Revert "macvlan : fix checksums error when we are in bridge mode"
[firefly-linux-kernel-4.4.55.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         if (!skb)
122                 return NET_RX_DROP;
123
124         if (local)
125                 return vlan->forward(dev, skb);
126
127         skb->dev = dev;
128         if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
129                 skb->pkt_type = PACKET_BROADCAST;
130         else
131                 skb->pkt_type = PACKET_MULTICAST;
132
133         return vlan->receive(skb);
134 }
135
136 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
137 {
138         return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
139 }
140
141
142 static unsigned int mc_hash(const struct macvlan_dev *vlan,
143                             const unsigned char *addr)
144 {
145         u32 val = __get_unaligned_cpu32(addr + 2);
146
147         val ^= macvlan_hash_mix(vlan);
148         return hash_32(val, MACVLAN_MC_FILTER_BITS);
149 }
150
151 static void macvlan_broadcast(struct sk_buff *skb,
152                               const struct macvlan_port *port,
153                               struct net_device *src,
154                               enum macvlan_mode mode)
155 {
156         const struct ethhdr *eth = eth_hdr(skb);
157         const struct macvlan_dev *vlan;
158         struct sk_buff *nskb;
159         unsigned int i;
160         int err;
161         unsigned int hash;
162
163         if (skb->protocol == htons(ETH_P_PAUSE))
164                 return;
165
166         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
167                 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
168                         if (vlan->dev == src || !(vlan->mode & mode))
169                                 continue;
170
171                         hash = mc_hash(vlan, eth->h_dest);
172                         if (!test_bit(hash, vlan->mc_filter))
173                                 continue;
174                         nskb = skb_clone(skb, GFP_ATOMIC);
175                         err = macvlan_broadcast_one(nskb, vlan, eth,
176                                          mode == MACVLAN_MODE_BRIDGE);
177                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
178                                          err == NET_RX_SUCCESS, 1);
179                 }
180         }
181 }
182
183 /* called under rcu_read_lock() from netif_receive_skb */
184 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
185 {
186         struct macvlan_port *port;
187         struct sk_buff *skb = *pskb;
188         const struct ethhdr *eth = eth_hdr(skb);
189         const struct macvlan_dev *vlan;
190         const struct macvlan_dev *src;
191         struct net_device *dev;
192         unsigned int len = 0;
193         int ret = NET_RX_DROP;
194
195         port = macvlan_port_get_rcu(skb->dev);
196         if (is_multicast_ether_addr(eth->h_dest)) {
197                 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
198                 if (!skb)
199                         return RX_HANDLER_CONSUMED;
200                 eth = eth_hdr(skb);
201                 src = macvlan_hash_lookup(port, eth->h_source);
202                 if (!src)
203                         /* frame comes from an external address */
204                         macvlan_broadcast(skb, port, NULL,
205                                           MACVLAN_MODE_PRIVATE |
206                                           MACVLAN_MODE_VEPA    |
207                                           MACVLAN_MODE_PASSTHRU|
208                                           MACVLAN_MODE_BRIDGE);
209                 else if (src->mode == MACVLAN_MODE_VEPA)
210                         /* flood to everyone except source */
211                         macvlan_broadcast(skb, port, src->dev,
212                                           MACVLAN_MODE_VEPA |
213                                           MACVLAN_MODE_BRIDGE);
214                 else if (src->mode == MACVLAN_MODE_BRIDGE)
215                         /*
216                          * flood only to VEPA ports, bridge ports
217                          * already saw the frame on the way out.
