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 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 | NETIF_F_LLTX)
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 netif_inherit_tso_max(dev, lowerdev);
936 dev->hard_header_len = lowerdev->hard_header_len;
937 macvlan_set_lockdep_class(dev);
938
939 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
940 if (!vlan->pcpu_stats)
941 return -ENOMEM;
942
943 port->count += 1;
944
945 /* Get macvlan's reference to lowerdev */
946 netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
947
948 return 0;
949}
950
951static void macvlan_uninit(struct net_device *dev)
952{
953 struct macvlan_dev *vlan = netdev_priv(dev);
954 struct macvlan_port *port = vlan->port;
955
956 free_percpu(vlan->pcpu_stats);
957
958 macvlan_flush_sources(port, vlan);
959 port->count -= 1;
960 if (!port->count)
961 macvlan_port_destroy(port->dev);
962}
963
964static void macvlan_dev_get_stats64(struct net_device *dev,
965 struct rtnl_link_stats64 *stats)
966{
967 struct macvlan_dev *vlan = netdev_priv(dev);
968
969 if (vlan->pcpu_stats) {
970 struct vlan_pcpu_stats *p;
971 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
972 u32 rx_errors = 0, tx_dropped = 0;
973 unsigned int start;
974 int i;
975
976 for_each_possible_cpu(i) {
977 p = per_cpu_ptr(vlan->pcpu_stats, i);
978 do {
979 start = u64_stats_fetch_begin(&p->syncp);
980 rx_packets = u64_stats_read(&p->rx_packets);
981 rx_bytes = u64_stats_read(&p->rx_bytes);
982 rx_multicast = u64_stats_read(&p->rx_multicast);
983 tx_packets = u64_stats_read(&p->tx_packets);
984 tx_bytes = u64_stats_read(&p->tx_bytes);
985 } while (u64_stats_fetch_retry(&p->syncp, start));
986
987 stats->rx_packets += rx_packets;
988 stats->rx_bytes += rx_bytes;
989 stats->multicast += rx_multicast;
990 stats->tx_packets += tx_packets;
991 stats->tx_bytes += tx_bytes;
992 /* rx_errors & tx_dropped are u32, updated
993 * without syncp protection.
994 */
995 rx_errors += READ_ONCE(p->rx_errors);
996 tx_dropped += READ_ONCE(p->tx_dropped);
997 }
998 stats->rx_errors = rx_errors;
999 stats->rx_dropped = rx_errors;
1000 stats->tx_dropped = tx_dropped;
1001 }
1002}
1003
1004static int macvlan_vlan_rx_add_vid(struct net_device *dev,
1005 __be16 proto, u16 vid)
1006{
1007 struct macvlan_dev *vlan = netdev_priv(dev);
1008 struct net_device *lowerdev = vlan->lowerdev;
1009
1010 return vlan_vid_add(lowerdev, proto, vid);
1011}
1012
1013static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
1014 __be16 proto, u16 vid)
1015{
1016 struct macvlan_dev *vlan = netdev_priv(dev);
1017 struct net_device *lowerdev = vlan->lowerdev;
1018
1019 vlan_vid_del(lowerdev, proto, vid);
1020 return 0;
1021}
1022
1023static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1024 struct net_device *dev,
1025 const unsigned char *addr, u16 vid,
1026 u16 flags,
1027 struct netlink_ext_ack *extack)
1028{
1029 struct macvlan_dev *vlan = netdev_priv(dev);
1030 int err = -EINVAL;
1031
1032 /* Support unicast filter only on passthru devices.
1033 * Multicast filter should be allowed on all devices.
1034 */
1035 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1036 return -EOPNOTSUPP;
1037
1038 if (flags & NLM_F_REPLACE)
1039 return -EOPNOTSUPP;
1040
1041 if (is_unicast_ether_addr(addr))
1042 err = dev_uc_add_excl(dev, addr);
1043 else if (is_multicast_ether_addr(addr))
1044 err = dev_mc_add_excl(dev, addr);
1045
1046 return err;
1047}
1048
1049static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1050 struct net_device *dev,
1051 const unsigned char *addr, u16 vid,
1052 struct netlink_ext_ack *extack)
1053{
1054 struct macvlan_dev *vlan = netdev_priv(dev);
1055 int err = -EINVAL;
1056
1057 /* Support unicast filter only on passthru devices.
1058 * Multicast filter should be allowed on all devices.
1059 */
1060 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1061 return -EOPNOTSUPP;
1062
1063 if (is_unicast_ether_addr(addr))
1064 err = dev_uc_del(dev, addr);
1065 else if (is_multicast_ether_addr(addr))
1066 err = dev_mc_del(dev, addr);
1067
1068 return err;
1069}
1070
1071static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1072 struct ethtool_drvinfo *drvinfo)
1073{
1074 strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1075 strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1076}
1077
1078static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1079 struct ethtool_link_ksettings *cmd)
1080{
1081 const struct macvlan_dev *vlan = netdev_priv(dev);
1082
1083 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1084}
1085
1086static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1087 struct ethtool_ts_info *info)
1088{
1089 struct net_device *real_dev = macvlan_dev_real_dev(dev);
1090
1091 return ethtool_get_ts_info_by_layer(real_dev, info);
1092}
1093
1094static netdev_features_t macvlan_fix_features(struct net_device *dev,
1095 netdev_features_t features)
1096{
1097 struct macvlan_dev *vlan = netdev_priv(dev);
1098 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1099 netdev_features_t mask;
1100
1101 features |= NETIF_F_ALL_FOR_ALL;
1102 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1103 mask = features;
1104
1105 lowerdev_features &= (features | ~NETIF_F_LRO);
1106 features = netdev_increment_features(lowerdev_features, features, mask);
1107 features |= ALWAYS_ON_FEATURES;
1108 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1109
1110 return features;
1111}
1112
1113#ifdef CONFIG_NET_POLL_CONTROLLER
1114static void macvlan_dev_poll_controller(struct net_device *dev)
1115{
1116 return;
1117}
1118
1119static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1120{
1121 struct macvlan_dev *vlan = netdev_priv(dev);
1122 struct net_device *real_dev = vlan->lowerdev;
1123 struct netpoll *netpoll;
1124 int err;
1125
1126 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1127 err = -ENOMEM;
1128 if (!netpoll)
1129 goto out;
1130
1131 err = __netpoll_setup(netpoll, real_dev);
1132 if (err) {
1133 kfree(netpoll);
1134 goto out;
1135 }
1136
1137 vlan->netpoll = netpoll;
1138
1139out:
1140 return err;
1141}
1142
1143static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1144{
1145 struct macvlan_dev *vlan = netdev_priv(dev);
1146 struct netpoll *netpoll = vlan->netpoll;
1147
1148 if (!netpoll)
1149 return;
1150
1151 vlan->netpoll = NULL;
1152
1153 __netpoll_free(netpoll);
1154}
1155#endif /* CONFIG_NET_POLL_CONTROLLER */
1156
1157static int macvlan_dev_get_iflink(const struct net_device *dev)
1158{
1159 struct macvlan_dev *vlan = netdev_priv(dev);
1160
1161 return vlan->lowerdev->ifindex;
1162}
1163
1164static const struct ethtool_ops macvlan_ethtool_ops = {
1165 .get_link = ethtool_op_get_link,
1166 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1167 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1168 .get_ts_info = macvlan_ethtool_get_ts_info,
1169};
1170
1171static const struct net_device_ops macvlan_netdev_ops = {
1172 .ndo_init = macvlan_init,
1173 .ndo_uninit = macvlan_uninit,
1174 .ndo_open = macvlan_open,
1175 .ndo_stop = macvlan_stop,
1176 .ndo_start_xmit = macvlan_start_xmit,
1177 .ndo_change_mtu = macvlan_change_mtu,
1178 .ndo_fix_features = macvlan_fix_features,
1179 .ndo_change_rx_flags = macvlan_change_rx_flags,
1180 .ndo_set_mac_address = macvlan_set_mac_address,
1181 .ndo_set_rx_mode = macvlan_set_mac_lists,
1182 .ndo_get_stats64 = macvlan_dev_get_stats64,
1183 .ndo_validate_addr = eth_validate_addr,
1184 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1185 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1186 .ndo_fdb_add = macvlan_fdb_add,
1187 .ndo_fdb_del = macvlan_fdb_del,
1188 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1189#ifdef CONFIG_NET_POLL_CONTROLLER
1190 .ndo_poll_controller = macvlan_dev_poll_controller,
1191 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1192 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1193#endif
1194 .ndo_get_iflink = macvlan_dev_get_iflink,
1195 .ndo_features_check = passthru_features_check,
1196 .ndo_hwtstamp_get = macvlan_hwtstamp_get,
1197 .ndo_hwtstamp_set = macvlan_hwtstamp_set,
1198};
1199
1200static void macvlan_dev_free(struct net_device *dev)
1201{
1202 struct macvlan_dev *vlan = netdev_priv(dev);
1203
1204 /* Get rid of the macvlan's reference to lowerdev */
1205 netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1206}
1207
1208void macvlan_common_setup(struct net_device *dev)
1209{
1210 ether_setup(dev);
1211
1212 /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1213 dev->max_mtu = ETH_MAX_MTU;
1214 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1215 netif_keep_dst(dev);
1216 dev->priv_flags |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN;
1217 dev->netdev_ops = &macvlan_netdev_ops;
1218 dev->needs_free_netdev = true;
1219 dev->priv_destructor = macvlan_dev_free;
1220 dev->header_ops = &macvlan_hard_header_ops;
1221 dev->ethtool_ops = &macvlan_ethtool_ops;
1222}
1223EXPORT_SYMBOL_GPL(macvlan_common_setup);
1224
1225static void macvlan_setup(struct net_device *dev)
1226{
1227 macvlan_common_setup(dev);
1228 dev->priv_flags |= IFF_NO_QUEUE;
1229}
1230
1231static int macvlan_port_create(struct net_device *dev)
1232{
1233 struct macvlan_port *port;
1234 unsigned int i;
1235 int err;
1236
1237 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1238 return -EINVAL;
1239
1240 if (netdev_is_rx_handler_busy(dev))
1241 return -EBUSY;
1242
1243 port = kzalloc(sizeof(*port), GFP_KERNEL);
1244 if (port == NULL)
1245 return -ENOMEM;
1246
1247 port->dev = dev;
1248 ether_addr_copy(port->perm_addr, dev->dev_addr);
1249 INIT_LIST_HEAD(&port->vlans);
1250 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1251 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1252 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1253 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1254
1255 port->bc_queue_len_used = 0;
1256 port->bc_cutoff = 1;
1257 skb_queue_head_init(&port->bc_queue);
1258 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1259
1260 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1261 if (err)
1262 kfree(port);
1263 else
1264 dev->priv_flags |= IFF_MACVLAN_PORT;
1265 return err;
1266}
1267
1268static void macvlan_port_destroy(struct net_device *dev)
1269{
1270 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1271 struct sk_buff *skb;
1272
1273 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1274 netdev_rx_handler_unregister(dev);
1275
1276 /* After this point, no packet can schedule bc_work anymore,
1277 * but we need to cancel it and purge left skbs if any.
