZE;
1594}
1595
1596static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1597 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1598 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1599 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1600 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1601 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1602 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1603};
1604
1605int macvlan_link_register(struct rtnl_link_ops *ops)
1606{
1607 /* common fields */
1608 ops->validate = macvlan_validate;
1609 ops->maxtype = IFLA_MACVLAN_MAX;
1610 ops->policy = macvlan_policy;
1611 ops->changelink = macvlan_changelink;
1612 ops->get_size = macvlan_get_size;
1613 ops->fill_info = macvlan_fill_info;
1614
1615 return rtnl_link_register(ops);
1616};
1617EXPORT_SYMBOL_GPL(macvlan_link_register);
1618
1619static struct net *macvlan_get_link_net(const struct net_device *dev)
1620{
1621 return dev_net(macvlan_dev_real_dev(dev));
1622}
1623
1624static struct rtnl_link_ops macvlan_link_ops = {
1625 .kind = "macvlan",
1626 .setup = macvlan_setup,
1627 .newlink = macvlan_newlink,
1628 .dellink = macvlan_dellink,
1629 .get_link_net = macvlan_get_link_net,
1630 .priv_size = sizeof(struct macvlan_dev),
1631};
1632
1633static int macvlan_device_event(struct notifier_block *unused,
1634 unsigned long event, void *ptr)
1635{
1636 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1637 struct macvlan_dev *vlan, *next;
1638 struct macvlan_port *port;
1639 LIST_HEAD(list_kill);
1640
1641 if (!macvlan_port_exists(dev))
1642 return NOTIFY_DONE;
1643
1644 port = macvlan_port_get_rtnl(dev);
1645
1646 switch (event) {
1647 case NETDEV_UP:
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");
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4 *
5 * The code this is based on carried the following copyright notice:
6 * ---
7 * (C) Copyright 2001-2006
8 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
9 * Re-worked by Ben Greear <greearb@candelatech.com>
10 * ---
11 */
12#include <linux/kernel.h>
13#include <linux/types.h>
14#include <linux/module.h>
15#include <linux/init.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/string.h>
19#include <linux/rculist.h>
20#include <linux/notifier.h>
21#include <linux/netdevice.h>
22#include <linux/etherdevice.h>
23#include <linux/net_tstamp.h>
24#include <linux/ethtool.h>
25#include <linux/if_arp.h>
26#include <linux/if_vlan.h>
27#include <linux/if_link.h>
28#include <linux/if_macvlan.h>
29#include <linux/hash.h>
30#include <linux/workqueue.h>
31#include <net/rtnetlink.h>
32#include <net/xfrm.h>
33#include <linux/netpoll.h>
34#include <linux/phy.h>
35
36#define MACVLAN_HASH_BITS 8
37#define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS)
38#define MACVLAN_BC_QUEUE_LEN 1000
39
40#define MACVLAN_F_PASSTHRU 1
41#define MACVLAN_F_ADDRCHANGE 2
42
43struct macvlan_port {
44 struct net_device *dev;
45 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
46 struct list_head vlans;
47 struct sk_buff_head bc_queue;
48 struct work_struct bc_work;
49 u32 flags;
50 int count;
51 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE];
52 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
53 unsigned char perm_addr[ETH_ALEN];
54};
55
56struct macvlan_source_entry {
57 struct hlist_node hlist;
58 struct macvlan_dev *vlan;
59 unsigned char addr[6+2] __aligned(sizeof(u16));
60 struct rcu_head rcu;
61};
62
63struct macvlan_skb_cb {
64 const struct macvlan_dev *src;
65};
66
67#define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
68
69static void macvlan_port_destroy(struct net_device *dev);
70
71static inline bool macvlan_passthru(const struct macvlan_port *port)
72{
73 return port->flags & MACVLAN_F_PASSTHRU;
74}
75
76static inline void macvlan_set_passthru(struct macvlan_port *port)
77{
78 port->flags |= MACVLAN_F_PASSTHRU;
79}
80
81static inline bool macvlan_addr_change(const struct macvlan_port *port)
82{
83 return port->flags & MACVLAN_F_ADDRCHANGE;
84}
85
86static inline void macvlan_set_addr_change(struct macvlan_port *port)
87{
88 port->flags |= MACVLAN_F_ADDRCHANGE;
89}
90
91static inline void macvlan_clear_addr_change(struct macvlan_port *port)
92{
93 port->flags &= ~MACVLAN_F_ADDRCHANGE;
94}
95
96/* Hash Ethernet address */
97static u32 macvlan_eth_hash(const unsigned char *addr)
98{
99 u64 value = get_unaligned((u64 *)addr);
100
101 /* only want 6 bytes */
102#ifdef __BIG_ENDIAN
103 value >>= 16;
104#else
105 value <<= 16;
106#endif
107 return hash_64(value, MACVLAN_HASH_BITS);
108}
109
110static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
111{
112 return rcu_dereference(dev->rx_handler_data);
113}
114
115static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
116{
117 return rtnl_dereference(dev->rx_handler_data);
118}
119
120static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
121 const unsigned char *addr)
122{
123 struct macvlan_dev *vlan;
124 u32 idx = macvlan_eth_hash(addr);
125
126 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
127 lockdep_rtnl_is_held()) {
128 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
129 return vlan;
130 }
131 return NULL;
132}
133
134static struct macvlan_source_entry *macvlan_hash_lookup_source(
135 const struct macvlan_dev *vlan,
136 const unsigned char *addr)
137{
138 struct macvlan_source_entry *entry;
139 u32 idx = macvlan_eth_hash(addr);
140 struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
141
142 hlist_for_each_entry_rcu(entry, h, hlist) {
143 if (ether_addr_equal_64bits(entry->addr, addr) &&
144 entry->vlan == vlan)
145 return entry;
146 }
147 return NULL;
148}
149
150static int macvlan_hash_add_source(struct macvlan_dev *vlan,
151 const unsigned char *addr)
152{
153 struct macvlan_port *port = vlan->port;
154 struct macvlan_source_entry *entry;
155 struct hlist_head *h;
156
157 entry = macvlan_hash_lookup_source(vlan, addr);
158 if (entry)
159 return 0;
160
161 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
162 if (!entry)
163 return -ENOMEM;
164
165 ether_addr_copy(entry->addr, addr);
166 entry->vlan = vlan;
167 h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
168 hlist_add_head_rcu(&entry->hlist, h);
169 vlan->macaddr_count++;
170
171 return 0;
172}
173
174static void macvlan_hash_add(struct macvlan_dev *vlan)
175{
176 struct macvlan_port *port = vlan->port;
177 const unsigned char *addr = vlan->dev->dev_addr;
178 u32 idx = macvlan_eth_hash(addr);
179
180 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
181}
182
183static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
184{
185 hlist_del_rcu(&entry->hlist);
186 kfree_rcu(entry, rcu);
187}
188
189static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
190{
191 hlist_del_rcu(&vlan->hlist);
192 if (sync)
193 synchronize_rcu();
194}
195
196static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
197 const unsigned char *addr)
198{
199 macvlan_hash_del(vlan, true);
200 /* Now that we are unhashed it is safe to change the device
201 * address without confusing packet delivery.
