projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'vfio-v3.17-rc1' of git://github.com/awilliam/linux-vfio
[deliverable/linux.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 <linux/workqueue.h>
34 #include <net/rtnetlink.h>
35 #include <net/xfrm.h>
36 #include <linux/netpoll.h>
37
38 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE)
39
40 struct macvlan_port {
41 struct net_device *dev;
42 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
43 struct list_head vlans;
44 struct rcu_head rcu;
45 struct sk_buff_head bc_queue;
46 struct work_struct bc_work;
47 bool passthru;
48 };
49
50 #define MACVLAN_PORT_IS_EMPTY(port) list_empty(&port->vlans)
51
52 struct macvlan_skb_cb {
53 const struct macvlan_dev *src;
54 };
55
56 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
57
58 static void macvlan_port_destroy(struct net_device *dev);
59
60 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
61 {
62 return rcu_dereference(dev->rx_handler_data);
63 }
64
65 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
66 {
67 return rtnl_dereference(dev->rx_handler_data);
68 }
69
70 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
71
72 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
73 const unsigned char *addr)
74 {
75 struct macvlan_dev *vlan;
76
77 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) {
78 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
79 return vlan;
80 }
81 return NULL;
82 }
83
84 static void macvlan_hash_add(struct macvlan_dev *vlan)
85 {
86 struct macvlan_port *port = vlan->port;
87 const unsigned char *addr = vlan->dev->dev_addr;
88
89 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
90 }
91
92 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
93 {
94 hlist_del_rcu(&vlan->hlist);
95 if (sync)
96 synchronize_rcu();
97 }
98
99 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
100 const unsigned char *addr)
101 {
102 macvlan_hash_del(vlan, true);
103 /* Now that we are unhashed it is safe to change the device
104 * address without confusing packet delivery.
105 */
106 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
107 macvlan_hash_add(vlan);
108 }
109
110 static int macvlan_addr_busy(const struct macvlan_port *port,
111 const unsigned char *addr)
112 {
113 /* Test to see if the specified multicast address is
114 * currently in use by the underlying device or
115 * another macvlan.
116 */
117 if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
118 return 1;
119
120 if (macvlan_hash_lookup(port, addr))
121 return 1;
122
123 return 0;
124 }
125
126
127 static int macvlan_broadcast_one(struct sk_buff *skb,
128 const struct macvlan_dev *vlan,
129 const struct ethhdr *eth, bool local)
130 {
131 struct net_device *dev = vlan->dev;
132
133 if (local)
134 return __dev_forward_skb(dev, skb);
135
136 skb->dev = dev;
137 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
138 skb->pkt_type = PACKET_BROADCAST;
139 else
140 skb->pkt_type = PACKET_MULTICAST;
141
142 return 0;
143 }
144
145 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
146 {
147 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
148 }
149
150
151 static unsigned int mc_hash(const struct macvlan_dev *vlan,
152 const unsigned char *addr)
153 {
154 u32 val = __get_unaligned_cpu32(addr + 2);
155
156 val ^= macvlan_hash_mix(vlan);
157 return hash_32(val, MACVLAN_MC_FILTER_BITS);
158 }
159
160 static void macvlan_broadcast(struct sk_buff *skb,
161 const struct macvlan_port *port,
162 struct net_device *src,
163 enum macvlan_mode mode)
164 {
165 const struct ethhdr *eth = eth_hdr(skb);
166 const struct macvlan_dev *vlan;
167 struct sk_buff *nskb;
168 unsigned int i;
169 int err;
170 unsigned int hash;
171
172 if (skb->protocol == htons(ETH_P_PAUSE))
173 return;
174
175 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
176 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
177 if (vlan->dev == src || !(vlan->mode & mode))
178 continue;
179
180 hash = mc_hash(vlan, eth->h_dest);
181 if (!test_bit(hash, vlan->mc_filter))
182 continue;
183
184 err = NET_RX_DROP;
185 nskb = skb_clone(skb, GFP_ATOMIC);
186 if (likely(nskb))
187 err = macvlan_broadcast_one(
188 nskb, vlan, eth,
189 mode == MACVLAN_MODE_BRIDGE) ?:
190 netif_rx_ni(nskb);
191 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
192 err == NET_RX_SUCCESS, 1);
193 }
194 }
195 }
196
197 static void macvlan_process_broadcast(struct work_struct *w)
198 {
199 struct macvlan_port *port = container_of(w, struct macvlan_port,
200 bc_work);
201 struct sk_buff *skb;
202 struct sk_buff_head list;
203
204 skb_queue_head_init(&list);
205
206 spin_lock_bh(&port->bc_queue.lock);
207 skb_queue_splice_tail_init(&port->bc_queue, &list);
208 spin_unlock_bh(&port->bc_queue.lock);
209
210 while ((skb = __skb_dequeue(&list))) {
211 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
212
213 rcu_read_lock();
214
215 if (!src)
216 /* frame comes from an external address */
217 macvlan_broadcast(skb, port, NULL,
218 MACVLAN_MODE_PRIVATE |
219 MACVLAN_MODE_VEPA |
220 MACVLAN_MODE_PASSTHRU|
221 MACVLAN_MODE_BRIDGE);
222 else if (src->mode == MACVLAN_MODE_VEPA)
223 /* flood to everyone except source */
224 macvlan_broadcast(skb, port, src->dev,
225 MACVLAN_MODE_VEPA |
226 MACVLAN_MODE_BRIDGE);
227 else
228 /*
229 * flood only to VEPA ports, bridge ports
230 * already saw the frame on the way out.