1278 */
1279 cancel_work_sync(&port->bc_work);
1280
1281 while ((skb = __skb_dequeue(&port->bc_queue))) {
1282 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1283
1284 if (src)
1285 dev_put(src->dev);
1286
1287 kfree_skb(skb);
1288 }
1289
1290 /* If the lower device address has been changed by passthru
1291 * macvlan, put it back.
1292 */
1293 if (macvlan_passthru(port) &&
1294 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1295 struct sockaddr sa;
1296
1297 sa.sa_family = port->dev->type;
1298 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1299 dev_set_mac_address(port->dev, &sa, NULL);
1300 }
1301
1302 kfree(port);
1303}
1304
1305static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1306 struct netlink_ext_ack *extack)
1307{
1308 struct nlattr *nla, *head;
1309 int rem, len;
1310
1311 if (tb[IFLA_ADDRESS]) {
1312 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1313 return -EINVAL;
1314 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1315 return -EADDRNOTAVAIL;
1316 }
1317
1318 if (!data)
1319 return 0;
1320
1321 if (data[IFLA_MACVLAN_FLAGS] &&
1322 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1323 MACVLAN_FLAG_NODST))
1324 return -EINVAL;
1325
1326 if (data[IFLA_MACVLAN_MODE]) {
1327 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1328 case MACVLAN_MODE_PRIVATE:
1329 case MACVLAN_MODE_VEPA:
1330 case MACVLAN_MODE_BRIDGE:
1331 case MACVLAN_MODE_PASSTHRU:
1332 case MACVLAN_MODE_SOURCE:
1333 break;
1334 default:
1335 return -EINVAL;
1336 }
1337 }
1338
1339 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1340 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1341 case MACVLAN_MACADDR_ADD:
1342 case MACVLAN_MACADDR_DEL:
1343 case MACVLAN_MACADDR_FLUSH:
1344 case MACVLAN_MACADDR_SET:
1345 break;
1346 default:
1347 return -EINVAL;
1348 }
1349 }
1350
1351 if (data[IFLA_MACVLAN_MACADDR]) {
1352 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1353 return -EINVAL;
1354
1355 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1356 return -EADDRNOTAVAIL;
1357 }
1358
1359 if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1360 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1361 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1362
1363 nla_for_each_attr(nla, head, len, rem) {
1364 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1365 nla_len(nla) != ETH_ALEN)
1366 return -EINVAL;
1367
1368 if (!is_valid_ether_addr(nla_data(nla)))
1369 return -EADDRNOTAVAIL;
1370 }
1371 }
1372
1373 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1374 return -EINVAL;
1375
1376 return 0;
1377}
1378
1379/*
1380 * reconfigure list of remote source mac address
1381 * (only for macvlan devices in source mode)
1382 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1383 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1384 */
1385static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1386 struct nlattr *data[])
1387{
1388 char *addr = NULL;
1389 int ret, rem, len;
1390 struct nlattr *nla, *head;
1391 struct macvlan_source_entry *entry;
1392
1393 if (data[IFLA_MACVLAN_MACADDR])
1394 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1395
1396 if (mode == MACVLAN_MACADDR_ADD) {
1397 if (!addr)
1398 return -EINVAL;
1399
1400 return macvlan_hash_add_source(vlan, addr);
1401
1402 } else if (mode == MACVLAN_MACADDR_DEL) {
1403 if (!addr)
1404 return -EINVAL;
1405
1406 entry = macvlan_hash_lookup_source(vlan, addr);
1407 if (entry) {
1408 macvlan_hash_del_source(entry);
1409 vlan->macaddr_count--;
1410 }
1411 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1412 macvlan_flush_sources(vlan->port, vlan);
1413 } else if (mode == MACVLAN_MACADDR_SET) {
1414 macvlan_flush_sources(vlan->port, vlan);
1415
1416 if (addr) {
1417 ret = macvlan_hash_add_source(vlan, addr);
1418 if (ret)
1419 return ret;
1420 }
1421
1422 if (!data[IFLA_MACVLAN_MACADDR_DATA])
1423 return 0;
1424
1425 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1426 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1427
1428 nla_for_each_attr(nla, head, len, rem) {
1429 addr = nla_data(nla);
1430 ret = macvlan_hash_add_source(vlan, addr);
1431 if (ret)
1432 return ret;
1433 }
1434 } else {
1435 return -EINVAL;
1436 }
1437
1438 return 0;
1439}
1440
1441int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1442 struct nlattr *tb[], struct nlattr *data[],
1443 struct netlink_ext_ack *extack)
1444{
1445 struct macvlan_dev *vlan = netdev_priv(dev);
1446 struct macvlan_port *port;
1447 struct net_device *lowerdev;
1448 int err;
1449 int macmode;
1450 bool create = false;
1451
1452 if (!tb[IFLA_LINK])
1453 return -EINVAL;
1454
1455 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1456 if (lowerdev == NULL)
1457 return -ENODEV;
1458
1459 /* When creating macvlans or macvtaps on top of other macvlans - use
1460 * the real device as the lowerdev.
1461 */
1462 if (netif_is_macvlan(lowerdev))
1463 lowerdev = macvlan_dev_real_dev(lowerdev);
1464
1465 if (!tb[IFLA_MTU])
1466 dev->mtu = lowerdev->mtu;
1467 else if (dev->mtu > lowerdev->mtu)
1468 return -EINVAL;
1469
1470 /* MTU range: 68 - lowerdev->max_mtu */
1471 dev->min_mtu = ETH_MIN_MTU;
1472 dev->max_mtu = lowerdev->max_mtu;
1473
1474 if (!tb[IFLA_ADDRESS])
1475 eth_hw_addr_random(dev);
1476
1477 if (!netif_is_macvlan_port(lowerdev)) {
1478 err = macvlan_port_create(lowerdev);
1479 if (err < 0)
1480 return err;
1481 create = true;
1482 }
1483 port = macvlan_port_get_rtnl(lowerdev);
1484
1485 /* Only 1 macvlan device can be created in passthru mode */
1486 if (macvlan_passthru(port)) {
1487 /* The macvlan port must be not created this time,
1488 * still goto destroy_macvlan_port for readability.
1489 */
1490 err = -EINVAL;
1491 goto destroy_macvlan_port;
1492 }
1493
1494 vlan->lowerdev = lowerdev;
1495 vlan->dev = dev;
1496 vlan->port = port;
1497 vlan->set_features = MACVLAN_FEATURES;
1498
1499 vlan->mode = MACVLAN_MODE_VEPA;
1500 if (data && data[IFLA_MACVLAN_MODE])
1501 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1502
1503 if (data && data[IFLA_MACVLAN_FLAGS])
1504 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1505
1506 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1507 if (port->count) {
1508 err = -EINVAL;
1509 goto destroy_macvlan_port;
1510 }
1511 macvlan_set_passthru(port);
1512 eth_hw_addr_inherit(dev, lowerdev);
1513 }
1514
1515 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1516 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1517 err = -EINVAL;
1518 goto destroy_macvlan_port;
1519 }
1520 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1521 err = macvlan_changelink_sources(vlan, macmode, data);
1522 if (err)
1523 goto destroy_macvlan_port;
1524 }
1525
1526 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1527 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1528 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1529
1530 if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1531 update_port_bc_cutoff(
1532 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1533
1534 err = register_netdevice(dev);
1535 if (err < 0)
1536 goto destroy_macvlan_port;
1537
1538 dev->priv_flags |= IFF_MACVLAN;
1539 err = netdev_upper_dev_link(lowerdev, dev, extack);
1540 if (err)
1541 goto unregister_netdev;
1542
1543 list_add_tail_rcu(&vlan->list, &port->vlans);
1544 update_port_bc_queue_len(vlan->port);
1545 netif_stacked_transfer_operstate(lowerdev, dev);
1546 linkwatch_fire_event(dev);
1547
1548 return 0;
1549
1550unregister_netdev:
1551 /* macvlan_uninit would free the macvlan port */
1552 unregister_netdevice(dev);
1553 return err;
1554destroy_macvlan_port:
1555 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1556 * so we destroy the macvlan port only when it's valid.
