Merge tag 'iio-fixes-for-6.1c' of https://git.kernel.org/pub/scm/linux/kernel/git...
[platform/kernel/linux-starfive.git] / drivers / net / macvlan.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4  *
5  * The code this is based on carried the following copyright notice:
6  * ---
7  * (C) Copyright 2001-2006
8  * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
9  * Re-worked by Ben Greear <greearb@candelatech.com>
10  * ---
11  */
12 #include <linux/kernel.h>
13 #include <linux/types.h>
14 #include <linux/module.h>
15 #include <linux/init.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/string.h>
19 #include <linux/rculist.h>
20 #include <linux/notifier.h>
21 #include <linux/netdevice.h>
22 #include <linux/etherdevice.h>
23 #include <linux/net_tstamp.h>
24 #include <linux/ethtool.h>
25 #include <linux/if_arp.h>
26 #include <linux/if_vlan.h>
27 #include <linux/if_link.h>
28 #include <linux/if_macvlan.h>
29 #include <linux/hash.h>
30 #include <linux/workqueue.h>
31 #include <net/rtnetlink.h>
32 #include <net/xfrm.h>
33 #include <linux/netpoll.h>
34 #include <linux/phy.h>
35
36 #define MACVLAN_HASH_BITS       8
37 #define MACVLAN_HASH_SIZE       (1<<MACVLAN_HASH_BITS)
38 #define MACVLAN_DEFAULT_BC_QUEUE_LEN    1000
39
40 #define MACVLAN_F_PASSTHRU      1
41 #define MACVLAN_F_ADDRCHANGE    2
42
43 struct macvlan_port {
44         struct net_device       *dev;
45         struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
46         struct list_head        vlans;
47         struct sk_buff_head     bc_queue;
48         struct work_struct      bc_work;
49         u32                     bc_queue_len_used;
50         u32                     flags;
51         int                     count;
52         struct hlist_head       vlan_source_hash[MACVLAN_HASH_SIZE];
53         DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
54         unsigned char           perm_addr[ETH_ALEN];
55 };
56
57 struct macvlan_source_entry {
58         struct hlist_node       hlist;
59         struct macvlan_dev      *vlan;
60         unsigned char           addr[6+2] __aligned(sizeof(u16));
61         struct rcu_head         rcu;
62 };
63
64 struct macvlan_skb_cb {
65         const struct macvlan_dev *src;
66 };
67
68 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
69
70 static void macvlan_port_destroy(struct net_device *dev);
71 static void update_port_bc_queue_len(struct macvlan_port *port);
72
73 static inline bool macvlan_passthru(const struct macvlan_port *port)
74 {
75         return port->flags & MACVLAN_F_PASSTHRU;
76 }
77
78 static inline void macvlan_set_passthru(struct macvlan_port *port)
79 {
80         port->flags |= MACVLAN_F_PASSTHRU;
81 }
82
83 static inline bool macvlan_addr_change(const struct macvlan_port *port)
84 {
85         return port->flags & MACVLAN_F_ADDRCHANGE;
86 }
87
88 static inline void macvlan_set_addr_change(struct macvlan_port *port)
89 {
90         port->flags |= MACVLAN_F_ADDRCHANGE;
91 }
92
93 static inline void macvlan_clear_addr_change(struct macvlan_port *port)
94 {
95         port->flags &= ~MACVLAN_F_ADDRCHANGE;
96 }
97
98 /* Hash Ethernet address */
99 static u32 macvlan_eth_hash(const unsigned char *addr)
100 {
101         u64 value = get_unaligned((u64 *)addr);
102
103         /* only want 6 bytes */
104 #ifdef __BIG_ENDIAN
105         value >>= 16;
106 #else
107         value <<= 16;
108 #endif
109         return hash_64(value, MACVLAN_HASH_BITS);
110 }
111
112 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
113 {
114         return rcu_dereference(dev->rx_handler_data);
115 }
116
117 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
118 {
119         return rtnl_dereference(dev->rx_handler_data);
120 }
121
122 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
123                                                const unsigned char *addr)
124 {
125         struct macvlan_dev *vlan;
126         u32 idx = macvlan_eth_hash(addr);
127
128         hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
129                                  lockdep_rtnl_is_held()) {
130                 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
131                         return vlan;
132         }
133         return NULL;
134 }
135
136 static struct macvlan_source_entry *macvlan_hash_lookup_source(
137         const struct macvlan_dev *vlan,
138         const unsigned char *addr)
139 {
140         struct macvlan_source_entry *entry;
141         u32 idx = macvlan_eth_hash(addr);
142         struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
143
144         hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
145                 if (ether_addr_equal_64bits(entry->addr, addr) &&
146                     entry->vlan == vlan)
147                         return entry;
148         }
149         return NULL;
150 }
151
152 static int macvlan_hash_add_source(struct macvlan_dev *vlan,
153                                    const unsigned char *addr)
154 {
155         struct macvlan_port *port = vlan->port;
156         struct macvlan_source_entry *entry;
157         struct hlist_head *h;
158
159         entry = macvlan_hash_lookup_source(vlan, addr);
160         if (entry)
161                 return 0;
162
163         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
164         if (!entry)
165                 return -ENOMEM;
166
167         ether_addr_copy(entry->addr, addr);
168         entry->vlan = vlan;
169         h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
170         hlist_add_head_rcu(&entry->hlist, h);
171         vlan->macaddr_count++;
172
173         return 0;
174 }
175
176 static void macvlan_hash_add(struct macvlan_dev *vlan)
177 {
178         struct macvlan_port *port = vlan->port;
179         const unsigned char *addr = vlan->dev->dev_addr;
180         u32 idx = macvlan_eth_hash(addr);
181
182         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
183 }
184
185 static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
186 {
187         hlist_del_rcu(&entry->hlist);
188         kfree_rcu(entry, rcu);
189 }
190
191 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
192 {
193         hlist_del_rcu(&vlan->hlist);
194         if (sync)
195                 synchronize_rcu();
196 }
197
198 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
199                                         const unsigned char *addr)
200 {
201         macvlan_hash_del(vlan, true);
202         /* Now that we are unhashed it is safe to change the device
203          * address without confusing packet delivery.
204          */
205         eth_hw_addr_set(vlan->dev, addr);
206         macvlan_hash_add(vlan);
207 }
208
209 static bool macvlan_addr_busy(const struct macvlan_port *port,
210                               const unsigned char *addr)
211 {
212         /* Test to see if the specified address is
213          * currently in use by the underlying device or
214          * another macvlan.
