Loading...
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_DEFAULT_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 bc_queue_len_used;
50 int bc_cutoff;
51 u32 flags;
52 int count;
53 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE];
54 DECLARE_BITMAP(bc_filter, MACVLAN_MC_FILTER_SZ);
55 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
56 unsigned char perm_addr[ETH_ALEN];
57};
58
59struct macvlan_source_entry {
60 struct hlist_node hlist;
61 struct macvlan_dev *vlan;
62 unsigned char addr[6+2] __aligned(sizeof(u16));
63 struct rcu_head rcu;
64};
65
66struct macvlan_skb_cb {
67 const struct macvlan_dev *src;
68};
69
70#define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
71
72static void macvlan_port_destroy(struct net_device *dev);
73static void update_port_bc_queue_len(struct macvlan_port *port);
74
75static inline bool macvlan_passthru(const struct macvlan_port *port)
76{
77 return port->flags & MACVLAN_F_PASSTHRU;
78}
79
80static inline void macvlan_set_passthru(struct macvlan_port *port)
81{
82 port->flags |= MACVLAN_F_PASSTHRU;
83}
84
85static inline bool macvlan_addr_change(const struct macvlan_port *port)
86{
87 return port->flags & MACVLAN_F_ADDRCHANGE;
88}
89
90static inline void macvlan_set_addr_change(struct macvlan_port *port)
91{
92 port->flags |= MACVLAN_F_ADDRCHANGE;
93}
94
95static inline void macvlan_clear_addr_change(struct macvlan_port *port)
96{
97 port->flags &= ~MACVLAN_F_ADDRCHANGE;
98}
99
100/* Hash Ethernet address */
101static u32 macvlan_eth_hash(const unsigned char *addr)
102{
103 u64 value = get_unaligned((u64 *)addr);
104
105 /* only want 6 bytes */
106#ifdef __BIG_ENDIAN
107 value >>= 16;
108#else
109 value <<= 16;
110#endif
111 return hash_64(value, MACVLAN_HASH_BITS);
112}
113
114static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
115{
116 return rcu_dereference(dev->rx_handler_data);
117}
118
119static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
120{
121 return rtnl_dereference(dev->rx_handler_data);
122}
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 lockdep_rtnl_is_held()) {
132 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
133 return vlan;
134 }
135 return NULL;
136}
137
138static struct macvlan_source_entry *macvlan_hash_lookup_source(
139 const struct macvlan_dev *vlan,
140 const unsigned char *addr)
141{
142 struct macvlan_source_entry *entry;
143 u32 idx = macvlan_eth_hash(addr);
144 struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
145
146 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
147 if (ether_addr_equal_64bits(entry->addr, addr) &&
148 entry->vlan == vlan)
149 return entry;
150 }
151 return NULL;
152}
153
154static int macvlan_hash_add_source(struct macvlan_dev *vlan,
155 const unsigned char *addr)
156{
157 struct macvlan_port *port = vlan->port;
158 struct macvlan_source_entry *entry;
159 struct hlist_head *h;
160
161 entry = macvlan_hash_lookup_source(vlan, addr);
162 if (entry)
163 return 0;
164
165 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
166 if (!entry)
167 return -ENOMEM;
168
169 ether_addr_copy(entry->addr, addr);
170 entry->vlan = vlan;
171 h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
172 hlist_add_head_rcu(&entry->hlist, h);
173 vlan->macaddr_count++;
174
175 return 0;
176}
177
178static void macvlan_hash_add(struct macvlan_dev *vlan)
179{
180 struct macvlan_port *port = vlan->port;
181 const unsigned char *addr = vlan->dev->dev_addr;
182 u32 idx = macvlan_eth_hash(addr);
183
184 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
185}
186
187static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
188{
189 hlist_del_rcu(&entry->hlist);
190 kfree_rcu(entry, rcu);
191}
192
193static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
194{
195 hlist_del_rcu(&vlan->hlist);
196 if (sync)
197 synchronize_rcu();
198}
199
200static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
201 const unsigned char *addr)
202{
203 macvlan_hash_del(vlan, true);
204 /* Now that we are unhashed it is safe to change the device
205 * address without confusing packet delivery.
206 */
207 eth_hw_addr_set(vlan->dev, addr);
208 macvlan_hash_add(vlan);
209}
210
211static bool macvlan_addr_busy(const struct macvlan_port *port,
212 const unsigned char *addr)
213{
214 /* Test to see if the specified address is
215 * currently in use by the underlying device or
216 * another macvlan.
217 */
218 if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
219 ether_addr_equal_64bits(port->dev->dev_addr, addr))
220 return true;
221
222 if (macvlan_hash_lookup(port, addr))
223 return true;
224
225 return false;
226}
227
228
229static int macvlan_broadcast_one(struct sk_buff *skb,
230 const struct macvlan_dev *vlan,
231 const struct ethhdr *eth, bool local)
232{
233 struct net_device *dev = vlan->dev;
234
235 if (local)
236 return __dev_forward_skb(dev, skb);
237
238 skb->dev = dev;
239 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
240 skb->pkt_type = PACKET_BROADCAST;
241 else
242 skb->pkt_type = PACKET_MULTICAST;
243
244 return 0;
245}
246
247static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
248{
249 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
250}
251
252
253static unsigned int mc_hash(const struct macvlan_dev *vlan,
254 const unsigned char *addr)
255{
256 u32 val = __get_unaligned_cpu32(addr + 2);
257
258 val ^= macvlan_hash_mix(vlan);
259 return hash_32(val, MACVLAN_MC_FILTER_BITS);
260}
261
262static void macvlan_broadcast(struct sk_buff *skb,
263 const struct macvlan_port *port,
264 struct net_device *src,
265 enum macvlan_mode mode)
266{
267 const struct ethhdr *eth = eth_hdr(skb);
268 const struct macvlan_dev *vlan;
269 struct sk_buff *nskb;
270 unsigned int i;
271 int err;
272 unsigned int hash;
273
274 if (skb->protocol == htons(ETH_P_PAUSE))
275 return;
276
277 hash_for_each_rcu(port->vlan_hash, i, vlan, 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(nskb, vlan, eth,
289 mode == MACVLAN_MODE_BRIDGE) ?:
290 netif_rx(nskb);
291 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
292 err == NET_RX_SUCCESS, true);
293 }
294}
295
296static void macvlan_multicast_rx(const struct macvlan_port *port,
297 const struct macvlan_dev *src,
298 struct sk_buff *skb)
299{
300 if (!src)
301 /* frame comes from an external address */
302 macvlan_broadcast(skb, port, NULL,
303 MACVLAN_MODE_PRIVATE |
304 MACVLAN_MODE_VEPA |
305 MACVLAN_MODE_PASSTHRU|
306 MACVLAN_MODE_BRIDGE);
307 else if (src->mode == MACVLAN_MODE_VEPA)
308 /* flood to everyone except source */
309 macvlan_broadcast(skb, port, src->dev,
310 MACVLAN_MODE_VEPA |
311 MACVLAN_MODE_BRIDGE);
312 else
313 /*
314 * flood only to VEPA ports, bridge ports
315 * already saw the frame on the way out.