1557 */
1558 if (create && macvlan_port_get_rtnl(lowerdev)) {
1559 macvlan_flush_sources(port, vlan);
1560 macvlan_port_destroy(port->dev);
1561 }
1562 return err;
1563}
1564EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1565
1566static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1567 struct nlattr *tb[], struct nlattr *data[],
1568 struct netlink_ext_ack *extack)
1569{
1570 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1571}
1572
1573void macvlan_dellink(struct net_device *dev, struct list_head *head)
1574{
1575 struct macvlan_dev *vlan = netdev_priv(dev);
1576
1577 if (vlan->mode == MACVLAN_MODE_SOURCE)
1578 macvlan_flush_sources(vlan->port, vlan);
1579 list_del_rcu(&vlan->list);
1580 update_port_bc_queue_len(vlan->port);
1581 unregister_netdevice_queue(dev, head);
1582 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1583}
1584EXPORT_SYMBOL_GPL(macvlan_dellink);
1585
1586static int macvlan_changelink(struct net_device *dev,
1587 struct nlattr *tb[], struct nlattr *data[],
1588 struct netlink_ext_ack *extack)
1589{
1590 struct macvlan_dev *vlan = netdev_priv(dev);
1591 enum macvlan_mode mode;
1592 bool set_mode = false;
1593 enum macvlan_macaddr_mode macmode;
1594 int ret;
1595
1596 /* Validate mode, but don't set yet: setting flags may fail. */
1597 if (data && data[IFLA_MACVLAN_MODE]) {
1598 set_mode = true;
1599 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1600 /* Passthrough mode can't be set or cleared dynamically */
1601 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1602 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1603 return -EINVAL;
1604 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1605 vlan->mode != mode)
1606 macvlan_flush_sources(vlan->port, vlan);
1607 }
1608
1609 if (data && data[IFLA_MACVLAN_FLAGS]) {
1610 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1611 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1612 if (macvlan_passthru(vlan->port) && promisc) {
1613 int err;
1614
1615 if (flags & MACVLAN_FLAG_NOPROMISC)
1616 err = dev_set_promiscuity(vlan->lowerdev, -1);
1617 else
1618 err = dev_set_promiscuity(vlan->lowerdev, 1);
1619 if (err < 0)
1620 return err;
1621 }
1622 vlan->flags = flags;
1623 }
1624
1625 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1626 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1627 update_port_bc_queue_len(vlan->port);
1628 }
1629
1630 if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1631 update_port_bc_cutoff(
1632 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1633
1634 if (set_mode)
1635 vlan->mode = mode;
1636 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1637 if (vlan->mode != MACVLAN_MODE_SOURCE)
1638 return -EINVAL;
1639 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1640 ret = macvlan_changelink_sources(vlan, macmode, data);
1641 if (ret)
1642 return ret;
1643 }
1644 return 0;
1645}
1646
1647static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1648{
1649 if (vlan->macaddr_count == 0)
1650 return 0;
1651 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1652 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1653}
1654
1655static size_t macvlan_get_size(const struct net_device *dev)
1656{
1657 struct macvlan_dev *vlan = netdev_priv(dev);
1658
1659 return (0
1660 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1661 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1662 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1663 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1664 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1665 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1666 );
1667}
1668
1669static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1670 const struct macvlan_dev *vlan,
1671 const int i)
1672{
1673 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1674 struct macvlan_source_entry *entry;
1675
1676 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1677 if (entry->vlan != vlan)
1678 continue;
1679 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1680 return 1;
1681 }
1682 return 0;
1683}
1684
1685static int macvlan_fill_info(struct sk_buff *skb,
1686 const struct net_device *dev)
1687{
1688 struct macvlan_dev *vlan = netdev_priv(dev);
1689 struct macvlan_port *port = vlan->port;
1690 int i;
1691 struct nlattr *nest;
1692
1693 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1694 goto nla_put_failure;
1695 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1696 goto nla_put_failure;
1697 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1698 goto nla_put_failure;
1699 if (vlan->macaddr_count > 0) {
1700 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1701 if (nest == NULL)
1702 goto nla_put_failure;
1703
1704 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1705 if (macvlan_fill_info_macaddr(skb, vlan, i))
1706 goto nla_put_failure;
1707 }
1708 nla_nest_end(skb, nest);
1709 }
1710 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1711 goto nla_put_failure;
1712 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1713 goto nla_put_failure;
1714 if (port->bc_cutoff != 1 &&
1715 nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff))
1716 goto nla_put_failure;
1717 return 0;
1718
1719nla_put_failure:
1720 return -EMSGSIZE;
1721}
1722
1723static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1724 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1725 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1726 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1727 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1728 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1729 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1730 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1731 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1732 [IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 },
1733};
1734
1735int macvlan_link_register(struct rtnl_link_ops *ops)
1736{
1737 /* common fields */
1738 ops->validate = macvlan_validate;
1739 ops->maxtype = IFLA_MACVLAN_MAX;
1740 ops->policy = macvlan_policy;
1741 ops->changelink = macvlan_changelink;
1742 ops->get_size = macvlan_get_size;
1743 ops->fill_info = macvlan_fill_info;
1744
1745 return rtnl_link_register(ops);
1746};
1747EXPORT_SYMBOL_GPL(macvlan_link_register);
1748
1749static struct net *macvlan_get_link_net(const struct net_device *dev)
1750{
1751 return dev_net(macvlan_dev_real_dev(dev));
1752}
1753
1754static struct rtnl_link_ops macvlan_link_ops = {
1755 .kind = "macvlan",
1756 .setup = macvlan_setup,
1757 .newlink = macvlan_newlink,
1758 .dellink = macvlan_dellink,
1759 .get_link_net = macvlan_get_link_net,
1760 .priv_size = sizeof(struct macvlan_dev),
1761};
1762
1763static void update_port_bc_queue_len(struct macvlan_port *port)
1764{
1765 u32 max_bc_queue_len_req = 0;
1766 struct macvlan_dev *vlan;
1767
1768 list_for_each_entry(vlan, &port->vlans, list) {
1769 if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1770 max_bc_queue_len_req = vlan->bc_queue_len_req;
1771 }
1772 port->bc_queue_len_used = max_bc_queue_len_req;
1773}
1774
1775static int macvlan_device_event(struct notifier_block *unused,
1776 unsigned long event, void *ptr)
1777{
1778 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1779 struct macvlan_dev *vlan, *next;
1780 struct macvlan_port *port;
1781 LIST_HEAD(list_kill);
1782
1783 if (!netif_is_macvlan_port(dev))
1784 return NOTIFY_DONE;
1785
1786 port = macvlan_port_get_rtnl(dev);
1787
1788 switch (event) {
1789 case NETDEV_UP:
1790 case NETDEV_DOWN:
1791 case NETDEV_CHANGE:
1792 list_for_each_entry(vlan, &port->vlans, list)
1793 netif_stacked_transfer_operstate(vlan->lowerdev,
1794 vlan->dev);
1795 break;
1796 case NETDEV_FEAT_CHANGE:
1797 list_for_each_entry(vlan, &port->vlans, list) {
1798 netif_inherit_tso_max(vlan->dev, dev);
1799 netdev_update_features(vlan->dev);
1800 }
1801 break;
1802 case NETDEV_CHANGEMTU:
1803 list_for_each_entry(vlan, &port->vlans, list) {
1804 if (vlan->dev->mtu <= dev->mtu)
1805 continue;
1806 dev_set_mtu(vlan->dev, dev->mtu);
1807 }
1808 break;
1809 case NETDEV_CHANGEADDR:
1810 if (!macvlan_passthru(port))
1811 return NOTIFY_DONE;
1812
1813 vlan = list_first_entry_or_null(&port->vlans,
1814 struct macvlan_dev,
1815 list);
1816
1817 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1818 return NOTIFY_BAD;
1819
1820 break;
1821 case NETDEV_UNREGISTER:
1822 /* twiddle thumbs on netns device moves */
1823 if (dev->reg_state != NETREG_UNREGISTERING)
1824 break;
1825
1826 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1827 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1828 unregister_netdevice_many(&list_kill);
1829 break;
1830 case NETDEV_PRE_TYPE_CHANGE:
1831 /* Forbid underlying device to change its type. */
1832 return NOTIFY_BAD;
1833
1834 case NETDEV_NOTIFY_PEERS:
1835 case NETDEV_BONDING_FAILOVER:
1836 case NETDEV_RESEND_IGMP:
1837 /* Propagate to all vlans */
1838 list_for_each_entry(vlan, &port->vlans, list)
1839 call_netdevice_notifiers(event, vlan->dev);
1840 }
1841 return NOTIFY_DONE;
1842}
1843
1844static struct notifier_block macvlan_notifier_block __read_mostly = {
1845 .notifier_call = macvlan_device_event,
1846};
1847
1848static int __init macvlan_init_module(void)
1849{
1850 int err;
1851
1852 register_netdevice_notifier(&macvlan_notifier_block);
1853
1854 err = macvlan_link_register(&macvlan_link_ops);
1855 if (err < 0)
1856 goto err1;
1857 return 0;
1858err1:
1859 unregister_netdevice_notifier(&macvlan_notifier_block);
1860 return err;
1861}
1862
1863static void __exit macvlan_cleanup_module(void)
1864{
1865 rtnl_link_unregister(&macvlan_link_ops);
1866 unregister_netdevice_notifier(&macvlan_notifier_block);
1867}
1868
1869module_init(macvlan_init_module);
1870module_exit(macvlan_cleanup_module);
1871
1872MODULE_LICENSE("GPL");
1873MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1874MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1875MODULE_ALIAS_RTNL_LINK("macvlan");
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4 *
5 * The code this is based on carried the following copyright notice:
6 * ---
7 * (C) Copyright 2001-2006
8 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
9 * Re-worked by Ben Greear <greearb@candelatech.com>
10 * ---
11 */
12#include <linux/kernel.h>
13#include <linux/types.h>
14#include <linux/module.h>
15#include <linux/init.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/string.h>
19#include <linux/rculist.h>
20#include <linux/notifier.h>
21#include <linux/netdevice.h>
22#include <linux/etherdevice.h>
23#include <linux/net_tstamp.h>
24#include <linux/ethtool.h>
25#include <linux/if_arp.h>
26#include <linux/if_vlan.h>
27#include <linux/if_link.h>
28#include <linux/if_macvlan.h>
29#include <linux/hash.h>
30#include <linux/workqueue.h>
31#include <net/rtnetlink.h>
32#include <net/xfrm.h>
33#include <linux/netpoll.h>
34#include <linux/phy.h>
35
36#define MACVLAN_HASH_BITS 8
37#define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS)
38#define MACVLAN_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 u32 flags;
51 int count;
52 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE];
53 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
54 unsigned char perm_addr[ETH_ALEN];
55};
56
57struct macvlan_source_entry {
58 struct hlist_node hlist;
59 struct macvlan_dev *vlan;
60 unsigned char addr[6+2] __aligned(sizeof(u16));
61 struct rcu_head rcu;
62};
63
64struct macvlan_skb_cb {
65 const struct macvlan_dev *src;
66};
67
68#define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
69
70static void macvlan_port_destroy(struct net_device *dev);
71static void update_port_bc_queue_len(struct macvlan_port *port);
72
73static inline bool macvlan_passthru(const struct macvlan_port *port)
74{
75 return port->flags & MACVLAN_F_PASSTHRU;
76}
77
78static inline void macvlan_set_passthru(struct macvlan_port *port)
79{
80 port->flags |= MACVLAN_F_PASSTHRU;
81}
82
83static inline bool macvlan_addr_change(const struct macvlan_port *port)
84{
85 return port->flags & MACVLAN_F_ADDRCHANGE;
86}
87
88static inline void macvlan_set_addr_change(struct macvlan_port *port)
89{
90 port->flags |= MACVLAN_F_ADDRCHANGE;
91}
92
93static inline void macvlan_clear_addr_change(struct macvlan_port *port)
94{
95 port->flags &= ~MACVLAN_F_ADDRCHANGE;
96}
97
98/* Hash Ethernet address */
99static u32 macvlan_eth_hash(const unsigned char *addr)
100{
101 u64 value = get_unaligned((u64 *)addr);
102
103 /* only want 6 bytes */
104#ifdef __BIG_ENDIAN
105 value >>= 16;
106#else
107 value <<= 16;
108#endif
109 return hash_64(value, MACVLAN_HASH_BITS);
110}
111
112static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
113{
114 return rcu_dereference(dev->rx_handler_data);
115}
116
117static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
118{
119 return rtnl_dereference(dev->rx_handler_data);
120}
121
122static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
123 const unsigned char *addr)
124{
125 struct macvlan_dev *vlan;
126 u32 idx = macvlan_eth_hash(addr);
127
128 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
129 lockdep_rtnl_is_held()) {
130 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
131 return vlan;
132 }
133 return NULL;
134}
135
136static struct macvlan_source_entry *macvlan_hash_lookup_source(
137 const struct macvlan_dev *vlan,
138 const unsigned char *addr)
139{
140 struct macvlan_source_entry *entry;
141 u32 idx = macvlan_eth_hash(addr);
142 struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
143
144 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
145 if (ether_addr_equal_64bits(entry->addr, addr) &&
146 entry->vlan == vlan)
147 return entry;
148 }
149 return NULL;
150}
151
152static int macvlan_hash_add_source(struct macvlan_dev *vlan,
153 const unsigned char *addr)
154{
155 struct macvlan_port *port = vlan->port;
156 struct macvlan_source_entry *entry;
157 struct hlist_head *h;
158
159 entry = macvlan_hash_lookup_source(vlan, addr);
160 if (entry)
161 return 0;
162
163 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
164 if (!entry)
165 return -ENOMEM;
166
167 ether_addr_copy(entry->addr, addr);
168 entry->vlan = vlan;
169 h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
170 hlist_add_head_rcu(&entry->hlist, h);
171 vlan->macaddr_count++;
172
173 return 0;
174}
175
176static void macvlan_hash_add(struct macvlan_dev *vlan)
177{
178 struct macvlan_port *port = vlan->port;
179 const unsigned char *addr = vlan->dev->dev_addr;
180 u32 idx = macvlan_eth_hash(addr);
181
182 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
183}
184
185static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
186{
187 hlist_del_rcu(&entry->hlist);
188 kfree_rcu(entry, rcu);
189}
190
191static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
192{
193 hlist_del_rcu(&vlan->hlist);
194 if (sync)
195 synchronize_rcu();
196}
197
198static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
199 const unsigned char *addr)
200{
201 macvlan_hash_del(vlan, true);
202 /* Now that we are unhashed it is safe to change the device
203 * address without confusing packet delivery.
