]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - drivers/net/macvlan.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'batman-adv-fix-for-davem' of git://git.open-mesh.org/linux-merge
[mirror_ubuntu-jammy-kernel.git] / drivers / net / macvlan.c
1 /*
2 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of
7 * the License, or (at your option) any later version.
8 *
9 * The code this is based on carried the following copyright notice:
10 * ---
11 * (C) Copyright 2001-2006
12 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
13 * Re-worked by Ben Greear <greearb@candelatech.com>
14 * ---
15 */
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/init.h>
20 #include <linux/errno.h>
21 #include <linux/slab.h>
22 #include <linux/string.h>
23 #include <linux/rculist.h>
24 #include <linux/notifier.h>
25 #include <linux/netdevice.h>
26 #include <linux/etherdevice.h>
27 #include <linux/ethtool.h>
28 #include <linux/if_arp.h>
29 #include <linux/if_vlan.h>
30 #include <linux/if_link.h>
31 #include <linux/if_macvlan.h>
32 #include <linux/hash.h>
33 #include <net/rtnetlink.h>
34 #include <net/xfrm.h>
35
36 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE)
37
38 struct macvlan_port {
39 struct net_device *dev;
40 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
41 struct list_head vlans;
42 struct rcu_head rcu;
43 bool passthru;
44 int count;
45 };
46
47 static void macvlan_port_destroy(struct net_device *dev);
48
49 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
50 {
51 return rcu_dereference(dev->rx_handler_data);
52 }
53
54 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
55 {
56 return rtnl_dereference(dev->rx_handler_data);
57 }
58
59 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
60
61 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
62 const unsigned char *addr)
63 {
64 struct macvlan_dev *vlan;
65
66 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) {
67 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
68 return vlan;
69 }
70 return NULL;
71 }
72
73 static void macvlan_hash_add(struct macvlan_dev *vlan)
74 {
75 struct macvlan_port *port = vlan->port;
76 const unsigned char *addr = vlan->dev->dev_addr;
77
78 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
79 }
80
81 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
82 {
83 hlist_del_rcu(&vlan->hlist);
84 if (sync)
85 synchronize_rcu();
86 }
87
88 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
89 const unsigned char *addr)
90 {
91 macvlan_hash_del(vlan, true);
92 /* Now that we are unhashed it is safe to change the device
93 * address without confusing packet delivery.
94 */
95 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
96 macvlan_hash_add(vlan);
97 }
98
99 static int macvlan_addr_busy(const struct macvlan_port *port,
100 const unsigned char *addr)
101 {
102 /* Test to see if the specified multicast address is
103 * currently in use by the underlying device or
104 * another macvlan.
105 */
106 if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
107 return 1;
108
109 if (macvlan_hash_lookup(port, addr))
110 return 1;
111
112 return 0;
113 }
114
115
116 static int macvlan_broadcast_one(struct sk_buff *skb,
117 const struct macvlan_dev *vlan,
118 const struct ethhdr *eth, bool local)
119 {
120 struct net_device *dev = vlan->dev;
121
122 if (local)
123 return dev_forward_skb(dev, skb);
124
125 skb->dev = dev;
126 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
127 skb->pkt_type = PACKET_BROADCAST;
128 else
129 skb->pkt_type = PACKET_MULTICAST;
130
131 return netif_rx(skb);
132 }
133
134 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
135 {
136 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
137 }
138
139
140 static unsigned int mc_hash(const struct macvlan_dev *vlan,
141 const unsigned char *addr)
142 {