316 */
317 macvlan_broadcast(skb, port, src->dev,
318 MACVLAN_MODE_VEPA);
319}
320
321static void macvlan_process_broadcast(struct work_struct *w)
322{
323 struct macvlan_port *port = container_of(w, struct macvlan_port,
324 bc_work);
325 struct sk_buff *skb;
326 struct sk_buff_head list;
327
328 __skb_queue_head_init(&list);
329
330 spin_lock_bh(&port->bc_queue.lock);
331 skb_queue_splice_tail_init(&port->bc_queue, &list);
332 spin_unlock_bh(&port->bc_queue.lock);
333
334 while ((skb = __skb_dequeue(&list))) {
335 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
336
337 rcu_read_lock();
338 macvlan_multicast_rx(port, src, skb);
339 rcu_read_unlock();
340
341 if (src)
342 dev_put(src->dev);
343 consume_skb(skb);
344
345 cond_resched();
346 }
347}
348
349static void macvlan_broadcast_enqueue(struct macvlan_port *port,
350 const struct macvlan_dev *src,
351 struct sk_buff *skb)
352{
353 struct sk_buff *nskb;
354 int err = -ENOMEM;
355
356 nskb = skb_clone(skb, GFP_ATOMIC);
357 if (!nskb)
358 goto err;
359
360 MACVLAN_SKB_CB(nskb)->src = src;
361
362 spin_lock(&port->bc_queue.lock);
363 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) {
364 if (src)
365 dev_hold(src->dev);
366 __skb_queue_tail(&port->bc_queue, nskb);
367 err = 0;
368 }
369 spin_unlock(&port->bc_queue.lock);
370
371 queue_work(system_unbound_wq, &port->bc_work);
372
373 if (err)
374 goto free_nskb;
375
376 return;
377
378free_nskb:
379 kfree_skb(nskb);
380err:
381 dev_core_stats_rx_dropped_inc(skb->dev);
382}
383
384static void macvlan_flush_sources(struct macvlan_port *port,
385 struct macvlan_dev *vlan)
386{
387 struct macvlan_source_entry *entry;
388 struct hlist_node *next;
389 int i;
390
391 hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
392 if (entry->vlan == vlan)
393 macvlan_hash_del_source(entry);
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 bool 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 bool consume = false;
432
433 hlist_for_each_entry_rcu(entry, h, hlist) {
434 if (ether_addr_equal_64bits(entry->addr, addr)) {
435 if (entry->vlan->flags & MACVLAN_FLAG_NODST)
436 consume = true;
437 macvlan_forward_source_one(skb, entry->vlan);
438 }
439 }
440
441 return consume;
442}
443
444/* called under rcu_read_lock() from netif_receive_skb */
445static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
446{
447 struct macvlan_port *port;
448 struct sk_buff *skb = *pskb;
449 const struct ethhdr *eth = eth_hdr(skb);
450 const struct macvlan_dev *vlan;
451 const struct macvlan_dev *src;
452 struct net_device *dev;
453 unsigned int len = 0;
454 int ret;
455 rx_handler_result_t handle_res;
456
457 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */
458 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
459 return RX_HANDLER_PASS;
460
461 port = macvlan_port_get_rcu(skb->dev);
462 if (is_multicast_ether_addr(eth->h_dest)) {
463 unsigned int hash;
464
465 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
466 if (!skb)
467 return RX_HANDLER_CONSUMED;
468 *pskb = skb;
469 eth = eth_hdr(skb);
470 if (macvlan_forward_source(skb, port, eth->h_source)) {
471 kfree_skb(skb);
472 return RX_HANDLER_CONSUMED;
473 }
474 src = macvlan_hash_lookup(port, eth->h_source);
475 if (src && src->mode != MACVLAN_MODE_VEPA &&
476 src->mode != MACVLAN_MODE_BRIDGE) {
477 /* forward to original port. */
478 vlan = src;
479 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
480 __netif_rx(skb);
481 handle_res = RX_HANDLER_CONSUMED;
482 goto out;
483 }
484
485 hash = mc_hash(NULL, eth->h_dest);
486 if (test_bit(hash, port->bc_filter))
487 macvlan_broadcast_enqueue(port, src, skb);
488 else if (test_bit(hash, port->mc_filter))
489 macvlan_multicast_rx(port, src, skb);
490
491 return RX_HANDLER_PASS;
492 }
493
494 if (macvlan_forward_source(skb, port, eth->h_source)) {
495 kfree_skb(skb);
496 return RX_HANDLER_CONSUMED;
497 }
498 if (macvlan_passthru(port))
499 vlan = list_first_or_null_rcu(&port->vlans,
500 struct macvlan_dev, list);
501 else
502 vlan = macvlan_hash_lookup(port, eth->h_dest);
503 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
504 return RX_HANDLER_PASS;
505
506 dev = vlan->dev;
507 if (unlikely(!(dev->flags & IFF_UP))) {
508 kfree_skb(skb);
509 return RX_HANDLER_CONSUMED;
510 }
511 len = skb->len + ETH_HLEN;
512 skb = skb_share_check(skb, GFP_ATOMIC);
513 if (!skb) {
514 ret = NET_RX_DROP;
515 handle_res = RX_HANDLER_CONSUMED;
516 goto out;
517 }
518
519 *pskb = skb;
520 skb->dev = dev;
521 skb->pkt_type = PACKET_HOST;
522
523 ret = NET_RX_SUCCESS;
524 handle_res = RX_HANDLER_ANOTHER;
525out:
526 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
527 return handle_res;
528}
529
530static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
531{
532 const struct macvlan_dev *vlan = netdev_priv(dev);
533 const struct macvlan_port *port = vlan->port;
534 const struct macvlan_dev *dest;
535
536 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
537 const struct ethhdr *eth = skb_eth_hdr(skb);
538
539 /* send to other bridge ports directly */
540 if (is_multicast_ether_addr(eth->h_dest)) {
541 skb_reset_mac_header(skb);
542 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
543 goto xmit_world;
544 }
545
546 dest = macvlan_hash_lookup(port, eth->h_dest);
547 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