204 */
205 eth_hw_addr_set(vlan->dev, addr);
206 macvlan_hash_add(vlan);
207}
208
209static bool macvlan_addr_busy(const struct macvlan_port *port,
210 const unsigned char *addr)
211{
212 /* Test to see if the specified address is
213 * currently in use by the underlying device or
214 * another macvlan.
215 */
216 if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
217 ether_addr_equal_64bits(port->dev->dev_addr, addr))
218 return true;
219
220 if (macvlan_hash_lookup(port, addr))
221 return true;
222
223 return false;
224}
225
226
227static int macvlan_broadcast_one(struct sk_buff *skb,
228 const struct macvlan_dev *vlan,
229 const struct ethhdr *eth, bool local)
230{
231 struct net_device *dev = vlan->dev;
232
233 if (local)
234 return __dev_forward_skb(dev, skb);
235
236 skb->dev = dev;
237 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
238 skb->pkt_type = PACKET_BROADCAST;
239 else
240 skb->pkt_type = PACKET_MULTICAST;
241
242 return 0;
243}
244
245static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
246{
247 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
248}
249
250
251static unsigned int mc_hash(const struct macvlan_dev *vlan,
252 const unsigned char *addr)
253{
254 u32 val = __get_unaligned_cpu32(addr + 2);
255
256 val ^= macvlan_hash_mix(vlan);
257 return hash_32(val, MACVLAN_MC_FILTER_BITS);
258}
259
260static void macvlan_broadcast(struct sk_buff *skb,
261 const struct macvlan_port *port,
262 struct net_device *src,
263 enum macvlan_mode mode)
264{
265 const struct ethhdr *eth = eth_hdr(skb);
266 const struct macvlan_dev *vlan;
267 struct sk_buff *nskb;
268 unsigned int i;
269 int err;
270 unsigned int hash;
271
272 if (skb->protocol == htons(ETH_P_PAUSE))
273 return;
274
275 hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) {
276 if (vlan->dev == src || !(vlan->mode & mode))
277 continue;
278
279 hash = mc_hash(vlan, eth->h_dest);
280 if (!test_bit(hash, vlan->mc_filter))
281 continue;
282
283 err = NET_RX_DROP;
284 nskb = skb_clone(skb, GFP_ATOMIC);
285 if (likely(nskb))
286 err = macvlan_broadcast_one(nskb, vlan, eth,
287 mode == MACVLAN_MODE_BRIDGE) ?:
288 netif_rx(nskb);
289 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
290 err == NET_RX_SUCCESS, true);
291 }
292}
293
294static void macvlan_process_broadcast(struct work_struct *w)
295{
296 struct macvlan_port *port = container_of(w, struct macvlan_port,
297 bc_work);
298 struct sk_buff *skb;
299 struct sk_buff_head list;
300
301 __skb_queue_head_init(&list);
302
303 spin_lock_bh(&port->bc_queue.lock);
304 skb_queue_splice_tail_init(&port->bc_queue, &list);
305 spin_unlock_bh(&port->bc_queue.lock);
306
307 while ((skb = __skb_dequeue(&list))) {
308 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
309
310 rcu_read_lock();
311
312 if (!src)
313 /* frame comes from an external address */
314 macvlan_broadcast(skb, port, NULL,
315 MACVLAN_MODE_PRIVATE |
316 MACVLAN_MODE_VEPA |
317 MACVLAN_MODE_PASSTHRU|
318 MACVLAN_MODE_BRIDGE);
319 else if (src->mode == MACVLAN_MODE_VEPA)
320 /* flood to everyone except source */
321 macvlan_broadcast(skb, port, src->dev,
322 MACVLAN_MODE_VEPA |
323 MACVLAN_MODE_BRIDGE);
324 else
325 /*
326 * flood only to VEPA ports, bridge ports
327 * already saw the frame on the way out.
328 */
329 macvlan_broadcast(skb, port, src->dev,
330 MACVLAN_MODE_VEPA);
331
332 rcu_read_unlock();
333
334 if (src)
335 dev_put(src->dev);
336 consume_skb(skb);
337
338 cond_resched();
339 }
340}
341
342static void macvlan_broadcast_enqueue(struct macvlan_port *port,
343 const struct macvlan_dev *src,
344 struct sk_buff *skb)
345{
346 struct sk_buff *nskb;
347 int err = -ENOMEM;
348
349 nskb = skb_clone(skb, GFP_ATOMIC);
350 if (!nskb)
351 goto err;
352
353 MACVLAN_SKB_CB(nskb)->src = src;
354
355 spin_lock(&port->bc_queue.lock);
356 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) {
357 if (src)
358 dev_hold(src->dev);
359 __skb_queue_tail(&port->bc_queue, nskb);
360 err = 0;
361 }
362 spin_unlock(&port->bc_queue.lock);
363
364 queue_work(system_unbound_wq, &port->bc_work);
365
366 if (err)
367 goto free_nskb;
368
369 return;
370
371free_nskb:
372 kfree_skb(nskb);
373err:
374 dev_core_stats_rx_dropped_inc(skb->dev);
375}
376
377static void macvlan_flush_sources(struct macvlan_port *port,
378 struct macvlan_dev *vlan)
379{
380 struct macvlan_source_entry *entry;
381 struct hlist_node *next;
382 int i;
383
384 hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
385 if (entry->vlan == vlan)
386 macvlan_hash_del_source(entry);
387
388 vlan->macaddr_count = 0;
389}
390
391static void macvlan_forward_source_one(struct sk_buff *skb,
392 struct macvlan_dev *vlan)
393{
394 struct sk_buff *nskb;
395 struct net_device *dev;
396 int len;
397 int ret;
398
399 dev = vlan->dev;
400 if (unlikely(!(dev->flags & IFF_UP)))
401 return;
402
403 nskb = skb_clone(skb, GFP_ATOMIC);
404 if (!nskb)
405 return;
406
407 len = nskb->len + ETH_HLEN;
408 nskb->dev = dev;
409
410 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
411 nskb->pkt_type = PACKET_HOST;
412
413 ret = __netif_rx(nskb);
414 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
415}
416
417static bool macvlan_forward_source(struct sk_buff *skb,
418 struct macvlan_port *port,
419 const unsigned char *addr)
420{
421 struct macvlan_source_entry *entry;
422 u32 idx = macvlan_eth_hash(addr);
423 struct hlist_head *h = &port->vlan_source_hash[idx];
424 bool consume = false;
425
426 hlist_for_each_entry_rcu(entry, h, hlist) {
427 if (ether_addr_equal_64bits(entry->addr, addr)) {
428 if (entry->vlan->flags & MACVLAN_FLAG_NODST)
429 consume = true;
430 macvlan_forward_source_one(skb, entry->vlan);
431 }
432 }
433
434 return consume;
435}
436
437/* called under rcu_read_lock() from netif_receive_skb */
438static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
439{
440 struct macvlan_port *port;
441 struct sk_buff *skb = *pskb;
442 const struct ethhdr *eth = eth_hdr(skb);
443 const struct macvlan_dev *vlan;
444 const struct macvlan_dev *src;
445 struct net_device *dev;
446 unsigned int len = 0;
447 int ret;
448 rx_handler_result_t handle_res;
449
450 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */
451 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
452 return RX_HANDLER_PASS;
453
454 port = macvlan_port_get_rcu(skb->dev);
455 if (is_multicast_ether_addr(eth->h_dest)) {
456 unsigned int hash;
457
458 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
459 if (!skb)
460 return RX_HANDLER_CONSUMED;
461 *pskb = skb;
462 eth = eth_hdr(skb);
463 if (macvlan_forward_source(skb, port, eth->h_source)) {
464 kfree_skb(skb);
465 return RX_HANDLER_CONSUMED;
466 }
467 src = macvlan_hash_lookup(port, eth->h_source);
468 if (src && src->mode != MACVLAN_MODE_VEPA &&
469 src->mode != MACVLAN_MODE_BRIDGE) {
470 /* forward to original port. */
471 vlan = src;
472 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
473 __netif_rx(skb);
474 handle_res = RX_HANDLER_CONSUMED;
475 goto out;
476 }
477
478 hash = mc_hash(NULL, eth->h_dest);
479 if (test_bit(hash, port->mc_filter))
480 macvlan_broadcast_enqueue(port, src, skb);
481
482 return RX_HANDLER_PASS;
483 }
484
485 if (macvlan_forward_source(skb, port, eth->h_source)) {
486 kfree_skb(skb);
487 return RX_HANDLER_CONSUMED;
488 }
489 if (macvlan_passthru(port))
490 vlan = list_first_or_null_rcu(&port->vlans,
491 struct macvlan_dev, list);
492 else
493 vlan = macvlan_hash_lookup(port, eth->h_dest);
494 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
495 return RX_HANDLER_PASS;
496
497 dev = vlan->dev;
498 if (unlikely(!(dev->flags & IFF_UP))) {
499 kfree_skb(skb);
500 return RX_HANDLER_CONSUMED;
501 }
502 len = skb->len + ETH_HLEN;
503 skb = skb_share_check(skb, GFP_ATOMIC);
504 if (!skb) {
505 ret = NET_RX_DROP;
506 handle_res = RX_HANDLER_CONSUMED;
507 goto out;
508 }
509
510 *pskb = skb;
511 skb->dev = dev;
512 skb->pkt_type = PACKET_HOST;
513
514 ret = NET_RX_SUCCESS;
515 handle_res = RX_HANDLER_ANOTHER;
516out:
517 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
518 return handle_res;
519}
520
521static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
522{
523 const struct macvlan_dev *vlan = netdev_priv(dev);
524 const struct macvlan_port *port = vlan->port;
525 const struct macvlan_dev *dest;
526