143 u32 val = __get_unaligned_cpu32(addr + 2);
144
145 val ^= macvlan_hash_mix(vlan);
146 return hash_32(val, MACVLAN_MC_FILTER_BITS);
147 }
148
149 static void macvlan_broadcast(struct sk_buff *skb,
150 const struct macvlan_port *port,
151 struct net_device *src,
152 enum macvlan_mode mode)
153 {
154 const struct ethhdr *eth = eth_hdr(skb);
155 const struct macvlan_dev *vlan;
156 struct sk_buff *nskb;
157 unsigned int i;
158 int err;
159 unsigned int hash;
160
161 if (skb->protocol == htons(ETH_P_PAUSE))
162 return;
163
164 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
165 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
166 if (vlan->dev == src || !(vlan->mode & mode))
167 continue;
168
169 hash = mc_hash(vlan, eth->h_dest);
170 if (!test_bit(hash, vlan->mc_filter))
171 continue;
172
173 err = NET_RX_DROP;
174 nskb = skb_clone(skb, GFP_ATOMIC);
175 if (likely(nskb))
176 err = macvlan_broadcast_one(
177 nskb, vlan, eth,
178 mode == MACVLAN_MODE_BRIDGE);
179 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
180 err == NET_RX_SUCCESS, 1);
181 }
182 }
183 }
184
185 /* called under rcu_read_lock() from netif_receive_skb */
186 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
187 {
188 struct macvlan_port *port;
189 struct sk_buff *skb = *pskb;
190 const struct ethhdr *eth = eth_hdr(skb);
191 const struct macvlan_dev *vlan;
192 const struct macvlan_dev *src;
193 struct net_device *dev;
194 unsigned int len = 0;
195 int ret = NET_RX_DROP;
196
197 port = macvlan_port_get_rcu(skb->dev);
198 if (is_multicast_ether_addr(eth->h_dest)) {
199 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
200 if (!skb)
201 return RX_HANDLER_CONSUMED;
202 eth = eth_hdr(skb);
203 src = macvlan_hash_lookup(port, eth->h_source);
204 if (!src)
205 /* frame comes from an external address */
206 macvlan_broadcast(skb, port, NULL,
207 MACVLAN_MODE_PRIVATE |
208 MACVLAN_MODE_VEPA |
209 MACVLAN_MODE_PASSTHRU|
210 MACVLAN_MODE_BRIDGE);
211 else if (src->mode == MACVLAN_MODE_VEPA)
212 /* flood to everyone except source */
213 macvlan_broadcast(skb, port, src->dev,
214 MACVLAN_MODE_VEPA |
215 MACVLAN_MODE_BRIDGE);
216 else if (src->mode == MACVLAN_MODE_BRIDGE)
217 /*
218 * flood only to VEPA ports, bridge ports
219 * already saw the frame on the way out.
220 */
221 macvlan_broadcast(skb, port, src->dev,
222 MACVLAN_MODE_VEPA);
223 else {
224 /* forward to original port. */
225 vlan = src;
226 ret = macvlan_broadcast_one(skb, vlan, eth, 0);
227 goto out;
228 }
229
230 return RX_HANDLER_PASS;
231 }
232
233 if (port->passthru)
234 vlan = list_first_or_null_rcu(&port->vlans,
235 struct macvlan_dev, list);
236 else
237 vlan = macvlan_hash_lookup(port, eth->h_dest);
238 if (vlan == NULL)
239 return RX_HANDLER_PASS;
240
241 dev = vlan->dev;
242 if (unlikely(!(dev->flags & IFF_UP))) {
243 kfree_skb(skb);
244 return RX_HANDLER_CONSUMED;
245 }
246 len = skb->len + ETH_HLEN;
247 skb = skb_share_check(skb, GFP_ATOMIC);
248 if (!skb)
249 goto out;
250
251 skb->dev = dev;
252 skb->pkt_type = PACKET_HOST;
253
254 ret = netif_rx(skb);
255
256 out:
257 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
258 return RX_HANDLER_CONSUMED;
259 }
260
261 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
262 {
263 const struct macvlan_dev *vlan = netdev_priv(dev);
264 const struct macvlan_port *port = vlan->port;
265 const struct macvlan_dev *dest;
266
267 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
268 const struct ethhdr *eth = (void *)skb->data;
269
270 /* send to other bridge ports directly */
271 if (is_multicast_ether_addr(eth->h_dest)) {
272 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
273 goto xmit_world;
274 }
275
276 dest = macvlan_hash_lookup(port, eth->h_dest);
277 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
278 /* send to lowerdev first for its network taps */