231 */
232 macvlan_broadcast(skb, port, src->dev,
233 MACVLAN_MODE_VEPA);
234
235 rcu_read_unlock();
236
237 kfree_skb(skb);
238 }
239 }
240
241 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
242 struct sk_buff *skb)
243 {
244 struct sk_buff *nskb;
245 int err = -ENOMEM;
246
247 nskb = skb_clone(skb, GFP_ATOMIC);
248 if (!nskb)
249 goto err;
250
251 spin_lock(&port->bc_queue.lock);
252 if (skb_queue_len(&port->bc_queue) < skb->dev->tx_queue_len) {
253 __skb_queue_tail(&port->bc_queue, nskb);
254 err = 0;
255 }
256 spin_unlock(&port->bc_queue.lock);
257
258 if (err)
259 goto free_nskb;
260
261 schedule_work(&port->bc_work);
262 return;
263
264 free_nskb:
265 kfree_skb(nskb);
266 err:
267 atomic_long_inc(&skb->dev->rx_dropped);
268 }
269
270 /* called under rcu_read_lock() from netif_receive_skb */
271 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
272 {
273 struct macvlan_port *port;
274 struct sk_buff *skb = *pskb;
275 const struct ethhdr *eth = eth_hdr(skb);
276 const struct macvlan_dev *vlan;
277 const struct macvlan_dev *src;
278 struct net_device *dev;
279 unsigned int len = 0;
280 int ret = NET_RX_DROP;
281
282 port = macvlan_port_get_rcu(skb->dev);
283 if (is_multicast_ether_addr(eth->h_dest)) {
284 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
285 if (!skb)
286 return RX_HANDLER_CONSUMED;
287 eth = eth_hdr(skb);
288 src = macvlan_hash_lookup(port, eth->h_source);
289 if (src && src->mode != MACVLAN_MODE_VEPA &&
290 src->mode != MACVLAN_MODE_BRIDGE) {
291 /* forward to original port. */
292 vlan = src;
293 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
294 netif_rx(skb);
295 goto out;
296 }
297
298 MACVLAN_SKB_CB(skb)->src = src;
299 macvlan_broadcast_enqueue(port, skb);
300
301 return RX_HANDLER_PASS;
302 }
303
304 if (port->passthru)
305 vlan = list_first_or_null_rcu(&port->vlans,
306 struct macvlan_dev, list);
307 else
308 vlan = macvlan_hash_lookup(port, eth->h_dest);
309 if (vlan == NULL)
310 return RX_HANDLER_PASS;
311
312 dev = vlan->dev;
313 if (unlikely(!(dev->flags & IFF_UP))) {
314 kfree_skb(skb);
315 return RX_HANDLER_CONSUMED;
316 }
317 len = skb->len + ETH_HLEN;
318 skb = skb_share_check(skb, GFP_ATOMIC);
319 if (!skb)
320 goto out;
321
322 skb->dev = dev;
323 skb->pkt_type = PACKET_HOST;
324
325 ret = netif_rx(skb);
326
327 out:
328 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
329 return RX_HANDLER_CONSUMED;
330 }
331
332 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
333 {
334 const struct macvlan_dev *vlan = netdev_priv(dev);
335 const struct macvlan_port *port = vlan->port;
336 const struct macvlan_dev *dest;
337
338 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
339 const struct ethhdr *eth = (void *)skb->data;
340
341 /* send to other bridge ports directly */
342 if (is_multicast_ether_addr(eth->h_dest)) {
343 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
344 goto xmit_world;
345 }
346
347 dest = macvlan_hash_lookup(port, eth->h_dest);
348 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
349 /* send to lowerdev first for its network taps */
350 dev_forward_skb(vlan->lowerdev, skb);
351
352 return NET_XMIT_SUCCESS;
353 }
354 }
355
356 xmit_world:
357 skb->dev = vlan->lowerdev;
358 return dev_queue_xmit(skb);
359 }
360
361 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
362 {
363 #ifdef CONFIG_NET_POLL_CONTROLLER
364 if (vlan->netpoll)
365 netpoll_send_skb(vlan->netpoll, skb);
366 #else
367 BUG();
368 #endif
369 return NETDEV_TX_OK;
370 }
371
372 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
373 struct net_device *dev)
374 {
375 unsigned int len = skb->len;
376 int ret;
377 struct macvlan_dev *vlan = netdev_priv(dev);
378
379 if (unlikely(netpoll_tx_running(dev)))
380 return macvlan_netpoll_send_skb(vlan, skb);
381
382 if (vlan->fwd_priv) {
383 skb->dev = vlan->lowerdev;
384 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
385 } else {
386 ret = macvlan_queue_xmit(skb, dev);
387 }
388
389 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
390 struct vlan_pcpu_stats *pcpu_stats;
391
392 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
393 u64_stats_update_begin(&pcpu_stats->syncp);
394 pcpu_stats->tx_packets++;
395 pcpu_stats->tx_bytes += len;
396 u64_stats_update_end(&pcpu_stats->syncp);
397 } else {