527 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
528 const struct ethhdr *eth = skb_eth_hdr(skb);
529
530 /* send to other bridge ports directly */
531 if (is_multicast_ether_addr(eth->h_dest)) {
532 skb_reset_mac_header(skb);
533 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
534 goto xmit_world;
535 }
536
537 dest = macvlan_hash_lookup(port, eth->h_dest);
538 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
539 /* send to lowerdev first for its network taps */
540 dev_forward_skb(vlan->lowerdev, skb);
541
542 return NET_XMIT_SUCCESS;
543 }
544 }
545xmit_world:
546 skb->dev = vlan->lowerdev;
547 return dev_queue_xmit_accel(skb,
548 netdev_get_sb_channel(dev) ? dev : NULL);
549}
550
551static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
552{
553#ifdef CONFIG_NET_POLL_CONTROLLER
554 return netpoll_send_skb(vlan->netpoll, skb);
555#else
556 BUG();
557 return NETDEV_TX_OK;
558#endif
559}
560
561static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
562 struct net_device *dev)
563{
564 struct macvlan_dev *vlan = netdev_priv(dev);
565 unsigned int len = skb->len;
566 int ret;
567
568 if (unlikely(netpoll_tx_running(dev)))
569 return macvlan_netpoll_send_skb(vlan, skb);
570
571 ret = macvlan_queue_xmit(skb, dev);
572
573 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
574 struct vlan_pcpu_stats *pcpu_stats;
575
576 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
577 u64_stats_update_begin(&pcpu_stats->syncp);
578 u64_stats_inc(&pcpu_stats->tx_packets);
579 u64_stats_add(&pcpu_stats->tx_bytes, len);
580 u64_stats_update_end(&pcpu_stats->syncp);
581 } else {
582 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
583 }
584 return ret;
585}
586
587static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
588 unsigned short type, const void *daddr,
589 const void *saddr, unsigned len)
590{
591 const struct macvlan_dev *vlan = netdev_priv(dev);
592 struct net_device *lowerdev = vlan->lowerdev;
593
594 return dev_hard_header(skb, lowerdev, type, daddr,
595 saddr ? : dev->dev_addr, len);
596}
597
598static const struct header_ops macvlan_hard_header_ops = {
599 .create = macvlan_hard_header,
600 .parse = eth_header_parse,
601 .cache = eth_header_cache,
602 .cache_update = eth_header_cache_update,
603};
604
605static int macvlan_open(struct net_device *dev)
606{
607 struct macvlan_dev *vlan = netdev_priv(dev);
608 struct net_device *lowerdev = vlan->lowerdev;
609 int err;
610
611 if (macvlan_passthru(vlan->port)) {
612 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
613 err = dev_set_promiscuity(lowerdev, 1);
614 if (err < 0)
615 goto out;
616 }
617 goto hash_add;
618 }
619
620 err = -EADDRINUSE;
621 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
622 goto out;
623
624 /* Attempt to populate accel_priv which is used to offload the L2
625 * forwarding requests for unicast packets.
626 */
627 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
628 vlan->accel_priv =
629 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
630
631 /* If earlier attempt to offload failed, or accel_priv is not
632 * populated we must add the unicast address to the lower device.
633 */
634 if (IS_ERR_OR_NULL(vlan->accel_priv)) {
635 vlan->accel_priv = NULL;
636 err = dev_uc_add(lowerdev, dev->dev_addr);
637 if (err < 0)
638 goto out;
639 }
640
641 if (dev->flags & IFF_ALLMULTI) {
642 err = dev_set_allmulti(lowerdev, 1);
643 if (err < 0)
644 goto del_unicast;
645 }
646
647 if (dev->flags & IFF_PROMISC) {
648 err = dev_set_promiscuity(lowerdev, 1);
649 if (err < 0)
650 goto clear_multi;
651 }
652
653hash_add:
654 macvlan_hash_add(vlan);
655 return 0;
656
657clear_multi:
658 if (dev->flags & IFF_ALLMULTI)
659 dev_set_allmulti(lowerdev, -1);
660del_unicast:
661 if (vlan->accel_priv) {
662 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
663 vlan->accel_priv);
664 vlan->accel_priv = NULL;
665 } else {
666 dev_uc_del(lowerdev, dev->dev_addr);
667 }
668out:
669 return err;
670}
671
672static int macvlan_stop(struct net_device *dev)
673{
674 struct macvlan_dev *vlan = netdev_priv(dev);
675 struct net_device *lowerdev = vlan->lowerdev;
676
677 if (vlan->accel_priv) {
678 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
679 vlan->accel_priv);
680 vlan->accel_priv = NULL;
681 }
682
683 dev_uc_unsync(lowerdev, dev);
684 dev_mc_unsync(lowerdev, dev);
685
686 if (macvlan_passthru(vlan->port)) {
687 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
688 dev_set_promiscuity(lowerdev, -1);
689 goto hash_del;
690 }
691
692 if (dev->flags & IFF_ALLMULTI)
693 dev_set_allmulti(lowerdev, -1);
694
695 if (dev->flags & IFF_PROMISC)
696 dev_set_promiscuity(lowerdev, -1);
697
698 dev_uc_del(lowerdev, dev->dev_addr);
699
700hash_del:
701 macvlan_hash_del(vlan, !dev->dismantle);
702 return 0;
703}
704
705static int macvlan_sync_address(struct net_device *dev,
706 const unsigned char *addr)
707{
708 struct macvlan_dev *vlan = netdev_priv(dev);
709 struct net_device *lowerdev = vlan->lowerdev;
710 struct macvlan_port *port = vlan->port;
711 int err;
712
713 if (!(dev->flags & IFF_UP)) {
714 /* Just copy in the new address */
715 eth_hw_addr_set(dev, addr);
716 } else {
717 /* Rehash and update the device filters */
718 if (macvlan_addr_busy(vlan->port, addr))
719 return -EADDRINUSE;
720
721 if (!macvlan_passthru(port)) {
722 err = dev_uc_add(lowerdev, addr);
723 if (err)
724 return err;
725
726 dev_uc_del(lowerdev, dev->dev_addr);
727 }
728
729 macvlan_hash_change_addr(vlan, addr);
730 }
731 if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
732 /* Since addr_change isn't set, we are here due to lower
733 * device change. Save the lower-dev address so we can
734 * restore it later.
735 */
736 ether_addr_copy(vlan->port->perm_addr,
737 lowerdev->dev_addr);
738 }
739 macvlan_clear_addr_change(port);
740 return 0;
741}
742
743static int macvlan_set_mac_address(struct net_device *dev, void *p)
744{
745 struct macvlan_dev *vlan = netdev_priv(dev);
746 struct sockaddr *addr = p;
747
748 if (!is_valid_ether_addr(addr->sa_data))
749 return -EADDRNOTAVAIL;
750
751 /* If the addresses are the same, this is a no-op */
752 if (ether_addr_equal(dev->dev_addr, addr->sa_data))
753 return 0;
754
755 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
756 macvlan_set_addr_change(vlan->port);
757 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
758 }
759
760 if (macvlan_addr_busy(vlan->port, addr->sa_data))
761 return -EADDRINUSE;
762
763 return macvlan_sync_address(dev, addr->sa_data);
764}
765
766static void macvlan_change_rx_flags(struct net_device *dev, int change)
767{
768 struct macvlan_dev *vlan = netdev_priv(dev);
769 struct net_device *lowerdev = vlan->lowerdev;
770
771 if (dev->flags & IFF_UP) {
772 if (change & IFF_ALLMULTI)
773 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
774 if (change & IFF_PROMISC)
775 dev_set_promiscuity(lowerdev,
776 dev->flags & IFF_PROMISC ? 1 : -1);
777
778 }
779}
780
781static void macvlan_compute_filter(unsigned long *mc_filter,
782 struct net_device *dev,
783 struct macvlan_dev *vlan)
784{
785 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
786 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
787 } else {
788 struct netdev_hw_addr *ha;
789 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
790
791 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
792 netdev_for_each_mc_addr(ha, dev) {
793 __set_bit(mc_hash(vlan, ha->addr), filter);
794 }
795
796 __set_bit(mc_hash(vlan, dev->broadcast), filter);
797
798 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
799 }
800}
801
802static void macvlan_set_mac_lists(struct net_device *dev)
803{
804 struct macvlan_dev *vlan = netdev_priv(dev);
805
806 macvlan_compute_filter(vlan->mc_filter, dev, vlan);
807
808 dev_uc_sync(vlan->lowerdev, dev);
809 dev_mc_sync(vlan->lowerdev, dev);
810
811 /* This is slightly inaccurate as we're including the subscription
812 * list of vlan->lowerdev too.
813 *
814 * Bug alert: This only works if everyone has the same broadcast
815 * address as lowerdev. As soon as someone changes theirs this
816 * will break.
817 *
818 * However, this is already broken as when you change your broadcast
819 * address we don't get called.
820 *
821 * The solution is to maintain a list of broadcast addresses like
822 * we do for uc/mc, if you care.