548 /* send to lowerdev first for its network taps */
549 dev_forward_skb(vlan->lowerdev, skb);
550
551 return NET_XMIT_SUCCESS;
552 }
553 }
554xmit_world:
555 skb->dev = vlan->lowerdev;
556 return dev_queue_xmit_accel(skb,
557 netdev_get_sb_channel(dev) ? dev : NULL);
558}
559
560static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
561{
562#ifdef CONFIG_NET_POLL_CONTROLLER
563 return netpoll_send_skb(vlan->netpoll, skb);
564#else
565 BUG();
566 return NETDEV_TX_OK;
567#endif
568}
569
570static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
571 struct net_device *dev)
572{
573 struct macvlan_dev *vlan = netdev_priv(dev);
574 unsigned int len = skb->len;
575 int ret;
576
577 if (unlikely(netpoll_tx_running(dev)))
578 return macvlan_netpoll_send_skb(vlan, skb);
579
580 ret = macvlan_queue_xmit(skb, dev);
581
582 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
583 struct vlan_pcpu_stats *pcpu_stats;
584
585 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
586 u64_stats_update_begin(&pcpu_stats->syncp);
587 u64_stats_inc(&pcpu_stats->tx_packets);
588 u64_stats_add(&pcpu_stats->tx_bytes, len);
589 u64_stats_update_end(&pcpu_stats->syncp);
590 } else {
591 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
592 }
593 return ret;
594}
595
596static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
597 unsigned short type, const void *daddr,
598 const void *saddr, unsigned len)
599{
600 const struct macvlan_dev *vlan = netdev_priv(dev);
601 struct net_device *lowerdev = vlan->lowerdev;
602
603 return dev_hard_header(skb, lowerdev, type, daddr,
604 saddr ? : dev->dev_addr, len);
605}
606
607static const struct header_ops macvlan_hard_header_ops = {
608 .create = macvlan_hard_header,
609 .parse = eth_header_parse,
610 .cache = eth_header_cache,
611 .cache_update = eth_header_cache_update,
612 .parse_protocol = eth_header_parse_protocol,
613};
614
615static int macvlan_open(struct net_device *dev)
616{
617 struct macvlan_dev *vlan = netdev_priv(dev);
618 struct net_device *lowerdev = vlan->lowerdev;
619 int err;
620
621 if (macvlan_passthru(vlan->port)) {
622 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
623 err = dev_set_promiscuity(lowerdev, 1);
624 if (err < 0)
625 goto out;
626 }
627 goto hash_add;
628 }
629
630 err = -EADDRINUSE;
631 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
632 goto out;
633
634 /* Attempt to populate accel_priv which is used to offload the L2
635 * forwarding requests for unicast packets.
636 */
637 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
638 vlan->accel_priv =
639 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
640
641 /* If earlier attempt to offload failed, or accel_priv is not
642 * populated we must add the unicast address to the lower device.
643 */
644 if (IS_ERR_OR_NULL(vlan->accel_priv)) {
645 vlan->accel_priv = NULL;
646 err = dev_uc_add(lowerdev, dev->dev_addr);
647 if (err < 0)
648 goto out;
649 }
650
651 if (dev->flags & IFF_ALLMULTI) {
652 err = dev_set_allmulti(lowerdev, 1);
653 if (err < 0)
654 goto del_unicast;
655 }
656
657 if (dev->flags & IFF_PROMISC) {
658 err = dev_set_promiscuity(lowerdev, 1);
659 if (err < 0)
660 goto clear_multi;
661 }
662
663hash_add:
664 macvlan_hash_add(vlan);
665 return 0;
666
667clear_multi:
668 if (dev->flags & IFF_ALLMULTI)
669 dev_set_allmulti(lowerdev, -1);
670del_unicast:
671 if (vlan->accel_priv) {
672 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
673 vlan->accel_priv);
674 vlan->accel_priv = NULL;
675 } else {
676 dev_uc_del(lowerdev, dev->dev_addr);
677 }
678out:
679 return err;
680}
681
682static int macvlan_stop(struct net_device *dev)
683{
684 struct macvlan_dev *vlan = netdev_priv(dev);
685 struct net_device *lowerdev = vlan->lowerdev;
686
687 if (vlan->accel_priv) {
688 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
689 vlan->accel_priv);
690 vlan->accel_priv = NULL;
691 }
692
693 dev_uc_unsync(lowerdev, dev);
694 dev_mc_unsync(lowerdev, dev);
695
696 if (macvlan_passthru(vlan->port)) {
697 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
698 dev_set_promiscuity(lowerdev, -1);
699 goto hash_del;
700 }
701
702 if (dev->flags & IFF_ALLMULTI)
703 dev_set_allmulti(lowerdev, -1);
704
705 if (dev->flags & IFF_PROMISC)
706 dev_set_promiscuity(lowerdev, -1);
707
708 dev_uc_del(lowerdev, dev->dev_addr);
709
710hash_del:
711 macvlan_hash_del(vlan, !dev->dismantle);
712 return 0;
713}
714
715static int macvlan_sync_address(struct net_device *dev,
716 const unsigned char *addr)
717{
718 struct macvlan_dev *vlan = netdev_priv(dev);
719 struct net_device *lowerdev = vlan->lowerdev;
720 struct macvlan_port *port = vlan->port;
721 int err;
722
723 if (!(dev->flags & IFF_UP)) {
724 /* Just copy in the new address */
725 eth_hw_addr_set(dev, addr);
726 } else {
727 /* Rehash and update the device filters */
728 if (macvlan_addr_busy(vlan->port, addr))
729 return -EADDRINUSE;
730
731 if (!macvlan_passthru(port)) {
732 err = dev_uc_add(lowerdev, addr);
733 if (err)
734 return err;
735
736 dev_uc_del(lowerdev, dev->dev_addr);
737 }
738
739 macvlan_hash_change_addr(vlan, addr);
740 }
741 if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
742 /* Since addr_change isn't set, we are here due to lower
743 * device change. Save the lower-dev address so we can
744 * restore it later.