398 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
399 }
400 return ret;
401 }
402
403 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
404 unsigned short type, const void *daddr,
405 const void *saddr, unsigned len)
406 {
407 const struct macvlan_dev *vlan = netdev_priv(dev);
408 struct net_device *lowerdev = vlan->lowerdev;
409
410 return dev_hard_header(skb, lowerdev, type, daddr,
411 saddr ? : dev->dev_addr, len);
412 }
413
414 static const struct header_ops macvlan_hard_header_ops = {
415 .create = macvlan_hard_header,
416 .rebuild = eth_rebuild_header,
417 .parse = eth_header_parse,
418 .cache = eth_header_cache,
419 .cache_update = eth_header_cache_update,
420 };
421
422 static struct rtnl_link_ops macvlan_link_ops;
423
424 static int macvlan_open(struct net_device *dev)
425 {
426 struct macvlan_dev *vlan = netdev_priv(dev);
427 struct net_device *lowerdev = vlan->lowerdev;
428 int err;
429
430 if (vlan->port->passthru) {
431 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
432 err = dev_set_promiscuity(lowerdev, 1);
433 if (err < 0)
434 goto out;
435 }
436 goto hash_add;
437 }
438
439 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
440 dev->rtnl_link_ops == &macvlan_link_ops) {
441 vlan->fwd_priv =
442 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
443
444 /* If we get a NULL pointer back, or if we get an error
445 * then we should just fall through to the non accelerated path
446 */
447 if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
448 vlan->fwd_priv = NULL;
449 } else
450 return 0;
451 }
452
453 err = -EBUSY;
454 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
455 goto out;
456
457 err = dev_uc_add(lowerdev, dev->dev_addr);
458 if (err < 0)
459 goto out;
460 if (dev->flags & IFF_ALLMULTI) {
461 err = dev_set_allmulti(lowerdev, 1);
462 if (err < 0)
463 goto del_unicast;
464 }
465
466 hash_add:
467 macvlan_hash_add(vlan);
468 return 0;
469
470 del_unicast:
471 dev_uc_del(lowerdev, dev->dev_addr);
472 out:
473 if (vlan->fwd_priv) {
474 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
475 vlan->fwd_priv);
476 vlan->fwd_priv = NULL;
477 }
478 return err;
479 }
480
481 static int macvlan_stop(struct net_device *dev)
482 {
483 struct macvlan_dev *vlan = netdev_priv(dev);
484 struct net_device *lowerdev = vlan->lowerdev;
485
486 if (vlan->fwd_priv) {
487 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
488 vlan->fwd_priv);
489 vlan->fwd_priv = NULL;
490 return 0;
491 }
492
493 dev_uc_unsync(lowerdev, dev);
494 dev_mc_unsync(lowerdev, dev);
495
496 if (vlan->port->passthru) {
497 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
498 dev_set_promiscuity(lowerdev, -1);
499 goto hash_del;
500 }
501
502 if (dev->flags & IFF_ALLMULTI)
503 dev_set_allmulti(lowerdev, -1);
504
505 dev_uc_del(lowerdev, dev->dev_addr);
506
507 hash_del:
508 macvlan_hash_del(vlan, !dev->dismantle);
509 return 0;
510 }
511
512 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
513 {
514 struct macvlan_dev *vlan = netdev_priv(dev);
515 struct net_device *lowerdev = vlan->lowerdev;
516 int err;
517
518 if (!(dev->flags & IFF_UP)) {
519 /* Just copy in the new address */
520 ether_addr_copy(dev->dev_addr, addr);
521 } else {
522 /* Rehash and update the device filters */
523 if (macvlan_addr_busy(vlan->port, addr))
524 return -EBUSY;
525
526 if (!vlan->port->passthru) {
527 err = dev_uc_add(lowerdev, addr);
528 if (err)
529 return err;
530
531 dev_uc_del(lowerdev, dev->dev_addr);
532 }
533
534 macvlan_hash_change_addr(vlan, addr);
535 }
536 return 0;
537 }
538
539 static int macvlan_set_mac_address(struct net_device *dev, void *p)
540 {
541 struct macvlan_dev *vlan = netdev_priv(dev);
542 struct sockaddr *addr = p;
543
544 if (!is_valid_ether_addr(addr->sa_data))
545 return -EADDRNOTAVAIL;
546
547 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
548 dev_set_mac_address(vlan->lowerdev, addr);
549 return 0;
550 }
551
552 return macvlan_sync_address(dev, addr->sa_data);
553 }
554
555 static void macvlan_change_rx_flags(struct net_device *dev, int change)
556 {
557 struct macvlan_dev *vlan = netdev_priv(dev);
558 struct net_device *lowerdev = vlan->lowerdev;
559
560 if (dev->flags & IFF_UP) {
561 if (change & IFF_ALLMULTI)
562 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
563 }
564 }
565