202 */
203 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
204 macvlan_hash_add(vlan);
205}
206
207static bool macvlan_addr_busy(const struct macvlan_port *port,
208 const unsigned char *addr)
209{
210 /* Test to see if the specified address is
211 * currently in use by the underlying device or
212 * another macvlan.
213 */
214 if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
215 ether_addr_equal_64bits(port->dev->dev_addr, addr))
216 return true;
217
218 if (macvlan_hash_lookup(port, addr))
219 return true;
220
221 return false;
222}
223
224
225static int macvlan_broadcast_one(struct sk_buff *skb,
226 const struct macvlan_dev *vlan,
227 const struct ethhdr *eth, bool local)
228{
229 struct net_device *dev = vlan->dev;
230
231 if (local)
232 return __dev_forward_skb(dev, skb);
233
234 skb->dev = dev;
235 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
236 skb->pkt_type = PACKET_BROADCAST;
237 else
238 skb->pkt_type = PACKET_MULTICAST;
239
240 return 0;
241}
242
243static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
244{
245 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
246}
247
248
249static unsigned int mc_hash(const struct macvlan_dev *vlan,
250 const unsigned char *addr)
251{
252 u32 val = __get_unaligned_cpu32(addr + 2);
253
254 val ^= macvlan_hash_mix(vlan);
255 return hash_32(val, MACVLAN_MC_FILTER_BITS);
256}
257
258static void macvlan_broadcast(struct sk_buff *skb,
259 const struct macvlan_port *port,
260 struct net_device *src,
261 enum macvlan_mode mode)
262{
263 const struct ethhdr *eth = eth_hdr(skb);
264 const struct macvlan_dev *vlan;
265 struct sk_buff *nskb;
266 unsigned int i;
267 int err;
268 unsigned int hash;
269
270 if (skb->protocol == htons(ETH_P_PAUSE))
271 return;
272
273 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
274 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
275 if (vlan->dev == src || !(vlan->mode & mode))
276 continue;
277
278 hash = mc_hash(vlan, eth->h_dest);
279 if (!test_bit(hash, vlan->mc_filter))
280 continue;
281
282 err = NET_RX_DROP;
283 nskb = skb_clone(skb, GFP_ATOMIC);
284 if (likely(nskb))
285 err = macvlan_broadcast_one(
286 nskb, vlan, eth,
287 mode == MACVLAN_MODE_BRIDGE) ?:
288 netif_rx_ni(nskb);
289 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
290 err == NET_RX_SUCCESS, true);
291 }
292 }
293}
294
295static void macvlan_process_broadcast(struct work_struct *w)
296{
297 struct macvlan_port *port = container_of(w, struct macvlan_port,
298 bc_work);
299 struct sk_buff *skb;
300 struct sk_buff_head list;
301
302 __skb_queue_head_init(&list);
303
304 spin_lock_bh(&port->bc_queue.lock);
305 skb_queue_splice_tail_init(&port->bc_queue, &list);
306 spin_unlock_bh(&port->bc_queue.lock);
307
308 while ((skb = __skb_dequeue(&list))) {
309 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
310
311 rcu_read_lock();
312
313 if (!src)
314 /* frame comes from an external address */
315 macvlan_broadcast(skb, port, NULL,
316 MACVLAN_MODE_PRIVATE |
317 MACVLAN_MODE_VEPA |
318 MACVLAN_MODE_PASSTHRU|
319 MACVLAN_MODE_BRIDGE);
320 else if (src->mode == MACVLAN_MODE_VEPA)
321 /* flood to everyone except source */
322 macvlan_broadcast(skb, port, src->dev,
323 MACVLAN_MODE_VEPA |
324 MACVLAN_MODE_BRIDGE);
325 else
326 /*
327 * flood only to VEPA ports, bridge ports
328 * already saw the frame on the way out.
329 */
330 macvlan_broadcast(skb, port, src->dev,
331 MACVLAN_MODE_VEPA);
332
333 rcu_read_unlock();
334
335 if (src)
336 dev_put(src->dev);
337 consume_skb(skb);
338
339 cond_resched();
340 }
341}
342
343static void macvlan_broadcast_enqueue(struct macvlan_port *port,
344 const struct macvlan_dev *src,
345 struct sk_buff *skb)
346{
347 struct sk_buff *nskb;
348 int err = -ENOMEM;
349
350 nskb = skb_clone(skb, GFP_ATOMIC);
351 if (!nskb)
352 goto err;
353
354 MACVLAN_SKB_CB(nskb)->src = src;
355
356 spin_lock(&port->bc_queue.lock);
357 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
358 if (src)
359 dev_hold(src->dev);
360 __skb_queue_tail(&port->bc_queue, nskb);
361 err = 0;
362 }
363 spin_unlock(&port->bc_queue.lock);
364
365 schedule_work(&port->bc_work);
366
367 if (err)
368 goto free_nskb;
369
370 return;
371
372free_nskb:
373 kfree_skb(nskb);
374err:
375 atomic_long_inc(&skb->dev->rx_dropped);
376}
377
378static void macvlan_flush_sources(struct macvlan_port *port,
379 struct macvlan_dev *vlan)
380{
381 int i;
382
383 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
384 struct hlist_node *h, *n;
385
386 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
387 struct macvlan_source_entry *entry;
388
389 entry = hlist_entry(h, struct macvlan_source_entry,
390 hlist);
391 if (entry->vlan == vlan)
392 macvlan_hash_del_source(entry);
393 }
394 }
395 vlan->macaddr_count = 0;
396}
397
398static void macvlan_forward_source_one(struct sk_buff *skb,
399 struct macvlan_dev *vlan)
400{
401 struct sk_buff *nskb;
402 struct net_device *dev;
403 int len;
404 int ret;
405
406 dev = vlan->dev;
407 if (unlikely(!(dev->flags & IFF_UP)))
408 return;
409
410 nskb = skb_clone(skb, GFP_ATOMIC);
411 if (!nskb)
412 return;
413
414 len = nskb->len + ETH_HLEN;
415 nskb->dev = dev;
416
417 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
418 nskb->pkt_type = PACKET_HOST;
419
420 ret = netif_rx(nskb);
421 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
422}
423
424static void macvlan_forward_source(struct sk_buff *skb,
425 struct macvlan_port *port,
426 const unsigned char *addr)
427{
428 struct macvlan_source_entry *entry;
429 u32 idx = macvlan_eth_hash(addr);
430 struct hlist_head *h = &port->vlan_source_hash[idx];
431
432 hlist_for_each_entry_rcu(entry, h, hlist) {
