net/mlx5: E-Switch, properly handle ingress tagged packets on VST
[platform/kernel/linux-rpi.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         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
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_ni(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         schedule_work(&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         atomic_long_inc(&skb->dev->rx_dropped);
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                 pcpu_stats->tx_packets++;
579                 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, unsigned char *addr)
706 {
707         struct macvlan_dev *vlan = netdev_priv(dev);
708         struct net_device *lowerdev = vlan->lowerdev;
709         struct macvlan_port *port = vlan->port;
710         int err;
711
712         if (!(dev->flags & IFF_UP)) {
713                 /* Just copy in the new address */
714                 eth_hw_addr_set(dev, addr);
715         } else {
716                 /* Rehash and update the device filters */
717                 if (macvlan_addr_busy(vlan->port, addr))
718                         return -EADDRINUSE;
719
720                 if (!macvlan_passthru(port)) {
721                         err = dev_uc_add(lowerdev, addr);
722                         if (err)
723                                 return err;
724
725                         dev_uc_del(lowerdev, dev->dev_addr);
726                 }
727
728                 macvlan_hash_change_addr(vlan, addr);
729         }
730         if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
731                 /* Since addr_change isn't set, we are here due to lower
732                  * device change.  Save the lower-dev address so we can
733                  * restore it later.
734                  */
735                 ether_addr_copy(vlan->port->perm_addr,
736                                 lowerdev->dev_addr);
737         }
738         macvlan_clear_addr_change(port);
739         return 0;
740 }
741
742 static int macvlan_set_mac_address(struct net_device *dev, void *p)
743 {
744         struct macvlan_dev *vlan = netdev_priv(dev);
745         struct sockaddr *addr = p;
746
747         if (!is_valid_ether_addr(addr->sa_data))
748                 return -EADDRNOTAVAIL;
749
750         /* If the addresses are the same, this is a no-op */
751         if (ether_addr_equal(dev->dev_addr, addr->sa_data))
752                 return 0;
753
754         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
755                 macvlan_set_addr_change(vlan->port);
756                 return dev_set_mac_address(vlan->lowerdev, addr, NULL);
757         }
758
759         if (macvlan_addr_busy(vlan->port, addr->sa_data))
760                 return -EADDRINUSE;
761
762         return macvlan_sync_address(dev, addr->sa_data);
763 }
764
765 static void macvlan_change_rx_flags(struct net_device *dev, int change)
766 {
767         struct macvlan_dev *vlan = netdev_priv(dev);
768         struct net_device *lowerdev = vlan->lowerdev;
769
770         if (dev->flags & IFF_UP) {
771                 if (change & IFF_ALLMULTI)
772                         dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
773                 if (change & IFF_PROMISC)
774                         dev_set_promiscuity(lowerdev,
775                                             dev->flags & IFF_PROMISC ? 1 : -1);
776
777         }
778 }
779
780 static void macvlan_compute_filter(unsigned long *mc_filter,
781                                    struct net_device *dev,
782                                    struct macvlan_dev *vlan)
783 {
784         if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
785                 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
786         } else {
787                 struct netdev_hw_addr *ha;
788                 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
789
790                 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
791                 netdev_for_each_mc_addr(ha, dev) {
792                         __set_bit(mc_hash(vlan, ha->addr), filter);
793                 }
794
795                 __set_bit(mc_hash(vlan, dev->broadcast), filter);
796
797                 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
798         }
799 }
800
801 static void macvlan_set_mac_lists(struct net_device *dev)
802 {
803         struct macvlan_dev *vlan = netdev_priv(dev);
804
805         macvlan_compute_filter(vlan->mc_filter, dev, vlan);
806
807         dev_uc_sync(vlan->lowerdev, dev);
808         dev_mc_sync(vlan->lowerdev, dev);
809
810         /* This is slightly inaccurate as we're including the subscription
811          * list of vlan->lowerdev too.
812          *
813          * Bug alert: This only works if everyone has the same broadcast
814          * address as lowerdev.  As soon as someone changes theirs this
815          * will break.
816          *
817          * However, this is already broken as when you change your broadcast
818          * address we don't get called.
819          *
820          * The solution is to maintain a list of broadcast addresses like
821          * we do for uc/mc, if you care.