215          */
216         if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
217             ether_addr_equal_64bits(port->dev->dev_addr, addr))
218                 return true;
219
220         if (macvlan_hash_lookup(port, addr))
221                 return true;
222
223         return false;
224 }
225
226
227 static int macvlan_broadcast_one(struct sk_buff *skb,
228                                  const struct macvlan_dev *vlan,
229                                  const struct ethhdr *eth, bool local)
230 {
231         struct net_device *dev = vlan->dev;
232
233         if (local)
234                 return __dev_forward_skb(dev, skb);
235
236         skb->dev = dev;
237         if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
238                 skb->pkt_type = PACKET_BROADCAST;
239         else
240                 skb->pkt_type = PACKET_MULTICAST;
241
242         return 0;
243 }
244
245 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
246 {
247         return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
248 }
249
250
251 static unsigned int mc_hash(const struct macvlan_dev *vlan,
252                             const unsigned char *addr)
253 {
254         u32 val = __get_unaligned_cpu32(addr + 2);
255
256         val ^= macvlan_hash_mix(vlan);
257         return hash_32(val, MACVLAN_MC_FILTER_BITS);
258 }
259
260 static void macvlan_broadcast(struct sk_buff *skb,
261                               const struct macvlan_port *port,
262                               struct net_device *src,
263                               enum macvlan_mode mode)
264 {
265         const struct ethhdr *eth = eth_hdr(skb);
266         const struct macvlan_dev *vlan;
267         struct sk_buff *nskb;
268         unsigned int i;
269         int err;
270         unsigned int hash;
271
272         if (skb->protocol == htons(ETH_P_PAUSE))
273                 return;
274
275         hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) {
276                 if (vlan->dev == src || !(vlan->mode & mode))
277                         continue;
278
279                 hash = mc_hash(vlan, eth->h_dest);
280                 if (!test_bit(hash, vlan->mc_filter))
281                         continue;
282
283                 err = NET_RX_DROP;
284                 nskb = skb_clone(skb, GFP_ATOMIC);
285                 if (likely(nskb))
286                         err = macvlan_broadcast_one(nskb, vlan, eth,
287                                         mode == MACVLAN_MODE_BRIDGE) ?:
288                               netif_rx(nskb);
289                 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
290                                  err == NET_RX_SUCCESS, true);
291         }
292 }
293
294 static void macvlan_process_broadcast(struct work_struct *w)
295 {
296         struct macvlan_port *port = container_of(w, struct macvlan_port,
297                                                  bc_work);
298         struct sk_buff *skb;
299         struct sk_buff_head list;
300
301         __skb_queue_head_init(&list);
302
303         spin_lock_bh(&port->bc_queue.lock);
304         skb_queue_splice_tail_init(&port->bc_queue, &list);
305         spin_unlock_bh(&port->bc_queue.lock);
306
307         while ((skb = __skb_dequeue(&list))) {
308                 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
309
310                 rcu_read_lock();
311
312                 if (!src)
313                         /* frame comes from an external address */
314                         macvlan_broadcast(skb, port, NULL,
315                                           MACVLAN_MODE_PRIVATE |
316                                           MACVLAN_MODE_VEPA    |
317                                           MACVLAN_MODE_PASSTHRU|
318                                           MACVLAN_MODE_BRIDGE);
319                 else if (src->mode == MACVLAN_MODE_VEPA)
320                         /* flood to everyone except source */
321                         macvlan_broadcast(skb, port, src->dev,
322                                           MACVLAN_MODE_VEPA |
323                                           MACVLAN_MODE_BRIDGE);
324                 else
325                         /*
326                          * flood only to VEPA ports, bridge ports
327                          * already saw the frame on the way out.
328                          */
329                         macvlan_broadcast(skb, port, src->dev,
330                                           MACVLAN_MODE_VEPA);
331
332                 rcu_read_unlock();
333
334                 if (src)
335                         dev_put(src->dev);
336                 consume_skb(skb);
337
338                 cond_resched();
339         }
340 }
341
342 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
343                                       const struct macvlan_dev *src,
344                                       struct sk_buff *skb)
345 {
346         struct sk_buff *nskb;
347         int err = -ENOMEM;
348
349         nskb = skb_clone(skb, GFP_ATOMIC);
350         if (!nskb)
351                 goto err;
352
353         MACVLAN_SKB_CB(nskb)->src = src;
354
355         spin_lock(&port->bc_queue.lock);
356         if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) {
357                 if (src)
358                         dev_hold(src->dev);
359                 __skb_queue_tail(&port->bc_queue, nskb);
360                 err = 0;
361         }
362         spin_unlock(&port->bc_queue.lock);
363
364         queue_work(system_unbound_wq, &port->bc_work);
365
366         if (err)
367                 goto free_nskb;
368
369         return;
370
371 free_nskb:
372         kfree_skb(nskb);
373 err:
374         dev_core_stats_rx_dropped_inc(skb->dev);
375 }
376
377 static void macvlan_flush_sources(struct macvlan_port *port,
378                                   struct macvlan_dev *vlan)
379 {
380         struct macvlan_source_entry *entry;
381         struct hlist_node *next;
382         int i;
383
384         hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
385                 if (entry->vlan == vlan)
386                         macvlan_hash_del_source(entry);
387
388         vlan->macaddr_count = 0;
389 }
390
391 static void macvlan_forward_source_one(struct sk_buff *skb,
392                                        struct macvlan_dev *vlan)
393 {
394         struct sk_buff *nskb;
395         struct net_device *dev;
396         int len;
397         int ret;
398
399         dev = vlan->dev;
400         if (unlikely(!(dev->flags & IFF_UP)))
401                 return;
402
403         nskb = skb_clone(skb, GFP_ATOMIC);
404         if (!nskb)
405                 return;
406
407         len = nskb->len + ETH_HLEN;
408         nskb->dev = dev;
409
410         if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
411                 nskb->pkt_type = PACKET_HOST;
412
413         ret = __netif_rx(nskb);
414         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
415 }
416
417 static bool macvlan_forward_source(struct sk_buff *skb,
418                                    struct macvlan_port *port,
419                                    const unsigned char *addr)
420 {
421         struct macvlan_source_entry *entry;
422         u32 idx = macvlan_eth_hash(addr);
423         struct hlist_head *h = &port->vlan_source_hash[idx];
424         bool consume = false;
425
426         hlist_for_each_entry_rcu(entry, h, hlist) {
427                 if (ether_addr_equal_64bits(entry->addr, addr)) {
428                         if (entry->vlan->flags & MACVLAN_FLAG_NODST)
429                                 consume = true;
430                         macvlan_forward_source_one(skb, entry->vlan);
431                 }
432         }
433
434         return consume;
435 }
436
437 /* called under rcu_read_lock() from netif_receive_skb */
438 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
439 {
440         struct macvlan_port *port;
441         struct sk_buff *skb = *pskb;
442         const struct ethhdr *eth = eth_hdr(skb);
443         const struct macvlan_dev *vlan;
444         const struct macvlan_dev *src;
445         struct net_device *dev;
446         unsigned int len = 0;
447         int ret;
448         rx_handler_result_t handle_res;
449
450         /* Packets from dev_loopback_xmit() do not have L2 header, bail out */
451         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
452                 return RX_HANDLER_PASS;
453
454         port = macvlan_port_get_rcu(skb->dev);
455         if (is_multicast_ether_addr(eth->h_dest)) {
456                 unsigned int hash;
457
458                 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
459                 if (!skb)
460                         return RX_HANDLER_CONSUMED;
461                 *pskb = skb;
462                 eth = eth_hdr(skb);
463                 if (macvlan_forward_source(skb, port, eth->h_source)) {
464                         kfree_skb(skb);
465                         return RX_HANDLER_CONSUMED;
466                 }
467                 src = macvlan_hash_lookup(port, eth->h_source);
468                 if (src && src->mode != MACVLAN_MODE_VEPA &&
469                     src->mode != MACVLAN_MODE_BRIDGE) {
470                         /* forward to original port. */
471                         vlan = src;
472                         ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
473                               __netif_rx(skb);
474                         handle_res = RX_HANDLER_CONSUMED;
475                         goto out;