218                          */
219                         macvlan_broadcast(skb, port, src->dev,
220                                           MACVLAN_MODE_VEPA);
221                 else {
222                         /* forward to original port. */
223                         vlan = src;
224                         ret = macvlan_broadcast_one(skb, vlan, eth, 0);
225                         goto out;
226                 }
227
228                 return RX_HANDLER_PASS;
229         }
230
231         if (port->passthru)
232                 vlan = list_first_or_null_rcu(&port->vlans,
233                                               struct macvlan_dev, list);
234         else
235                 vlan = macvlan_hash_lookup(port, eth->h_dest);
236         if (vlan == NULL)
237                 return RX_HANDLER_PASS;
238
239         dev = vlan->dev;
240         if (unlikely(!(dev->flags & IFF_UP))) {
241                 kfree_skb(skb);
242                 return RX_HANDLER_CONSUMED;
243         }
244         len = skb->len + ETH_HLEN;
245         skb = skb_share_check(skb, GFP_ATOMIC);
246         if (!skb)
247                 goto out;
248
249         skb->dev = dev;
250         skb->pkt_type = PACKET_HOST;
251
252         ret = vlan->receive(skb);
253
254 out:
255         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
256         return RX_HANDLER_CONSUMED;
257 }
258
259 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
260 {
261         const struct macvlan_dev *vlan = netdev_priv(dev);
262         const struct macvlan_port *port = vlan->port;
263         const struct macvlan_dev *dest;
264
265         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
266                 const struct ethhdr *eth = (void *)skb->data;
267
268                 /* send to other bridge ports directly */
269                 if (is_multicast_ether_addr(eth->h_dest)) {
270                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
271                         goto xmit_world;
272                 }
273
274                 dest = macvlan_hash_lookup(port, eth->h_dest);
275                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
276                         /* send to lowerdev first for its network taps */
277                         dev_forward_skb(vlan->lowerdev, skb);
278
279                         return NET_XMIT_SUCCESS;
280                 }
281         }
282
283 xmit_world:
284         skb->dev = vlan->lowerdev;
285         return dev_queue_xmit(skb);
286 }
287
288 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
289                                struct net_device *dev)
290 {
291         unsigned int len = skb->len;
292         int ret;
293         const struct macvlan_dev *vlan = netdev_priv(dev);
294
295         ret = macvlan_queue_xmit(skb, dev);
296         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
297                 struct macvlan_pcpu_stats *pcpu_stats;
298
299                 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
300                 u64_stats_update_begin(&pcpu_stats->syncp);
301                 pcpu_stats->tx_packets++;
302                 pcpu_stats->tx_bytes += len;
303                 u64_stats_update_end(&pcpu_stats->syncp);
304         } else {
305                 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
306         }
307         return ret;
308 }
309 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
310
311 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
312                                unsigned short type, const void *daddr,
313                                const void *saddr, unsigned len)
314 {
315         const struct macvlan_dev *vlan = netdev_priv(dev);
316         struct net_device *lowerdev = vlan->lowerdev;
317
318         return dev_hard_header(skb, lowerdev, type, daddr,
319                                saddr ? : dev->dev_addr, len);
320 }
321
322 static const struct header_ops macvlan_hard_header_ops = {
323         .create         = macvlan_hard_header,
324         .rebuild        = eth_rebuild_header,
325         .parse          = eth_header_parse,
326         .cache          = eth_header_cache,
327         .cache_update   = eth_header_cache_update,
328 };
329
330 static int macvlan_open(struct net_device *dev)
331 {
332         struct macvlan_dev *vlan = netdev_priv(dev);
333         struct net_device *lowerdev = vlan->lowerdev;