745 */
746 ether_addr_copy(vlan->port->perm_addr,
747 lowerdev->dev_addr);
748 }
749 macvlan_clear_addr_change(port);
750 return 0;
751}
752
753static int macvlan_set_mac_address(struct net_device *dev, void *p)
754{
755 struct macvlan_dev *vlan = netdev_priv(dev);
756 struct sockaddr *addr = p;
757
758 if (!is_valid_ether_addr(addr->sa_data))
759 return -EADDRNOTAVAIL;
760
761 /* If the addresses are the same, this is a no-op */
762 if (ether_addr_equal(dev->dev_addr, addr->sa_data))
763 return 0;
764
765 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
766 macvlan_set_addr_change(vlan->port);
767 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
768 }
769
770 if (macvlan_addr_busy(vlan->port, addr->sa_data))
771 return -EADDRINUSE;
772
773 return macvlan_sync_address(dev, addr->sa_data);
774}
775
776static void macvlan_change_rx_flags(struct net_device *dev, int change)
777{
778 struct macvlan_dev *vlan = netdev_priv(dev);
779 struct net_device *lowerdev = vlan->lowerdev;
780
781 if (dev->flags & IFF_UP) {
782 if (change & IFF_ALLMULTI)
783 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
784 if (!macvlan_passthru(vlan->port) && change & IFF_PROMISC)
785 dev_set_promiscuity(lowerdev,
786 dev->flags & IFF_PROMISC ? 1 : -1);
787
788 }
789}
790
791static void macvlan_compute_filter(unsigned long *mc_filter,
792 struct net_device *dev,
793 struct macvlan_dev *vlan, int cutoff)
794{
795 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
796 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
797 } else {
798 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
799 struct netdev_hw_addr *ha;
800
801 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
802 netdev_for_each_mc_addr(ha, dev) {
803 if (!vlan && ha->synced <= cutoff)
804 continue;
805
806 __set_bit(mc_hash(vlan, ha->addr), filter);
807 }
808
809 __set_bit(mc_hash(vlan, dev->broadcast), filter);
810
811 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
812 }
813}
814
815static void macvlan_recompute_bc_filter(struct macvlan_dev *vlan)
816{
817 if (vlan->port->bc_cutoff < 0) {
818 bitmap_zero(vlan->port->bc_filter, MACVLAN_MC_FILTER_SZ);
819 return;
820 }
821
822 macvlan_compute_filter(vlan->port->bc_filter, vlan->lowerdev, NULL,
823 vlan->port->bc_cutoff);
824}
825
826static void macvlan_set_mac_lists(struct net_device *dev)
827{
828 struct macvlan_dev *vlan = netdev_priv(dev);
829
830 macvlan_compute_filter(vlan->mc_filter, dev, vlan, 0);
831
832 dev_uc_sync(vlan->lowerdev, dev);
833 dev_mc_sync(vlan->lowerdev, dev);
834
835 /* This is slightly inaccurate as we're including the subscription
836 * list of vlan->lowerdev too.
837 *
838 * Bug alert: This only works if everyone has the same broadcast
839 * address as lowerdev. As soon as someone changes theirs this
840 * will break.
841 *
842 * However, this is already broken as when you change your broadcast
843 * address we don't get called.
844 *
845 * The solution is to maintain a list of broadcast addresses like
846 * we do for uc/mc, if you care.
847 */
848 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL,
849 0);
850 macvlan_recompute_bc_filter(vlan);
851}
852
853static void update_port_bc_cutoff(struct macvlan_dev *vlan, int cutoff)
854{
855 if (vlan->port->bc_cutoff == cutoff)
856 return;
857
858 vlan->port->bc_cutoff = cutoff;
859 macvlan_recompute_bc_filter(vlan);
860}
861
862static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
863{
864 struct macvlan_dev *vlan = netdev_priv(dev);
865
866 if (vlan->lowerdev->mtu < new_mtu)
867 return -EINVAL;
868 WRITE_ONCE(dev->mtu, new_mtu);
869 return 0;
870}
871
872static int macvlan_hwtstamp_get(struct net_device *dev,
873 struct kernel_hwtstamp_config *cfg)
874{
875 struct net_device *real_dev = macvlan_dev_real_dev(dev);
876
877 return generic_hwtstamp_get_lower(real_dev, cfg);
878}
879
880static int macvlan_hwtstamp_set(struct net_device *dev,
881 struct kernel_hwtstamp_config *cfg,
882 struct netlink_ext_ack *extack)
883{
884 struct net_device *real_dev = macvlan_dev_real_dev(dev);
885
886 if (!net_eq(dev_net(dev), &init_net))
887 return -EOPNOTSUPP;
888
889 return generic_hwtstamp_set_lower(real_dev, cfg, extack);
890}
891
892/*
893 * macvlan network devices have devices nesting below it and are a special
894 * "super class" of normal network devices; split their locks off into a
895 * separate class since they always nest.
896 */
897static struct lock_class_key macvlan_netdev_addr_lock_key;
898
899#define ALWAYS_ON_OFFLOADS \
900 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
901 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
902
903#define ALWAYS_ON_FEATURES ALWAYS_ON_OFFLOADS
904
905#define MACVLAN_FEATURES \
906 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
907 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
908 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
909 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
910
911#define MACVLAN_STATE_MASK \
912 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
913
914static void macvlan_set_lockdep_class(struct net_device *dev)
915{
916 netdev_lockdep_set_classes(dev);
917 lockdep_set_class(&dev->addr_list_lock,
918 &macvlan_netdev_addr_lock_key);
919}
920
921static int macvlan_init(struct net_device *dev)
922{
923 struct macvlan_dev *vlan = netdev_priv(dev);
924 struct net_device *lowerdev = vlan->lowerdev;
925 struct macvlan_port *port = vlan->port;
926
927 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
928 (lowerdev->state & MACVLAN_STATE_MASK);
929 dev->features = lowerdev->features & MACVLAN_FEATURES;
930 dev->features |= ALWAYS_ON_FEATURES;
931 dev->hw_features |= NETIF_F_LRO;
932 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
933 dev->vlan_features |= ALWAYS_ON_OFFLOADS;
934 dev->hw_enc_features |= dev->features;
935 dev->lltx = true;
936 netif_inherit_tso_max(dev, lowerdev);
937 dev->hard_header_len = lowerdev->hard_header_len;
938 macvlan_set_lockdep_class(dev);
939
940 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
941 if (!vlan->pcpu_stats)
942 return -ENOMEM;
943
944 port->count += 1;
945
946 /* Get macvlan's reference to lowerdev */
947 netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
948
949 return 0;
950}
951
952static void macvlan_uninit(struct net_device *dev)
953{
954 struct macvlan_dev *vlan = netdev_priv(dev);
955 struct macvlan_port *port = vlan->port;
956
957 free_percpu(vlan->pcpu_stats);
958
959 macvlan_flush_sources(port, vlan);
960 port->count -= 1;
961 if (!port->count)
962 macvlan_port_destroy(port->dev);
963}
964
965static void macvlan_dev_get_stats64(struct net_device *dev,
966 struct rtnl_link_stats64 *stats)
967{
968 struct macvlan_dev *vlan = netdev_priv(dev);
969
970 if (vlan->pcpu_stats) {
971 struct vlan_pcpu_stats *p;
972 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
973 u32 rx_errors = 0, tx_dropped = 0;
974 unsigned int start;
975 int i;
976
977 for_each_possible_cpu(i) {
978 p = per_cpu_ptr(vlan->pcpu_stats, i);
979 do {
980 start = u64_stats_fetch_begin(&p->syncp);
981 rx_packets = u64_stats_read(&p->rx_packets);
982 rx_bytes = u64_stats_read(&p->rx_bytes);
983 rx_multicast = u64_stats_read(&p->rx_multicast);
984 tx_packets = u64_stats_read(&p->tx_packets);
985 tx_bytes = u64_stats_read(&p->tx_bytes);
986 } while (u64_stats_fetch_retry(&p->syncp, start));
987
988 stats->rx_packets += rx_packets;
989 stats->rx_bytes += rx_bytes;
990 stats->multicast += rx_multicast;
991 stats->tx_packets += tx_packets;
992 stats->tx_bytes += tx_bytes;
993 /* rx_errors & tx_dropped are u32, updated
994 * without syncp protection.