476                 }
477
478                 hash = mc_hash(NULL, eth->h_dest);
479                 if (test_bit(hash, port->mc_filter))
480                         macvlan_broadcast_enqueue(port, src, skb);
481
482                 return RX_HANDLER_PASS;
483         }
484
485         if (macvlan_forward_source(skb, port, eth->h_source)) {
486                 kfree_skb(skb);
487                 return RX_HANDLER_CONSUMED;
488         }
489         if (macvlan_passthru(port))
490                 vlan = list_first_or_null_rcu(&port->vlans,
491                                               struct macvlan_dev, list);
492         else
493                 vlan = macvlan_hash_lookup(port, eth->h_dest);
494         if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
495                 return RX_HANDLER_PASS;
496
497         dev = vlan->dev;
498         if (unlikely(!(dev->flags & IFF_UP))) {
499                 kfree_skb(skb);
500                 return RX_HANDLER_CONSUMED;
501         }
502         len = skb->len + ETH_HLEN;
503         skb = skb_share_check(skb, GFP_ATOMIC);
504         if (!skb) {
505                 ret = NET_RX_DROP;
506                 handle_res = RX_HANDLER_CONSUMED;
507                 goto out;
508         }
509
510         *pskb = skb;
511         skb->dev = dev;
512         skb->pkt_type = PACKET_HOST;
513
514         ret = NET_RX_SUCCESS;
515         handle_res = RX_HANDLER_ANOTHER;
516 out:
517         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
518         return handle_res;
519 }
520
521 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
522 {
523         const struct macvlan_dev *vlan = netdev_priv(dev);
524         const struct macvlan_port *port = vlan->port;
525         const struct macvlan_dev *dest;
526
527         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
528                 const struct ethhdr *eth = skb_eth_hdr(skb);
529
530                 /* send to other bridge ports directly */
531                 if (is_multicast_ether_addr(eth->h_dest)) {
532                         skb_reset_mac_header(skb);
533                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
534                         goto xmit_world;
535                 }
536
537                 dest = macvlan_hash_lookup(port, eth->h_dest);
538                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
539                         /* send to lowerdev first for its network taps */
540                         dev_forward_skb(vlan->lowerdev, skb);
541
542                         return NET_XMIT_SUCCESS;
543                 }
544         }
545 xmit_world:
546         skb->dev = vlan->lowerdev;
547         return dev_queue_xmit_accel(skb,
548                                     netdev_get_sb_channel(dev) ? dev : NULL);
549 }
550
551 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
552 {
553 #ifdef CONFIG_NET_POLL_CONTROLLER
554         return netpoll_send_skb(vlan->netpoll, skb);
555 #else
556         BUG();
557         return NETDEV_TX_OK;
558 #endif
559 }
560
561 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
562                                       struct net_device *dev)
563 {
564         struct macvlan_dev *vlan = netdev_priv(dev);
565         unsigned int len = skb->len;
566         int ret;
567
568         if (unlikely(netpoll_tx_running(dev)))
569                 return macvlan_netpoll_send_skb(vlan, skb);
570
571         ret = macvlan_queue_xmit(skb, dev);
572
573         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
574                 struct vlan_pcpu_stats *pcpu_stats;
575
576                 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
577                 u64_stats_update_begin(&pcpu_stats->syncp);
578                 u64_stats_inc(&pcpu_stats->tx_packets);
579                 u64_stats_add(&pcpu_stats->tx_bytes, len);
580                 u64_stats_update_end(&pcpu_stats->syncp);
581         } else {
582                 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
583         }
584         return ret;
585 }
586
587 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
588                                unsigned short type, const void *daddr,
589                                const void *saddr, unsigned len)
590 {
591         const struct macvlan_dev *vlan = netdev_priv(dev);
592         struct net_device *lowerdev = vlan->lowerdev;
593
594         return dev_hard_header(skb, lowerdev, type, daddr,
595                                saddr ? : dev->dev_addr, len);
596 }
597
598 static const struct header_ops macvlan_hard_header_ops = {
599         .create         = macvlan_hard_header,
600         .parse          = eth_header_parse,
601         .cache          = eth_header_cache,
602         .cache_update   = eth_header_cache_update,
603 };
604
605 static int macvlan_open(struct net_device *dev)
606 {
607         struct macvlan_dev *vlan = netdev_priv(dev);
608         struct net_device *lowerdev = vlan->lowerdev;
609         int err;
610
611         if (macvlan_passthru(vlan->port)) {
612                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
613                         err = dev_set_promiscuity(lowerdev, 1);
614                         if (err < 0)
615                                 goto out;
616                 }
617                 goto hash_add;
618         }
619
620         err = -EADDRINUSE;
621         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
622                 goto out;
623
624         /* Attempt to populate accel_priv which is used to offload the L2
625          * forwarding requests for unicast packets.
626          */
627         if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
628                 vlan->accel_priv =
629                       lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
630
631         /* If earlier attempt to offload failed, or accel_priv is not
632          * populated we must add the unicast address to the lower device.
633          */
634         if (IS_ERR_OR_NULL(vlan->accel_priv)) {
635                 vlan->accel_priv = NULL;
636                 err = dev_uc_add(lowerdev, dev->dev_addr);
637                 if (err < 0)
638                         goto out;
639         }
640
641         if (dev->flags & IFF_ALLMULTI) {
642                 err = dev_set_allmulti(lowerdev, 1);
643                 if (err < 0)
644                         goto del_unicast;
645         }
646
647         if (dev->flags & IFF_PROMISC) {
648                 err = dev_set_promiscuity(lowerdev, 1);
649                 if (err < 0)
650                         goto clear_multi;
651         }
652
653 hash_add:
654         macvlan_hash_add(vlan);
655         return 0;
656
657 clear_multi:
658         if (dev->flags & IFF_ALLMULTI)
659                 dev_set_allmulti(lowerdev, -1);
660 del_unicast:
661         if (vlan->accel_priv) {
662                 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
663                                                            vlan->accel_priv);
664                 vlan->accel_priv = NULL;
665         } else {
666                 dev_uc_del(lowerdev, dev->dev_addr);
667         }
668 out:
669         return err;
670 }
671
672 static int macvlan_stop(struct net_device *dev)
673 {
674         struct macvlan_dev *vlan = netdev_priv(dev);
675         struct net_device *lowerdev = vlan->lowerdev;
676
677         if (vlan->accel_priv) {
678                 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
679                                                            vlan->accel_priv);
680                 vlan->accel_priv = NULL;
681         }
682
683         dev_uc_unsync(lowerdev, dev);
684         dev_mc_unsync(lowerdev, dev);
685
686         if (macvlan_passthru(vlan->port)) {
687                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
688                         dev_set_promiscuity(lowerdev, -1);
689                 goto hash_del;
690         }
691
692         if (dev->flags & IFF_ALLMULTI)
693                 dev_set_allmulti(lowerdev, -1);
694
695         if (dev->flags & IFF_PROMISC)
696                 dev_set_promiscuity(lowerdev, -1);
697
698         dev_uc_del(lowerdev, dev->dev_addr);
699
700 hash_del:
701         macvlan_hash_del(vlan, !dev->dismantle);
702         return 0;
703 }
704
705 static int macvlan_sync_address(struct net_device *dev,
706                                 const unsigned char *addr)
707 {
708         struct macvlan_dev *vlan = netdev_priv(dev);
709         struct net_device *lowerdev = vlan->lowerdev;
710         struct macvlan_port *port = vlan->port;
711         int err;
712
713         if (!(dev->flags & IFF_UP)) {
714                 /* Just copy in the new address */
715                 eth_hw_addr_set(dev, addr);
716         } else {
717                 /* Rehash and update the device filters */
718                 if (macvlan_addr_busy(vlan->port, addr))
719                         return -EADDRINUSE;
720
721                 if (!macvlan_passthru(port)) {
722                         err = dev_uc_add(lowerdev, addr);
723                         if (err)
724                                 return err;
725
726                         dev_uc_del(lowerdev, dev->dev_addr);
727                 }
728
729                 macvlan_hash_change_addr(vlan, addr);
730         }
731         if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
732                 /* Since addr_change isn't set, we are here due to lower
733                  * device change.  Save the lower-dev address so we can
734                  * restore it later.