334         int err;
335
336         if (vlan->port->passthru) {
337                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
338                         dev_set_promiscuity(lowerdev, 1);
339                 goto hash_add;
340         }
341
342         err = -EBUSY;
343         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
344                 goto out;
345
346         err = dev_uc_add(lowerdev, dev->dev_addr);
347         if (err < 0)
348                 goto out;
349         if (dev->flags & IFF_ALLMULTI) {
350                 err = dev_set_allmulti(lowerdev, 1);
351                 if (err < 0)
352                         goto del_unicast;
353         }
354
355 hash_add:
356         macvlan_hash_add(vlan);
357         return 0;
358
359 del_unicast:
360         dev_uc_del(lowerdev, dev->dev_addr);
361 out:
362         return err;
363 }
364
365 static int macvlan_stop(struct net_device *dev)
366 {
367         struct macvlan_dev *vlan = netdev_priv(dev);
368         struct net_device *lowerdev = vlan->lowerdev;
369
370         dev_uc_unsync(lowerdev, dev);
371         dev_mc_unsync(lowerdev, dev);
372
373         if (vlan->port->passthru) {
374                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
375                         dev_set_promiscuity(lowerdev, -1);
376                 goto hash_del;
377         }
378
379         if (dev->flags & IFF_ALLMULTI)
380                 dev_set_allmulti(lowerdev, -1);
381
382         dev_uc_del(lowerdev, dev->dev_addr);
383
384 hash_del:
385         macvlan_hash_del(vlan, !dev->dismantle);
386         return 0;
387 }
388
389 static int macvlan_set_mac_address(struct net_device *dev, void *p)
390 {
391         struct macvlan_dev *vlan = netdev_priv(dev);
392         struct net_device *lowerdev = vlan->lowerdev;
393         struct sockaddr *addr = p;
394         int err;
395
396         if (!is_valid_ether_addr(addr->sa_data))
397                 return -EADDRNOTAVAIL;
398
399         if (!(dev->flags & IFF_UP)) {
400                 /* Just copy in the new address */
401                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
402         } else {
403                 /* Rehash and update the device filters */
404                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
405                         return -EBUSY;
406
407                 err = dev_uc_add(lowerdev, addr->sa_data);
408                 if (err)
409                         return err;
410
411                 dev_uc_del(lowerdev, dev->dev_addr);
412
413                 macvlan_hash_change_addr(vlan, addr->sa_data);
414         }
415         return 0;
416 }
417
418 static void macvlan_change_rx_flags(struct net_device *dev, int change)
419 {
420         struct macvlan_dev *vlan = netdev_priv(dev);
421         struct net_device *lowerdev = vlan->lowerdev;
422
423         if (change & IFF_ALLMULTI)
424                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
425 }
426
427 static void macvlan_set_mac_lists(struct net_device *dev)
428 {
429         struct macvlan_dev *vlan = netdev_priv(dev);
430
431         if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
432                 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
433         } else {
434                 struct netdev_hw_addr *ha;
435                 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
436
437                 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
438                 netdev_for_each_mc_addr(ha, dev) {
439                         __set_bit(mc_hash(vlan, ha->addr), filter);
440                 }
441
442                 __set_bit(mc_hash(vlan, dev->broadcast), filter);
443
444                 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
445         }
446         dev_uc_sync(vlan->lowerdev, dev);
447         dev_mc_sync(vlan->lowerdev, dev);
448 }
449
450 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
451 {
452         struct macvlan_dev *vlan = netdev_priv(dev);
453
454         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
455                 return -EINVAL;
456         dev->mtu = new_mtu;
457         return 0;
458 }
459
460 /*
461  * macvlan network devices have devices nesting below it and are a special
462  * "super class" of normal network devices; split their locks off into a
463  * separate class since they always nest.