433 if (ether_addr_equal_64bits(entry->addr, addr))
434 macvlan_forward_source_one(skb, entry->vlan);
435 }
436}
437
438/* called under rcu_read_lock() from netif_receive_skb */
439static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
440{
441 struct macvlan_port *port;
442 struct sk_buff *skb = *pskb;
443 const struct ethhdr *eth = eth_hdr(skb);
444 const struct macvlan_dev *vlan;
445 const struct macvlan_dev *src;
446 struct net_device *dev;
447 unsigned int len = 0;
448 int ret;
449 rx_handler_result_t handle_res;
450
451 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */
452 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
453 return RX_HANDLER_PASS;
454
455 port = macvlan_port_get_rcu(skb->dev);
456 if (is_multicast_ether_addr(eth->h_dest)) {
457 unsigned int hash;
458
459 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
460 if (!skb)
461 return RX_HANDLER_CONSUMED;
462 *pskb = skb;
463 eth = eth_hdr(skb);
464 macvlan_forward_source(skb, port, eth->h_source);
465 src = macvlan_hash_lookup(port, eth->h_source);
466 if (src && src->mode != MACVLAN_MODE_VEPA &&
467 src->mode != MACVLAN_MODE_BRIDGE) {
468 /* forward to original port. */
469 vlan = src;
470 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
471 netif_rx(skb);
472 handle_res = RX_HANDLER_CONSUMED;
473 goto out;
474 }
475
476 hash = mc_hash(NULL, eth->h_dest);
477 if (test_bit(hash, port->mc_filter))
478 macvlan_broadcast_enqueue(port, src, skb);
479
480 return RX_HANDLER_PASS;
481 }
482
483 macvlan_forward_source(skb, port, eth->h_source);
484 if (macvlan_passthru(port))
485 vlan = list_first_or_null_rcu(&port->vlans,
486 struct macvlan_dev, list);
487 else
488 vlan = macvlan_hash_lookup(port, eth->h_dest);
489 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
490 return RX_HANDLER_PASS;
491
492 dev = vlan->dev;
493 if (unlikely(!(dev->flags & IFF_UP))) {
494 kfree_skb(skb);
495 return RX_HANDLER_CONSUMED;
496 }
497 len = skb->len + ETH_HLEN;
498 skb = skb_share_check(skb, GFP_ATOMIC);
499 if (!skb) {
500 ret = NET_RX_DROP;
501 handle_res = RX_HANDLER_CONSUMED;
502 goto out;
503 }
504
505 *pskb = skb;
506 skb->dev = dev;
507 skb->pkt_type = PACKET_HOST;
508
509 ret = NET_RX_SUCCESS;
510 handle_res = RX_HANDLER_ANOTHER;
511out:
512 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
513 return handle_res;
514}
515
516static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
517{
518 const struct macvlan_dev *vlan = netdev_priv(dev);
519 const struct macvlan_port *port = vlan->port;
520 const struct macvlan_dev *dest;
521
522 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
523 const struct ethhdr *eth = skb_eth_hdr(skb);
524
525 /* send to other bridge ports directly */
526 if (is_multicast_ether_addr(eth->h_dest)) {
527 skb_reset_mac_header(skb);
528 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
529 goto xmit_world;
530 }
531
532 dest = macvlan_hash_lookup(port, eth->h_dest);
533 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
534 /* send to lowerdev first for its network taps */
535 dev_forward_skb(vlan->lowerdev, skb);
536
537 return NET_XMIT_SUCCESS;
538 }
539 }
540xmit_world:
541 skb->dev = vlan->lowerdev;
542 return dev_queue_xmit_accel(skb,
543 netdev_get_sb_channel(dev) ? dev : NULL);
544}
545
546static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
547{
548#ifdef CONFIG_NET_POLL_CONTROLLER
549 return netpoll_send_skb(vlan->netpoll, skb);
550#else
551 BUG();
552 return NETDEV_TX_OK;
553#endif
554}
555
556static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
557 struct net_device *dev)
558{
559 struct macvlan_dev *vlan = netdev_priv(dev);
560 unsigned int len = skb->len;
561 int ret;
562
563 if (unlikely(netpoll_tx_running(dev)))
564 return macvlan_netpoll_send_skb(vlan, skb);
565
566 ret = macvlan_queue_xmit(skb, dev);
567
568 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
569 struct vlan_pcpu_stats *pcpu_stats;
570
571 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
572 u64_stats_update_begin(&pcpu_stats->syncp);
573 pcpu_stats->tx_packets++;
574 pcpu_stats->tx_bytes += len;
575 u64_stats_update_end(&pcpu_stats->syncp);
576 } else {
577 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
578 }
579 return ret;
580}
581
582static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
583 unsigned short type, const void *daddr,
584 const void *saddr, unsigned len)
585{
586 const struct macvlan_dev *vlan = netdev_priv(dev);
587 struct net_device *lowerdev = vlan->lowerdev;
588
589 return dev_hard_header(skb, lowerdev, type, daddr,
590 saddr ? : dev->dev_addr, len);
591}
592
593static const struct header_ops macvlan_hard_header_ops = {
594 .create = macvlan_hard_header,
595 .parse = eth_header_parse,
596 .cache = eth_header_cache,
597 .cache_update = eth_header_cache_update,
598};
599
600static int macvlan_open(struct net_device *dev)
601{
602 struct macvlan_dev *vlan = netdev_priv(dev);
603 struct net_device *lowerdev = vlan->lowerdev;
604 int err;
605
606 if (macvlan_passthru(vlan->port)) {
607 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
608 err = dev_set_promiscuity(lowerdev, 1);
609 if (err < 0)
610 goto out;
611 }
612 goto hash_add;
613 }
614
615 err = -EADDRINUSE;
616 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
617 goto out;
618
619 /* Attempt to populate accel_priv which is used to offload the L2
620 * forwarding requests for unicast packets.
621 */
622 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
623 vlan->accel_priv =
624 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
625
626 /* If earlier attempt to offload failed, or accel_priv is not
627 * populated we must add the unicast address to the lower device.