735                  */
736                 ether_addr_copy(vlan->port->perm_addr,
737                                 lowerdev->dev_addr);
738         }
739         macvlan_clear_addr_change(port);
740         return 0;
741 }
742
743 static int macvlan_set_mac_address(struct net_device *dev, void *p)
744 {
745         struct macvlan_dev *vlan = netdev_priv(dev);
746         struct sockaddr *addr = p;
747
748         if (!is_valid_ether_addr(addr->sa_data))
749                 return -EADDRNOTAVAIL;
750
751         /* If the addresses are the same, this is a no-op */
752         if (ether_addr_equal(dev->dev_addr, addr->sa_data))
753                 return 0;
754
755         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
756                 macvlan_set_addr_change(vlan->port);
757                 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
758         }
759
760         if (macvlan_addr_busy(vlan->port, addr->sa_data))
761                 return -EADDRINUSE;
762
763         return macvlan_sync_address(dev, addr->sa_data);
764 }
765
766 static void macvlan_change_rx_flags(struct net_device *dev, int change)
767 {
768         struct macvlan_dev *vlan = netdev_priv(dev);
769         struct net_device *lowerdev = vlan->lowerdev;
770
771         if (dev->flags & IFF_UP) {
772                 if (change & IFF_ALLMULTI)
773                         dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
774                 if (change & IFF_PROMISC)
775                         dev_set_promiscuity(lowerdev,
776                                             dev->flags & IFF_PROMISC ? 1 : -1);
777
778         }
779 }
780
781 static void macvlan_compute_filter(unsigned long *mc_filter,
782                                    struct net_device *dev,
783                                    struct macvlan_dev *vlan)
784 {
785         if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
786                 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
787         } else {
788                 struct netdev_hw_addr *ha;
789                 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
790
791                 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
792                 netdev_for_each_mc_addr(ha, dev) {
793                         __set_bit(mc_hash(vlan, ha->addr), filter);
794                 }
795
796                 __set_bit(mc_hash(vlan, dev->broadcast), filter);
797
798                 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
799         }
800 }
801
802 static void macvlan_set_mac_lists(struct net_device *dev)
803 {
804         struct macvlan_dev *vlan = netdev_priv(dev);
805
806         macvlan_compute_filter(vlan->mc_filter, dev, vlan);
807
808         dev_uc_sync(vlan->lowerdev, dev);
809         dev_mc_sync(vlan->lowerdev, dev);
810
811         /* This is slightly inaccurate as we're including the subscription
812          * list of vlan->lowerdev too.
813          *
814          * Bug alert: This only works if everyone has the same broadcast
815          * address as lowerdev.  As soon as someone changes theirs this
816          * will break.
817          *
818          * However, this is already broken as when you change your broadcast
819          * address we don't get called.
820          *
821          * The solution is to maintain a list of broadcast addresses like
822          * we do for uc/mc, if you care.
823          */
824         macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
825 }
826
827 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
828 {
829         struct macvlan_dev *vlan = netdev_priv(dev);
830
831         if (vlan->lowerdev->mtu < new_mtu)
832                 return -EINVAL;
833         dev->mtu = new_mtu;
834         return 0;
835 }
836
837 static int macvlan_eth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
838 {
839         struct net_device *real_dev = macvlan_dev_real_dev(dev);
840         const struct net_device_ops *ops = real_dev->netdev_ops;
841         struct ifreq ifrr;
842         int err = -EOPNOTSUPP;
843
844         strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
845         ifrr.ifr_ifru = ifr->ifr_ifru;
846
847         switch (cmd) {
848         case SIOCSHWTSTAMP:
849                 if (!net_eq(dev_net(dev), &init_net))
850                         break;
851                 fallthrough;
852         case SIOCGHWTSTAMP:
853                 if (netif_device_present(real_dev) && ops->ndo_eth_ioctl)
854                         err = ops->ndo_eth_ioctl(real_dev, &ifrr, cmd);
855                 break;
856         }
857
858         if (!err)
859                 ifr->ifr_ifru = ifrr.ifr_ifru;
860
861         return err;
862 }
863
864 /*
865  * macvlan network devices have devices nesting below it and are a special
866  * "super class" of normal network devices; split their locks off into a
867  * separate class since they always nest.
868  */
869 static struct lock_class_key macvlan_netdev_addr_lock_key;
870
871 #define ALWAYS_ON_OFFLOADS \
872         (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
873          NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
874
875 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
876
877 #define MACVLAN_FEATURES \
878         (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
879          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
880          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
881          NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
882
883 #define MACVLAN_STATE_MASK \
884         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
885
886 static void macvlan_set_lockdep_class(struct net_device *dev)
887 {
888         netdev_lockdep_set_classes(dev);
889         lockdep_set_class(&dev->addr_list_lock,
890                           &macvlan_netdev_addr_lock_key);
891 }
892
893 static int macvlan_init(struct net_device *dev)
894 {
895         struct macvlan_dev *vlan = netdev_priv(dev);
896         struct net_device *lowerdev = vlan->lowerdev;
897         struct macvlan_port *port = vlan->port;
898
899         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
900                                   (lowerdev->state & MACVLAN_STATE_MASK);
901         dev->features           = lowerdev->features & MACVLAN_FEATURES;
902         dev->features           |= ALWAYS_ON_FEATURES;
903         dev->hw_features        |= NETIF_F_LRO;
904         dev->vlan_features      = lowerdev->vlan_features & MACVLAN_FEATURES;
905         dev->vlan_features      |= ALWAYS_ON_OFFLOADS;
906         dev->hw_enc_features    |= dev->features;
907         netif_inherit_tso_max(dev, lowerdev);
908         dev->hard_header_len    = lowerdev->hard_header_len;
909         macvlan_set_lockdep_class(dev);
910
911         vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
912         if (!vlan->pcpu_stats)
913                 return -ENOMEM;
914
915         port->count += 1;
916
917         /* Get macvlan's reference to lowerdev */
918         netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
919
920         return 0;
921 }
922
923 static void macvlan_uninit(struct net_device *dev)
924 {
925         struct macvlan_dev *vlan = netdev_priv(dev);
926         struct macvlan_port *port = vlan->port;
927
928         free_percpu(vlan->pcpu_stats);
929
930         macvlan_flush_sources(port, vlan);
931         port->count -= 1;
932         if (!port->count)
933                 macvlan_port_destroy(port->dev);
934 }
935
936 static void macvlan_dev_get_stats64(struct net_device *dev,
937                                     struct rtnl_link_stats64 *stats)
938 {
939         struct macvlan_dev *vlan = netdev_priv(dev);
940
941         if (vlan->pcpu_stats) {
942                 struct vlan_pcpu_stats *p;
943                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
944                 u32 rx_errors = 0, tx_dropped = 0;
945                 unsigned int start;
946                 int i;
947
948                 for_each_possible_cpu(i) {
949                         p = per_cpu_ptr(vlan->pcpu_stats, i);
950                         do {
951                                 start = u64_stats_fetch_begin_irq(&p->syncp);
952                                 rx_packets      = u64_stats_read(&p->rx_packets);
953                                 rx_bytes        = u64_stats_read(&p->rx_bytes);
954                                 rx_multicast    = u64_stats_read(&p->rx_multicast);
955                                 tx_packets      = u64_stats_read(&p->tx_packets);
956                                 tx_bytes        = u64_stats_read(&p->tx_bytes);
957                         } while (u64_stats_fetch_retry_irq(&p->syncp, start));
958
959                         stats->rx_packets       += rx_packets;
960                         stats->rx_bytes         += rx_bytes;
961                         stats->multicast        += rx_multicast;
962                         stats->tx_packets       += tx_packets;
963                         stats->tx_bytes         += tx_bytes;
964                         /* rx_errors & tx_dropped are u32, updated
965                          * without syncp protection.