566 static void macvlan_set_mac_lists(struct net_device *dev)
567 {
568 struct macvlan_dev *vlan = netdev_priv(dev);
569
570 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
571 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
572 } else {
573 struct netdev_hw_addr *ha;
574 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
575
576 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
577 netdev_for_each_mc_addr(ha, dev) {
578 __set_bit(mc_hash(vlan, ha->addr), filter);
579 }
580
581 __set_bit(mc_hash(vlan, dev->broadcast), filter);
582
583 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
584 }
585 dev_uc_sync(vlan->lowerdev, dev);
586 dev_mc_sync(vlan->lowerdev, dev);
587 }
588
589 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
590 {
591 struct macvlan_dev *vlan = netdev_priv(dev);
592
593 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
594 return -EINVAL;
595 dev->mtu = new_mtu;
596 return 0;
597 }
598
599 /*
600 * macvlan network devices have devices nesting below it and are a special
601 * "super class" of normal network devices; split their locks off into a
602 * separate class since they always nest.
603 */
604 static struct lock_class_key macvlan_netdev_xmit_lock_key;
605 static struct lock_class_key macvlan_netdev_addr_lock_key;
606
607 #define ALWAYS_ON_FEATURES \
608 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX)
609
610 #define MACVLAN_FEATURES \
611 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
612 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
613 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
614 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
615
616 #define MACVLAN_STATE_MASK \
617 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
618
619 static int macvlan_get_nest_level(struct net_device *dev)
620 {
621 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
622 }
623
624 static void macvlan_set_lockdep_class_one(struct net_device *dev,
625 struct netdev_queue *txq,
626 void *_unused)
627 {
628 lockdep_set_class(&txq->_xmit_lock,
629 &macvlan_netdev_xmit_lock_key);
630 }
631
632 static void macvlan_set_lockdep_class(struct net_device *dev)
633 {
634 lockdep_set_class_and_subclass(&dev->addr_list_lock,
635 &macvlan_netdev_addr_lock_key,
636 macvlan_get_nest_level(dev));
637 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
638 }
639
640 static int macvlan_init(struct net_device *dev)
641 {
642 struct macvlan_dev *vlan = netdev_priv(dev);
643 const struct net_device *lowerdev = vlan->lowerdev;
644
645 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
646 (lowerdev->state & MACVLAN_STATE_MASK);
647 dev->features = lowerdev->features & MACVLAN_FEATURES;
648 dev->features |= ALWAYS_ON_FEATURES;
649 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
650 dev->gso_max_size = lowerdev->gso_max_size;
651 dev->iflink = lowerdev->ifindex;
652 dev->hard_header_len = lowerdev->hard_header_len;
653
654 macvlan_set_lockdep_class(dev);
655
656 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
657 if (!vlan->pcpu_stats)
658 return -ENOMEM;
659
660 return 0;
661 }
662
663 static void macvlan_uninit(struct net_device *dev)
664 {
665 struct macvlan_dev *vlan = netdev_priv(dev);
666 struct macvlan_port *port = vlan->port;
667
668 free_percpu(vlan->pcpu_stats);
669
670 if (MACVLAN_PORT_IS_EMPTY(port))
671 macvlan_port_destroy(port->dev);
672 }
673
674 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
675 struct rtnl_link_stats64 *stats)
676 {
677 struct macvlan_dev *vlan = netdev_priv(dev);
678
679 if (vlan->pcpu_stats) {
680 struct vlan_pcpu_stats *p;
681 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
682 u32 rx_errors = 0, tx_dropped = 0;
683 unsigned int start;
684 int i;
685
686 for_each_possible_cpu(i) {
687 p = per_cpu_ptr(vlan->pcpu_stats, i);
688 do {
689 start = u64_stats_fetch_begin_irq(&p->syncp);
690 rx_packets = p->rx_packets;
691 rx_bytes = p->rx_bytes;
692 rx_multicast = p->rx_multicast;
693 tx_packets = p->tx_packets;
694 tx_bytes = p->tx_bytes;
695 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
696
697 stats->rx_packets += rx_packets;
698 stats->rx_bytes += rx_bytes;
699 stats->multicast += rx_multicast;
700 stats->tx_packets += tx_packets;
701 stats->tx_bytes += tx_bytes;
702 /* rx_errors & tx_dropped are u32, updated
703 * without syncp protection.