464  */
465 static struct lock_class_key macvlan_netdev_xmit_lock_key;
466 static struct lock_class_key macvlan_netdev_addr_lock_key;
467
468 #define MACVLAN_FEATURES \
469         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
470          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
471          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
472          NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
473
474 #define MACVLAN_STATE_MASK \
475         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
476
477 static void macvlan_set_lockdep_class_one(struct net_device *dev,
478                                           struct netdev_queue *txq,
479                                           void *_unused)
480 {
481         lockdep_set_class(&txq->_xmit_lock,
482                           &macvlan_netdev_xmit_lock_key);
483 }
484
485 static void macvlan_set_lockdep_class(struct net_device *dev)
486 {
487         lockdep_set_class(&dev->addr_list_lock,
488                           &macvlan_netdev_addr_lock_key);
489         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
490 }
491
492 static int macvlan_init(struct net_device *dev)
493 {
494         struct macvlan_dev *vlan = netdev_priv(dev);
495         const struct net_device *lowerdev = vlan->lowerdev;
496
497         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
498                                   (lowerdev->state & MACVLAN_STATE_MASK);
499         dev->features           = lowerdev->features & MACVLAN_FEATURES;
500         dev->features           |= NETIF_F_LLTX;
501         dev->gso_max_size       = lowerdev->gso_max_size;
502         dev->iflink             = lowerdev->ifindex;
503         dev->hard_header_len    = lowerdev->hard_header_len;
504
505         macvlan_set_lockdep_class(dev);
506
507         vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
508         if (!vlan->pcpu_stats)
509                 return -ENOMEM;
510
511         return 0;
512 }
513
514 static void macvlan_uninit(struct net_device *dev)
515 {
516         struct macvlan_dev *vlan = netdev_priv(dev);
517         struct macvlan_port *port = vlan->port;
518
519         free_percpu(vlan->pcpu_stats);
520
521         port->count -= 1;
522         if (!port->count)
523                 macvlan_port_destroy(port->dev);
524 }
525
526 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
527                                                          struct rtnl_link_stats64 *stats)
528 {
529         struct macvlan_dev *vlan = netdev_priv(dev);
530
531         if (vlan->pcpu_stats) {
532                 struct macvlan_pcpu_stats *p;
533                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
534                 u32 rx_errors = 0, tx_dropped = 0;
535                 unsigned int start;
536                 int i;
537
538                 for_each_possible_cpu(i) {
539                         p = per_cpu_ptr(vlan->pcpu_stats, i);
540                         do {
541                                 start = u64_stats_fetch_begin_bh(&p->syncp);
542                                 rx_packets      = p->rx_packets;
543                                 rx_bytes        = p->rx_bytes;
544                                 rx_multicast    = p->rx_multicast;
545                                 tx_packets      = p->tx_packets;
546                                 tx_bytes        = p->tx_bytes;
547                         } while (u64_stats_fetch_retry_bh(&p->syncp, start));
548
549                         stats->rx_packets       += rx_packets;
550                         stats->rx_bytes         += rx_bytes;
551                         stats->multicast        += rx_multicast;
552                         stats->tx_packets       += tx_packets;
553                         stats->tx_bytes         += tx_bytes;
554                         /* rx_errors & tx_dropped are u32, updated
555                          * without syncp protection.
556                          */
557                         rx_errors       += p->rx_errors;
558                         tx_dropped      += p->tx_dropped;
559                 }
560                 stats->rx_errors        = rx_errors;
561                 stats->rx_dropped       = rx_errors;
562                 stats->tx_dropped       = tx_dropped;
563         }
564         return stats;
565 }
566
567 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
568                                    __be16 proto, u16 vid)
569 {
570         struct macvlan_dev *vlan = netdev_priv(dev);
571         struct net_device *lowerdev = vlan->lowerdev;
572
573         return vlan_vid_add(lowerdev, proto, vid);
574 }
575
576 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
577                                     __be16 proto, u16 vid)
578 {
579         struct macvlan_dev *vlan = netdev_priv(dev);
580         struct net_device *lowerdev = vlan->lowerdev;
581
582         vlan_vid_del(lowerdev, proto, vid);
583         return 0;
584 }
585
586 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
587                            struct net_device *dev,
588                            const unsigned char *addr,
589                            u16 flags)
590 {
591         struct macvlan_dev *vlan = netdev_priv(dev);
592         int err = -EINVAL;
593
594         if (!vlan->port->passthru)
595                 return -EOPNOTSUPP;
596
597         if (is_unicast_ether_addr(addr))
598                 err = dev_uc_add_excl(dev, addr);
599         else if (is_multicast_ether_addr(addr))
600                 err = dev_mc_add_excl(dev, addr);
601
602         return err;
603 }
604
605 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
606                            struct net_device *dev,
607                            const unsigned char *addr)
608 {
609         struct macvlan_dev *vlan = netdev_priv(dev);