966                          */
967                         rx_errors       += READ_ONCE(p->rx_errors);
968                         tx_dropped      += READ_ONCE(p->tx_dropped);
969                 }
970                 stats->rx_errors        = rx_errors;
971                 stats->rx_dropped       = rx_errors;
972                 stats->tx_dropped       = tx_dropped;
973         }
974 }
975
976 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
977                                    __be16 proto, u16 vid)
978 {
979         struct macvlan_dev *vlan = netdev_priv(dev);
980         struct net_device *lowerdev = vlan->lowerdev;
981
982         return vlan_vid_add(lowerdev, proto, vid);
983 }
984
985 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
986                                     __be16 proto, u16 vid)
987 {
988         struct macvlan_dev *vlan = netdev_priv(dev);
989         struct net_device *lowerdev = vlan->lowerdev;
990
991         vlan_vid_del(lowerdev, proto, vid);
992         return 0;
993 }
994
995 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
996                            struct net_device *dev,
997                            const unsigned char *addr, u16 vid,
998                            u16 flags,
999                            struct netlink_ext_ack *extack)
1000 {
1001         struct macvlan_dev *vlan = netdev_priv(dev);
1002         int err = -EINVAL;
1003
1004         /* Support unicast filter only on passthru devices.
1005          * Multicast filter should be allowed on all devices.
1006          */
1007         if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1008                 return -EOPNOTSUPP;
1009
1010         if (flags & NLM_F_REPLACE)
1011                 return -EOPNOTSUPP;
1012
1013         if (is_unicast_ether_addr(addr))
1014                 err = dev_uc_add_excl(dev, addr);
1015         else if (is_multicast_ether_addr(addr))
1016                 err = dev_mc_add_excl(dev, addr);
1017
1018         return err;
1019 }
1020
1021 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1022                            struct net_device *dev,
1023                            const unsigned char *addr, u16 vid,
1024                            struct netlink_ext_ack *extack)
1025 {
1026         struct macvlan_dev *vlan = netdev_priv(dev);
1027         int err = -EINVAL;
1028
1029         /* Support unicast filter only on passthru devices.
1030          * Multicast filter should be allowed on all devices.
1031          */
1032         if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1033                 return -EOPNOTSUPP;
1034
1035         if (is_unicast_ether_addr(addr))
1036                 err = dev_uc_del(dev, addr);
1037         else if (is_multicast_ether_addr(addr))
1038                 err = dev_mc_del(dev, addr);
1039
1040         return err;
1041 }
1042
1043 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1044                                         struct ethtool_drvinfo *drvinfo)
1045 {
1046         strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1047         strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1048 }
1049
1050 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1051                                               struct ethtool_link_ksettings *cmd)
1052 {
1053         const struct macvlan_dev *vlan = netdev_priv(dev);
1054
1055         return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1056 }
1057
1058 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1059                                        struct ethtool_ts_info *info)
1060 {
1061         struct net_device *real_dev = macvlan_dev_real_dev(dev);
1062         const struct ethtool_ops *ops = real_dev->ethtool_ops;
1063         struct phy_device *phydev = real_dev->phydev;
1064
1065         if (phy_has_tsinfo(phydev)) {
1066                 return phy_ts_info(phydev, info);
1067         } else if (ops->get_ts_info) {
1068                 return ops->get_ts_info(real_dev, info);
1069         } else {
1070                 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1071                         SOF_TIMESTAMPING_SOFTWARE;
1072                 info->phc_index = -1;
1073         }
1074
1075         return 0;
1076 }
1077
1078 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1079                                               netdev_features_t features)
1080 {
1081         struct macvlan_dev *vlan = netdev_priv(dev);
1082         netdev_features_t lowerdev_features = vlan->lowerdev->features;
1083         netdev_features_t mask;
1084
1085         features |= NETIF_F_ALL_FOR_ALL;
1086         features &= (vlan->set_features | ~MACVLAN_FEATURES);
1087         mask = features;
1088
1089         lowerdev_features &= (features | ~NETIF_F_LRO);
1090         features = netdev_increment_features(lowerdev_features, features, mask);
1091         features |= ALWAYS_ON_FEATURES;
1092         features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1093
1094         return features;
1095 }
1096
1097 #ifdef CONFIG_NET_POLL_CONTROLLER
1098 static void macvlan_dev_poll_controller(struct net_device *dev)
1099 {
1100         return;
1101 }
1102
1103 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1104 {
1105         struct macvlan_dev *vlan = netdev_priv(dev);
1106         struct net_device *real_dev = vlan->lowerdev;
1107         struct netpoll *netpoll;
1108         int err;
1109
1110         netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1111         err = -ENOMEM;
1112         if (!netpoll)
1113                 goto out;
1114
1115         err = __netpoll_setup(netpoll, real_dev);
1116         if (err) {
1117                 kfree(netpoll);
1118                 goto out;
1119         }
1120
1121         vlan->netpoll = netpoll;
1122
1123 out:
1124         return err;
1125 }
1126
1127 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1128 {
1129         struct macvlan_dev *vlan = netdev_priv(dev);
1130         struct netpoll *netpoll = vlan->netpoll;
1131
1132         if (!netpoll)
1133                 return;
1134
1135         vlan->netpoll = NULL;
1136
1137         __netpoll_free(netpoll);
1138 }
1139 #endif  /* CONFIG_NET_POLL_CONTROLLER */
1140
1141 static int macvlan_dev_get_iflink(const struct net_device *dev)
1142 {
1143         struct macvlan_dev *vlan = netdev_priv(dev);
1144
1145         return vlan->lowerdev->ifindex;
1146 }
1147
1148 static const struct ethtool_ops macvlan_ethtool_ops = {
1149         .get_link               = ethtool_op_get_link,
1150         .get_link_ksettings     = macvlan_ethtool_get_link_ksettings,
1151         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
1152         .get_ts_info            = macvlan_ethtool_get_ts_info,
1153 };
1154
1155 static const struct net_device_ops macvlan_netdev_ops = {
1156         .ndo_init               = macvlan_init,
1157         .ndo_uninit             = macvlan_uninit,
1158         .ndo_open               = macvlan_open,
1159         .ndo_stop               = macvlan_stop,
1160         .ndo_start_xmit         = macvlan_start_xmit,
1161         .ndo_change_mtu         = macvlan_change_mtu,
1162         .ndo_eth_ioctl          = macvlan_eth_ioctl,
1163         .ndo_fix_features       = macvlan_fix_features,
1164         .ndo_change_rx_flags    = macvlan_change_rx_flags,
1165         .ndo_set_mac_address    = macvlan_set_mac_address,
1166         .ndo_set_rx_mode        = macvlan_set_mac_lists,
1167         .ndo_get_stats64        = macvlan_dev_get_stats64,
1168         .ndo_validate_addr      = eth_validate_addr,
1169         .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
1170         .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
1171         .ndo_fdb_add            = macvlan_fdb_add,
1172         .ndo_fdb_del            = macvlan_fdb_del,
1173         .ndo_fdb_dump           = ndo_dflt_fdb_dump,
1174 #ifdef CONFIG_NET_POLL_CONTROLLER
1175         .ndo_poll_controller    = macvlan_dev_poll_controller,
1176         .ndo_netpoll_setup      = macvlan_dev_netpoll_setup,
1177         .ndo_netpoll_cleanup    = macvlan_dev_netpoll_cleanup,
1178 #endif
1179         .ndo_get_iflink         = macvlan_dev_get_iflink,
1180         .ndo_features_check     = passthru_features_check,
1181 };
1182
1183 static void macvlan_dev_free(struct net_device *dev)
1184 {
1185         struct macvlan_dev *vlan = netdev_priv(dev);
1186
1187         /* Get rid of the macvlan's reference to lowerdev */
1188         netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1189 }
1190
1191 void macvlan_common_setup(struct net_device *dev)
1192 {
1193         ether_setup(dev);
1194
1195         /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1196         dev->max_mtu            = ETH_MAX_MTU;
1197         dev->priv_flags        &= ~IFF_TX_SKB_SHARING;
1198         netif_keep_dst(dev);
1199         dev->priv_flags        |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN;
1200         dev->netdev_ops         = &macvlan_netdev_ops;
1201         dev->needs_free_netdev  = true;