823 */
824 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
825}
826
827static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
828{
829 struct macvlan_dev *vlan = netdev_priv(dev);
830
831 if (vlan->lowerdev->mtu < new_mtu)
832 return -EINVAL;
833 dev->mtu = new_mtu;
834 return 0;
835}
836
837static int macvlan_eth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
838{
839 struct net_device *real_dev = macvlan_dev_real_dev(dev);
840 const struct net_device_ops *ops = real_dev->netdev_ops;
841 struct ifreq ifrr;
842 int err = -EOPNOTSUPP;
843
844 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
845 ifrr.ifr_ifru = ifr->ifr_ifru;
846
847 switch (cmd) {
848 case SIOCSHWTSTAMP:
849 if (!net_eq(dev_net(dev), &init_net))
850 break;
851 fallthrough;
852 case SIOCGHWTSTAMP:
853 if (netif_device_present(real_dev) && ops->ndo_eth_ioctl)
854 err = ops->ndo_eth_ioctl(real_dev, &ifrr, cmd);
855 break;
856 }
857
858 if (!err)
859 ifr->ifr_ifru = ifrr.ifr_ifru;
860
861 return err;
862}
863
864/*
865 * macvlan network devices have devices nesting below it and are a special
866 * "super class" of normal network devices; split their locks off into a
867 * separate class since they always nest.
868 */
869static struct lock_class_key macvlan_netdev_addr_lock_key;
870
871#define ALWAYS_ON_OFFLOADS \
872 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
873 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
874
875#define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
876
877#define MACVLAN_FEATURES \
878 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
879 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
880 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
881 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
882
883#define MACVLAN_STATE_MASK \
884 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
885
886static void macvlan_set_lockdep_class(struct net_device *dev)
887{
888 netdev_lockdep_set_classes(dev);
889 lockdep_set_class(&dev->addr_list_lock,
890 &macvlan_netdev_addr_lock_key);
891}
892
893static int macvlan_init(struct net_device *dev)
894{
895 struct macvlan_dev *vlan = netdev_priv(dev);
896 struct net_device *lowerdev = vlan->lowerdev;
897 struct macvlan_port *port = vlan->port;
898
899 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
900 (lowerdev->state & MACVLAN_STATE_MASK);
901 dev->features = lowerdev->features & MACVLAN_FEATURES;
902 dev->features |= ALWAYS_ON_FEATURES;
903 dev->hw_features |= NETIF_F_LRO;
904 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
905 dev->vlan_features |= ALWAYS_ON_OFFLOADS;
906 dev->hw_enc_features |= dev->features;
907 netif_inherit_tso_max(dev, lowerdev);
908 dev->hard_header_len = lowerdev->hard_header_len;
909 macvlan_set_lockdep_class(dev);
910
911 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
912 if (!vlan->pcpu_stats)
913 return -ENOMEM;
914
915 port->count += 1;
916
917 /* Get macvlan's reference to lowerdev */
918 netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
919
920 return 0;
921}
922
923static void macvlan_uninit(struct net_device *dev)
924{
925 struct macvlan_dev *vlan = netdev_priv(dev);
926 struct macvlan_port *port = vlan->port;
927
928 free_percpu(vlan->pcpu_stats);
929
930 macvlan_flush_sources(port, vlan);
931 port->count -= 1;
932 if (!port->count)
933 macvlan_port_destroy(port->dev);
934}
935
936static void macvlan_dev_get_stats64(struct net_device *dev,
937 struct rtnl_link_stats64 *stats)
938{
939 struct macvlan_dev *vlan = netdev_priv(dev);
940
941 if (vlan->pcpu_stats) {
942 struct vlan_pcpu_stats *p;
943 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
944 u32 rx_errors = 0, tx_dropped = 0;
945 unsigned int start;
946 int i;
947
948 for_each_possible_cpu(i) {
949 p = per_cpu_ptr(vlan->pcpu_stats, i);
950 do {
951 start = u64_stats_fetch_begin(&p->syncp);
952 rx_packets = u64_stats_read(&p->rx_packets);
953 rx_bytes = u64_stats_read(&p->rx_bytes);
954 rx_multicast = u64_stats_read(&p->rx_multicast);
955 tx_packets = u64_stats_read(&p->tx_packets);
956 tx_bytes = u64_stats_read(&p->tx_bytes);
957 } while (u64_stats_fetch_retry(&p->syncp, start));
958
959 stats->rx_packets += rx_packets;
960 stats->rx_bytes += rx_bytes;
961 stats->multicast += rx_multicast;
962 stats->tx_packets += tx_packets;
963 stats->tx_bytes += tx_bytes;
964 /* rx_errors & tx_dropped are u32, updated
965 * without syncp protection.
966 */
967 rx_errors += READ_ONCE(p->rx_errors);
968 tx_dropped += READ_ONCE(p->tx_dropped);
969 }
970 stats->rx_errors = rx_errors;
971 stats->rx_dropped = rx_errors;
972 stats->tx_dropped = tx_dropped;
973 }
974}
975
976static int macvlan_vlan_rx_add_vid(struct net_device *dev,
977 __be16 proto, u16 vid)
978{
979 struct macvlan_dev *vlan = netdev_priv(dev);
980 struct net_device *lowerdev = vlan->lowerdev;
981
982 return vlan_vid_add(lowerdev, proto, vid);
983}
984
985static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
986 __be16 proto, u16 vid)
987{
988 struct macvlan_dev *vlan = netdev_priv(dev);
989 struct net_device *lowerdev = vlan->lowerdev;
990
991 vlan_vid_del(lowerdev, proto, vid);
992 return 0;
993}
994
995static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
996 struct net_device *dev,
997 const unsigned char *addr, u16 vid,
998 u16 flags,
999 struct netlink_ext_ack *extack)
1000{
1001 struct macvlan_dev *vlan = netdev_priv(dev);
1002 int err = -EINVAL;
1003
1004 /* Support unicast filter only on passthru devices.
1005 * Multicast filter should be allowed on all devices.
1006 */
1007 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1008 return -EOPNOTSUPP;
1009
1010 if (flags & NLM_F_REPLACE)
1011 return -EOPNOTSUPP;
1012
1013 if (is_unicast_ether_addr(addr))
1014 err = dev_uc_add_excl(dev, addr);
1015 else if (is_multicast_ether_addr(addr))
1016 err = dev_mc_add_excl(dev, addr);
1017
1018 return err;
1019}
1020
1021static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1022 struct net_device *dev,
1023 const unsigned char *addr, u16 vid,
1024 struct netlink_ext_ack *extack)
1025{
1026 struct macvlan_dev *vlan = netdev_priv(dev);
1027 int err = -EINVAL;
1028
1029 /* Support unicast filter only on passthru devices.
1030 * Multicast filter should be allowed on all devices.
1031 */
1032 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1033 return -EOPNOTSUPP;
1034
1035 if (is_unicast_ether_addr(addr))
1036 err = dev_uc_del(dev, addr);
1037 else if (is_multicast_ether_addr(addr))
1038 err = dev_mc_del(dev, addr);
1039
1040 return err;
1041}
1042
1043static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1044 struct ethtool_drvinfo *drvinfo)
1045{
1046 strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1047 strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1048}
1049
1050static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1051 struct ethtool_link_ksettings *cmd)
1052{
1053 const struct macvlan_dev *vlan = netdev_priv(dev);
1054
1055 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1056}
1057
1058static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1059 struct ethtool_ts_info *info)
1060{
1061 struct net_device *real_dev = macvlan_dev_real_dev(dev);
1062 const struct ethtool_ops *ops = real_dev->ethtool_ops;
1063 struct phy_device *phydev = real_dev->phydev;
1064
1065 if (phy_has_tsinfo(phydev)) {
1066 return phy_ts_info(phydev, info);
1067 } else if (ops->get_ts_info) {
1068 return ops->get_ts_info(real_dev, info);
1069 } else {
1070 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1071 SOF_TIMESTAMPING_SOFTWARE;
1072 info->phc_index = -1;
1073 }
1074
1075 return 0;
1076}
1077
1078static netdev_features_t macvlan_fix_features(struct net_device *dev,
1079 netdev_features_t features)
1080{
1081 struct macvlan_dev *vlan = netdev_priv(dev);
1082 netdev_features_t lowerdev_features = vlan->lowerdev->features;
1083 netdev_features_t mask;
1084
1085 features |= NETIF_F_ALL_FOR_ALL;
1086 features &= (vlan->set_features | ~MACVLAN_FEATURES);
1087 mask = features;
1088
1089 lowerdev_features &= (features | ~NETIF_F_LRO);
1090 features = netdev_increment_features(lowerdev_features, features, mask);
1091 features |= ALWAYS_ON_FEATURES;
1092 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1093
1094 return features;
1095}
1096
1097#ifdef CONFIG_NET_POLL_CONTROLLER
1098static void macvlan_dev_poll_controller(struct net_device *dev)
1099{
1100 return;
1101}
1102
1103static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1104{
1105 struct macvlan_dev *vlan = netdev_priv(dev);
1106 struct net_device *real_dev = vlan->lowerdev;
1107 struct netpoll *netpoll;
1108 int err;
1109
1110 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1111 err = -ENOMEM;
1112 if (!netpoll)
1113 goto out;
1114
1115 err = __netpoll_setup(netpoll, real_dev);
1116 if (err) {
1117 kfree(netpoll);
1118 goto out;
1119 }
1120
1121 vlan->netpoll = netpoll;
1122
1123out:
1124 return err;
1125}
1126
1127static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1128{
1129 struct macvlan_dev *vlan = netdev_priv(dev);
1130 struct netpoll *netpoll = vlan->netpoll;
1131
1132 if (!netpoll)
1133 return;
1134
1135 vlan->netpoll = NULL;
1136
1137 __netpoll_free(netpoll);
1138}
1139#endif /* CONFIG_NET_POLL_CONTROLLER */
1140
1141static int macvlan_dev_get_iflink(const struct net_device *dev)
1142{
1143 struct macvlan_dev *vlan = netdev_priv(dev);
1144
1145 return vlan->lowerdev->ifindex;
1146}
1147
1148static const struct ethtool_ops macvlan_ethtool_ops = {