628 */
629 if (IS_ERR_OR_NULL(vlan->accel_priv)) {
630 vlan->accel_priv = NULL;
631 err = dev_uc_add(lowerdev, dev->dev_addr);
632 if (err < 0)
633 goto out;
634 }
635
636 if (dev->flags & IFF_ALLMULTI) {
637 err = dev_set_allmulti(lowerdev, 1);
638 if (err < 0)
639 goto del_unicast;
640 }
641
642 if (dev->flags & IFF_PROMISC) {
643 err = dev_set_promiscuity(lowerdev, 1);
644 if (err < 0)
645 goto clear_multi;
646 }
647
648hash_add:
649 macvlan_hash_add(vlan);
650 return 0;
651
652clear_multi:
653 if (dev->flags & IFF_ALLMULTI)
654 dev_set_allmulti(lowerdev, -1);
655del_unicast:
656 if (vlan->accel_priv) {
657 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
658 vlan->accel_priv);
659 vlan->accel_priv = NULL;
660 } else {
661 dev_uc_del(lowerdev, dev->dev_addr);
662 }
663out:
664 return err;
665}
666
667static int macvlan_stop(struct net_device *dev)
668{
669 struct macvlan_dev *vlan = netdev_priv(dev);
670 struct net_device *lowerdev = vlan->lowerdev;
671
672 if (vlan->accel_priv) {
673 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
674 vlan->accel_priv);
675 vlan->accel_priv = NULL;
676 }
677
678 dev_uc_unsync(lowerdev, dev);
679 dev_mc_unsync(lowerdev, dev);
680
681 if (macvlan_passthru(vlan->port)) {
682 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
683 dev_set_promiscuity(lowerdev, -1);
684 goto hash_del;
685 }
686
687 if (dev->flags & IFF_ALLMULTI)
688 dev_set_allmulti(lowerdev, -1);
689
690 if (dev->flags & IFF_PROMISC)
691 dev_set_promiscuity(lowerdev, -1);
692
693 dev_uc_del(lowerdev, dev->dev_addr);
694
695hash_del:
696 macvlan_hash_del(vlan, !dev->dismantle);
697 return 0;
698}
699
700static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
701{
702 struct macvlan_dev *vlan = netdev_priv(dev);
703 struct net_device *lowerdev = vlan->lowerdev;
704 struct macvlan_port *port = vlan->port;
705 int err;
706
707 if (!(dev->flags & IFF_UP)) {
708 /* Just copy in the new address */
709 ether_addr_copy(dev->dev_addr, addr);
710 } else {
711 /* Rehash and update the device filters */
712 if (macvlan_addr_busy(vlan->port, addr))
713 return -EADDRINUSE;
714
715 if (!macvlan_passthru(port)) {
716 err = dev_uc_add(lowerdev, addr);
717 if (err)
718 return err;
719
720 dev_uc_del(lowerdev, dev->dev_addr);
721 }
722
723 macvlan_hash_change_addr(vlan, addr);
724 }
725 if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
726 /* Since addr_change isn't set, we are here due to lower
727 * device change. Save the lower-dev address so we can
728 * restore it later.
729 */
730 ether_addr_copy(vlan->port->perm_addr,
731 lowerdev->dev_addr);
732 }
733 macvlan_clear_addr_change(port);
734 return 0;
735}
736
737static int macvlan_set_mac_address(struct net_device *dev, void *p)
738{
739 struct macvlan_dev *vlan = netdev_priv(dev);
740 struct sockaddr *addr = p;
741
742 if (!is_valid_ether_addr(addr->sa_data))
743 return -EADDRNOTAVAIL;
744
745 /* If the addresses are the same, this is a no-op */
746 if (ether_addr_equal(dev->dev_addr, addr->sa_data))
747 return 0;
748
749 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
750 macvlan_set_addr_change(vlan->port);
751 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
752 }
753
754 if (macvlan_addr_busy(vlan->port, addr->sa_data))
755 return -EADDRINUSE;
756
757 return macvlan_sync_address(dev, addr->sa_data);
758}
759
760static void macvlan_change_rx_flags(struct net_device *dev, int change)
761{
762 struct macvlan_dev *vlan = netdev_priv(dev);
763 struct net_device *lowerdev = vlan->lowerdev;
764
765 if (dev->flags & IFF_UP) {
766 if (change & IFF_ALLMULTI)
767 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
768 if (change & IFF_PROMISC)
769 dev_set_promiscuity(lowerdev,
770 dev->flags & IFF_PROMISC ? 1 : -1);
771
772 }
773}
774
775static void macvlan_compute_filter(unsigned long *mc_filter,
776 struct net_device *dev,
777 struct macvlan_dev *vlan)
778{
779 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
780 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
781 } else {
782 struct netdev_hw_addr *ha;
783 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
784
785 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
786 netdev_for_each_mc_addr(ha, dev) {
787 __set_bit(mc_hash(vlan, ha->addr), filter);
788 }
789
790 __set_bit(mc_hash(vlan, dev->broadcast), filter);
791
792 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
793 }
794}
795
796static void macvlan_set_mac_lists(struct net_device *dev)
797{
798 struct macvlan_dev *vlan = netdev_priv(dev);
799
800 macvlan_compute_filter(vlan->mc_filter, dev, vlan);
801
802 dev_uc_sync(vlan->lowerdev, dev);
803 dev_mc_sync(vlan->lowerdev, dev);
804
805 /* This is slightly inaccurate as we're including the subscription
806 * list of vlan->lowerdev too.
807 *
808 * Bug alert: This only works if everyone has the same broadcast
809 * address as lowerdev. As soon as someone changes theirs this
810 * will break.
811 *
812 * However, this is already broken as when you change your broadcast
813 * address we don't get called.
814 *
815 * The solution is to maintain a list of broadcast addresses like
816 * we do for uc/mc, if you care.
817 */
818 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
819}
820
821static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
822{
823 struct macvlan_dev *vlan = netdev_priv(dev);
824
825 if (vlan->lowerdev->mtu < new_mtu)
826 return -EINVAL;
827 dev->mtu = new_mtu;
828 return 0;
829}
830
831static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
832{
833 struct net_device *real_dev = macvlan_dev_real_dev(dev);
834 const struct net_device_ops *ops = real_dev->netdev_ops;
835 struct ifreq ifrr;
836 int err = -EOPNOTSUPP;
837
838 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
839 ifrr.ifr_ifru = ifr->ifr_ifru;
840
841 switch (cmd) {
842 case SIOCSHWTSTAMP:
843 if (!net_eq(dev_net(dev), &init_net))
844 break;
845 fallthrough;
846 case SIOCGHWTSTAMP:
847 if (netif_device_present(real_dev) && ops->ndo_do_ioctl)
848 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd);
849 break;
850 }
851
852 if (!err)
853 ifr->ifr_ifru = ifrr.ifr_ifru;
854
855 return err;
856}
857
858/*
859 * macvlan network devices have devices nesting below it and are a special
860 * "super class" of normal network devices; split their locks off into a
861 * separate class since they always nest.