1202         dev->priv_destructor    = macvlan_dev_free;
1203         dev->header_ops         = &macvlan_hard_header_ops;
1204         dev->ethtool_ops        = &macvlan_ethtool_ops;
1205 }
1206 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1207
1208 static void macvlan_setup(struct net_device *dev)
1209 {
1210         macvlan_common_setup(dev);
1211         dev->priv_flags |= IFF_NO_QUEUE;
1212 }
1213
1214 static int macvlan_port_create(struct net_device *dev)
1215 {
1216         struct macvlan_port *port;
1217         unsigned int i;
1218         int err;
1219
1220         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1221                 return -EINVAL;
1222
1223         if (netdev_is_rx_handler_busy(dev))
1224                 return -EBUSY;
1225
1226         port = kzalloc(sizeof(*port), GFP_KERNEL);
1227         if (port == NULL)
1228                 return -ENOMEM;
1229
1230         port->dev = dev;
1231         ether_addr_copy(port->perm_addr, dev->dev_addr);
1232         INIT_LIST_HEAD(&port->vlans);
1233         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1234                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1235         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1236                 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1237
1238         port->bc_queue_len_used = 0;
1239         skb_queue_head_init(&port->bc_queue);
1240         INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1241
1242         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1243         if (err)
1244                 kfree(port);
1245         else
1246                 dev->priv_flags |= IFF_MACVLAN_PORT;
1247         return err;
1248 }
1249
1250 static void macvlan_port_destroy(struct net_device *dev)
1251 {
1252         struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1253         struct sk_buff *skb;
1254
1255         dev->priv_flags &= ~IFF_MACVLAN_PORT;
1256         netdev_rx_handler_unregister(dev);
1257
1258         /* After this point, no packet can schedule bc_work anymore,
1259          * but we need to cancel it and purge left skbs if any.
1260          */
1261         cancel_work_sync(&port->bc_work);
1262
1263         while ((skb = __skb_dequeue(&port->bc_queue))) {
1264                 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1265
1266                 if (src)
1267                         dev_put(src->dev);
1268
1269                 kfree_skb(skb);
1270         }
1271
1272         /* If the lower device address has been changed by passthru
1273          * macvlan, put it back.
1274          */
1275         if (macvlan_passthru(port) &&
1276             !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1277                 struct sockaddr sa;
1278
1279                 sa.sa_family = port->dev->type;
1280                 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1281                 dev_set_mac_address(port->dev, &sa, NULL);
1282         }
1283
1284         kfree(port);
1285 }
1286
1287 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1288                             struct netlink_ext_ack *extack)
1289 {
1290         struct nlattr *nla, *head;
1291         int rem, len;
1292
1293         if (tb[IFLA_ADDRESS]) {
1294                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1295                         return -EINVAL;
1296                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1297                         return -EADDRNOTAVAIL;
1298         }
1299
1300         if (!data)
1301                 return 0;
1302
1303         if (data[IFLA_MACVLAN_FLAGS] &&
1304             nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1305                                                       MACVLAN_FLAG_NODST))
1306                 return -EINVAL;
1307
1308         if (data[IFLA_MACVLAN_MODE]) {
1309                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1310                 case MACVLAN_MODE_PRIVATE:
1311                 case MACVLAN_MODE_VEPA:
1312                 case MACVLAN_MODE_BRIDGE:
1313                 case MACVLAN_MODE_PASSTHRU:
1314                 case MACVLAN_MODE_SOURCE:
1315                         break;
1316                 default:
1317                         return -EINVAL;
1318                 }
1319         }
1320
1321         if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1322                 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1323                 case MACVLAN_MACADDR_ADD:
1324                 case MACVLAN_MACADDR_DEL:
1325                 case MACVLAN_MACADDR_FLUSH:
1326                 case MACVLAN_MACADDR_SET:
1327                         break;
1328                 default:
1329                         return -EINVAL;
1330                 }
1331         }
1332
1333         if (data[IFLA_MACVLAN_MACADDR]) {
1334                 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1335                         return -EINVAL;
1336
1337                 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1338                         return -EADDRNOTAVAIL;
1339         }
1340
1341         if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1342                 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1343                 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1344
1345                 nla_for_each_attr(nla, head, len, rem) {
1346                         if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1347                             nla_len(nla) != ETH_ALEN)
1348                                 return -EINVAL;
1349
1350                         if (!is_valid_ether_addr(nla_data(nla)))
1351                                 return -EADDRNOTAVAIL;
1352                 }
1353         }
1354
1355         if (data[IFLA_MACVLAN_MACADDR_COUNT])
1356                 return -EINVAL;
1357
1358         return 0;
1359 }
1360
1361 /*
1362  * reconfigure list of remote source mac address
1363  * (only for macvlan devices in source mode)
1364  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1365  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1366  */
1367 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1368                                       struct nlattr *data[])
1369 {
1370         char *addr = NULL;
1371         int ret, rem, len;
1372         struct nlattr *nla, *head;
1373         struct macvlan_source_entry *entry;
1374
1375         if (data[IFLA_MACVLAN_MACADDR])
1376                 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1377
1378         if (mode == MACVLAN_MACADDR_ADD) {
1379                 if (!addr)
1380                         return -EINVAL;
1381
1382                 return macvlan_hash_add_source(vlan, addr);
1383
1384         } else if (mode == MACVLAN_MACADDR_DEL) {
1385                 if (!addr)
1386                         return -EINVAL;
1387
1388                 entry = macvlan_hash_lookup_source(vlan, addr);
1389                 if (entry) {
1390                         macvlan_hash_del_source(entry);
1391                         vlan->macaddr_count--;
1392                 }
1393         } else if (mode == MACVLAN_MACADDR_FLUSH) {
1394                 macvlan_flush_sources(vlan->port, vlan);
1395         } else if (mode == MACVLAN_MACADDR_SET) {
1396                 macvlan_flush_sources(vlan->port, vlan);
1397
1398                 if (addr) {
1399                         ret = macvlan_hash_add_source(vlan, addr);
1400                         if (ret)
1401                                 return ret;
1402                 }
1403
1404                 if (!data[IFLA_MACVLAN_MACADDR_DATA])
1405                         return 0;
1406
1407                 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1408                 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1409
1410                 nla_for_each_attr(nla, head, len, rem) {
1411                         addr = nla_data(nla);
1412                         ret = macvlan_hash_add_source(vlan, addr);
1413                         if (ret)
1414                                 return ret;
1415                 }
1416         } else {
1417                 return -EINVAL;
1418         }
1419
1420         return 0;
1421 }
1422
1423 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1424                            struct nlattr *tb[], struct nlattr *data[],
1425                            struct netlink_ext_ack *extack)
1426 {
1427         struct macvlan_dev *vlan = netdev_priv(dev);
1428         struct macvlan_port *port;
1429         struct net_device *lowerdev;
1430         int err;
1431         int macmode;
1432         bool create = false;
1433
1434         if (!tb[IFLA_LINK])
1435                 return -EINVAL;
1436
1437         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1438         if (lowerdev == NULL)
1439                 return -ENODEV;
1440
1441         /* When creating macvlans or macvtaps on top of other macvlans - use
1442          * the real device as the lowerdev.