995 */
996 rx_errors += READ_ONCE(p->rx_errors);
997 tx_dropped += READ_ONCE(p->tx_dropped);
998 }
999 stats->rx_errors = rx_errors;
1000 stats->rx_dropped = rx_errors;
1001 stats->tx_dropped = tx_dropped;
1002 }
1003}
1004
1005static int macvlan_vlan_rx_add_vid(struct net_device *dev,
1006 __be16 proto, u16 vid)
1007{
1008 struct macvlan_dev *vlan = netdev_priv(dev);
1009 struct net_device *lowerdev = vlan->lowerdev;
1010
1011 return vlan_vid_add(lowerdev, proto, vid);
1012}
1013
1014static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
1015 __be16 proto, u16 vid)
1016{
1017 struct macvlan_dev *vlan = netdev_priv(dev);
1018 struct net_device *lowerdev = vlan->lowerdev;
1019
1020 vlan_vid_del(lowerdev, proto, vid);
1021 return 0;
1022}
1023
1024static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1025 struct net_device *dev,
1026 const unsigned char *addr, u16 vid,
1027 u16 flags, bool *notified,
1028 struct netlink_ext_ack *extack)
1029{
1030 struct macvlan_dev *vlan = netdev_priv(dev);
1031 int err = -EINVAL;
1032
1033 /* Support unicast filter only on passthru devices.
1034 * Multicast filter should be allowed on all devices.
1035 */
1036 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1037 return -EOPNOTSUPP;
1038
1039 if (flags & NLM_F_REPLACE)
1040 return -EOPNOTSUPP;
1041
1042 if (is_unicast_ether_addr(addr))
1043 err = dev_uc_add_excl(dev, addr);
1044 else if (is_multicast_ether_addr(addr))
1045 err = dev_mc_add_excl(dev, addr);
1046
1047 return err;
1048}
1049
1050static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1051 struct net_device *dev,
1052 const unsigned char *addr, u16 vid, bool *notified,
1053 struct netlink_ext_ack *extack)
1054{
1055 struct macvlan_dev *vlan = netdev_priv(dev);
1056 int err = -EINVAL;
1057
1058 /* Support unicast filter only on passthru devices.
1059 * Multicast filter should be allowed on all devices.
1060 */
1061 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1062 return -EOPNOTSUPP;
1063
1064 if (is_unicast_ether_addr(addr))
1065 err = dev_uc_del(dev, addr);
1066 else if (is_multicast_ether_addr(addr))
1067 err = dev_mc_del(dev, addr);
1068
1069 return err;
1070}
1071
1072static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1073 struct ethtool_drvinfo *drvinfo)
1074{
1075 strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1076 strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1077}
1078
1079static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1080 struct ethtool_link_ksettings *cmd)
1081{
1082 const struct macvlan_dev *vlan = netdev_priv(dev);
1083
1084 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1085}
1086
1087static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1088 struct kernel_ethtool_ts_info *info)
1089{
1090 struct net_device *real_dev = macvlan_dev_real_dev(dev);
1091
1092 return ethtool_get_ts_info_by_layer(real_dev, info);
1093}
1094
1095static netdev_features_t macvlan_fix_features(struct net_device *dev,
1096 netdev_features_t features)
1097{
1098 struct macvlan_dev *vlan = netdev_priv(dev);
1099 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1100 netdev_features_t mask;
1101
1102 features |= NETIF_F_ALL_FOR_ALL;
1103 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1104 mask = features;
1105
1106 lowerdev_features &= (features | ~NETIF_F_LRO);
1107 features = netdev_increment_features(lowerdev_features, features, mask);
1108 features |= ALWAYS_ON_FEATURES;
1109 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1110
1111 return features;
1112}
1113
1114#ifdef CONFIG_NET_POLL_CONTROLLER
1115static void macvlan_dev_poll_controller(struct net_device *dev)
1116{
1117 return;
1118}
1119
1120static int macvlan_dev_netpoll_setup(struct net_device *dev)
1121{
1122 struct macvlan_dev *vlan = netdev_priv(dev);
1123 struct net_device *real_dev = vlan->lowerdev;
1124 struct netpoll *netpoll;
1125 int err;
1126
1127 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1128 err = -ENOMEM;
1129 if (!netpoll)
1130 goto out;
1131
1132 err = __netpoll_setup(netpoll, real_dev);
1133 if (err) {
1134 kfree(netpoll);
1135 goto out;
1136 }
1137
1138 vlan->netpoll = netpoll;
1139
1140out:
1141 return err;
1142}
1143
1144static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1145{
1146 struct macvlan_dev *vlan = netdev_priv(dev);
1147 struct netpoll *netpoll = vlan->netpoll;
1148
1149 if (!netpoll)
1150 return;
1151
1152 vlan->netpoll = NULL;
1153
1154 __netpoll_free(netpoll);
1155}
1156#endif /* CONFIG_NET_POLL_CONTROLLER */
1157
1158static int macvlan_dev_get_iflink(const struct net_device *dev)
1159{
1160 struct macvlan_dev *vlan = netdev_priv(dev);
1161
1162 return READ_ONCE(vlan->lowerdev->ifindex);
1163}
1164
1165static const struct ethtool_ops macvlan_ethtool_ops = {
1166 .get_link = ethtool_op_get_link,
1167 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1168 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1169 .get_ts_info = macvlan_ethtool_get_ts_info,
1170};
1171
1172static const struct net_device_ops macvlan_netdev_ops = {
1173 .ndo_init = macvlan_init,
1174 .ndo_uninit = macvlan_uninit,
1175 .ndo_open = macvlan_open,
1176 .ndo_stop = macvlan_stop,
1177 .ndo_start_xmit = macvlan_start_xmit,
1178 .ndo_change_mtu = macvlan_change_mtu,
1179 .ndo_fix_features = macvlan_fix_features,
1180 .ndo_change_rx_flags = macvlan_change_rx_flags,
1181 .ndo_set_mac_address = macvlan_set_mac_address,
1182 .ndo_set_rx_mode = macvlan_set_mac_lists,
1183 .ndo_get_stats64 = macvlan_dev_get_stats64,
1184 .ndo_validate_addr = eth_validate_addr,
1185 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1186 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1187 .ndo_fdb_add = macvlan_fdb_add,
1188 .ndo_fdb_del = macvlan_fdb_del,
1189 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1190#ifdef CONFIG_NET_POLL_CONTROLLER
1191 .ndo_poll_controller = macvlan_dev_poll_controller,
1192 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1193 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1194#endif
1195 .ndo_get_iflink = macvlan_dev_get_iflink,
1196 .ndo_features_check = passthru_features_check,
1197 .ndo_hwtstamp_get = macvlan_hwtstamp_get,
1198 .ndo_hwtstamp_set = macvlan_hwtstamp_set,
1199};
1200
1201static void macvlan_dev_free(struct net_device *dev)
1202{
1203 struct macvlan_dev *vlan = netdev_priv(dev);
1204
1205 /* Get rid of the macvlan's reference to lowerdev */
1206 netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1207}
1208
1209void macvlan_common_setup(struct net_device *dev)
1210{
1211 ether_setup(dev);
1212
1213 /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1214 dev->max_mtu = ETH_MAX_MTU;
1215 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1216 netif_keep_dst(dev);
1217 dev->priv_flags |= IFF_UNICAST_FLT;