862 */
863static struct lock_class_key macvlan_netdev_addr_lock_key;
864
865#define ALWAYS_ON_OFFLOADS \
866 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
867 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
868
869#define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
870
871#define MACVLAN_FEATURES \
872 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
873 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
874 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
875 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
876
877#define MACVLAN_STATE_MASK \
878 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
879
880static void macvlan_set_lockdep_class(struct net_device *dev)
881{
882 netdev_lockdep_set_classes(dev);
883 lockdep_set_class(&dev->addr_list_lock,
884 &macvlan_netdev_addr_lock_key);
885}
886
887static int macvlan_init(struct net_device *dev)
888{
889 struct macvlan_dev *vlan = netdev_priv(dev);
890 const struct net_device *lowerdev = vlan->lowerdev;
891 struct macvlan_port *port = vlan->port;
892
893 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
894 (lowerdev->state & MACVLAN_STATE_MASK);
895 dev->features = lowerdev->features & MACVLAN_FEATURES;
896 dev->features |= ALWAYS_ON_FEATURES;
897 dev->hw_features |= NETIF_F_LRO;
898 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
899 dev->vlan_features |= ALWAYS_ON_OFFLOADS;
900 dev->hw_enc_features |= dev->features;
901 dev->gso_max_size = lowerdev->gso_max_size;
902 dev->gso_max_segs = lowerdev->gso_max_segs;
903 dev->hard_header_len = lowerdev->hard_header_len;
904 macvlan_set_lockdep_class(dev);
905
906 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
907 if (!vlan->pcpu_stats)
908 return -ENOMEM;
909
910 port->count += 1;
911
912 return 0;
913}
914
915static void macvlan_uninit(struct net_device *dev)
916{
917 struct macvlan_dev *vlan = netdev_priv(dev);
918 struct macvlan_port *port = vlan->port;
919
920 free_percpu(vlan->pcpu_stats);
921
922 macvlan_flush_sources(port, vlan);
923 port->count -= 1;
924 if (!port->count)
925 macvlan_port_destroy(port->dev);
926}
927
928static void macvlan_dev_get_stats64(struct net_device *dev,
929 struct rtnl_link_stats64 *stats)
930{
931 struct macvlan_dev *vlan = netdev_priv(dev);
932
933 if (vlan->pcpu_stats) {
934 struct vlan_pcpu_stats *p;
935 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
936 u32 rx_errors = 0, tx_dropped = 0;
937 unsigned int start;
938 int i;
939
940 for_each_possible_cpu(i) {
941 p = per_cpu_ptr(vlan->pcpu_stats, i);
942 do {
943 start = u64_stats_fetch_begin_irq(&p->syncp);
944 rx_packets = p->rx_packets;
945 rx_bytes = p->rx_bytes;
946 rx_multicast = p->rx_multicast;
947 tx_packets = p->tx_packets;
948 tx_bytes = p->tx_bytes;
949 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
950
951 stats->rx_packets += rx_packets;
952 stats->rx_bytes += rx_bytes;
953 stats->multicast += rx_multicast;
954 stats->tx_packets += tx_packets;
955 stats->tx_bytes += tx_bytes;
956 /* rx_errors & tx_dropped are u32, updated
957 * without syncp protection.
958 */
959 rx_errors += p->rx_errors;
960 tx_dropped += p->tx_dropped;
961 }
962 stats->rx_errors = rx_errors;
963 stats->rx_dropped = rx_errors;
964 stats->tx_dropped = tx_dropped;
965 }
966}
967
968static int macvlan_vlan_rx_add_vid(struct net_device *dev,
969 __be16 proto, u16 vid)
970{
971 struct macvlan_dev *vlan = netdev_priv(dev);
972 struct net_device *lowerdev = vlan->lowerdev;
973
974 return vlan_vid_add(lowerdev, proto, vid);
975}
976
977static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
978 __be16 proto, u16 vid)
979{
980 struct macvlan_dev *vlan = netdev_priv(dev);
981 struct net_device *lowerdev = vlan->lowerdev;
982
983 vlan_vid_del(lowerdev, proto, vid);
984 return 0;
985}
986
987static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
988 struct net_device *dev,
989 const unsigned char *addr, u16 vid,
990 u16 flags,
991 struct netlink_ext_ack *extack)
992{
993 struct macvlan_dev *vlan = netdev_priv(dev);
994 int err = -EINVAL;
995
996 /* Support unicast filter only on passthru devices.
997 * Multicast filter should be allowed on all devices.
998 */
999 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1000 return -EOPNOTSUPP;
1001
1002 if (flags & NLM_F_REPLACE)
1003 return -EOPNOTSUPP;
1004
1005 if (is_unicast_ether_addr(addr))
1006 err = dev_uc_add_excl(dev, addr);
1007 else if (is_multicast_ether_addr(addr))
1008 err = dev_mc_add_excl(dev, addr);
1009
1010 return err;
1011}
1012
1013static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1014 struct net_device *dev,
1015 const unsigned char *addr, u16 vid)
1016{
1017 struct macvlan_dev *vlan = netdev_priv(dev);
1018 int err = -EINVAL;
1019
1020 /* Support unicast filter only on passthru devices.
1021 * Multicast filter should be allowed on all devices.
1022 */
1023 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1024 return -EOPNOTSUPP;
1025
1026 if (is_unicast_ether_addr(addr))
1027 err = dev_uc_del(dev, addr);
1028 else if (is_multicast_ether_addr(addr))
1029 err = dev_mc_del(dev, addr);
1030
1031 return err;
1032}
1033
1034static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1035 struct ethtool_drvinfo *drvinfo)
1036{
1037 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1038 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1039}
1040
1041static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1042 struct ethtool_link_ksettings *cmd)
1043{
1044 const struct macvlan_dev *vlan = netdev_priv(dev);
1045
1046 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1047}
1048
1049static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1050 struct ethtool_ts_info *info)
1051{
1052 struct net_device *real_dev = macvlan_dev_real_dev(dev);
1053 const struct ethtool_ops *ops = real_dev->ethtool_ops;
1054 struct phy_device *phydev = real_dev->phydev;
1055
1056 if (phy_has_tsinfo(phydev)) {
1057 return phy_ts_info(phydev, info);
1058 } else if (ops->get_ts_info) {
1059 return ops->get_ts_info(real_dev, info);
1060 } else {
1061 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1062 SOF_TIMESTAMPING_SOFTWARE;
1063 info->phc_index = -1;
1064 }
1065
1066 return 0;
1067}
1068
1069static netdev_features_t macvlan_fix_features(struct net_device *dev,
1070 netdev_features_t features)