279 dev_forward_skb(vlan->lowerdev, skb);
280
281 return NET_XMIT_SUCCESS;
282 }
283 }
284
285 xmit_world:
286 skb->dev = vlan->lowerdev;
287 return dev_queue_xmit(skb);
288 }
289
290 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
291 struct net_device *dev)
292 {
293 unsigned int len = skb->len;
294 int ret;
295 const struct macvlan_dev *vlan = netdev_priv(dev);
296
297 if (vlan->fwd_priv) {
298 skb->dev = vlan->lowerdev;
299 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
300 } else {
301 ret = macvlan_queue_xmit(skb, dev);
302 }
303
304 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
305 struct vlan_pcpu_stats *pcpu_stats;
306
307 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
308 u64_stats_update_begin(&pcpu_stats->syncp);
309 pcpu_stats->tx_packets++;
310 pcpu_stats->tx_bytes += len;
311 u64_stats_update_end(&pcpu_stats->syncp);
312 } else {
313 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
314 }
315 return ret;
316 }
317
318 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
319 unsigned short type, const void *daddr,
320 const void *saddr, unsigned len)
321 {
322 const struct macvlan_dev *vlan = netdev_priv(dev);
323 struct net_device *lowerdev = vlan->lowerdev;
324
325 return dev_hard_header(skb, lowerdev, type, daddr,
326 saddr ? : dev->dev_addr, len);
327 }
328
329 static const struct header_ops macvlan_hard_header_ops = {
330 .create = macvlan_hard_header,
331 .rebuild = eth_rebuild_header,
332 .parse = eth_header_parse,
333 .cache = eth_header_cache,
334 .cache_update = eth_header_cache_update,
335 };
336
337 static struct rtnl_link_ops macvlan_link_ops;
338
339 static int macvlan_open(struct net_device *dev)
340 {
341 struct macvlan_dev *vlan = netdev_priv(dev);
342 struct net_device *lowerdev = vlan->lowerdev;
343 int err;
344
345 if (vlan->port->passthru) {
346 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
347 err = dev_set_promiscuity(lowerdev, 1);
348 if (err < 0)
349 goto out;
350 }
351 goto hash_add;
352 }
353
354 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
355 dev->rtnl_link_ops == &macvlan_link_ops) {
356 vlan->fwd_priv =
357 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
358
359 /* If we get a NULL pointer back, or if we get an error
360 * then we should just fall through to the non accelerated path
361 */
362 if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
363 vlan->fwd_priv = NULL;
364 } else
365 return 0;
366 }
367
368 err = -EBUSY;
369 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
370 goto out;
371
372 err = dev_uc_add(lowerdev, dev->dev_addr);
373 if (err < 0)
374 goto out;
375 if (dev->flags & IFF_ALLMULTI) {
376 err = dev_set_allmulti(lowerdev, 1);
377 if (err < 0)
378 goto del_unicast;
379 }
380
381 hash_add:
382 macvlan_hash_add(vlan);
383 return 0;
384
385 del_unicast:
386 dev_uc_del(lowerdev, dev->dev_addr);
387 out:
388 if (vlan->fwd_priv) {
389 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
390 vlan->fwd_priv);
391 vlan->fwd_priv = NULL;
392 }
393 return err;
394 }
395
396 static int macvlan_stop(struct net_device *dev)
397 {
398 struct macvlan_dev *vlan = netdev_priv(dev);
399 struct net_device *lowerdev = vlan->lowerdev;
400
401 if (vlan->fwd_priv) {
402 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
403 vlan->fwd_priv);
404 vlan->fwd_priv = NULL;
405 return 0;
406 }
407
408 dev_uc_unsync(lowerdev, dev);
409 dev_mc_unsync(lowerdev, dev);
410
411 if (vlan->port->passthru) {
412 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
413 dev_set_promiscuity(lowerdev, -1);
414 goto hash_del;
415 }
416
417 if (dev->flags & IFF_ALLMULTI)
418 dev_set_allmulti(lowerdev, -1);
419
420 dev_uc_del(lowerdev, dev->dev_addr);
421
422 hash_del:
423 macvlan_hash_del(vlan, !dev->dismantle);
424 return 0;
425 }
426