822          */
823         macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
824 }
825
826 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
827 {
828         struct macvlan_dev *vlan = netdev_priv(dev);
829
830         if (vlan->lowerdev->mtu < new_mtu)
831                 return -EINVAL;
832         dev->mtu = new_mtu;
833         return 0;
834 }
835
836 static int macvlan_eth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
837 {
838         struct net_device *real_dev = macvlan_dev_real_dev(dev);
839         const struct net_device_ops *ops = real_dev->netdev_ops;
840         struct ifreq ifrr;
841         int err = -EOPNOTSUPP;
842
843         strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
844         ifrr.ifr_ifru = ifr->ifr_ifru;
845
846         switch (cmd) {
847         case SIOCSHWTSTAMP:
848                 if (!net_eq(dev_net(dev), &init_net))
849                         break;
850                 fallthrough;
851         case SIOCGHWTSTAMP:
852                 if (netif_device_present(real_dev) && ops->ndo_eth_ioctl)
853                         err = ops->ndo_eth_ioctl(real_dev, &ifrr, cmd);
854                 break;
855         }
856
857         if (!err)
858                 ifr->ifr_ifru = ifrr.ifr_ifru;
859
860         return err;
861 }
862
863 /*
864  * macvlan network devices have devices nesting below it and are a special
865  * "super class" of normal network devices; split their locks off into a
866  * separate class since they always nest.
867  */
868 static struct lock_class_key macvlan_netdev_addr_lock_key;
869
870 #define ALWAYS_ON_OFFLOADS \
871         (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
872          NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
873
874 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
875
876 #define MACVLAN_FEATURES \
877         (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
878          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
879          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
880          NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
881
882 #define MACVLAN_STATE_MASK \
883         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
884
885 static void macvlan_set_lockdep_class(struct net_device *dev)
886 {
887         netdev_lockdep_set_classes(dev);
888         lockdep_set_class(&dev->addr_list_lock,
889                           &macvlan_netdev_addr_lock_key);
890 }
891
892 static int macvlan_init(struct net_device *dev)
893 {
894         struct macvlan_dev *vlan = netdev_priv(dev);
895         const struct net_device *lowerdev = vlan->lowerdev;
896         struct macvlan_port *port = vlan->port;
897
898         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
899                                   (lowerdev->state & MACVLAN_STATE_MASK);
900         dev->features           = lowerdev->features & MACVLAN_FEATURES;
901         dev->features           |= ALWAYS_ON_FEATURES;
902         dev->hw_features        |= NETIF_F_LRO;
903         dev->vlan_features      = lowerdev->vlan_features & MACVLAN_FEATURES;
904         dev->vlan_features      |= ALWAYS_ON_OFFLOADS;
905         dev->hw_enc_features    |= dev->features;
906         dev->gso_max_size       = lowerdev->gso_max_size;
907         dev->gso_max_segs       = lowerdev->gso_max_segs;
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         return 0;
918 }
919
920 static void macvlan_uninit(struct net_device *dev)
921 {
922         struct macvlan_dev *vlan = netdev_priv(dev);
923         struct macvlan_port *port = vlan->port;
924
925         free_percpu(vlan->pcpu_stats);
926
927         macvlan_flush_sources(port, vlan);
928         port->count -= 1;
929         if (!port->count)
930                 macvlan_port_destroy(port->dev);
931 }
932
933 static void macvlan_dev_get_stats64(struct net_device *dev,
934                                     struct rtnl_link_stats64 *stats)
935 {
936         struct macvlan_dev *vlan = netdev_priv(dev);
937
938         if (vlan->pcpu_stats) {
939                 struct vlan_pcpu_stats *p;
940                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
941                 u32 rx_errors = 0, tx_dropped = 0;
942                 unsigned int start;
943                 int i;
944
945                 for_each_possible_cpu(i) {
946                         p = per_cpu_ptr(vlan->pcpu_stats, i);
947                         do {
948                                 start = u64_stats_fetch_begin_irq(&p->syncp);
949                                 rx_packets      = p->rx_packets;
950                                 rx_bytes        = p->rx_bytes;
951                                 rx_multicast    = p->rx_multicast;
952                                 tx_packets      = p->tx_packets;
953                                 tx_bytes        = p->tx_bytes;
954                         } while (u64_stats_fetch_retry_irq(&p->syncp, start));
955
956                         stats->rx_packets       += rx_packets;
957                         stats->rx_bytes         += rx_bytes;
958                         stats->multicast        += rx_multicast;
959                         stats->tx_packets       += tx_packets;
960                         stats->tx_bytes         += tx_bytes;
961                         /* rx_errors & tx_dropped are u32, updated
962                          * without syncp protection.