704 */
705 rx_errors += p->rx_errors;
706 tx_dropped += p->tx_dropped;
707 }
708 stats->rx_errors = rx_errors;
709 stats->rx_dropped = rx_errors;
710 stats->tx_dropped = tx_dropped;
711 }
712 return stats;
713 }
714
715 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
716 __be16 proto, u16 vid)
717 {
718 struct macvlan_dev *vlan = netdev_priv(dev);
719 struct net_device *lowerdev = vlan->lowerdev;
720
721 return vlan_vid_add(lowerdev, proto, vid);
722 }
723
724 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
725 __be16 proto, u16 vid)
726 {
727 struct macvlan_dev *vlan = netdev_priv(dev);
728 struct net_device *lowerdev = vlan->lowerdev;
729
730 vlan_vid_del(lowerdev, proto, vid);
731 return 0;
732 }
733
734 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
735 struct net_device *dev,
736 const unsigned char *addr,
737 u16 flags)
738 {
739 struct macvlan_dev *vlan = netdev_priv(dev);
740 int err = -EINVAL;
741
742 if (!vlan->port->passthru)
743 return -EOPNOTSUPP;
744
745 if (flags & NLM_F_REPLACE)
746 return -EOPNOTSUPP;
747
748 if (is_unicast_ether_addr(addr))
749 err = dev_uc_add_excl(dev, addr);
750 else if (is_multicast_ether_addr(addr))
751 err = dev_mc_add_excl(dev, addr);
752
753 return err;
754 }
755
756 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
757 struct net_device *dev,
758 const unsigned char *addr)
759 {
760 struct macvlan_dev *vlan = netdev_priv(dev);
761 int err = -EINVAL;
762
763 if (!vlan->port->passthru)
764 return -EOPNOTSUPP;
765
766 if (is_unicast_ether_addr(addr))
767 err = dev_uc_del(dev, addr);
768 else if (is_multicast_ether_addr(addr))
769 err = dev_mc_del(dev, addr);
770
771 return err;
772 }
773
774 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
775 struct ethtool_drvinfo *drvinfo)
776 {
777 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
778 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
779 }
780
781 static int macvlan_ethtool_get_settings(struct net_device *dev,
782 struct ethtool_cmd *cmd)
783 {
784 const struct macvlan_dev *vlan = netdev_priv(dev);
785
786 return __ethtool_get_settings(vlan->lowerdev, cmd);
787 }
788
789 static netdev_features_t macvlan_fix_features(struct net_device *dev,
790 netdev_features_t features)
791 {
792 struct macvlan_dev *vlan = netdev_priv(dev);
793 netdev_features_t mask;
794
795 features |= NETIF_F_ALL_FOR_ALL;
796 features &= (vlan->set_features | ~MACVLAN_FEATURES);
797 mask = features;
798
799 features = netdev_increment_features(vlan->lowerdev->features,
800 features,
801 mask);
802 features |= ALWAYS_ON_FEATURES;
803
804 return features;
805 }
806
807 #ifdef CONFIG_NET_POLL_CONTROLLER
808 static void macvlan_dev_poll_controller(struct net_device *dev)
809 {
810 return;
811 }
812
813 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
814 {
815 struct macvlan_dev *vlan = netdev_priv(dev);
816 struct net_device *real_dev = vlan->lowerdev;
817 struct netpoll *netpoll;
818 int err = 0;
819
820 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
821 err = -ENOMEM;
822 if (!netpoll)
823 goto out;
824
825 err = __netpoll_setup(netpoll, real_dev);
826 if (err) {
827 kfree(netpoll);
828 goto out;
829 }
830
831 vlan->netpoll = netpoll;
832
833 out:
834 return err;
835 }
836
837 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
838 {
839 struct macvlan_dev *vlan = netdev_priv(dev);
840 struct netpoll *netpoll = vlan->netpoll;
841
842 if (!netpoll)
843 return;
844
845 vlan->netpoll = NULL;
846
847 __netpoll_free_async(netpoll);
848 }
849 #endif /* CONFIG_NET_POLL_CONTROLLER */
850
851 static const struct ethtool_ops macvlan_ethtool_ops = {
852 .get_link = ethtool_op_get_link,
853 .get_settings = macvlan_ethtool_get_settings,
854 .get_drvinfo = macvlan_ethtool_get_drvinfo,
855 };
856
857 static const struct net_device_ops macvlan_netdev_ops = {