427 static int macvlan_set_mac_address(struct net_device *dev, void *p)
428 {
429 struct macvlan_dev *vlan = netdev_priv(dev);
430 struct net_device *lowerdev = vlan->lowerdev;
431 struct sockaddr *addr = p;
432 int err;
433
434 if (!is_valid_ether_addr(addr->sa_data))
435 return -EADDRNOTAVAIL;
436
437 if (!(dev->flags & IFF_UP)) {
438 /* Just copy in the new address */
439 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
440 } else {
441 /* Rehash and update the device filters */
442 if (macvlan_addr_busy(vlan->port, addr->sa_data))
443 return -EBUSY;
444
445 err = dev_uc_add(lowerdev, addr->sa_data);
446 if (err)
447 return err;
448
449 dev_uc_del(lowerdev, dev->dev_addr);
450
451 macvlan_hash_change_addr(vlan, addr->sa_data);
452 }
453 return 0;
454 }
455
456 static void macvlan_change_rx_flags(struct net_device *dev, int change)
457 {
458 struct macvlan_dev *vlan = netdev_priv(dev);
459 struct net_device *lowerdev = vlan->lowerdev;
460
461 if (dev->flags & IFF_UP) {
462 if (change & IFF_ALLMULTI)
463 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
464 }
465 }
466
467 static void macvlan_set_mac_lists(struct net_device *dev)
468 {
469 struct macvlan_dev *vlan = netdev_priv(dev);
470
471 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
472 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
473 } else {
474 struct netdev_hw_addr *ha;
475 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
476
477 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
478 netdev_for_each_mc_addr(ha, dev) {
479 __set_bit(mc_hash(vlan, ha->addr), filter);
480 }
481
482 __set_bit(mc_hash(vlan, dev->broadcast), filter);
483
484 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
485 }
486 dev_uc_sync(vlan->lowerdev, dev);
487 dev_mc_sync(vlan->lowerdev, dev);
488 }
489
490 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
491 {
492 struct macvlan_dev *vlan = netdev_priv(dev);
493
494 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
495 return -EINVAL;
496 dev->mtu = new_mtu;
497 return 0;
498 }
499
500 /*
501 * macvlan network devices have devices nesting below it and are a special
502 * "super class" of normal network devices; split their locks off into a
503 * separate class since they always nest.
504 */
505 static struct lock_class_key macvlan_netdev_xmit_lock_key;
506 static struct lock_class_key macvlan_netdev_addr_lock_key;
507
508 #define ALWAYS_ON_FEATURES \
509 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX)
510
511 #define MACVLAN_FEATURES \
512 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
513 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
514 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
515 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
516
517 #define MACVLAN_STATE_MASK \
518 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
519
520 static void macvlan_set_lockdep_class_one(struct net_device *dev,
521 struct netdev_queue *txq,
522 void *_unused)
523 {
524 lockdep_set_class(&txq->_xmit_lock,
525 &macvlan_netdev_xmit_lock_key);
526 }
527
528 static void macvlan_set_lockdep_class(struct net_device *dev)
529 {
530 lockdep_set_class(&dev->addr_list_lock,
531 &macvlan_netdev_addr_lock_key);
532 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
533 }
534
535 static int macvlan_init(struct net_device *dev)
536 {
537 struct macvlan_dev *vlan = netdev_priv(dev);
538 const struct net_device *lowerdev = vlan->lowerdev;
539
540 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
541 (lowerdev->state & MACVLAN_STATE_MASK);
542 dev->features = lowerdev->features & MACVLAN_FEATURES;
543 dev->features |= ALWAYS_ON_FEATURES;
544 dev->gso_max_size = lowerdev->gso_max_size;
545 dev->iflink = lowerdev->ifindex;
546 dev->hard_header_len = lowerdev->hard_header_len;
547
548 macvlan_set_lockdep_class(dev);