963                          */
964                         rx_errors       += p->rx_errors;
965                         tx_dropped      += p->tx_dropped;
966                 }
967                 stats->rx_errors        = rx_errors;
968                 stats->rx_dropped       = rx_errors;
969                 stats->tx_dropped       = tx_dropped;
970         }
971 }
972
973 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
974                                    __be16 proto, u16 vid)
975 {
976         struct macvlan_dev *vlan = netdev_priv(dev);
977         struct net_device *lowerdev = vlan->lowerdev;
978
979         return vlan_vid_add(lowerdev, proto, vid);
980 }
981
982 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
983                                     __be16 proto, u16 vid)
984 {
985         struct macvlan_dev *vlan = netdev_priv(dev);
986         struct net_device *lowerdev = vlan->lowerdev;
987
988         vlan_vid_del(lowerdev, proto, vid);
989         return 0;
990 }
991
992 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
993                            struct net_device *dev,
994                            const unsigned char *addr, u16 vid,
995                            u16 flags,
996                            struct netlink_ext_ack *extack)
997 {
998         struct macvlan_dev *vlan = netdev_priv(dev);
999         int err = -EINVAL;
1000
1001         /* Support unicast filter only on passthru devices.
1002          * Multicast filter should be allowed on all devices.
1003          */
1004         if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1005                 return -EOPNOTSUPP;
1006
1007         if (flags & NLM_F_REPLACE)
1008                 return -EOPNOTSUPP;
1009
1010         if (is_unicast_ether_addr(addr))
1011                 err = dev_uc_add_excl(dev, addr);
1012         else if (is_multicast_ether_addr(addr))
1013                 err = dev_mc_add_excl(dev, addr);
1014
1015         return err;
1016 }
1017
1018 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1019                            struct net_device *dev,
1020                            const unsigned char *addr, u16 vid)
1021 {
1022         struct macvlan_dev *vlan = netdev_priv(dev);
1023         int err = -EINVAL;
1024
1025         /* Support unicast filter only on passthru devices.
1026          * Multicast filter should be allowed on all devices.
1027          */
1028         if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1029                 return -EOPNOTSUPP;
1030
1031         if (is_unicast_ether_addr(addr))
1032                 err = dev_uc_del(dev, addr);
1033         else if (is_multicast_ether_addr(addr))
1034                 err = dev_mc_del(dev, addr);
1035
1036         return err;
1037 }
1038
1039 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1040                                         struct ethtool_drvinfo *drvinfo)
1041 {
1042         strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1043         strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1044 }
1045
1046 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1047                                               struct ethtool_link_ksettings *cmd)
1048 {
1049         const struct macvlan_dev *vlan = netdev_priv(dev);
1050
1051         return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1052 }
1053
1054 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1055                                        struct ethtool_ts_info *info)
1056 {
1057         struct net_device *real_dev = macvlan_dev_real_dev(dev);
1058         const struct ethtool_ops *ops = real_dev->ethtool_ops;
1059         struct phy_device *phydev = real_dev->phydev;
1060
1061         if (phy_has_tsinfo(phydev)) {
1062                 return phy_ts_info(phydev, info);
1063         } else if (ops->get_ts_info) {
1064                 return ops->get_ts_info(real_dev, info);
1065         } else {
1066                 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1067                         SOF_TIMESTAMPING_SOFTWARE;
1068                 info->phc_index = -1;
1069         }
1070
1071         return 0;
1072 }
1073
1074 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1075                                               netdev_features_t features)
1076 {
1077         struct macvlan_dev *vlan = netdev_priv(dev);
1078         netdev_features_t lowerdev_features = vlan->lowerdev->features;
1079         netdev_features_t mask;
1080
1081         features |= NETIF_F_ALL_FOR_ALL;
1082         features &= (vlan->set_features | ~MACVLAN_FEATURES);
1083         mask = features;
1084
1085         lowerdev_features &= (features | ~NETIF_F_LRO);
1086         features = netdev_increment_features(lowerdev_features, features, mask);
1087         features |= ALWAYS_ON_FEATURES;
1088         features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1089
1090         return features;
1091 }
1092
1093 #ifdef CONFIG_NET_POLL_CONTROLLER
1094 static void macvlan_dev_poll_controller(struct net_device *dev)
1095 {
1096         return;
1097 }
1098
1099 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1100 {
1101         struct macvlan_dev *vlan = netdev_priv(dev);
1102         struct net_device *real_dev = vlan->lowerdev;
1103         struct netpoll *netpoll;
1104         int err;
1105
1106         netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1107         err = -ENOMEM;
1108         if (!netpoll)
1109                 goto out;
1110
1111         err = __netpoll_setup(netpoll, real_dev);
1112         if (err) {
1113                 kfree(netpoll);
1114                 goto out;