858 .ndo_init = macvlan_init,
859 .ndo_uninit = macvlan_uninit,
860 .ndo_open = macvlan_open,
861 .ndo_stop = macvlan_stop,
862 .ndo_start_xmit = macvlan_start_xmit,
863 .ndo_change_mtu = macvlan_change_mtu,
864 .ndo_fix_features = macvlan_fix_features,
865 .ndo_change_rx_flags = macvlan_change_rx_flags,
866 .ndo_set_mac_address = macvlan_set_mac_address,
867 .ndo_set_rx_mode = macvlan_set_mac_lists,
868 .ndo_get_stats64 = macvlan_dev_get_stats64,
869 .ndo_validate_addr = eth_validate_addr,
870 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
871 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
872 .ndo_fdb_add = macvlan_fdb_add,
873 .ndo_fdb_del = macvlan_fdb_del,
874 .ndo_fdb_dump = ndo_dflt_fdb_dump,
875 .ndo_get_lock_subclass = macvlan_get_nest_level,
876 #ifdef CONFIG_NET_POLL_CONTROLLER
877 .ndo_poll_controller = macvlan_dev_poll_controller,
878 .ndo_netpoll_setup = macvlan_dev_netpoll_setup,
879 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup,
880 #endif
881 };
882
883 void macvlan_common_setup(struct net_device *dev)
884 {
885 ether_setup(dev);
886
887 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
888 dev->priv_flags |= IFF_UNICAST_FLT;
889 dev->netdev_ops = &macvlan_netdev_ops;
890 dev->destructor = free_netdev;
891 dev->header_ops = &macvlan_hard_header_ops;
892 dev->ethtool_ops = &macvlan_ethtool_ops;
893 }
894 EXPORT_SYMBOL_GPL(macvlan_common_setup);
895
896 static void macvlan_setup(struct net_device *dev)
897 {
898 macvlan_common_setup(dev);
899 dev->tx_queue_len = 0;
900 }
901
902 static int macvlan_port_create(struct net_device *dev)
903 {
904 struct macvlan_port *port;
905 unsigned int i;
906 int err;
907
908 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
909 return -EINVAL;
910
911 port = kzalloc(sizeof(*port), GFP_KERNEL);
912 if (port == NULL)
913 return -ENOMEM;
914
915 port->passthru = false;
916 port->dev = dev;
917 INIT_LIST_HEAD(&port->vlans);
918 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
919 INIT_HLIST_HEAD(&port->vlan_hash[i]);
920
921 skb_queue_head_init(&port->bc_queue);
922 INIT_WORK(&port->bc_work, macvlan_process_broadcast);
923
924 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
925 if (err)
926 kfree(port);
927 else
928 dev->priv_flags |= IFF_MACVLAN_PORT;
929 return err;
930 }
931
932 static void macvlan_port_destroy(struct net_device *dev)
933 {
934 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
935
936 cancel_work_sync(&port->bc_work);
937 dev->priv_flags &= ~IFF_MACVLAN_PORT;
938 netdev_rx_handler_unregister(dev);
939 kfree_rcu(port, rcu);
940 }
941
942 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
943 {
944 if (tb[IFLA_ADDRESS]) {
945 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
946 return -EINVAL;
947 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
948 return -EADDRNOTAVAIL;
949 }
950
951 if (data && data[IFLA_MACVLAN_FLAGS] &&
952 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
953 return -EINVAL;
954
955 if (data && data[IFLA_MACVLAN_MODE]) {
956 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
957 case MACVLAN_MODE_PRIVATE:
958 case MACVLAN_MODE_VEPA:
959 case MACVLAN_MODE_BRIDGE:
960 case MACVLAN_MODE_PASSTHRU:
961 break;
962 default:
963 return -EINVAL;
964 }
965 }
966 return 0;
967 }
968
969 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
970 struct nlattr *tb[], struct nlattr *data[])
971 {
972 struct macvlan_dev *vlan = netdev_priv(dev);
973 struct macvlan_port *port;
974 struct net_device *lowerdev;
975 int err;
976
977 if (!tb[IFLA_LINK])
978 return -EINVAL;
979
980 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
981 if (lowerdev == NULL)
982 return -ENODEV;
983
984 /* When creating macvlans or macvtaps on top of other macvlans - use
985 * the real device as the lowerdev.