549
550 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
551 if (!vlan->pcpu_stats)
552 return -ENOMEM;
553
554 return 0;
555 }
556
557 static void macvlan_uninit(struct net_device *dev)
558 {
559 struct macvlan_dev *vlan = netdev_priv(dev);
560 struct macvlan_port *port = vlan->port;
561
562 free_percpu(vlan->pcpu_stats);
563
564 port->count -= 1;
565 if (!port->count)
566 macvlan_port_destroy(port->dev);
567 }
568
569 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
570 struct rtnl_link_stats64 *stats)
571 {
572 struct macvlan_dev *vlan = netdev_priv(dev);
573
574 if (vlan->pcpu_stats) {
575 struct vlan_pcpu_stats *p;
576 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
577 u32 rx_errors = 0, tx_dropped = 0;
578 unsigned int start;
579 int i;
580
581 for_each_possible_cpu(i) {
582 p = per_cpu_ptr(vlan->pcpu_stats, i);
583 do {
584 start = u64_stats_fetch_begin_irq(&p->syncp);
585 rx_packets = p->rx_packets;
586 rx_bytes = p->rx_bytes;
587 rx_multicast = p->rx_multicast;
588 tx_packets = p->tx_packets;
589 tx_bytes = p->tx_bytes;
590 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
591
592 stats->rx_packets += rx_packets;
593 stats->rx_bytes += rx_bytes;
594 stats->multicast += rx_multicast;
595 stats->tx_packets += tx_packets;
596 stats->tx_bytes += tx_bytes;
597 /* rx_errors & tx_dropped are u32, updated
598 * without syncp protection.
599 */
600 rx_errors += p->rx_errors;
601 tx_dropped += p->tx_dropped;
602 }
603 stats->rx_errors = rx_errors;
604 stats->rx_dropped = rx_errors;
605 stats->tx_dropped = tx_dropped;
606 }
607 return stats;
608 }
609
610 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
611 __be16 proto, u16 vid)
612 {
613 struct macvlan_dev *vlan = netdev_priv(dev);
614 struct net_device *lowerdev = vlan->lowerdev;
615
616 return vlan_vid_add(lowerdev, proto, vid);
617 }
618
619 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
620 __be16 proto, u16 vid)
621 {
622 struct macvlan_dev *vlan = netdev_priv(dev);
623 struct net_device *lowerdev = vlan->lowerdev;
624
625 vlan_vid_del(lowerdev, proto, vid);
626 return 0;
627 }
628
629 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
630 struct net_device *dev,
631 const unsigned char *addr,
632 u16 flags)
633 {
634 struct macvlan_dev *vlan = netdev_priv(dev);
635 int err = -EINVAL;
636
637 if (!vlan->port->passthru)
638 return -EOPNOTSUPP;
639
640 if (flags & NLM_F_REPLACE)
641 return -EOPNOTSUPP;
642
643 if (is_unicast_ether_addr(addr))
644 err = dev_uc_add_excl(dev, addr);
645 else if (is_multicast_ether_addr(addr))
646 err = dev_mc_add_excl(dev, addr);
647
648 return err;
649 }
650
651 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
652 struct net_device *dev,
653 const unsigned char *addr)
654 {
655 struct macvlan_dev *vlan = netdev_priv(dev);
656 int err = -EINVAL;
657
658 if (!vlan->port->passthru)
659 return -EOPNOTSUPP;
660
661 if (is_unicast_ether_addr(addr))
662 err = dev_uc_del(dev, addr);
663 else if (is_multicast_ether_addr(addr))
664 err = dev_mc_del(dev, addr);
665
666 return err;
667 }
668
669 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
670 struct ethtool_drvinfo *drvinfo)
671 {
672 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
673 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
674 }
675
676 static int macvlan_ethtool_get_settings(struct net_device *dev,
677 struct ethtool_cmd *cmd)
678 {
679 const struct macvlan_dev *vlan = netdev_priv(dev);
680
681 return __ethtool_get_settings(vlan->lowerdev, cmd);
682 }
683
684 static netdev_features_t macvlan_fix_features(struct net_device *dev,
685 netdev_features_t features)
686 {
687 struct macvlan_dev *vlan = netdev_priv(dev);
688 netdev_features_t mask;
689
690 features |= NETIF_F_ALL_FOR_ALL;
691 features &= (vlan->set_features | ~MACVLAN_FEATURES);
692 mask = features;