1115         }
1116
1117         vlan->netpoll = netpoll;
1118
1119 out:
1120         return err;
1121 }
1122
1123 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1124 {
1125         struct macvlan_dev *vlan = netdev_priv(dev);
1126         struct netpoll *netpoll = vlan->netpoll;
1127
1128         if (!netpoll)
1129                 return;
1130
1131         vlan->netpoll = NULL;
1132
1133         __netpoll_free(netpoll);
1134 }
1135 #endif  /* CONFIG_NET_POLL_CONTROLLER */
1136
1137 static int macvlan_dev_get_iflink(const struct net_device *dev)
1138 {
1139         struct macvlan_dev *vlan = netdev_priv(dev);
1140
1141         return vlan->lowerdev->ifindex;
1142 }
1143
1144 static const struct ethtool_ops macvlan_ethtool_ops = {
1145         .get_link               = ethtool_op_get_link,
1146         .get_link_ksettings     = macvlan_ethtool_get_link_ksettings,
1147         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
1148         .get_ts_info            = macvlan_ethtool_get_ts_info,
1149 };
1150
1151 static const struct net_device_ops macvlan_netdev_ops = {
1152         .ndo_init               = macvlan_init,
1153         .ndo_uninit             = macvlan_uninit,
1154         .ndo_open               = macvlan_open,
1155         .ndo_stop               = macvlan_stop,
1156         .ndo_start_xmit         = macvlan_start_xmit,
1157         .ndo_change_mtu         = macvlan_change_mtu,
1158         .ndo_eth_ioctl          = macvlan_eth_ioctl,
1159         .ndo_fix_features       = macvlan_fix_features,
1160         .ndo_change_rx_flags    = macvlan_change_rx_flags,
1161         .ndo_set_mac_address    = macvlan_set_mac_address,
1162         .ndo_set_rx_mode        = macvlan_set_mac_lists,
1163         .ndo_get_stats64        = macvlan_dev_get_stats64,
1164         .ndo_validate_addr      = eth_validate_addr,
1165         .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
1166         .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
1167         .ndo_fdb_add            = macvlan_fdb_add,
1168         .ndo_fdb_del            = macvlan_fdb_del,
1169         .ndo_fdb_dump           = ndo_dflt_fdb_dump,
1170 #ifdef CONFIG_NET_POLL_CONTROLLER
1171         .ndo_poll_controller    = macvlan_dev_poll_controller,
1172         .ndo_netpoll_setup      = macvlan_dev_netpoll_setup,
1173         .ndo_netpoll_cleanup    = macvlan_dev_netpoll_cleanup,
1174 #endif
1175         .ndo_get_iflink         = macvlan_dev_get_iflink,
1176         .ndo_features_check     = passthru_features_check,
1177         .ndo_change_proto_down  = dev_change_proto_down_generic,
1178 };
1179
1180 void macvlan_common_setup(struct net_device *dev)
1181 {
1182         ether_setup(dev);
1183
1184         /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1185         dev->max_mtu            = ETH_MAX_MTU;
1186         dev->priv_flags        &= ~IFF_TX_SKB_SHARING;
1187         netif_keep_dst(dev);
1188         dev->priv_flags        |= IFF_UNICAST_FLT;
1189         dev->netdev_ops         = &macvlan_netdev_ops;
1190         dev->needs_free_netdev  = true;
1191         dev->header_ops         = &macvlan_hard_header_ops;
1192         dev->ethtool_ops        = &macvlan_ethtool_ops;
1193 }
1194 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1195
1196 static void macvlan_setup(struct net_device *dev)
1197 {
1198         macvlan_common_setup(dev);
1199         dev->priv_flags |= IFF_NO_QUEUE;
1200 }
1201
1202 static int macvlan_port_create(struct net_device *dev)
1203 {
1204         struct macvlan_port *port;
1205         unsigned int i;
1206         int err;
1207
1208         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1209                 return -EINVAL;
1210
1211         if (netdev_is_rx_handler_busy(dev))
1212                 return -EBUSY;
1213
1214         port = kzalloc(sizeof(*port), GFP_KERNEL);
1215         if (port == NULL)
1216                 return -ENOMEM;
1217
1218         port->dev = dev;
1219         ether_addr_copy(port->perm_addr, dev->dev_addr);
1220         INIT_LIST_HEAD(&port->vlans);
1221         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1222                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
1223         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1224                 INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1225
1226         port->bc_queue_len_used = 0;
1227         skb_queue_head_init(&port->bc_queue);
1228         INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1229
1230         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1231         if (err)
1232                 kfree(port);
1233         else
1234                 dev->priv_flags |= IFF_MACVLAN_PORT;
1235         return err;
1236 }
1237
1238 static void macvlan_port_destroy(struct net_device *dev)
1239 {
1240         struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1241         struct sk_buff *skb;
1242
1243         dev->priv_flags &= ~IFF_MACVLAN_PORT;
1244         netdev_rx_handler_unregister(dev);
1245
1246         /* After this point, no packet can schedule bc_work anymore,
1247          * but we need to cancel it and purge left skbs if any.
1248          */
1249         cancel_work_sync(&port->bc_work);
1250
1251         while ((skb = __skb_dequeue(&port->bc_queue))) {
1252                 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1253
1254                 if (src)
1255                         dev_put(src->dev);
1256
1257                 kfree_skb(skb);
1258         }
1259
1260         /* If the lower device address has been changed by passthru
1261          * macvlan, put it back.