1149 .get_link = ethtool_op_get_link,
1150 .get_link_ksettings = macvlan_ethtool_get_link_ksettings,
1151 .get_drvinfo = macvlan_ethtool_get_drvinfo,
1152 .get_ts_info = macvlan_ethtool_get_ts_info,
1153};
1154
1155static const struct net_device_ops macvlan_netdev_ops = {
1156 .ndo_init = macvlan_init,
1157 .ndo_uninit = macvlan_uninit,
1158 .ndo_open = macvlan_open,
1159 .ndo_stop = macvlan_stop,
1160 .ndo_start_xmit = macvlan_start_xmit,
1161 .ndo_change_mtu = macvlan_change_mtu,
1162 .ndo_eth_ioctl = macvlan_eth_ioctl,
1163 .ndo_fix_features = macvlan_fix_features,
1164 .ndo_change_rx_flags = macvlan_change_rx_flags,
1165 .ndo_set_mac_address = macvlan_set_mac_address,
1166 .ndo_set_rx_mode = macvlan_set_mac_lists,
1167 .ndo_get_stats64 = macvlan_dev_get_stats64,
1168 .ndo_validate_addr = eth_validate_addr,
1169 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
1170 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
1171 .ndo_fdb_add = macvlan_fdb_add,
1172 .ndo_fdb_del = macvlan_fdb_del,
1173 .ndo_fdb_dump = ndo_dflt_fdb_dump,
1174#ifdef CONFIG_NET_POLL_CONTROLLER
1175 .ndo_poll_controller = macvlan_dev_poll_controller,
1176 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
1177 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
1178#endif
1179 .ndo_get_iflink = macvlan_dev_get_iflink,
1180 .ndo_features_check = passthru_features_check,
1181};
1182
1183static void macvlan_dev_free(struct net_device *dev)
1184{
1185 struct macvlan_dev *vlan = netdev_priv(dev);
1186
1187 /* Get rid of the macvlan's reference to lowerdev */
1188 netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1189}
1190
1191void macvlan_common_setup(struct net_device *dev)
1192{
1193 ether_setup(dev);
1194
1195 /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1196 dev->max_mtu = ETH_MAX_MTU;
1197 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1198 netif_keep_dst(dev);
1199 dev->priv_flags |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN;
1200 dev->netdev_ops = &macvlan_netdev_ops;
1201 dev->needs_free_netdev = true;
1202 dev->priv_destructor = macvlan_dev_free;
1203 dev->header_ops = &macvlan_hard_header_ops;
1204 dev->ethtool_ops = &macvlan_ethtool_ops;
1205}
1206EXPORT_SYMBOL_GPL(macvlan_common_setup);
1207
1208static void macvlan_setup(struct net_device *dev)
1209{
1210 macvlan_common_setup(dev);
1211 dev->priv_flags |= IFF_NO_QUEUE;
1212}
1213
1214static int macvlan_port_create(struct net_device *dev)
1215{
1216 struct macvlan_port *port;
1217 unsigned int i;
1218 int err;
1219
1220 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1221 return -EINVAL;
1222
1223 if (netdev_is_rx_handler_busy(dev))
1224 return -EBUSY;
1225
1226 port = kzalloc(sizeof(*port), GFP_KERNEL);
1227 if (port == NULL)
1228 return -ENOMEM;
1229
1230 port->dev = dev;
1231 ether_addr_copy(port->perm_addr, dev->dev_addr);
1232 INIT_LIST_HEAD(&port->vlans);
1233 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1234 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1235 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1236 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1237
1238 port->bc_queue_len_used = 0;
1239 skb_queue_head_init(&port->bc_queue);
1240 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1241
1242 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1243 if (err)
1244 kfree(port);
1245 else
1246 dev->priv_flags |= IFF_MACVLAN_PORT;
1247 return err;
1248}
1249
1250static void macvlan_port_destroy(struct net_device *dev)
1251{
1252 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1253 struct sk_buff *skb;
1254
1255 dev->priv_flags &= ~IFF_MACVLAN_PORT;
1256 netdev_rx_handler_unregister(dev);
1257
1258 /* After this point, no packet can schedule bc_work anymore,
1259 * but we need to cancel it and purge left skbs if any.
1260 */
1261 cancel_work_sync(&port->bc_work);
1262
1263 while ((skb = __skb_dequeue(&port->bc_queue))) {
1264 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1265
1266 if (src)
1267 dev_put(src->dev);
1268
1269 kfree_skb(skb);
1270 }
1271
1272 /* If the lower device address has been changed by passthru
1273 * macvlan, put it back.
1274 */
1275 if (macvlan_passthru(port) &&
1276 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1277 struct sockaddr sa;
1278
1279 sa.sa_family = port->dev->type;
1280 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1281 dev_set_mac_address(port->dev, &sa, NULL);
1282 }
1283
1284 kfree(port);
1285}
1286
1287static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1288 struct netlink_ext_ack *extack)
1289{
1290 struct nlattr *nla, *head;
1291 int rem, len;
1292
1293 if (tb[IFLA_ADDRESS]) {
1294 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1295 return -EINVAL;
1296 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1297 return -EADDRNOTAVAIL;
1298 }
1299
1300 if (!data)
1301 return 0;
1302
1303 if (data[IFLA_MACVLAN_FLAGS] &&
1304 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1305 MACVLAN_FLAG_NODST))
1306 return -EINVAL;
1307
1308 if (data[IFLA_MACVLAN_MODE]) {
1309 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1310 case MACVLAN_MODE_PRIVATE:
1311 case MACVLAN_MODE_VEPA:
1312 case MACVLAN_MODE_BRIDGE:
1313 case MACVLAN_MODE_PASSTHRU:
1314 case MACVLAN_MODE_SOURCE:
1315 break;
1316 default:
1317 return -EINVAL;
1318 }
1319 }
1320
1321 if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1322 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1323 case MACVLAN_MACADDR_ADD:
1324 case MACVLAN_MACADDR_DEL:
1325 case MACVLAN_MACADDR_FLUSH:
1326 case MACVLAN_MACADDR_SET:
1327 break;
1328 default:
1329 return -EINVAL;
1330 }
1331 }
1332
1333 if (data[IFLA_MACVLAN_MACADDR]) {
1334 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1335 return -EINVAL;
1336
1337 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1338 return -EADDRNOTAVAIL;
1339 }
1340
1341 if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1342 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1343 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1344
1345 nla_for_each_attr(nla, head, len, rem) {
1346 if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1347 nla_len(nla) != ETH_ALEN)
1348 return -EINVAL;
1349
1350 if (!is_valid_ether_addr(nla_data(nla)))
1351 return -EADDRNOTAVAIL;
1352 }
1353 }
1354
1355 if (data[IFLA_MACVLAN_MACADDR_COUNT])
1356 return -EINVAL;
1357
1358 return 0;
1359}
1360
1361/*
1362 * reconfigure list of remote source mac address
1363 * (only for macvlan devices in source mode)
1364 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1365 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1366 */
1367static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1368 struct nlattr *data[])
1369{
1370 char *addr = NULL;
1371 int ret, rem, len;
1372 struct nlattr *nla, *head;
1373 struct macvlan_source_entry *entry;
1374
1375 if (data[IFLA_MACVLAN_MACADDR])
1376 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1377
1378 if (mode == MACVLAN_MACADDR_ADD) {
1379 if (!addr)
1380 return -EINVAL;
1381
1382 return macvlan_hash_add_source(vlan, addr);
1383
1384 } else if (mode == MACVLAN_MACADDR_DEL) {
1385 if (!addr)
1386 return -EINVAL;
1387
1388 entry = macvlan_hash_lookup_source(vlan, addr);
1389 if (entry) {
1390 macvlan_hash_del_source(entry);
1391 vlan->macaddr_count--;
1392 }
1393 } else if (mode == MACVLAN_MACADDR_FLUSH) {
1394 macvlan_flush_sources(vlan->port, vlan);
1395 } else if (mode == MACVLAN_MACADDR_SET) {
1396 macvlan_flush_sources(vlan->port, vlan);
1397
1398 if (addr) {
1399 ret = macvlan_hash_add_source(vlan, addr);
1400 if (ret)
1401 return ret;
1402 }
1403
1404 if (!data[IFLA_MACVLAN_MACADDR_DATA])
1405 return 0;
1406
1407 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1408 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1409
1410 nla_for_each_attr(nla, head, len, rem) {
1411 addr = nla_data(nla);
1412 ret = macvlan_hash_add_source(vlan, addr);
1413 if (ret)
1414 return ret;
1415 }
1416 } else {
1417 return -EINVAL;
1418 }
1419
1420 return 0;
1421}
1422
1423int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1424 struct nlattr *tb[], struct nlattr *data[],
1425 struct netlink_ext_ack *extack)
1426{
1427 struct macvlan_dev *vlan = netdev_priv(dev);
1428 struct macvlan_port *port;
1429 struct net_device *lowerdev;
1430 int err;
1431 int macmode;
1432 bool create = false;
1433
1434 if (!tb[IFLA_LINK])
1435 return -EINVAL;
1436
1437 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1438 if (lowerdev == NULL)
1439 return -ENODEV;
1440
1441 /* When creating macvlans or macvtaps on top of other macvlans - use
1442 * the real device as the lowerdev.
1443 */
1444 if (netif_is_macvlan(lowerdev))
1445 lowerdev = macvlan_dev_real_dev(lowerdev);
1446
1447 if (!tb[IFLA_MTU])
1448 dev->mtu = lowerdev->mtu;
1449 else if (dev->mtu > lowerdev->mtu)
1450 return -EINVAL;
1451
1452 /* MTU range: 68 - lowerdev->max_mtu */
1453 dev->min_mtu = ETH_MIN_MTU;
1454 dev->max_mtu = lowerdev->max_mtu;
1455
1456 if (!tb[IFLA_ADDRESS])
1457 eth_hw_addr_random(dev);
1458
1459 if (!netif_is_macvlan_port(lowerdev)) {
1460 err = macvlan_port_create(lowerdev);
1461 if (err < 0)
1462 return err;
1463 create = true;
1464 }
1465 port = macvlan_port_get_rtnl(lowerdev);
1466
1467 /* Only 1 macvlan device can be created in passthru mode */
1468 if (macvlan_passthru(port)) {
1469 /* The macvlan port must be not created this time,
1470 * still goto destroy_macvlan_port for readability.