693
694 features = netdev_increment_features(vlan->lowerdev->features,
695 features,
696 mask);
697 features |= ALWAYS_ON_FEATURES;
698
699 return features;
700 }
701
702 static const struct ethtool_ops macvlan_ethtool_ops = {
703 .get_link = ethtool_op_get_link,
704 .get_settings = macvlan_ethtool_get_settings,
705 .get_drvinfo = macvlan_ethtool_get_drvinfo,
706 };
707
708 static const struct net_device_ops macvlan_netdev_ops = {
709 .ndo_init = macvlan_init,
710 .ndo_uninit = macvlan_uninit,
711 .ndo_open = macvlan_open,
712 .ndo_stop = macvlan_stop,
713 .ndo_start_xmit = macvlan_start_xmit,
714 .ndo_change_mtu = macvlan_change_mtu,
715 .ndo_fix_features = macvlan_fix_features,
716 .ndo_change_rx_flags = macvlan_change_rx_flags,
717 .ndo_set_mac_address = macvlan_set_mac_address,
718 .ndo_set_rx_mode = macvlan_set_mac_lists,
719 .ndo_get_stats64 = macvlan_dev_get_stats64,
720 .ndo_validate_addr = eth_validate_addr,
721 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
722 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
723 .ndo_fdb_add = macvlan_fdb_add,
724 .ndo_fdb_del = macvlan_fdb_del,
725 .ndo_fdb_dump = ndo_dflt_fdb_dump,
726 };
727
728 void macvlan_common_setup(struct net_device *dev)
729 {
730 ether_setup(dev);
731
732 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
733 dev->priv_flags |= IFF_UNICAST_FLT;
734 dev->netdev_ops = &macvlan_netdev_ops;
735 dev->destructor = free_netdev;
736 dev->header_ops = &macvlan_hard_header_ops;
737 dev->ethtool_ops = &macvlan_ethtool_ops;
738 }
739 EXPORT_SYMBOL_GPL(macvlan_common_setup);
740
741 static void macvlan_setup(struct net_device *dev)
742 {
743 macvlan_common_setup(dev);
744 dev->tx_queue_len = 0;
745 }
746
747 static int macvlan_port_create(struct net_device *dev)
748 {
749 struct macvlan_port *port;
750 unsigned int i;
751 int err;
752
753 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
754 return -EINVAL;
755
756 port = kzalloc(sizeof(*port), GFP_KERNEL);
757 if (port == NULL)
758 return -ENOMEM;
759
760 port->passthru = false;
761 port->dev = dev;
762 INIT_LIST_HEAD(&port->vlans);
763 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
764 INIT_HLIST_HEAD(&port->vlan_hash[i]);
765
766 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
767 if (err)
768 kfree(port);
769 else
770 dev->priv_flags |= IFF_MACVLAN_PORT;
771 return err;
772 }
773
774 static void macvlan_port_destroy(struct net_device *dev)
775 {
776 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
777
778 dev->priv_flags &= ~IFF_MACVLAN_PORT;
779 netdev_rx_handler_unregister(dev);
780 kfree_rcu(port, rcu);
781 }
782
783 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
784 {
785 if (tb[IFLA_ADDRESS]) {
786 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
787 return -EINVAL;
788 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
789 return -EADDRNOTAVAIL;
790 }
791
792 if (data && data[IFLA_MACVLAN_FLAGS] &&
793 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
794 return -EINVAL;
795
796 if (data && data[IFLA_MACVLAN_MODE]) {
797 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
798 case MACVLAN_MODE_PRIVATE:
799 case MACVLAN_MODE_VEPA:
800 case MACVLAN_MODE_BRIDGE:
801 case MACVLAN_MODE_PASSTHRU:
802 break;
803 default:
804 return -EINVAL;
805 }
806 }
807 return 0;
808 }
809
810 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
811 struct nlattr *tb[], struct nlattr *data[])
812 {
813 struct macvlan_dev *vlan = netdev_priv(dev);
814 struct macvlan_port *port;
815 struct net_device *lowerdev;
816 int err;
817
818 if (!tb[IFLA_LINK])
819 return -EINVAL;
820
821 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
822 if (lowerdev == NULL)
823 return -ENODEV;
824
825 /* When creating macvlans or macvtaps on top of other macvlans - use
826 * the real device as the lowerdev.