1262          */
1263         if (macvlan_passthru(port) &&
1264             !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1265                 struct sockaddr sa;
1266
1267                 sa.sa_family = port->dev->type;
1268                 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1269                 dev_set_mac_address(port->dev, &sa, NULL);
1270         }
1271
1272         kfree(port);
1273 }
1274
1275 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1276                             struct netlink_ext_ack *extack)
1277 {
1278         struct nlattr *nla, *head;
1279         int rem, len;
1280
1281         if (tb[IFLA_ADDRESS]) {
1282                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1283                         return -EINVAL;
1284                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1285                         return -EADDRNOTAVAIL;
1286         }
1287
1288         if (!data)
1289                 return 0;
1290
1291         if (data[IFLA_MACVLAN_FLAGS] &&
1292             nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1293                                                       MACVLAN_FLAG_NODST))
1294                 return -EINVAL;
1295
1296         if (data[IFLA_MACVLAN_MODE]) {
1297                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1298                 case MACVLAN_MODE_PRIVATE:
1299                 case MACVLAN_MODE_VEPA:
1300                 case MACVLAN_MODE_BRIDGE:
1301                 case MACVLAN_MODE_PASSTHRU:
1302                 case MACVLAN_MODE_SOURCE:
1303                         break;
1304                 default:
1305                         return -EINVAL;
1306                 }
1307         }
1308
1309         if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1310                 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1311                 case MACVLAN_MACADDR_ADD:
1312                 case MACVLAN_MACADDR_DEL:
1313                 case MACVLAN_MACADDR_FLUSH:
1314                 case MACVLAN_MACADDR_SET:
1315                         break;
1316                 default:
1317                         return -EINVAL;
1318                 }
1319         }
1320
1321         if (data[IFLA_MACVLAN_MACADDR]) {
1322                 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1323                         return -EINVAL;
1324
1325                 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1326                         return -EADDRNOTAVAIL;
1327         }
1328
1329         if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1330                 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1331                 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1332
1333                 nla_for_each_attr(nla, head, len, rem) {
1334                         if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1335                             nla_len(nla) != ETH_ALEN)
1336                                 return -EINVAL;
1337
1338                         if (!is_valid_ether_addr(nla_data(nla)))
1339                                 return -EADDRNOTAVAIL;
1340                 }
1341         }
1342
1343         if (data[IFLA_MACVLAN_MACADDR_COUNT])
1344                 return -EINVAL;
1345
1346         return 0;
1347 }
1348
1349 /*
1350  * reconfigure list of remote source mac address
1351  * (only for macvlan devices in source mode)
1352  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1353  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1354  */
1355 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1356                                       struct nlattr *data[])
1357 {
1358         char *addr = NULL;
1359         int ret, rem, len;
1360         struct nlattr *nla, *head;
1361         struct macvlan_source_entry *entry;
1362
1363         if (data[IFLA_MACVLAN_MACADDR])
1364                 addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1365
1366         if (mode == MACVLAN_MACADDR_ADD) {
1367                 if (!addr)
1368                         return -EINVAL;
1369
1370                 return macvlan_hash_add_source(vlan, addr);
1371
1372         } else if (mode == MACVLAN_MACADDR_DEL) {
1373                 if (!addr)
1374                         return -EINVAL;
1375
1376                 entry = macvlan_hash_lookup_source(vlan, addr);
1377                 if (entry) {
1378                         macvlan_hash_del_source(entry);
1379                         vlan->macaddr_count--;
1380                 }
1381         } else if (mode == MACVLAN_MACADDR_FLUSH) {
1382                 macvlan_flush_sources(vlan->port, vlan);
1383         } else if (mode == MACVLAN_MACADDR_SET) {
1384                 macvlan_flush_sources(vlan->port, vlan);
1385
1386                 if (addr) {
1387                         ret = macvlan_hash_add_source(vlan, addr);
1388                         if (ret)
1389                                 return ret;
1390                 }
1391
1392                 if (!data[IFLA_MACVLAN_MACADDR_DATA])
1393                         return 0;
1394
1395                 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1396                 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1397
1398                 nla_for_each_attr(nla, head, len, rem) {
1399                         addr = nla_data(nla);
1400                         ret = macvlan_hash_add_source(vlan, addr);
1401                         if (ret)
1402                                 return ret;
1403                 }
1404         } else {
1405                 return -EINVAL;
1406         }
1407
1408         return 0;
1409 }
1410
1411 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1412                            struct nlattr *tb[], struct nlattr *data[],
1413                            struct netlink_ext_ack *extack)
1414 {
1415         struct macvlan_dev *vlan = netdev_priv(dev);
1416         struct macvlan_port *port;
1417         struct net_device *lowerdev;
1418         int err;
1419         int macmode;
1420         bool create = false;
1421
1422         if (!tb[IFLA_LINK])
1423                 return -EINVAL;
1424
1425         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1426         if (lowerdev == NULL)
1427                 return -ENODEV;
1428
1429         /* When creating macvlans or macvtaps on top of other macvlans - use
1430          * the real device as the lowerdev.