1218 dev->change_proto_down = true;
1219 dev->netdev_ops = &macvlan_netdev_ops;
1220 dev->needs_free_netdev = true;
1221 dev->priv_destructor = macvlan_dev_free;
1222 dev->header_ops = &macvlan_hard_header_ops;
1223 dev->ethtool_ops = &macvlan_ethtool_ops;
1224}
1225EXPORT_SYMBOL_GPL(macvlan_common_setup);
1226
1227static void macvlan_setup(struct net_device *dev)
1228{
1229 macvlan_common_setup(dev);
1230 dev->priv_flags |= IFF_NO_QUEUE;
1231}
1232
1233static int macvlan_port_create(struct net_device *dev)
1234{
1235 struct macvlan_port *port;
1236 unsigned int i;
1237 int err;
1238
1239 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1240 return -EINVAL;
1241
1242 if (netdev_is_rx_handler_busy(dev))
1243 return -EBUSY;
1244
1245 port = kzalloc(sizeof(*port), GFP_KERNEL);
1246 if (port == NULL)
1247 return -ENOMEM;
1248
1249 port->dev = dev;
1250 ether_addr_copy(port->perm_addr, dev->dev_addr);
1251 INIT_LIST_HEAD(&port->vlans);
1252 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1253 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1254 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1255 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1256
1257 port->bc_queue_len_used = 0;
1258 port->bc_cutoff = 1;
1259 skb_queue_head_init(&port->bc_queue);
1260 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1261
1262 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1263 if (err)
1264 kfree(port);
1265 else
1266 dev->priv_flags |= IFF_MACVLAN_PORT;
1267 return err;
1268}
1269
1270static void macvlan_port_destroy(struct net_device *dev)
1271{
1272 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1273 struct sk_buff *skb;
1274
1275 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1276 netdev_rx_handler_unregister(dev);
1277
1278 /* After this point, no packet can schedule bc_work anymore,
1279 * but we need to cancel it and purge left skbs if any.
1280 */
1281 cancel_work_sync(&port->bc_work);
1282
1283 while ((skb = __skb_dequeue(&port->bc_queue))) {
1284 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1285
1286 if (src)
1287 dev_put(src->dev);
1288
1289 kfree_skb(skb);
1290 }
1291
1292 /* If the lower device address has been changed by passthru
1293 * macvlan, put it back.
1294 */
1295 if (macvlan_passthru(port) &&
1296 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1297 struct sockaddr sa;
1298
1299 sa.sa_family = port->dev->type;
1300 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1301 dev_set_mac_address(port->dev, &sa, NULL);
1302 }
1303
1304 kfree(port);
1305}
1306
1307static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1308 struct netlink_ext_ack *extack)
1309{
1310 struct nlattr *nla, *head;
1311 int rem, len;
1312
1313 if (tb[IFLA_ADDRESS]) {
1314 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1315 return -EINVAL;
1316 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1317 return -EADDRNOTAVAIL;
1318 }
1319
1320 if (!data)
1321 return 0;
1322
1323 if (data[IFLA_MACVLAN_FLAGS] &&
1324 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1325 MACVLAN_FLAG_NODST))
1326 return -EINVAL;
1327
1328 if (data[IFLA_MACVLAN_MODE]) {
1329 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1330 case MACVLAN_MODE_PRIVATE:
1331 case MACVLAN_MODE_VEPA:
1332 case MACVLAN_MODE_BRIDGE:
1333 case MACVLAN_MODE_PASSTHRU:
1334 case MACVLAN_MODE_SOURCE:
1335 break;
1336 default:
1337 return -EINVAL;
1338 }
1339 }
1340
1341 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1342 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1343 case MACVLAN_MACADDR_ADD:
1344 case MACVLAN_MACADDR_DEL:
1345 case MACVLAN_MACADDR_FLUSH:
1346 case MACVLAN_MACADDR_SET:
1347 break;
1348 default:
1349 return -EINVAL;
1350 }
1351 }
1352
1353 if (data[IFLA_MACVLAN_MACADDR]) {
1354 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1355 return -EINVAL;
1356
1357 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1358 return -EADDRNOTAVAIL;
1359 }
1360
1361 if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1362 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1363 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1364
1365 nla_for_each_attr(nla, head, len, rem) {
1366 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1367 nla_len(nla) != ETH_ALEN)
1368 return -EINVAL;
1369
1370 if (!is_valid_ether_addr(nla_data(nla)))
1371 return -EADDRNOTAVAIL;
1372 }
1373 }
1374
1375 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1376 return -EINVAL;
1377
1378 return 0;
1379}
1380
1381/*
1382 * reconfigure list of remote source mac address
1383 * (only for macvlan devices in source mode)
1384 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1385 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1386 */
1387static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1388 struct nlattr *data[])
1389{
1390 char *addr = NULL;
1391 int ret, rem, len;
1392 struct nlattr *nla, *head;
1393 struct macvlan_source_entry *entry;
1394
1395 if (data[IFLA_MACVLAN_MACADDR])
1396 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1397
1398 if (mode == MACVLAN_MACADDR_ADD) {
1399 if (!addr)
1400 return -EINVAL;
1401
1402 return macvlan_hash_add_source(vlan, addr);
1403
1404 } else if (mode == MACVLAN_MACADDR_DEL) {
1405 if (!addr)
1406 return -EINVAL;
1407
1408 entry = macvlan_hash_lookup_source(vlan, addr);
1409 if (entry) {
1410 macvlan_hash_del_source(entry);
1411 vlan->macaddr_count--;
1412 }
1413 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1414 macvlan_flush_sources(vlan->port, vlan);
1415 } else if (mode == MACVLAN_MACADDR_SET) {
1416 macvlan_flush_sources(vlan->port, vlan);
1417
1418 if (addr) {
1419 ret = macvlan_hash_add_source(vlan, addr);
1420 if (ret)
1421 return ret;
1422 }
1423
1424 if (!data[IFLA_MACVLAN_MACADDR_DATA])
1425 return 0;
1426
1427 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1428 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1429
1430 nla_for_each_attr(nla, head, len, rem) {
1431 addr = nla_data(nla);
1432 ret = macvlan_hash_add_source(vlan, addr);
1433 if (ret)
1434 return ret;
1435 }
1436 } else {
1437 return -EINVAL;
1438 }
1439
1440 return 0;
1441}
1442
1443int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1444 struct nlattr *tb[], struct nlattr *data[],
1445 struct netlink_ext_ack *extack)
1446{
1447 struct macvlan_dev *vlan = netdev_priv(dev);
1448 struct macvlan_port *port;
1449 struct net_device *lowerdev;
1450 int err;
1451 int macmode;
1452 bool create = false;
1453
1454 if (!tb[IFLA_LINK])
1455 return -EINVAL;
1456
1457 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1458 if (lowerdev == NULL)
1459 return -ENODEV;
1460
1461 /* When creating macvlans or macvtaps on top of other macvlans - use
1462 * the real device as the lowerdev.