1471 */
1472 err = -EINVAL;
1473 goto destroy_macvlan_port;
1474 }
1475
1476 vlan->lowerdev = lowerdev;
1477 vlan->dev = dev;
1478 vlan->port = port;
1479 vlan->set_features = MACVLAN_FEATURES;
1480
1481 vlan->mode = MACVLAN_MODE_VEPA;
1482 if (data && data[IFLA_MACVLAN_MODE])
1483 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1484
1485 if (data && data[IFLA_MACVLAN_FLAGS])
1486 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1487
1488 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1489 if (port->count) {
1490 err = -EINVAL;
1491 goto destroy_macvlan_port;
1492 }
1493 macvlan_set_passthru(port);
1494 eth_hw_addr_inherit(dev, lowerdev);
1495 }
1496
1497 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1498 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1499 err = -EINVAL;
1500 goto destroy_macvlan_port;
1501 }
1502 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1503 err = macvlan_changelink_sources(vlan, macmode, data);
1504 if (err)
1505 goto destroy_macvlan_port;
1506 }
1507
1508 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1509 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1510 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1511
1512 err = register_netdevice(dev);
1513 if (err < 0)
1514 goto destroy_macvlan_port;
1515
1516 dev->priv_flags |= IFF_MACVLAN;
1517 err = netdev_upper_dev_link(lowerdev, dev, extack);
1518 if (err)
1519 goto unregister_netdev;
1520
1521 list_add_tail_rcu(&vlan->list, &port->vlans);
1522 update_port_bc_queue_len(vlan->port);
1523 netif_stacked_transfer_operstate(lowerdev, dev);
1524 linkwatch_fire_event(dev);
1525
1526 return 0;
1527
1528unregister_netdev:
1529 /* macvlan_uninit would free the macvlan port */
1530 unregister_netdevice(dev);
1531 return err;
1532destroy_macvlan_port:
1533 /* the macvlan port may be freed by macvlan_uninit when fail to register.
1534 * so we destroy the macvlan port only when it's valid.
1535 */
1536 if (create && macvlan_port_get_rtnl(lowerdev)) {
1537 macvlan_flush_sources(port, vlan);
1538 macvlan_port_destroy(port->dev);
1539 }
1540 return err;
1541}
1542EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1543
1544static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1545 struct nlattr *tb[], struct nlattr *data[],
1546 struct netlink_ext_ack *extack)
1547{
1548 return macvlan_common_newlink(src_net, dev, tb, data, extack);
1549}
1550
1551void macvlan_dellink(struct net_device *dev, struct list_head *head)
1552{
1553 struct macvlan_dev *vlan = netdev_priv(dev);
1554
1555 if (vlan->mode == MACVLAN_MODE_SOURCE)
1556 macvlan_flush_sources(vlan->port, vlan);
1557 list_del_rcu(&vlan->list);
1558 update_port_bc_queue_len(vlan->port);
1559 unregister_netdevice_queue(dev, head);
1560 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1561}
1562EXPORT_SYMBOL_GPL(macvlan_dellink);
1563
1564static int macvlan_changelink(struct net_device *dev,
1565 struct nlattr *tb[], struct nlattr *data[],
1566 struct netlink_ext_ack *extack)
1567{
1568 struct macvlan_dev *vlan = netdev_priv(dev);
1569 enum macvlan_mode mode;
1570 bool set_mode = false;
1571 enum macvlan_macaddr_mode macmode;
1572 int ret;
1573
1574 /* Validate mode, but don't set yet: setting flags may fail. */
1575 if (data && data[IFLA_MACVLAN_MODE]) {
1576 set_mode = true;
1577 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1578 /* Passthrough mode can't be set or cleared dynamically */
1579 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1580 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1581 return -EINVAL;
1582 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1583 vlan->mode != mode)
1584 macvlan_flush_sources(vlan->port, vlan);
1585 }
1586
1587 if (data && data[IFLA_MACVLAN_FLAGS]) {
1588 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1589 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1590 if (macvlan_passthru(vlan->port) && promisc) {
1591 int err;
1592
1593 if (flags & MACVLAN_FLAG_NOPROMISC)
1594 err = dev_set_promiscuity(vlan->lowerdev, -1);
1595 else
1596 err = dev_set_promiscuity(vlan->lowerdev, 1);
1597 if (err < 0)
1598 return err;
1599 }
1600 vlan->flags = flags;
1601 }
1602
1603 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1604 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1605 update_port_bc_queue_len(vlan->port);
1606 }
1607
1608 if (set_mode)
1609 vlan->mode = mode;
1610 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1611 if (vlan->mode != MACVLAN_MODE_SOURCE)
1612 return -EINVAL;
1613 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1614 ret = macvlan_changelink_sources(vlan, macmode, data);
1615 if (ret)
1616 return ret;
1617 }
1618 return 0;
1619}
1620
1621static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1622{
1623 if (vlan->macaddr_count == 0)
1624 return 0;
1625 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1626 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1627}
1628
1629static size_t macvlan_get_size(const struct net_device *dev)
1630{
1631 struct macvlan_dev *vlan = netdev_priv(dev);
1632
1633 return (0
1634 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1635 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1636 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1637 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1638 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1639 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1640 );
1641}
1642
1643static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1644 const struct macvlan_dev *vlan,
1645 const int i)
1646{
1647 struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1648 struct macvlan_source_entry *entry;
1649
1650 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1651 if (entry->vlan != vlan)
1652 continue;
1653 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1654 return 1;
1655 }
1656 return 0;
1657}
1658
1659static int macvlan_fill_info(struct sk_buff *skb,
1660 const struct net_device *dev)
1661{
1662 struct macvlan_dev *vlan = netdev_priv(dev);
1663 struct macvlan_port *port = vlan->port;
1664 int i;
1665 struct nlattr *nest;
1666
1667 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1668 goto nla_put_failure;
1669 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1670 goto nla_put_failure;
1671 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1672 goto nla_put_failure;
1673 if (vlan->macaddr_count > 0) {
1674 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1675 if (nest == NULL)
1676 goto nla_put_failure;
1677
1678 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1679 if (macvlan_fill_info_macaddr(skb, vlan, i))
1680 goto nla_put_failure;
1681 }
1682 nla_nest_end(skb, nest);
1683 }
1684 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1685 goto nla_put_failure;
1686 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1687 goto nla_put_failure;
1688 return 0;
1689
1690nla_put_failure:
1691 return -EMSGSIZE;
1692}
1693
1694static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1695 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1696 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1697 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1698 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1699 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1700 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1701 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1702 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1703};
1704
1705int macvlan_link_register(struct rtnl_link_ops *ops)
1706{
1707 /* common fields */
1708 ops->validate = macvlan_validate;
1709 ops->maxtype = IFLA_MACVLAN_MAX;
1710 ops->policy = macvlan_policy;
1711 ops->changelink = macvlan_changelink;
1712 ops->get_size = macvlan_get_size;
1713 ops->fill_info = macvlan_fill_info;
1714
1715 return rtnl_link_register(ops);
1716};
1717EXPORT_SYMBOL_GPL(macvlan_link_register);
1718
1719static struct net *macvlan_get_link_net(const struct net_device *dev)
1720{
1721 return dev_net(macvlan_dev_real_dev(dev));
1722}
1723
1724static struct rtnl_link_ops macvlan_link_ops = {
1725 .kind = "macvlan",
1726 .setup = macvlan_setup,
1727 .newlink = macvlan_newlink,
1728 .dellink = macvlan_dellink,
1729 .get_link_net = macvlan_get_link_net,
1730 .priv_size = sizeof(struct macvlan_dev),
1731};
1732
1733static void update_port_bc_queue_len(struct macvlan_port *port)
1734{
1735 u32 max_bc_queue_len_req = 0;
1736 struct macvlan_dev *vlan;
1737
1738 list_for_each_entry(vlan, &port->vlans, list) {
1739 if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1740 max_bc_queue_len_req = vlan->bc_queue_len_req;
1741 }
1742 port->bc_queue_len_used = max_bc_queue_len_req;
1743}
1744
1745static int macvlan_device_event(struct notifier_block *unused,
1746 unsigned long event, void *ptr)
1747{
1748 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1749 struct macvlan_dev *vlan, *next;
1750 struct macvlan_port *port;
1751 LIST_HEAD(list_kill);
1752
1753 if (!netif_is_macvlan_port(dev))
1754 return NOTIFY_DONE;
1755
1756 port = macvlan_port_get_rtnl(dev);
1757
1758 switch (event) {
1759 case NETDEV_UP:
1760 case NETDEV_DOWN:
1761 case NETDEV_CHANGE:
1762 list_for_each_entry(vlan, &port->vlans, list)
1763 netif_stacked_transfer_operstate(vlan->lowerdev,
1764 vlan->dev);
1765 break;
1766 case NETDEV_FEAT_CHANGE:
1767 list_for_each_entry(vlan, &port->vlans, list) {
1768 netif_inherit_tso_max(vlan->dev, dev);
1769 netdev_update_features(vlan->dev);
1770 }
1771 break;
1772 case NETDEV_CHANGEMTU:
1773 list_for_each_entry(vlan, &port->vlans, list) {
1774 if (vlan->dev->mtu <= dev->mtu)
1775 continue;
1776 dev_set_mtu(vlan->dev, dev->mtu);
1777 }
1778 break;
1779 case NETDEV_CHANGEADDR:
1780 if (!macvlan_passthru(port))
1781 return NOTIFY_DONE;
1782
1783 vlan = list_first_entry_or_null(&port->vlans,
1784 struct macvlan_dev,
1785 list);
1786
1787 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1788 return NOTIFY_BAD;
1789
1790 break;
1791 case NETDEV_UNREGISTER:
1792 /* twiddle thumbs on netns device moves */
1793 if (dev->reg_state != NETREG_UNREGISTERING)
1794 break;
1795
1796 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1797 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1798 unregister_netdevice_many(&list_kill);
1799 break;
1800 case NETDEV_PRE_TYPE_CHANGE:
1801 /* Forbid underlying device to change its type. */
1802 return NOTIFY_BAD;
1803
1804 case NETDEV_NOTIFY_PEERS:
1805 case NETDEV_BONDING_FAILOVER:
1806 case NETDEV_RESEND_IGMP:
1807 /* Propagate to all vlans */
1808 list_for_each_entry(vlan, &port->vlans, list)
1809 call_netdevice_notifiers(event, vlan->dev);
1810 }
1811 return NOTIFY_DONE;
1812}
1813
1814static struct notifier_block macvlan_notifier_block __read_mostly = {
1815 .notifier_call = macvlan_device_event,
1816};
1817
1818static int __init macvlan_init_module(void)
1819{
1820 int err;
1821
1822 register_netdevice_notifier(&macvlan_notifier_block);
1823
1824 err = macvlan_link_register(&macvlan_link_ops);
1825 if (err < 0)
1826 goto err1;
1827 return 0;
1828err1:
1829 unregister_netdevice_notifier(&macvlan_notifier_block);
1830 return err;
1831}
1832
1833static void __exit macvlan_cleanup_module(void)
1834{
1835 rtnl_link_unregister(&macvlan_link_ops);
1836 unregister_netdevice_notifier(&macvlan_notifier_block);
1837}
1838
1839module_init(macvlan_init_module);
1840module_exit(macvlan_cleanup_module);
1841
1842MODULE_LICENSE("GPL");
1843MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1844MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1845MODULE_ALIAS_RTNL_LINK("macvlan");