1431          */
1432         if (netif_is_macvlan(lowerdev))
1433                 lowerdev = macvlan_dev_real_dev(lowerdev);
1434
1435         if (!tb[IFLA_MTU])
1436                 dev->mtu = lowerdev->mtu;
1437         else if (dev->mtu > lowerdev->mtu)
1438                 return -EINVAL;
1439
1440         /* MTU range: 68 - lowerdev->max_mtu */
1441         dev->min_mtu = ETH_MIN_MTU;
1442         dev->max_mtu = lowerdev->max_mtu;
1443
1444         if (!tb[IFLA_ADDRESS])
1445                 eth_hw_addr_random(dev);
1446
1447         if (!netif_is_macvlan_port(lowerdev)) {
1448                 err = macvlan_port_create(lowerdev);
1449                 if (err < 0)
1450                         return err;
1451                 create = true;
1452         }
1453         port = macvlan_port_get_rtnl(lowerdev);
1454
1455         /* Only 1 macvlan device can be created in passthru mode */
1456         if (macvlan_passthru(port)) {
1457                 /* The macvlan port must be not created this time,
1458                  * still goto destroy_macvlan_port for readability.
1459                  */
1460                 err = -EINVAL;
1461                 goto destroy_macvlan_port;
1462         }
1463
1464         vlan->lowerdev = lowerdev;
1465         vlan->dev      = dev;
1466         vlan->port     = port;
1467         vlan->set_features = MACVLAN_FEATURES;
1468
1469         vlan->mode     = MACVLAN_MODE_VEPA;
1470         if (data && data[IFLA_MACVLAN_MODE])
1471                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1472
1473         if (data && data[IFLA_MACVLAN_FLAGS])
1474                 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1475
1476         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1477                 if (port->count) {
1478                         err = -EINVAL;
1479                         goto destroy_macvlan_port;
1480                 }
1481                 macvlan_set_passthru(port);
1482                 eth_hw_addr_inherit(dev, lowerdev);
1483         }
1484
1485         if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1486                 if (vlan->mode != MACVLAN_MODE_SOURCE) {
1487                         err = -EINVAL;
1488                         goto destroy_macvlan_port;
1489                 }
1490                 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1491                 err = macvlan_changelink_sources(vlan, macmode, data);
1492                 if (err)
1493                         goto destroy_macvlan_port;
1494         }
1495
1496         vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1497         if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1498                 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1499
1500         err = register_netdevice(dev);
1501         if (err < 0)
1502                 goto destroy_macvlan_port;
1503
1504         dev->priv_flags |= IFF_MACVLAN;
1505         err = netdev_upper_dev_link(lowerdev, dev, extack);
1506         if (err)
1507                 goto unregister_netdev;
1508
1509         list_add_tail_rcu(&vlan->list, &port->vlans);
1510         update_port_bc_queue_len(vlan->port);
1511         netif_stacked_transfer_operstate(lowerdev, dev);
1512         linkwatch_fire_event(dev);
1513
1514         return 0;
1515
1516 unregister_netdev:
1517         /* macvlan_uninit would free the macvlan port */
1518         unregister_netdevice(dev);
1519         return err;
1520 destroy_macvlan_port:
1521         /* the macvlan port may be freed by macvlan_uninit when fail to register.
1522          * so we destroy the macvlan port only when it's valid.