986 */
987 if (netif_is_macvlan(lowerdev))
988 lowerdev = macvlan_dev_real_dev(lowerdev);
989
990 if (!tb[IFLA_MTU])
991 dev->mtu = lowerdev->mtu;
992 else if (dev->mtu > lowerdev->mtu)
993 return -EINVAL;
994
995 if (!tb[IFLA_ADDRESS])
996 eth_hw_addr_random(dev);
997
998 if (!macvlan_port_exists(lowerdev)) {
999 err = macvlan_port_create(lowerdev);
1000 if (err < 0)
1001 return err;
1002 }
1003 port = macvlan_port_get_rtnl(lowerdev);
1004
1005 /* Only 1 macvlan device can be created in passthru mode */
1006 if (port->passthru)
1007 return -EINVAL;
1008
1009 vlan->lowerdev = lowerdev;
1010 vlan->dev = dev;
1011 vlan->port = port;
1012 vlan->set_features = MACVLAN_FEATURES;
1013 vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1;
1014
1015 vlan->mode = MACVLAN_MODE_VEPA;
1016 if (data && data[IFLA_MACVLAN_MODE])
1017 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1018
1019 if (data && data[IFLA_MACVLAN_FLAGS])
1020 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1021
1022 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1023 if (!MACVLAN_PORT_IS_EMPTY(port))
1024 return -EINVAL;
1025 port->passthru = true;
1026 eth_hw_addr_inherit(dev, lowerdev);
1027 }
1028
1029 err = register_netdevice(dev);
1030 if (err < 0)
1031 goto destroy_port;
1032
1033 dev->priv_flags |= IFF_MACVLAN;
1034 err = netdev_upper_dev_link(lowerdev, dev);
1035 if (err)
1036 goto unregister_netdev;
1037
1038 list_add_tail_rcu(&vlan->list, &port->vlans);
1039 netif_stacked_transfer_operstate(lowerdev, dev);
1040
1041 return 0;
1042
1043 unregister_netdev:
1044 unregister_netdevice(dev);
1045 destroy_port:
1046 if (MACVLAN_PORT_IS_EMPTY(port))
1047 macvlan_port_destroy(lowerdev);
1048
1049 return err;
1050 }
1051 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1052
1053 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1054 struct nlattr *tb[], struct nlattr *data[])
1055 {
1056 return macvlan_common_newlink(src_net, dev, tb, data);
1057 }
1058
1059 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1060 {
1061 struct macvlan_dev *vlan = netdev_priv(dev);
1062
1063 list_del_rcu(&vlan->list);
1064 unregister_netdevice_queue(dev, head);
1065 netdev_upper_dev_unlink(vlan->lowerdev, dev);
1066 }
1067 EXPORT_SYMBOL_GPL(macvlan_dellink);
1068
1069 static int macvlan_changelink(struct net_device *dev,
1070 struct nlattr *tb[], struct nlattr *data[])
1071 {
1072 struct macvlan_dev *vlan = netdev_priv(dev);
1073 enum macvlan_mode mode;
1074 bool set_mode = false;
1075
1076 /* Validate mode, but don't set yet: setting flags may fail. */
1077 if (data && data[IFLA_MACVLAN_MODE]) {
1078 set_mode = true;
1079 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1080 /* Passthrough mode can't be set or cleared dynamically */
1081 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1082 (vlan->mode == MACVLAN_MODE_PASSTHRU))
1083 return -EINVAL;
1084 }
1085
1086 if (data && data[IFLA_MACVLAN_FLAGS]) {
1087 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1088 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1089 if (vlan->port->passthru && promisc) {
1090 int err;
1091
1092 if (flags & MACVLAN_FLAG_NOPROMISC)
1093 err = dev_set_promiscuity(vlan->lowerdev, -1);
1094 else
1095 err = dev_set_promiscuity(vlan->lowerdev, 1);
1096 if (err < 0)
1097 return err;
1098 }
1099 vlan->flags = flags;
1100 }
1101 if (set_mode)
1102 vlan->mode = mode;
1103 return 0;
1104 }
1105
1106 static size_t macvlan_get_size(const struct net_device *dev)
1107 {
1108 return (0
1109 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1110 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1111 );
1112 }
1113
1114 static int macvlan_fill_info(struct sk_buff *skb,
1115 const struct net_device *dev)
1116 {
1117 struct macvlan_dev *vlan = netdev_priv(dev);
1118
1119 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1120 goto nla_put_failure;
1121 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1122 goto nla_put_failure;