1443          */
1444         if (netif_is_macvlan(lowerdev))
1445                 lowerdev = macvlan_dev_real_dev(lowerdev);
1446
1447         if (!tb[IFLA_MTU])
1448                 dev->mtu = lowerdev->mtu;
1449         else if (dev->mtu > lowerdev->mtu)
1450                 return -EINVAL;
1451
1452         /* MTU range: 68 - lowerdev->max_mtu */
1453         dev->min_mtu = ETH_MIN_MTU;
1454         dev->max_mtu = lowerdev->max_mtu;
1455
1456         if (!tb[IFLA_ADDRESS])
1457                 eth_hw_addr_random(dev);
1458
1459         if (!netif_is_macvlan_port(lowerdev)) {
1460                 err = macvlan_port_create(lowerdev);
1461                 if (err < 0)
1462                         return err;
1463                 create = true;
1464         }
1465         port = macvlan_port_get_rtnl(lowerdev);
1466
1467         /* Only 1 macvlan device can be created in passthru mode */
1468         if (macvlan_passthru(port)) {
1469                 /* The macvlan port must be not created this time,
1470                  * still goto destroy_macvlan_port for readability.
1471                  */
1472                 err = -EINVAL;
1473                 goto destroy_macvlan_port;
1474         }
1475
1476         vlan->lowerdev = lowerdev;
1477         vlan->dev      = dev;
1478         vlan->port     = port;
1479         vlan->set_features = MACVLAN_FEATURES;
1480
1481         vlan->mode     = MACVLAN_MODE_VEPA;
1482         if (data && data[IFLA_MACVLAN_MODE])
1483                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1484
1485         if (data && data[IFLA_MACVLAN_FLAGS])
1486                 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1487
1488         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1489                 if (port->count) {
1490                         err = -EINVAL;
1491                         goto destroy_macvlan_port;
1492                 }
1493                 macvlan_set_passthru(port);
1494                 eth_hw_addr_inherit(dev, lowerdev);
1495         }
1496
1497         if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1498                 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1499                         err = -EINVAL;
1500                         goto destroy_macvlan_port;
1501                 }
1502                 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1503                 err = macvlan_changelink_sources(vlan, macmode, data);
1504                 if (err)
1505                         goto destroy_macvlan_port;
1506         }
1507
1508         vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1509         if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1510                 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1511
1512         err = register_netdevice(dev);
1513         if (err < 0)
1514                 goto destroy_macvlan_port;
1515
1516         dev->priv_flags |= IFF_MACVLAN;
1517         err = netdev_upper_dev_link(lowerdev, dev, extack);
1518         if (err)
1519                 goto unregister_netdev;
1520
1521         list_add_tail_rcu(&vlan->list, &port->vlans);
1522         update_port_bc_queue_len(vlan->port);
1523         netif_stacked_transfer_operstate(lowerdev, dev);
1524         linkwatch_fire_event(dev);
1525
1526         return 0;
1527
1528 unregister_netdev:
1529         /* macvlan_uninit would free the macvlan port */
1530         unregister_netdevice(dev);
1531         return err;
1532 destroy_macvlan_port:
1533         /* the macvlan port may be freed by macvlan_uninit when fail to register.
1534          * so we destroy the macvlan port only when it's valid.
1535          */
1536         if (create && macvlan_port_get_rtnl(lowerdev)) {
1537                 macvlan_flush_sources(port, vlan);
1538                 macvlan_port_destroy(port->dev);
1539         }
1540         return err;
1541 }
1542 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1543
1544 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1545                            struct nlattr *tb[], struct nlattr *data[],
1546                            struct netlink_ext_ack *extack)
1547 {
1548         return macvlan_common_newlink(src_net, dev, tb, data, extack);
1549 }
1550
1551 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1552 {
1553         struct macvlan_dev *vlan = netdev_priv(dev);
1554
1555         if (vlan->mode == MACVLAN_MODE_SOURCE)
1556                 macvlan_flush_sources(vlan->port, vlan);
1557         list_del_rcu(&vlan->list);
1558         update_port_bc_queue_len(vlan->port);
1559         unregister_netdevice_queue(dev, head);
1560         netdev_upper_dev_unlink(vlan->lowerdev, dev);
1561 }
1562 EXPORT_SYMBOL_GPL(macvlan_dellink);
1563
1564 static int macvlan_changelink(struct net_device *dev,
1565                               struct nlattr *tb[], struct nlattr *data[],
1566                               struct netlink_ext_ack *extack)
1567 {
1568         struct macvlan_dev *vlan = netdev_priv(dev);
1569         enum macvlan_mode mode;
1570         bool set_mode = false;
1571         enum macvlan_macaddr_mode macmode;
1572         int ret;
1573
1574         /* Validate mode, but don't set yet: setting flags may fail. */
1575         if (data && data[IFLA_MACVLAN_MODE]) {
1576                 set_mode = true;
1577                 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1578                 /* Passthrough mode can't be set or cleared dynamically */
1579                 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1580                     (vlan->mode == MACVLAN_MODE_PASSTHRU))
1581                         return -EINVAL;
1582                 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1583                     vlan->mode != mode)
1584                         macvlan_flush_sources(vlan->port, vlan);
1585         }
1586
1587         if (data && data[IFLA_MACVLAN_FLAGS]) {
1588                 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1589                 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1590                 if (macvlan_passthru(vlan->port) && promisc) {
1591                         int err;
1592
1593                         if (flags & MACVLAN_FLAG_NOPROMISC)
1594                                 err = dev_set_promiscuity(vlan->lowerdev, -1);
1595                         else
1596                                 err = dev_set_promiscuity(vlan->lowerdev, 1);
1597                         if (err < 0)
1598                                 return err;
1599                 }
1600                 vlan->flags = flags;
1601         }
1602
1603         if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1604                 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1605                 update_port_bc_queue_len(vlan->port);
1606         }
1607
1608         if (set_mode)
1609                 vlan->mode = mode;
1610         if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1611                 if (vlan->mode != MACVLAN_MODE_SOURCE)
1612                         return -EINVAL;
1613                 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1614                 ret = macvlan_changelink_sources(vlan, macmode, data);
1615                 if (ret)
1616                         return ret;
1617         }
1618         return 0;
1619 }
1620
1621 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1622 {
1623         if (vlan->macaddr_count == 0)
1624                 return 0;
1625         return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1626                 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1627 }
1628
1629 static size_t macvlan_get_size(const struct net_device *dev)
1630 {
1631         struct macvlan_dev *vlan = netdev_priv(dev);
1632
1633         return (0
1634                 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1635                 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1636                 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1637                 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1638                 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1639                 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1640                 );