1523          */
1524         if (create && macvlan_port_get_rtnl(lowerdev)) {
1525                 macvlan_flush_sources(port, vlan);
1526                 macvlan_port_destroy(port->dev);
1527         }
1528         return err;
1529 }
1530 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1531
1532 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1533                            struct nlattr *tb[], struct nlattr *data[],
1534                            struct netlink_ext_ack *extack)
1535 {
1536         return macvlan_common_newlink(src_net, dev, tb, data, extack);
1537 }
1538
1539 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1540 {
1541         struct macvlan_dev *vlan = netdev_priv(dev);
1542
1543         if (vlan->mode == MACVLAN_MODE_SOURCE)
1544                 macvlan_flush_sources(vlan->port, vlan);
1545         list_del_rcu(&vlan->list);
1546         update_port_bc_queue_len(vlan->port);
1547         unregister_netdevice_queue(dev, head);
1548         netdev_upper_dev_unlink(vlan->lowerdev, dev);
1549 }
1550 EXPORT_SYMBOL_GPL(macvlan_dellink);
1551
1552 static int macvlan_changelink(struct net_device *dev,
1553                               struct nlattr *tb[], struct nlattr *data[],
1554                               struct netlink_ext_ack *extack)
1555 {
1556         struct macvlan_dev *vlan = netdev_priv(dev);
1557         enum macvlan_mode mode;
1558         bool set_mode = false;
1559         enum macvlan_macaddr_mode macmode;
1560         int ret;
1561
1562         /* Validate mode, but don't set yet: setting flags may fail. */
1563         if (data && data[IFLA_MACVLAN_MODE]) {
1564                 set_mode = true;
1565                 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1566                 /* Passthrough mode can't be set or cleared dynamically */
1567                 if ((mode == MACVLAN_MODE_PASSTHRU) !=
1568                     (vlan->mode == MACVLAN_MODE_PASSTHRU))
1569                         return -EINVAL;
1570                 if (vlan->mode == MACVLAN_MODE_SOURCE &&
1571                     vlan->mode != mode)
1572                         macvlan_flush_sources(vlan->port, vlan);
1573         }
1574
1575         if (data && data[IFLA_MACVLAN_FLAGS]) {
1576                 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1577                 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1578                 if (macvlan_passthru(vlan->port) && promisc) {
1579                         int err;
1580
1581                         if (flags & MACVLAN_FLAG_NOPROMISC)
1582                                 err = dev_set_promiscuity(vlan->lowerdev, -1);
1583                         else
1584                                 err = dev_set_promiscuity(vlan->lowerdev, 1);
1585                         if (err < 0)
1586                                 return err;
1587                 }
1588                 vlan->flags = flags;
1589         }
1590
1591         if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1592                 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1593                 update_port_bc_queue_len(vlan->port);
1594         }
1595
1596         if (set_mode)
1597                 vlan->mode = mode;
1598         if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1599                 if (vlan->mode != MACVLAN_MODE_SOURCE)
1600                         return -EINVAL;
1601                 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1602                 ret = macvlan_changelink_sources(vlan, macmode, data);
1603                 if (ret)
1604                         return ret;
1605         }
1606         return 0;
1607 }
1608
1609 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1610 {
1611         if (vlan->macaddr_count == 0)
1612                 return 0;
1613         return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1614                 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1615 }
1616
1617 static size_t macvlan_get_size(const struct net_device *dev)
1618 {
1619         struct macvlan_dev *vlan = netdev_priv(dev);
1620
1621         return (0
1622                 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1623                 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1624                 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1625                 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1626                 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1627                 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1628                 );
1629 }
1630
1631 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1632                                      const struct macvlan_dev *vlan,
1633                                      const int i)
1634 {
1635         struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1636         struct macvlan_source_entry *entry;
1637
1638         hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1639                 if (entry->vlan != vlan)
1640                         continue;
1641                 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1642                         return 1;
1643         }
1644         return 0;
1645 }
1646
1647 static int macvlan_fill_info(struct sk_buff *skb,
1648                                 const struct net_device *dev)
1649 {
1650         struct macvlan_dev *vlan = netdev_priv(dev);
1651         struct macvlan_port *port = vlan->port;
1652         int i;
1653         struct nlattr *nest;
1654
1655         if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1656                 goto nla_put_failure;
1657         if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1658                 goto nla_put_failure;
1659         if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1660                 goto nla_put_failure;
1661         if (vlan->macaddr_count > 0) {
1662                 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1663                 if (nest == NULL)
1664                         goto nla_put_failure;
1665
1666                 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1667                         if (macvlan_fill_info_macaddr(skb, vlan, i))
1668                                 goto nla_put_failure;
1669                 }
1670                 nla_nest_end(skb, nest);
1671         }