1071{
1072 struct macvlan_dev *vlan = netdev_priv(dev);
1073 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1074 netdev_features_t mask;
1075
1076 features |= NETIF_F_ALL_FOR_ALL;
1077 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1078 mask = features;
1079
1080 lowerdev_features &= (features | ~NETIF_F_LRO);
1081 features = netdev_increment_features(lowerdev_features, features, mask);
1082 features |= ALWAYS_ON_FEATURES;
1083 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1084
1085 return features;
1086}
1087
1088#ifdef CONFIG_NET_POLL_CONTROLLER
1089static void macvlan_dev_poll_controller(struct net_device *dev)
1090{
1091 return;
1092}
1093
1094static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1095{
1096 struct macvlan_dev *vlan = netdev_priv(dev);
1097 struct net_device *real_dev = vlan->lowerdev;
1098 struct netpoll *netpoll;
1099 int err = 0;
1100
1101 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1102 err = -ENOMEM;
1103 if (!netpoll)
1104 goto out;
1105
1106 err = __netpoll_setup(netpoll, real_dev);
1107 if (err) {
1108 kfree(netpoll);
1109 goto out;
1110 }
1111
1112 vlan->netpoll = netpoll;
1113
1114out:
1115 return err;
1116}
1117
1118static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1119{
1120 struct macvlan_dev *vlan = netdev_priv(dev);
1121 struct netpoll *netpoll = vlan->netpoll;
1122
1123 if (!netpoll)
1124 return;
1125
1126 vlan->netpoll = NULL;
1127
1128 __netpoll_free(netpoll);
1129}
1130#endif /* CONFIG_NET_POLL_CONTROLLER */
1131
1132static int macvlan_dev_get_iflink(const struct net_device *dev)
1133{
1134 struct macvlan_dev *vlan = netdev_priv(dev);
1135
1136 return vlan->lowerdev->ifindex;
1137}
1138
1139static const struct ethtool_ops macvlan_ethtool_ops = {
1140 .get_link = ethtool_op_get_link,
1141 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1142 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1143 .get_ts_info = macvlan_ethtool_get_ts_info,
1144};
1145
1146static const struct net_device_ops macvlan_netdev_ops = {
1147 .ndo_init = macvlan_init,
1148 .ndo_uninit = macvlan_uninit,
1149 .ndo_open = macvlan_open,
1150 .ndo_stop = macvlan_stop,
1151 .ndo_start_xmit = macvlan_start_xmit,
1152 .ndo_change_mtu = macvlan_change_mtu,
1153 .ndo_do_ioctl = macvlan_do_ioctl,
1154 .ndo_fix_features = macvlan_fix_features,
1155 .ndo_change_rx_flags = macvlan_change_rx_flags,
1156 .ndo_set_mac_address = macvlan_set_mac_address,
1157 .ndo_set_rx_mode = macvlan_set_mac_lists,
1158 .ndo_get_stats64 = macvlan_dev_get_stats64,
1159 .ndo_validate_addr = eth_validate_addr,
1160 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1161 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1162 .ndo_fdb_add = macvlan_fdb_add,
1163 .ndo_fdb_del = macvlan_fdb_del,
1164 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1165#ifdef CONFIG_NET_POLL_CONTROLLER
1166 .ndo_poll_controller = macvlan_dev_poll_controller,
1167 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1168 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1169#endif
1170 .ndo_get_iflink = macvlan_dev_get_iflink,
1171 .ndo_features_check = passthru_features_check,
1172 .ndo_change_proto_down = dev_change_proto_down_generic,
1173};
1174
1175void macvlan_common_setup(struct net_device *dev)
1176{
1177 ether_setup(dev);
1178
1179 dev->min_mtu = 0;
1180 dev->max_mtu = ETH_MAX_MTU;
1181 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1182 netif_keep_dst(dev);
1183 dev->priv_flags |= IFF_UNICAST_FLT;
1184 dev->netdev_ops = &macvlan_netdev_ops;
1185 dev->needs_free_netdev = true;
1186 dev->header_ops = &macvlan_hard_header_ops;
1187 dev->ethtool_ops = &macvlan_ethtool_ops;
1188}
1189EXPORT_SYMBOL_GPL(macvlan_common_setup);
1190
1191static void macvlan_setup(struct net_device *dev)
1192{
1193 macvlan_common_setup(dev);
1194 dev->priv_flags |= IFF_NO_QUEUE;
1195}
1196
1197static int macvlan_port_create(struct net_device *dev)
1198{
1199 struct macvlan_port *port;
1200 unsigned int i;
1201 int err;
1202
1203 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1204 return -EINVAL;
1205
1206 if (netdev_is_rx_handler_busy(dev))
1207 return -EBUSY;
1208
1209 port = kzalloc(sizeof(*port), GFP_KERNEL);
1210 if (port == NULL)
1211 return -ENOMEM;
1212
1213 port->dev = dev;
1214 ether_addr_copy(port->perm_addr, dev->dev_addr);
1215 INIT_LIST_HEAD(&port->vlans);
1216 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1217 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1218 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1219 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1220
1221 skb_queue_head_init(&port->bc_queue);
1222 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1223
1224 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1225 if (err)
1226 kfree(port);
1227 else
1228 dev->priv_flags |= IFF_MACVLAN_PORT;
1229 return err;
1230}
1231
1232static void macvlan_port_destroy(struct net_device *dev)
1233{
1234 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1235 struct sk_buff *skb;
1236
1237 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1238 netdev_rx_handler_unregister(dev);
1239
1240 /* After this point, no packet can schedule bc_work anymore,
1241 * but we need to cancel it and purge left skbs if any.
1242 */
1243 cancel_work_sync(&port->bc_work);
1244
1245 while ((skb = __skb_dequeue(&port->bc_queue))) {
1246 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1247
1248 if (src)
1249 dev_put(src->dev);
1250
1251 kfree_skb(skb);
1252 }
1253
1254 /* If the lower device address has been changed by passthru
1255 * macvlan, put it back.
1256 */
1257 if (macvlan_passthru(port) &&
1258 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1259 struct sockaddr sa;
1260
1261 sa.sa_family = port->dev->type;
1262 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1263 dev_set_mac_address(port->dev, &sa, NULL);
1264 }
1265
1266 kfree(port);
1267}
1268
1269static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1270 struct netlink_ext_ack *extack)
1271{
1272 struct nlattr *nla, *head;
1273 int rem, len;
1274
1275 if (tb[IFLA_ADDRESS]) {
1276 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1277 return -EINVAL;
1278 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1279 return -EADDRNOTAVAIL;
1280 }
1281
1282 if (!data)