1463 */
1464 if (netif_is_macvlan(lowerdev))
1465 lowerdev = macvlan_dev_real_dev(lowerdev);
1466
1467 if (!tb[IFLA_MTU])
1468 dev->mtu = lowerdev->mtu;
1469 else if (dev->mtu > lowerdev->mtu)
1470 return -EINVAL;
1471
1472 /* MTU range: 68 - lowerdev->max_mtu */
1473 dev->min_mtu = ETH_MIN_MTU;
1474 dev->max_mtu = lowerdev->max_mtu;
1475
1476 if (!tb[IFLA_ADDRESS])
1477 eth_hw_addr_random(dev);
1478
1479 if (!netif_is_macvlan_port(lowerdev)) {
1480 err = macvlan_port_create(lowerdev);
1481 if (err < 0)
1482 return err;
1483 create = true;
1484 }
1485 port = macvlan_port_get_rtnl(lowerdev);
1486
1487 /* Only 1 macvlan device can be created in passthru mode */
1488 if (macvlan_passthru(port)) {
1489 /* The macvlan port must be not created this time,
1490 * still goto destroy_macvlan_port for readability.
1491 */
1492 err = -EINVAL;
1493 goto destroy_macvlan_port;
1494 }
1495
1496 vlan->lowerdev = lowerdev;
1497 vlan->dev = dev;
1498 vlan->port = port;
1499 vlan->set_features = MACVLAN_FEATURES;
1500
1501 vlan->mode = MACVLAN_MODE_VEPA;
1502 if (data && data[IFLA_MACVLAN_MODE])
1503 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1504
1505 if (data && data[IFLA_MACVLAN_FLAGS])
1506 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1507
1508 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1509 if (port->count) {
1510 err = -EINVAL;
1511 goto destroy_macvlan_port;
1512 }
1513 macvlan_set_passthru(port);
1514 eth_hw_addr_inherit(dev, lowerdev);
1515 }
1516
1517 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1518 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1519 err = -EINVAL;
1520 goto destroy_macvlan_port;
1521 }
1522 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1523 err = macvlan_changelink_sources(vlan, macmode, data);
1524 if (err)
1525 goto destroy_macvlan_port;
1526 }
1527
1528 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1529 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1530 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1531
1532 if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1533 update_port_bc_cutoff(
1534 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1535
1536 err = register_netdevice(dev);
1537 if (err < 0)
1538 goto destroy_macvlan_port;
1539
1540 dev->priv_flags |= IFF_MACVLAN;
1541 err = netdev_upper_dev_link(lowerdev, dev, extack);
1542 if (err)
1543 goto unregister_netdev;
1544
1545 list_add_tail_rcu(&vlan->list, &port->vlans);
1546 update_port_bc_queue_len(vlan->port);
1547 netif_stacked_transfer_operstate(lowerdev, dev);
1548 linkwatch_fire_event(dev);
1549
1550 return 0;
1551
1552unregister_netdev:
1553 /* macvlan_uninit would free the macvlan port */
1554 unregister_netdevice(dev);
1555 return err;
1556destroy_macvlan_port:
1557 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1558 * so we destroy the macvlan port only when it's valid.
1559 */
1560 if (create && macvlan_port_get_rtnl(lowerdev)) {
1561 macvlan_flush_sources(port, vlan);
1562 macvlan_port_destroy(port->dev);
1563 }
1564 return err;
1565}
1566EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1567
1568static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1569 struct nlattr *tb[], struct nlattr *data[],
1570 struct netlink_ext_ack *extack)
1571{
1572 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1573}
1574
1575void macvlan_dellink(struct net_device *dev, struct list_head *head)
1576{
1577 struct macvlan_dev *vlan = netdev_priv(dev);
1578
1579 if (vlan->mode == MACVLAN_MODE_SOURCE)
1580 macvlan_flush_sources(vlan->port, vlan);
1581 list_del_rcu(&vlan->list);
1582 update_port_bc_queue_len(vlan->port);
1583 unregister_netdevice_queue(dev, head);
1584 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1585}
1586EXPORT_SYMBOL_GPL(macvlan_dellink);
1587
1588static int macvlan_changelink(struct net_device *dev,
1589 struct nlattr *tb[], struct nlattr *data[],
1590 struct netlink_ext_ack *extack)
1591{
1592 struct macvlan_dev *vlan = netdev_priv(dev);
1593 enum macvlan_mode mode;
1594 bool set_mode = false;
1595 enum macvlan_macaddr_mode macmode;
1596 int ret;
1597
1598 /* Validate mode, but don't set yet: setting flags may fail. */
1599 if (data && data[IFLA_MACVLAN_MODE]) {
1600 set_mode = true;
1601 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1602 /* Passthrough mode can't be set or cleared dynamically */
1603 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1604 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1605 return -EINVAL;
1606 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1607 vlan->mode != mode)
1608 macvlan_flush_sources(vlan->port, vlan);
1609 }
1610
1611 if (data && data[IFLA_MACVLAN_FLAGS]) {
1612 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1613 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1614 if (macvlan_passthru(vlan->port) && promisc) {
1615 int err;
1616
1617 if (flags & MACVLAN_FLAG_NOPROMISC)
1618 err = dev_set_promiscuity(vlan->lowerdev, -1);
1619 else
1620 err = dev_set_promiscuity(vlan->lowerdev, 1);
1621 if (err < 0)
1622 return err;
1623 }
1624 vlan->flags = flags;
1625 }
1626
1627 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1628 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1629 update_port_bc_queue_len(vlan->port);
1630 }
1631
1632 if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1633 update_port_bc_cutoff(
1634 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1635
1636 if (set_mode)
1637 vlan->mode = mode;
1638 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1639 if (vlan->mode != MACVLAN_MODE_SOURCE)
1640 return -EINVAL;
1641 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1642 ret = macvlan_changelink_sources(vlan, macmode, data);
1643 if (ret)
1644 return ret;
1645 }
1646 return 0;
1647}
1648
1649static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1650{
1651 if (vlan->macaddr_count == 0)
1652 return 0;
1653 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1654 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1655}
1656
1657static size_t macvlan_get_size(const struct net_device *dev)
1658{
1659 struct macvlan_dev *vlan = netdev_priv(dev);
1660
1661 return (0
1662 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1663 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1664 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1665 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1666 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1667 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1668 );