827 */
828 if (netif_is_macvlan(lowerdev))
829 lowerdev = macvlan_dev_real_dev(lowerdev);
830
831 if (!tb[IFLA_MTU])
832 dev->mtu = lowerdev->mtu;
833 else if (dev->mtu > lowerdev->mtu)
834 return -EINVAL;
835
836 if (!tb[IFLA_ADDRESS])
837 eth_hw_addr_random(dev);
838
839 if (!macvlan_port_exists(lowerdev)) {
840 err = macvlan_port_create(lowerdev);
841 if (err < 0)
842 return err;
843 }
844 port = macvlan_port_get_rtnl(lowerdev);
845
846 /* Only 1 macvlan device can be created in passthru mode */
847 if (port->passthru)
848 return -EINVAL;
849
850 vlan->lowerdev = lowerdev;
851 vlan->dev = dev;
852 vlan->port = port;
853 vlan->set_features = MACVLAN_FEATURES;
854
855 vlan->mode = MACVLAN_MODE_VEPA;
856 if (data && data[IFLA_MACVLAN_MODE])
857 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
858
859 if (data && data[IFLA_MACVLAN_FLAGS])
860 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
861
862 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
863 if (port->count)
864 return -EINVAL;
865 port->passthru = true;
866 eth_hw_addr_inherit(dev, lowerdev);
867 }
868
869 port->count += 1;
870 err = register_netdevice(dev);
871 if (err < 0)
872 goto destroy_port;
873
874 dev->priv_flags |= IFF_MACVLAN;
875 err = netdev_upper_dev_link(lowerdev, dev);
876 if (err)
877 goto unregister_netdev;
878
879 list_add_tail_rcu(&vlan->list, &port->vlans);
880 netif_stacked_transfer_operstate(lowerdev, dev);
881
882 return 0;
883
884 unregister_netdev:
885 unregister_netdevice(dev);
886 destroy_port:
887 port->count -= 1;
888 if (!port->count)
889 macvlan_port_destroy(lowerdev);
890
891 return err;
892 }
893 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
894
895 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
896 struct nlattr *tb[], struct nlattr *data[])
897 {
898 return macvlan_common_newlink(src_net, dev, tb, data);
899 }
900
901 void macvlan_dellink(struct net_device *dev, struct list_head *head)
902 {
903 struct macvlan_dev *vlan = netdev_priv(dev);
904
905 list_del_rcu(&vlan->list);
906 unregister_netdevice_queue(dev, head);
907 netdev_upper_dev_unlink(vlan->lowerdev, dev);
908 }
909 EXPORT_SYMBOL_GPL(macvlan_dellink);
910
911 static int macvlan_changelink(struct net_device *dev,
912 struct nlattr *tb[], struct nlattr *data[])
913 {
914 struct macvlan_dev *vlan = netdev_priv(dev);
915 enum macvlan_mode mode;
916 bool set_mode = false;
917
918 /* Validate mode, but don't set yet: setting flags may fail. */
919 if (data && data[IFLA_MACVLAN_MODE]) {
920 set_mode = true;
921 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
922 /* Passthrough mode can't be set or cleared dynamically */
923 if ((mode == MACVLAN_MODE_PASSTHRU) !=
924 (vlan->mode == MACVLAN_MODE_PASSTHRU))
925 return -EINVAL;
926 }
927
928 if (data && data[IFLA_MACVLAN_FLAGS]) {
929 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
930 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
931 if (vlan->port->passthru && promisc) {
932 int err;
933
934 if (flags & MACVLAN_FLAG_NOPROMISC)
935 err = dev_set_promiscuity(vlan->lowerdev, -1);
936 else
937 err = dev_set_promiscuity(vlan->lowerdev, 1);
938 if (err < 0)
939 return err;
940 }
941 vlan->flags = flags;
942 }
943 if (set_mode)
944 vlan->mode = mode;
945 return 0;
946 }
947
948 static size_t macvlan_get_size(const struct net_device *dev)
949 {
950 return (0
951 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
952 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