610         int err = -EINVAL;
611
612         if (!vlan->port->passthru)
613                 return -EOPNOTSUPP;
614
615         if (is_unicast_ether_addr(addr))
616                 err = dev_uc_del(dev, addr);
617         else if (is_multicast_ether_addr(addr))
618                 err = dev_mc_del(dev, addr);
619
620         return err;
621 }
622
623 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
624                                         struct ethtool_drvinfo *drvinfo)
625 {
626         strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
627         strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
628 }
629
630 static int macvlan_ethtool_get_settings(struct net_device *dev,
631                                         struct ethtool_cmd *cmd)
632 {
633         const struct macvlan_dev *vlan = netdev_priv(dev);
634
635         return __ethtool_get_settings(vlan->lowerdev, cmd);
636 }
637
638 static const struct ethtool_ops macvlan_ethtool_ops = {
639         .get_link               = ethtool_op_get_link,
640         .get_settings           = macvlan_ethtool_get_settings,
641         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
642 };
643
644 static const struct net_device_ops macvlan_netdev_ops = {
645         .ndo_init               = macvlan_init,
646         .ndo_uninit             = macvlan_uninit,
647         .ndo_open               = macvlan_open,
648         .ndo_stop               = macvlan_stop,
649         .ndo_start_xmit         = macvlan_start_xmit,
650         .ndo_change_mtu         = macvlan_change_mtu,
651         .ndo_change_rx_flags    = macvlan_change_rx_flags,
652         .ndo_set_mac_address    = macvlan_set_mac_address,
653         .ndo_set_rx_mode        = macvlan_set_mac_lists,
654         .ndo_get_stats64        = macvlan_dev_get_stats64,
655         .ndo_validate_addr      = eth_validate_addr,
656         .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
657         .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
658         .ndo_fdb_add            = macvlan_fdb_add,
659         .ndo_fdb_del            = macvlan_fdb_del,
660         .ndo_fdb_dump           = ndo_dflt_fdb_dump,
661 };
662
663 void macvlan_common_setup(struct net_device *dev)
664 {
665         ether_setup(dev);
666
667         dev->priv_flags        &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
668         dev->priv_flags        |= IFF_UNICAST_FLT;
669         dev->netdev_ops         = &macvlan_netdev_ops;
670         dev->destructor         = free_netdev;
671         dev->header_ops         = &macvlan_hard_header_ops,
672         dev->ethtool_ops        = &macvlan_ethtool_ops;
673 }
674 EXPORT_SYMBOL_GPL(macvlan_common_setup);
675
676 static void macvlan_setup(struct net_device *dev)
677 {
678         macvlan_common_setup(dev);
679         dev->tx_queue_len       = 0;
680 }
681
682 static int macvlan_port_create(struct net_device *dev)
683 {
684         struct macvlan_port *port;
685         unsigned int i;
686         int err;
687
688         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
689                 return -EINVAL;
690
691         port = kzalloc(sizeof(*port), GFP_KERNEL);
692         if (port == NULL)
693                 return -ENOMEM;
694
695         port->passthru = false;
696         port->dev = dev;
697         INIT_LIST_HEAD(&port->vlans);
698         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
699                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
700
701         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
702         if (err)
703                 kfree(port);
704         else
705                 dev->priv_flags |= IFF_MACVLAN_PORT;
706         return err;
707 }
708
709 static void macvlan_port_destroy(struct net_device *dev)
710 {
711         struct macvlan_port *port = macvlan_port_get_rtnl(dev);
712
713         dev->priv_flags &= ~IFF_MACVLAN_PORT;
714         netdev_rx_handler_unregister(dev);
715         kfree_rcu(port, rcu);
716 }
717
718 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
719 {
720         if (tb[IFLA_ADDRESS]) {
721                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
722                         return -EINVAL;
723                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
724                         return -EADDRNOTAVAIL;
725         }
726
727         if (data && data[IFLA_MACVLAN_FLAGS] &&
728             nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
729                 return -EINVAL;
730
731         if (data && data[IFLA_MACVLAN_MODE]) {
732                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
733                 case MACVLAN_MODE_PRIVATE:
734                 case MACVLAN_MODE_VEPA:
735                 case MACVLAN_MODE_BRIDGE:
736                 case MACVLAN_MODE_PASSTHRU:
737                         break;
738                 default:
739                         return -EINVAL;
740                 }
741         }
742         return 0;
743 }
744
745 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
746                            struct nlattr *tb[], struct nlattr *data[],
747                            int (*receive)(struct sk_buff *skb),
748                            int (*forward)(struct net_device *dev,
749                                           struct sk_buff *skb))
750 {
751         struct macvlan_dev *vlan = netdev_priv(dev);
752         struct macvlan_port *port;
753         struct net_device *lowerdev;
754         int err;
755
756         if (!tb[IFLA_LINK])
757                 return -EINVAL;
758
759         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
760         if (lowerdev == NULL)
761                 return -ENODEV;
762
763         /* When creating macvlans on top of other macvlans - use
764          * the real device as the lowerdev.