1283 return 0;
1284
1285 if (data[IFLA_MACVLAN_FLAGS] &&
1286 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1287 return -EINVAL;
1288
1289 if (data[IFLA_MACVLAN_MODE]) {
1290 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1291 case MACVLAN_MODE_PRIVATE:
1292 case MACVLAN_MODE_VEPA:
1293 case MACVLAN_MODE_BRIDGE:
1294 case MACVLAN_MODE_PASSTHRU:
1295 case MACVLAN_MODE_SOURCE:
1296 break;
1297 default:
1298 return -EINVAL;
1299 }
1300 }
1301
1302 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1303 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1304 case MACVLAN_MACADDR_ADD:
1305 case MACVLAN_MACADDR_DEL:
1306 case MACVLAN_MACADDR_FLUSH:
1307 case MACVLAN_MACADDR_SET:
1308 break;
1309 default:
1310 return -EINVAL;
1311 }
1312 }
1313
1314 if (data[IFLA_MACVLAN_MACADDR]) {
1315 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1316 return -EINVAL;
1317
1318 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1319 return -EADDRNOTAVAIL;
1320 }
1321
1322 if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1323 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1324 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1325
1326 nla_for_each_attr(nla, head, len, rem) {
1327 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1328 nla_len(nla) != ETH_ALEN)
1329 return -EINVAL;
1330
1331 if (!is_valid_ether_addr(nla_data(nla)))
1332 return -EADDRNOTAVAIL;
1333 }
1334 }
1335
1336 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1337 return -EINVAL;
1338
1339 return 0;
1340}
1341
1342/**
1343 * reconfigure list of remote source mac address
1344 * (only for macvlan devices in source mode)
1345 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1346 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1347 */
1348static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1349 struct nlattr *data[])
1350{
1351 char *addr = NULL;
1352 int ret, rem, len;
1353 struct nlattr *nla, *head;
1354 struct macvlan_source_entry *entry;
1355
1356 if (data[IFLA_MACVLAN_MACADDR])
1357 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1358
1359 if (mode == MACVLAN_MACADDR_ADD) {
1360 if (!addr)
1361 return -EINVAL;
1362
1363 return macvlan_hash_add_source(vlan, addr);
1364
1365 } else if (mode == MACVLAN_MACADDR_DEL) {
1366 if (!addr)
1367 return -EINVAL;
1368
1369 entry = macvlan_hash_lookup_source(vlan, addr);
1370 if (entry) {
1371 macvlan_hash_del_source(entry);
1372 vlan->macaddr_count--;
1373 }
1374 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1375 macvlan_flush_sources(vlan->port, vlan);
1376 } else if (mode == MACVLAN_MACADDR_SET) {
1377 macvlan_flush_sources(vlan->port, vlan);
1378
1379 if (addr) {
1380 ret = macvlan_hash_add_source(vlan, addr);
1381 if (ret)
1382 return ret;
1383 }
1384
1385 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1386 return 0;
1387
1388 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1389 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1390
1391 nla_for_each_attr(nla, head, len, rem) {
1392 addr = nla_data(nla);
1393 ret = macvlan_hash_add_source(vlan, addr);
1394 if (ret)
1395 return ret;
1396 }
1397 } else {
1398 return -EINVAL;
1399 }
1400
1401 return 0;
1402}
1403
1404int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1405 struct nlattr *tb[], struct nlattr *data[],
1406 struct netlink_ext_ack *extack)
1407{
1408 struct macvlan_dev *vlan = netdev_priv(dev);
1409 struct macvlan_port *port;
1410 struct net_device *lowerdev;
1411 int err;
1412 int macmode;
1413 bool create = false;
1414
1415 if (!tb[IFLA_LINK])
1416 return -EINVAL;
1417
1418 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1419 if (lowerdev == NULL)
1420 return -ENODEV;
1421
1422 /* When creating macvlans or macvtaps on top of other macvlans - use
1423 * the real device as the lowerdev.
1424 */
1425 if (netif_is_macvlan(lowerdev))
1426 lowerdev = macvlan_dev_real_dev(lowerdev);
1427
1428 if (!tb[IFLA_MTU])
1429 dev->mtu = lowerdev->mtu;
1430 else if (dev->mtu > lowerdev->mtu)
1431 return -EINVAL;
1432
1433 /* MTU range: 68 - lowerdev->max_mtu */
1434 dev->min_mtu = ETH_MIN_MTU;
1435 dev->max_mtu = lowerdev->max_mtu;
1436
1437 if (!tb[IFLA_ADDRESS])
1438 eth_hw_addr_random(dev);
1439
1440 if (!netif_is_macvlan_port(lowerdev)) {
1441 err = macvlan_port_create(lowerdev);
1442 if (err < 0)
1443 return err;
1444 create = true;
1445 }
1446 port = macvlan_port_get_rtnl(lowerdev);
1447
1448 /* Only 1 macvlan device can be created in passthru mode */
1449 if (macvlan_passthru(port)) {
1450 /* The macvlan port must be not created this time,
1451 * still goto destroy_macvlan_port for readability.
1452 */
1453 err = -EINVAL;
1454 goto destroy_macvlan_port;
1455 }
1456
1457 vlan->lowerdev = lowerdev;
1458 vlan->dev = dev;
1459 vlan->port = port;
1460 vlan->set_features = MACVLAN_FEATURES;
1461
1462 vlan->mode = MACVLAN_MODE_VEPA;
1463 if (data && data[IFLA_MACVLAN_MODE])
1464 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1465
1466 if (data && data[IFLA_MACVLAN_FLAGS])
1467 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1468
1469 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1470 if (port->count) {
1471 err = -EINVAL;
1472 goto destroy_macvlan_port;
1473 }
1474 macvlan_set_passthru(port);
1475 eth_hw_addr_inherit(dev, lowerdev);
1476 }
1477
1478 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1479 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1480 err = -EINVAL;
1481 goto destroy_macvlan_port;
1482 }
1483 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1484 err = macvlan_changelink_sources(vlan, macmode, data);
1485 if (err)
1486 goto destroy_macvlan_port;
1487 }
1488
1489 err = register_netdevice(dev);
1490 if (err < 0)
1491 goto destroy_macvlan_port;
1492
1493 dev->priv_flags |= IFF_MACVLAN;
1494 err = netdev_upper_dev_link(lowerdev, dev, extack);
1495 if (err)
1496 goto unregister_netdev;
1497
1498 list_add_tail_rcu(&vlan->list, &port->vlans);
1499 netif_stacked_transfer_operstate(lowerdev, dev);
1500 linkwatch_fire_event(dev);
1501
1502 return 0;
1503
1504unregister_netdev:
1505 /* macvlan_uninit would free the macvlan port */
1506 unregister_netdevice(dev);
1507 return err;
1508destroy_macvlan_port:
1509 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1510 * so we destroy the macvlan port only when it's valid.