1123 return 0;
1124
1125 nla_put_failure:
1126 return -EMSGSIZE;
1127 }
1128
1129 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1130 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
1131 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1132 };
1133
1134 int macvlan_link_register(struct rtnl_link_ops *ops)
1135 {
1136 /* common fields */
1137 ops->priv_size = sizeof(struct macvlan_dev);
1138 ops->validate = macvlan_validate;
1139 ops->maxtype = IFLA_MACVLAN_MAX;
1140 ops->policy = macvlan_policy;
1141 ops->changelink = macvlan_changelink;
1142 ops->get_size = macvlan_get_size;
1143 ops->fill_info = macvlan_fill_info;
1144
1145 return rtnl_link_register(ops);
1146 };
1147 EXPORT_SYMBOL_GPL(macvlan_link_register);
1148
1149 static struct rtnl_link_ops macvlan_link_ops = {
1150 .kind = "macvlan",
1151 .setup = macvlan_setup,
1152 .newlink = macvlan_newlink,
1153 .dellink = macvlan_dellink,
1154 };
1155
1156 static int macvlan_device_event(struct notifier_block *unused,
1157 unsigned long event, void *ptr)
1158 {
1159 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1160 struct macvlan_dev *vlan, *next;
1161 struct macvlan_port *port;
1162 LIST_HEAD(list_kill);
1163
1164 if (!macvlan_port_exists(dev))
1165 return NOTIFY_DONE;
1166
1167 port = macvlan_port_get_rtnl(dev);
1168
1169 switch (event) {
1170 case NETDEV_CHANGE:
1171 list_for_each_entry(vlan, &port->vlans, list)
1172 netif_stacked_transfer_operstate(vlan->lowerdev,
1173 vlan->dev);
1174 break;
1175 case NETDEV_FEAT_CHANGE:
1176 list_for_each_entry(vlan, &port->vlans, list) {
1177 vlan->dev->gso_max_size = dev->gso_max_size;
1178 netdev_update_features(vlan->dev);
1179 }
1180 break;
1181 case NETDEV_CHANGEMTU:
1182 list_for_each_entry(vlan, &port->vlans, list) {
1183 if (vlan->dev->mtu <= dev->mtu)
1184 continue;
1185 dev_set_mtu(vlan->dev, dev->mtu);
1186 }
1187 break;
1188 case NETDEV_CHANGEADDR:
1189 if (!port->passthru)
1190 return NOTIFY_DONE;
1191
1192 vlan = list_first_entry_or_null(&port->vlans,
1193 struct macvlan_dev,
1194 list);
1195
1196 if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1197 return NOTIFY_BAD;
1198
1199 break;
1200 case NETDEV_UNREGISTER:
1201 /* twiddle thumbs on netns device moves */
1202 if (dev->reg_state != NETREG_UNREGISTERING)
1203 break;
1204
1205 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1206 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1207 unregister_netdevice_many(&list_kill);
1208 break;
1209 case NETDEV_PRE_TYPE_CHANGE:
1210 /* Forbid underlaying device to change its type. */
1211 return NOTIFY_BAD;
1212
1213 case NETDEV_NOTIFY_PEERS:
1214 case NETDEV_BONDING_FAILOVER:
1215 case NETDEV_RESEND_IGMP:
1216 /* Propagate to all vlans */
1217 list_for_each_entry(vlan, &port->vlans, list)
1218 call_netdevice_notifiers(event, vlan->dev);
1219 }
1220 return NOTIFY_DONE;
1221 }
1222
1223 static struct notifier_block macvlan_notifier_block __read_mostly = {
1224 .notifier_call = macvlan_device_event,
1225 };
1226
1227 static int __init macvlan_init_module(void)
1228 {
1229 int err;
1230
1231 register_netdevice_notifier(&macvlan_notifier_block);
1232
1233 err = macvlan_link_register(&macvlan_link_ops);
1234 if (err < 0)
1235 goto err1;
1236 return 0;
1237 err1:
1238 unregister_netdevice_notifier(&macvlan_notifier_block);
1239 return err;
1240 }
1241
1242 static void __exit macvlan_cleanup_module(void)
1243 {
1244 rtnl_link_unregister(&macvlan_link_ops);
1245 unregister_netdevice_notifier(&macvlan_notifier_block);
1246 }
1247
1248 module_init(macvlan_init_module);
1249 module_exit(macvlan_cleanup_module);
1250
1251 MODULE_LICENSE("GPL");
1252 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1253 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1254 MODULE_ALIAS_RTNL_LINK("macvlan");
Linux kernel - Deliverables
This page took 0.056485 seconds and 5 git commands to generate.