765          */
766         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
767                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
768                 lowerdev = lowervlan->lowerdev;
769         }
770
771         if (!tb[IFLA_MTU])
772                 dev->mtu = lowerdev->mtu;
773         else if (dev->mtu > lowerdev->mtu)
774                 return -EINVAL;
775
776         if (!tb[IFLA_ADDRESS])
777                 eth_hw_addr_random(dev);
778
779         if (!macvlan_port_exists(lowerdev)) {
780                 err = macvlan_port_create(lowerdev);
781                 if (err < 0)
782                         return err;
783         }
784         port = macvlan_port_get_rtnl(lowerdev);
785
786         /* Only 1 macvlan device can be created in passthru mode */
787         if (port->passthru)
788                 return -EINVAL;
789
790         vlan->lowerdev = lowerdev;
791         vlan->dev      = dev;
792         vlan->port     = port;
793         vlan->receive  = receive;
794         vlan->forward  = forward;
795
796         vlan->mode     = MACVLAN_MODE_VEPA;
797         if (data && data[IFLA_MACVLAN_MODE])
798                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
799
800         if (data && data[IFLA_MACVLAN_FLAGS])
801                 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
802
803         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
804                 if (port->count)
805                         return -EINVAL;
806                 port->passthru = true;
807                 memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
808         }
809
810         err = netdev_upper_dev_link(lowerdev, dev);
811         if (err)
812                 goto destroy_port;
813
814         port->count += 1;
815         err = register_netdevice(dev);
816         if (err < 0)
817                 goto upper_dev_unlink;
818
819         list_add_tail_rcu(&vlan->list, &port->vlans);
820         netif_stacked_transfer_operstate(lowerdev, dev);
821
822         return 0;
823
824 upper_dev_unlink:
825         netdev_upper_dev_unlink(lowerdev, dev);
826 destroy_port:
827         port->count -= 1;
828         if (!port->count)
829                 macvlan_port_destroy(lowerdev);
830
831         return err;
832 }
833 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
834
835 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
836                            struct nlattr *tb[], struct nlattr *data[])
837 {
838         return macvlan_common_newlink(src_net, dev, tb, data,
839                                       netif_rx,
840                                       dev_forward_skb);
841 }
842
843 void macvlan_dellink(struct net_device *dev, struct list_head *head)
844 {
845         struct macvlan_dev *vlan = netdev_priv(dev);
846
847         list_del_rcu(&vlan->list);
848         unregister_netdevice_queue(dev, head);
849         netdev_upper_dev_unlink(vlan->lowerdev, dev);
850 }
851 EXPORT_SYMBOL_GPL(macvlan_dellink);
852
853 static int macvlan_changelink(struct net_device *dev,
854                 struct nlattr *tb[], struct nlattr *data[])
855 {
856         struct macvlan_dev *vlan = netdev_priv(dev);
857
858         if (data && data[IFLA_MACVLAN_FLAGS]) {
859                 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
860                 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
861                 if (vlan->port->passthru && promisc) {
862                         int err;
863
864                         if (flags & MACVLAN_FLAG_NOPROMISC)
865                                 err = dev_set_promiscuity(vlan->lowerdev, -1);
866                         else
867                                 err = dev_set_promiscuity(vlan->lowerdev, 1);
868                         if (err < 0)
869                                 return err;
870                 }
871                 vlan->flags = flags;
872         }
873         if (data && data[IFLA_MACVLAN_MODE])
874                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
875         return 0;
876 }
877
878 static size_t macvlan_get_size(const struct net_device *dev)
879 {
880         return (0
881                 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
882                 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
883                 );