1669}
1670
1671static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1672 const struct macvlan_dev *vlan,
1673 const int i)
1674{
1675 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1676 struct macvlan_source_entry *entry;
1677
1678 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1679 if (entry->vlan != vlan)
1680 continue;
1681 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1682 return 1;
1683 }
1684 return 0;
1685}
1686
1687static int macvlan_fill_info(struct sk_buff *skb,
1688 const struct net_device *dev)
1689{
1690 struct macvlan_dev *vlan = netdev_priv(dev);
1691 struct macvlan_port *port = vlan->port;
1692 int i;
1693 struct nlattr *nest;
1694
1695 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1696 goto nla_put_failure;
1697 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1698 goto nla_put_failure;
1699 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1700 goto nla_put_failure;
1701 if (vlan->macaddr_count > 0) {
1702 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1703 if (nest == NULL)
1704 goto nla_put_failure;
1705
1706 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1707 if (macvlan_fill_info_macaddr(skb, vlan, i))
1708 goto nla_put_failure;
1709 }
1710 nla_nest_end(skb, nest);
1711 }
1712 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1713 goto nla_put_failure;
1714 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1715 goto nla_put_failure;
1716 if (port->bc_cutoff != 1 &&
1717 nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff))
1718 goto nla_put_failure;
1719 return 0;
1720
1721nla_put_failure:
1722 return -EMSGSIZE;
1723}
1724
1725static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1726 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1727 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1728 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1729 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1730 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1731 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1732 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1733 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1734 [IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 },
1735};
1736
1737int macvlan_link_register(struct rtnl_link_ops *ops)
1738{
1739 /* common fields */
1740 ops->validate = macvlan_validate;
1741 ops->maxtype = IFLA_MACVLAN_MAX;
1742 ops->policy = macvlan_policy;
1743 ops->changelink = macvlan_changelink;
1744 ops->get_size = macvlan_get_size;
1745 ops->fill_info = macvlan_fill_info;
1746
1747 return rtnl_link_register(ops);
1748};
1749EXPORT_SYMBOL_GPL(macvlan_link_register);
1750
1751static struct net *macvlan_get_link_net(const struct net_device *dev)
1752{
1753 return dev_net(macvlan_dev_real_dev(dev));
1754}
1755
1756static struct rtnl_link_ops macvlan_link_ops = {
1757 .kind = "macvlan",
1758 .setup = macvlan_setup,
1759 .newlink = macvlan_newlink,
1760 .dellink = macvlan_dellink,
1761 .get_link_net = macvlan_get_link_net,
1762 .priv_size = sizeof(struct macvlan_dev),
1763};
1764
1765static void update_port_bc_queue_len(struct macvlan_port *port)
1766{
1767 u32 max_bc_queue_len_req = 0;
1768 struct macvlan_dev *vlan;
1769
1770 list_for_each_entry(vlan, &port->vlans, list) {
1771 if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1772 max_bc_queue_len_req = vlan->bc_queue_len_req;
1773 }
1774 port->bc_queue_len_used = max_bc_queue_len_req;
1775}
1776
1777static int macvlan_device_event(struct notifier_block *unused,
1778 unsigned long event, void *ptr)
1779{
1780 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1781 struct macvlan_dev *vlan, *next;
1782 struct macvlan_port *port;
1783 LIST_HEAD(list_kill);
1784
1785 if (!netif_is_macvlan_port(dev))
1786 return NOTIFY_DONE;
1787
1788 port = macvlan_port_get_rtnl(dev);
1789
1790 switch (event) {
1791 case NETDEV_UP:
1792 case NETDEV_DOWN:
1793 case NETDEV_CHANGE:
1794 list_for_each_entry(vlan, &port->vlans, list)
1795 netif_stacked_transfer_operstate(vlan->lowerdev,
1796 vlan->dev);
1797 break;
1798 case NETDEV_FEAT_CHANGE:
1799 list_for_each_entry(vlan, &port->vlans, list) {
1800 netif_inherit_tso_max(vlan->dev, dev);
1801 netdev_update_features(vlan->dev);
1802 }
1803 break;
1804 case NETDEV_CHANGEMTU:
1805 list_for_each_entry(vlan, &port->vlans, list) {
1806 if (vlan->dev->mtu <= dev->mtu)
1807 continue;
1808 dev_set_mtu(vlan->dev, dev->mtu);
1809 }
1810 break;
1811 case NETDEV_CHANGEADDR:
1812 if (!macvlan_passthru(port))
1813 return NOTIFY_DONE;
1814
1815 vlan = list_first_entry_or_null(&port->vlans,
1816 struct macvlan_dev,
1817 list);
1818
1819 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1820 return NOTIFY_BAD;
1821
1822 break;
1823 case NETDEV_UNREGISTER:
1824 /* twiddle thumbs on netns device moves */
1825 if (dev->reg_state != NETREG_UNREGISTERING)
1826 break;
1827
1828 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1829 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1830 unregister_netdevice_many(&list_kill);
1831 break;
1832 case NETDEV_PRE_TYPE_CHANGE:
1833 /* Forbid underlying device to change its type. */
1834 return NOTIFY_BAD;
1835
1836 case NETDEV_NOTIFY_PEERS:
1837 case NETDEV_BONDING_FAILOVER:
1838 case NETDEV_RESEND_IGMP:
1839 /* Propagate to all vlans */
1840 list_for_each_entry(vlan, &port->vlans, list)
1841 call_netdevice_notifiers(event, vlan->dev);
1842 }
1843 return NOTIFY_DONE;
1844}
1845
1846static struct notifier_block macvlan_notifier_block __read_mostly = {
1847 .notifier_call = macvlan_device_event,
1848};
1849
1850static int __init macvlan_init_module(void)
1851{
1852 int err;
1853
1854 register_netdevice_notifier(&macvlan_notifier_block);
1855
1856 err = macvlan_link_register(&macvlan_link_ops);
1857 if (err < 0)
1858 goto err1;
1859 return 0;
1860err1:
1861 unregister_netdevice_notifier(&macvlan_notifier_block);
1862 return err;
1863}
1864
1865static void __exit macvlan_cleanup_module(void)
1866{
1867 rtnl_link_unregister(&macvlan_link_ops);
1868 unregister_netdevice_notifier(&macvlan_notifier_block);
1869}
1870
1871module_init(macvlan_init_module);
1872module_exit(macvlan_cleanup_module);
1873
1874MODULE_LICENSE("GPL");
1875MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1876MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1877MODULE_ALIAS_RTNL_LINK("macvlan");
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");