953 );
954 }
955
956 static int macvlan_fill_info(struct sk_buff *skb,
957 const struct net_device *dev)
958 {
959 struct macvlan_dev *vlan = netdev_priv(dev);
960
961 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
962 goto nla_put_failure;
963 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
964 goto nla_put_failure;
965 return 0;
966
967 nla_put_failure:
968 return -EMSGSIZE;
969 }
970
971 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
972 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
973 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
974 };
975
976 int macvlan_link_register(struct rtnl_link_ops *ops)
977 {
978 /* common fields */
979 ops->priv_size = sizeof(struct macvlan_dev);
980 ops->validate = macvlan_validate;
981 ops->maxtype = IFLA_MACVLAN_MAX;
982 ops->policy = macvlan_policy;
983 ops->changelink = macvlan_changelink;
984 ops->get_size = macvlan_get_size;
985 ops->fill_info = macvlan_fill_info;
986
987 return rtnl_link_register(ops);
988 };
989 EXPORT_SYMBOL_GPL(macvlan_link_register);
990
991 static struct rtnl_link_ops macvlan_link_ops = {
992 .kind = "macvlan",
993 .setup = macvlan_setup,
994 .newlink = macvlan_newlink,
995 .dellink = macvlan_dellink,
996 };
997
998 static int macvlan_device_event(struct notifier_block *unused,
999 unsigned long event, void *ptr)
1000 {
1001 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1002 struct macvlan_dev *vlan, *next;
1003 struct macvlan_port *port;
1004 LIST_HEAD(list_kill);
1005
1006 if (!macvlan_port_exists(dev))
1007 return NOTIFY_DONE;
1008
1009 port = macvlan_port_get_rtnl(dev);
1010
1011 switch (event) {
1012 case NETDEV_CHANGE:
1013 list_for_each_entry(vlan, &port->vlans, list)
1014 netif_stacked_transfer_operstate(vlan->lowerdev,
1015 vlan->dev);
1016 break;
1017 case NETDEV_FEAT_CHANGE:
1018 list_for_each_entry(vlan, &port->vlans, list) {
1019 vlan->dev->gso_max_size = dev->gso_max_size;
1020 netdev_update_features(vlan->dev);
1021 }
1022 break;
1023 case NETDEV_UNREGISTER:
1024 /* twiddle thumbs on netns device moves */
1025 if (dev->reg_state != NETREG_UNREGISTERING)
1026 break;
1027
1028 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1029 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1030 unregister_netdevice_many(&list_kill);
1031 list_del(&list_kill);
1032 break;
1033 case NETDEV_PRE_TYPE_CHANGE:
1034 /* Forbid underlaying device to change its type. */
1035 return NOTIFY_BAD;
1036 }
1037 return NOTIFY_DONE;
1038 }
1039
1040 static struct notifier_block macvlan_notifier_block __read_mostly = {
1041 .notifier_call = macvlan_device_event,
1042 };
1043
1044 static int __init macvlan_init_module(void)
1045 {
1046 int err;
1047
1048 register_netdevice_notifier(&macvlan_notifier_block);
1049
1050 err = macvlan_link_register(&macvlan_link_ops);
1051 if (err < 0)
1052 goto err1;
1053 return 0;
1054 err1:
1055 unregister_netdevice_notifier(&macvlan_notifier_block);
1056 return err;
1057 }
1058
1059 static void __exit macvlan_cleanup_module(void)
1060 {
1061 rtnl_link_unregister(&macvlan_link_ops);
1062 unregister_netdevice_notifier(&macvlan_notifier_block);
1063 }
1064
1065 module_init(macvlan_init_module);
1066 module_exit(macvlan_cleanup_module);
1067
1068 MODULE_LICENSE("GPL");
1069 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1070 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1071 MODULE_ALIAS_RTNL_LINK("macvlan");
Ubuntu Jammy Linux kernel mirror