1672         if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1673                 goto nla_put_failure;
1674         if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1675                 goto nla_put_failure;
1676         return 0;
1677
1678 nla_put_failure:
1679         return -EMSGSIZE;
1680 }
1681
1682 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1683         [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1684         [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1685         [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1686         [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1687         [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1688         [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1689         [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1690         [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1691 };
1692
1693 int macvlan_link_register(struct rtnl_link_ops *ops)
1694 {
1695         /* common fields */
1696         ops->validate           = macvlan_validate;
1697         ops->maxtype            = IFLA_MACVLAN_MAX;
1698         ops->policy             = macvlan_policy;
1699         ops->changelink         = macvlan_changelink;
1700         ops->get_size           = macvlan_get_size;
1701         ops->fill_info          = macvlan_fill_info;
1702
1703         return rtnl_link_register(ops);
1704 };
1705 EXPORT_SYMBOL_GPL(macvlan_link_register);
1706
1707 static struct net *macvlan_get_link_net(const struct net_device *dev)
1708 {
1709         return dev_net(macvlan_dev_real_dev(dev));
1710 }
1711
1712 static struct rtnl_link_ops macvlan_link_ops = {
1713         .kind           = "macvlan",
1714         .setup          = macvlan_setup,
1715         .newlink        = macvlan_newlink,
1716         .dellink        = macvlan_dellink,
1717         .get_link_net   = macvlan_get_link_net,
1718         .priv_size      = sizeof(struct macvlan_dev),
1719 };
1720
1721 static void update_port_bc_queue_len(struct macvlan_port *port)
1722 {
1723         u32 max_bc_queue_len_req = 0;
1724         struct macvlan_dev *vlan;
1725
1726         list_for_each_entry(vlan, &port->vlans, list) {
1727                 if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1728                         max_bc_queue_len_req = vlan->bc_queue_len_req;
1729         }
1730         port->bc_queue_len_used = max_bc_queue_len_req;
1731 }
1732
1733 static int macvlan_device_event(struct notifier_block *unused,
1734                                 unsigned long event, void *ptr)
1735 {
1736         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1737         struct macvlan_dev *vlan, *next;
1738         struct macvlan_port *port;
1739         LIST_HEAD(list_kill);
1740
1741         if (!netif_is_macvlan_port(dev))
1742                 return NOTIFY_DONE;
1743
1744         port = macvlan_port_get_rtnl(dev);
1745
1746         switch (event) {
1747         case NETDEV_UP:
1748         case NETDEV_DOWN:
1749         case NETDEV_CHANGE:
1750                 list_for_each_entry(vlan, &port->vlans, list)
1751                         netif_stacked_transfer_operstate(vlan->lowerdev,
1752                                                          vlan->dev);
1753                 break;
1754         case NETDEV_FEAT_CHANGE:
1755                 list_for_each_entry(vlan, &port->vlans, list) {
1756                         vlan->dev->gso_max_size = dev->gso_max_size;
1757                         vlan->dev->gso_max_segs = dev->gso_max_segs;
1758                         netdev_update_features(vlan->dev);
1759                 }
1760                 break;
1761         case NETDEV_CHANGEMTU:
1762                 list_for_each_entry(vlan, &port->vlans, list) {
1763                         if (vlan->dev->mtu <= dev->mtu)
1764                                 continue;
1765                         dev_set_mtu(vlan->dev, dev->mtu);
1766                 }
1767                 break;
1768         case NETDEV_CHANGEADDR:
1769                 if (!macvlan_passthru(port))
1770                         return NOTIFY_DONE;
1771
1772                 vlan = list_first_entry_or_null(&port->vlans,
1773                                                 struct macvlan_dev,
1774                                                 list);
1775
1776                 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1777                         return NOTIFY_BAD;
1778
1779                 break;
1780         case NETDEV_UNREGISTER:
1781                 /* twiddle thumbs on netns device moves */
1782                 if (dev->reg_state != NETREG_UNREGISTERING)
1783                         break;
1784
1785                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1786                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1787                 unregister_netdevice_many(&list_kill);
1788                 break;
1789         case NETDEV_PRE_TYPE_CHANGE:
1790                 /* Forbid underlying device to change its type. */
1791                 return NOTIFY_BAD;
1792
1793         case NETDEV_NOTIFY_PEERS:
1794         case NETDEV_BONDING_FAILOVER:
1795         case NETDEV_RESEND_IGMP:
1796                 /* Propagate to all vlans */
1797                 list_for_each_entry(vlan, &port->vlans, list)
1798                         call_netdevice_notifiers(event, vlan->dev);
1799         }
1800         return NOTIFY_DONE;
1801 }
1802
1803 static struct notifier_block macvlan_notifier_block __read_mostly = {
1804         .notifier_call  = macvlan_device_event,
1805 };
1806
1807 static int __init macvlan_init_module(void)
1808 {
1809         int err;
1810
1811         register_netdevice_notifier(&macvlan_notifier_block);
1812
1813         err = macvlan_link_register(&macvlan_link_ops);
1814         if (err < 0)
1815                 goto err1;
1816         return 0;
1817 err1:
1818         unregister_netdevice_notifier(&macvlan_notifier_block);
1819         return err;
1820 }
1821
1822 static void __exit macvlan_cleanup_module(void)
1823 {
1824         rtnl_link_unregister(&macvlan_link_ops);
1825         unregister_netdevice_notifier(&macvlan_notifier_block);
1826 }
1827
1828 module_init(macvlan_init_module);
1829 module_exit(macvlan_cleanup_module);
1830
1831 MODULE_LICENSE("GPL");
1832 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1833 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1834 MODULE_ALIAS_RTNL_LINK("macvlan");