1511 */
1512 if (create && macvlan_port_get_rtnl(lowerdev))
1513 macvlan_port_destroy(port->dev);
1514 return err;
1515}
1516EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1517
1518static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1519 struct nlattr *tb[], struct nlattr *data[],
1520 struct netlink_ext_ack *extack)
1521{
1522 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1523}
1524
1525void macvlan_dellink(struct net_device *dev, struct list_head *head)
1526{
1527 struct macvlan_dev *vlan = netdev_priv(dev);
1528
1529 if (vlan->mode == MACVLAN_MODE_SOURCE)
1530 macvlan_flush_sources(vlan->port, vlan);
1531 list_del_rcu(&vlan->list);
1532 unregister_netdevice_queue(dev, head);
1533 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1534}
1535EXPORT_SYMBOL_GPL(macvlan_dellink);
1536
1537static int macvlan_changelink(struct net_device *dev,
1538 struct nlattr *tb[], struct nlattr *data[],
1539 struct netlink_ext_ack *extack)
1540{
1541 struct macvlan_dev *vlan = netdev_priv(dev);
1542 enum macvlan_mode mode;
1543 bool set_mode = false;
1544 enum macvlan_macaddr_mode macmode;
1545 int ret;
1546
1547 /* Validate mode, but don't set yet: setting flags may fail. */
1548 if (data && data[IFLA_MACVLAN_MODE]) {
1549 set_mode = true;
1550 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1551 /* Passthrough mode can't be set or cleared dynamically */
1552 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1553 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1554 return -EINVAL;
1555 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1556 vlan->mode != mode)
1557 macvlan_flush_sources(vlan->port, vlan);
1558 }
1559
1560 if (data && data[IFLA_MACVLAN_FLAGS]) {
1561 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1562 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1563 if (macvlan_passthru(vlan->port) && promisc) {
1564 int err;
1565
1566 if (flags & MACVLAN_FLAG_NOPROMISC)
1567 err = dev_set_promiscuity(vlan->lowerdev, -1);
1568 else
1569 err = dev_set_promiscuity(vlan->lowerdev, 1);
1570 if (err < 0)
1571 return err;
1572 }
1573 vlan->flags = flags;
1574 }
1575 if (set_mode)
1576 vlan->mode = mode;
1577 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1578 if (vlan->mode != MACVLAN_MODE_SOURCE)
1579 return -EINVAL;
1580 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1581 ret = macvlan_changelink_sources(vlan, macmode, data);
1582 if (ret)
1583 return ret;
1584 }
1585 return 0;
1586}
1587
1588static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1589{
1590 if (vlan->macaddr_count == 0)
1591 return 0;
1592 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1593 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1594}
1595
1596static size_t macvlan_get_size(const struct net_device *dev)
1597{
1598 struct macvlan_dev *vlan = netdev_priv(dev);
1599
1600 return (0
1601 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1602 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1603 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1604 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1605 );
1606}
1607
1608static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1609 const struct macvlan_dev *vlan,
1610 const int i)
1611{
1612 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1613 struct macvlan_source_entry *entry;
1614
1615 hlist_for_each_entry_rcu(entry, h, hlist) {
1616 if (entry->vlan != vlan)
1617 continue;
1618 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1619 return 1;
1620 }
1621 return 0;
1622}
1623
1624static int macvlan_fill_info(struct sk_buff *skb,
1625 const struct net_device *dev)
1626{
1627 struct macvlan_dev *vlan = netdev_priv(dev);
1628 int i;
1629 struct nlattr *nest;
1630
1631 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1632 goto nla_put_failure;
1633 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1634 goto nla_put_failure;
1635 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1636 goto nla_put_failure;
1637 if (vlan->macaddr_count > 0) {
1638 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1639 if (nest == NULL)
1640 goto nla_put_failure;
1641
1642 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1643 if (macvlan_fill_info_macaddr(skb, vlan, i))
1644 goto nla_put_failure;
1645 }
1646 nla_nest_end(skb, nest);
1647 }
1648 return 0;
1649
1650nla_put_failure:
1651 return -EMSGSIZE;
1652}
1653
1654static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1655 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1656 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1657 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1658 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1659 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1660 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1661};
1662
1663int macvlan_link_register(struct rtnl_link_ops *ops)
1664{
1665 /* common fields */
1666 ops->validate = macvlan_validate;
1667 ops->maxtype = IFLA_MACVLAN_MAX;
1668 ops->policy = macvlan_policy;
1669 ops->changelink = macvlan_changelink;
1670 ops->get_size = macvlan_get_size;
1671 ops->fill_info = macvlan_fill_info;
1672
1673 return rtnl_link_register(ops);
1674};
1675EXPORT_SYMBOL_GPL(macvlan_link_register);
1676
1677static struct net *macvlan_get_link_net(const struct net_device *dev)
1678{
1679 return dev_net(macvlan_dev_real_dev(dev));
1680}
1681
1682static struct rtnl_link_ops macvlan_link_ops = {
1683 .kind = "macvlan",
1684 .setup = macvlan_setup,
1685 .newlink = macvlan_newlink,
1686 .dellink = macvlan_dellink,
1687 .get_link_net = macvlan_get_link_net,
1688 .priv_size = sizeof(struct macvlan_dev),
1689};
1690
1691static int macvlan_device_event(struct notifier_block *unused,
1692 unsigned long event, void *ptr)
1693{
1694 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1695 struct macvlan_dev *vlan, *next;
1696 struct macvlan_port *port;
1697 LIST_HEAD(list_kill);
1698
1699 if (!netif_is_macvlan_port(dev))
1700 return NOTIFY_DONE;
1701
1702 port = macvlan_port_get_rtnl(dev);
1703
1704 switch (event) {
1705 case NETDEV_UP:
1706 case NETDEV_DOWN:
1707 case NETDEV_CHANGE:
1708 list_for_each_entry(vlan, &port->vlans, list)
1709 netif_stacked_transfer_operstate(vlan->lowerdev,
1710 vlan->dev);
1711 break;
1712 case NETDEV_FEAT_CHANGE:
1713 list_for_each_entry(vlan, &port->vlans, list) {
1714 vlan->dev->gso_max_size = dev->gso_max_size;
1715 vlan->dev->gso_max_segs = dev->gso_max_segs;
1716 netdev_update_features(vlan->dev);
1717 }
1718 break;
1719 case NETDEV_CHANGEMTU:
1720 list_for_each_entry(vlan, &port->vlans, list) {
1721 if (vlan->dev->mtu <= dev->mtu)
1722 continue;
1723 dev_set_mtu(vlan->dev, dev->mtu);
1724 }
1725 break;
1726 case NETDEV_CHANGEADDR:
1727 if (!macvlan_passthru(port))
1728 return NOTIFY_DONE;
1729
1730 vlan = list_first_entry_or_null(&port->vlans,
1731 struct macvlan_dev,
1732 list);
1733
1734 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1735 return NOTIFY_BAD;
1736
1737 break;
1738 case NETDEV_UNREGISTER:
1739 /* twiddle thumbs on netns device moves */
1740 if (dev->reg_state != NETREG_UNREGISTERING)
1741 break;
1742
1743 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1744 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1745 unregister_netdevice_many(&list_kill);
1746 break;
1747 case NETDEV_PRE_TYPE_CHANGE:
1748 /* Forbid underlaying device to change its type. */
1749 return NOTIFY_BAD;
1750
1751 case NETDEV_NOTIFY_PEERS:
1752 case NETDEV_BONDING_FAILOVER:
1753 case NETDEV_RESEND_IGMP:
1754 /* Propagate to all vlans */
1755 list_for_each_entry(vlan, &port->vlans, list)
1756 call_netdevice_notifiers(event, vlan->dev);
1757 }
1758 return NOTIFY_DONE;
1759}
1760
1761static struct notifier_block macvlan_notifier_block __read_mostly = {
1762 .notifier_call = macvlan_device_event,
1763};
1764
1765static int __init macvlan_init_module(void)
1766{
1767 int err;
1768
1769 register_netdevice_notifier(&macvlan_notifier_block);
1770
1771 err = macvlan_link_register(&macvlan_link_ops);
1772 if (err < 0)
1773 goto err1;
1774 return 0;
1775err1:
1776 unregister_netdevice_notifier(&macvlan_notifier_block);
1777 return err;
1778}
1779
1780static void __exit macvlan_cleanup_module(void)
1781{
1782 rtnl_link_unregister(&macvlan_link_ops);
1783 unregister_netdevice_notifier(&macvlan_notifier_block);
1784}
1785
1786module_init(macvlan_init_module);
1787module_exit(macvlan_cleanup_module);
1788
1789MODULE_LICENSE("GPL");
1790MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1791MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1792MODULE_ALIAS_RTNL_LINK("macvlan");