1641 }
1642
1643 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1644                                      const struct macvlan_dev *vlan,
1645                                      const int i)
1646 {
1647         struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1648         struct macvlan_source_entry *entry;
1649
1650         hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1651                 if (entry->vlan != vlan)
1652                         continue;
1653                 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1654                         return 1;
1655         }
1656         return 0;
1657 }
1658
1659 static int macvlan_fill_info(struct sk_buff *skb,
1660                                 const struct net_device *dev)
1661 {
1662         struct macvlan_dev *vlan = netdev_priv(dev);
1663         struct macvlan_port *port = vlan->port;
1664         int i;
1665         struct nlattr *nest;
1666
1667         if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1668                 goto nla_put_failure;
1669         if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1670                 goto nla_put_failure;
1671         if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1672                 goto nla_put_failure;
1673         if (vlan->macaddr_count > 0) {
1674                 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1675                 if (nest == NULL)
1676                         goto nla_put_failure;
1677
1678                 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1679                         if (macvlan_fill_info_macaddr(skb, vlan, i))
1680                                 goto nla_put_failure;
1681                 }
1682                 nla_nest_end(skb, nest);
1683         }
1684         if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1685                 goto nla_put_failure;
1686         if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1687                 goto nla_put_failure;
1688         return 0;
1689
1690 nla_put_failure:
1691         return -EMSGSIZE;
1692 }
1693
1694 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1695         [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1696         [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1697         [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1698         [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1699         [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1700         [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1701         [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1702         [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1703 };
1704
1705 int macvlan_link_register(struct rtnl_link_ops *ops)
1706 {
1707         /* common fields */
1708         ops->validate           = macvlan_validate;
1709         ops->maxtype            = IFLA_MACVLAN_MAX;
1710         ops->policy             = macvlan_policy;
1711         ops->changelink         = macvlan_changelink;
1712         ops->get_size           = macvlan_get_size;
1713         ops->fill_info          = macvlan_fill_info;
1714
1715         return rtnl_link_register(ops);
1716 };
1717 EXPORT_SYMBOL_GPL(macvlan_link_register);
1718
1719 static struct net *macvlan_get_link_net(const struct net_device *dev)
1720 {
1721         return dev_net(macvlan_dev_real_dev(dev));
1722 }
1723
1724 static struct rtnl_link_ops macvlan_link_ops = {
1725         .kind           = "macvlan",
1726         .setup          = macvlan_setup,
1727         .newlink        = macvlan_newlink,
1728         .dellink        = macvlan_dellink,
1729         .get_link_net   = macvlan_get_link_net,
1730         .priv_size      = sizeof(struct macvlan_dev),
1731 };
1732
1733 static void update_port_bc_queue_len(struct macvlan_port *port)
1734 {
1735         u32 max_bc_queue_len_req = 0;
1736         struct macvlan_dev *vlan;
1737
1738         list_for_each_entry(vlan, &port->vlans, list) {
1739                 if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1740                         max_bc_queue_len_req = vlan->bc_queue_len_req;
1741         }
1742         port->bc_queue_len_used = max_bc_queue_len_req;
1743 }
1744
1745 static int macvlan_device_event(struct notifier_block *unused,
1746                                 unsigned long event, void *ptr)
1747 {
1748         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1749         struct macvlan_dev *vlan, *next;
1750         struct macvlan_port *port;
1751         LIST_HEAD(list_kill);
1752
1753         if (!netif_is_macvlan_port(dev))
1754                 return NOTIFY_DONE;
1755
1756         port = macvlan_port_get_rtnl(dev);
1757
1758         switch (event) {
1759         case NETDEV_UP:
1760         case NETDEV_DOWN:
1761         case NETDEV_CHANGE:
1762                 list_for_each_entry(vlan, &port->vlans, list)
1763                         netif_stacked_transfer_operstate(vlan->lowerdev,
1764                                                          vlan->dev);
1765                 break;
1766         case NETDEV_FEAT_CHANGE:
1767                 list_for_each_entry(vlan, &port->vlans, list) {
1768                         netif_inherit_tso_max(vlan->dev, dev);
1769                         netdev_update_features(vlan->dev);
1770                 }
1771                 break;
1772         case NETDEV_CHANGEMTU:
1773                 list_for_each_entry(vlan, &port->vlans, list) {
1774                         if (vlan->dev->mtu <= dev->mtu)
1775                                 continue;
1776                         dev_set_mtu(vlan->dev, dev->mtu);
1777                 }
1778                 break;
1779         case NETDEV_CHANGEADDR:
1780                 if (!macvlan_passthru(port))
1781                         return NOTIFY_DONE;
1782
1783                 vlan = list_first_entry_or_null(&port->vlans,
1784                                                 struct macvlan_dev,
1785                                                 list);
1786
1787                 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1788                         return NOTIFY_BAD;
1789
1790                 break;
1791         case NETDEV_UNREGISTER:
1792                 /* twiddle thumbs on netns device moves */
1793                 if (dev->reg_state != NETREG_UNREGISTERING)
1794                         break;
1795
1796                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1797                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1798                 unregister_netdevice_many(&list_kill);
1799                 break;
1800         case NETDEV_PRE_TYPE_CHANGE:
1801                 /* Forbid underlying device to change its type. */
1802                 return NOTIFY_BAD;
1803
1804         case NETDEV_NOTIFY_PEERS:
1805         case NETDEV_BONDING_FAILOVER:
1806         case NETDEV_RESEND_IGMP:
1807                 /* Propagate to all vlans */
1808                 list_for_each_entry(vlan, &port->vlans, list)
1809                         call_netdevice_notifiers(event, vlan->dev);
1810         }
1811         return NOTIFY_DONE;
1812 }
1813
1814 static struct notifier_block macvlan_notifier_block __read_mostly = {
1815         .notifier_call  = macvlan_device_event,
1816 };
1817
1818 static int __init macvlan_init_module(void)
1819 {
1820         int err;
1821
1822         register_netdevice_notifier(&macvlan_notifier_block);
1823
1824         err = macvlan_link_register(&macvlan_link_ops);
1825         if (err < 0)
1826                 goto err1;
1827         return 0;
1828 err1:
1829         unregister_netdevice_notifier(&macvlan_notifier_block);
1830         return err;
1831 }
1832
1833 static void __exit macvlan_cleanup_module(void)
1834 {
1835         rtnl_link_unregister(&macvlan_link_ops);
1836         unregister_netdevice_notifier(&macvlan_notifier_block);
1837 }
1838
1839 module_init(macvlan_init_module);
1840 module_exit(macvlan_cleanup_module);
1841
1842 MODULE_LICENSE("GPL");
1843 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1844 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1845 MODULE_ALIAS_RTNL_LINK("macvlan");