884 }
885
886 static int macvlan_fill_info(struct sk_buff *skb,
887                                 const struct net_device *dev)
888 {
889         struct macvlan_dev *vlan = netdev_priv(dev);
890
891         if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
892                 goto nla_put_failure;
893         if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
894                 goto nla_put_failure;
895         return 0;
896
897 nla_put_failure:
898         return -EMSGSIZE;
899 }
900
901 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
902         [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
903         [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
904 };
905
906 int macvlan_link_register(struct rtnl_link_ops *ops)
907 {
908         /* common fields */
909         ops->priv_size          = sizeof(struct macvlan_dev);
910         ops->validate           = macvlan_validate;
911         ops->maxtype            = IFLA_MACVLAN_MAX;
912         ops->policy             = macvlan_policy;
913         ops->changelink         = macvlan_changelink;
914         ops->get_size           = macvlan_get_size;
915         ops->fill_info          = macvlan_fill_info;
916
917         return rtnl_link_register(ops);
918 };
919 EXPORT_SYMBOL_GPL(macvlan_link_register);
920
921 static struct rtnl_link_ops macvlan_link_ops = {
922         .kind           = "macvlan",
923         .setup          = macvlan_setup,
924         .newlink        = macvlan_newlink,
925         .dellink        = macvlan_dellink,
926 };
927
928 static int macvlan_device_event(struct notifier_block *unused,
929                                 unsigned long event, void *ptr)
930 {
931         struct net_device *dev = ptr;
932         struct macvlan_dev *vlan, *next;
933         struct macvlan_port *port;
934         LIST_HEAD(list_kill);
935
936         if (!macvlan_port_exists(dev))
937                 return NOTIFY_DONE;
938
939         port = macvlan_port_get_rtnl(dev);
940
941         switch (event) {
942         case NETDEV_CHANGE:
943                 list_for_each_entry(vlan, &port->vlans, list)
944                         netif_stacked_transfer_operstate(vlan->lowerdev,
945                                                          vlan->dev);
946                 break;
947         case NETDEV_FEAT_CHANGE:
948                 list_for_each_entry(vlan, &port->vlans, list) {
949                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
950                         vlan->dev->gso_max_size = dev->gso_max_size;
951                         netdev_features_change(vlan->dev);
952                 }
953                 break;
954         case NETDEV_UNREGISTER:
955                 /* twiddle thumbs on netns device moves */
956                 if (dev->reg_state != NETREG_UNREGISTERING)
957                         break;
958
959                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
960                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
961                 unregister_netdevice_many(&list_kill);
962                 list_del(&list_kill);
963                 break;
964         case NETDEV_PRE_TYPE_CHANGE:
965                 /* Forbid underlaying device to change its type. */
966                 return NOTIFY_BAD;
967         }
968         return NOTIFY_DONE;
969 }
970
971 static struct notifier_block macvlan_notifier_block __read_mostly = {
972         .notifier_call  = macvlan_device_event,
973 };
974
975 static int __init macvlan_init_module(void)
976 {
977         int err;
978
979         register_netdevice_notifier(&macvlan_notifier_block);
980
981         err = macvlan_link_register(&macvlan_link_ops);
982         if (err < 0)
983                 goto err1;
984         return 0;
985 err1:
986         unregister_netdevice_notifier(&macvlan_notifier_block);
987         return err;
988 }
989
990 static void __exit macvlan_cleanup_module(void)
991 {
992         rtnl_link_unregister(&macvlan_link_ops);
993         unregister_netdevice_notifier(&macvlan_notifier_block);
994 }
995
996 module_init(macvlan_init_module);
997 module_exit(macvlan_cleanup_module);
998
999 MODULE_LICENSE("GPL");
1000 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1001 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1002 MODULE_ALIAS_RTNL_LINK("macvlan");