xref: /linux/drivers/net/ipvlan/ipvlan_main.c (revision 1553a1c48281243359a9529a10ddb551f3b967ab)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
3  */
4 
5 #include <linux/ethtool.h>
6 
7 #include "ipvlan.h"
8 
9 static int ipvlan_set_port_mode(struct ipvl_port *port, u16 nval,
10 				struct netlink_ext_ack *extack)
11 {
12 	struct ipvl_dev *ipvlan;
13 	unsigned int flags;
14 	int err;
15 
16 	ASSERT_RTNL();
17 	if (port->mode != nval) {
18 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
19 			flags = ipvlan->dev->flags;
20 			if (nval == IPVLAN_MODE_L3 || nval == IPVLAN_MODE_L3S) {
21 				err = dev_change_flags(ipvlan->dev,
22 						       flags | IFF_NOARP,
23 						       extack);
24 			} else {
25 				err = dev_change_flags(ipvlan->dev,
26 						       flags & ~IFF_NOARP,
27 						       extack);
28 			}
29 			if (unlikely(err))
30 				goto fail;
31 		}
32 		if (nval == IPVLAN_MODE_L3S) {
33 			/* New mode is L3S */
34 			err = ipvlan_l3s_register(port);
35 			if (err)
36 				goto fail;
37 		} else if (port->mode == IPVLAN_MODE_L3S) {
38 			/* Old mode was L3S */
39 			ipvlan_l3s_unregister(port);
40 		}
41 		port->mode = nval;
42 	}
43 	return 0;
44 
45 fail:
46 	/* Undo the flags changes that have been done so far. */
47 	list_for_each_entry_continue_reverse(ipvlan, &port->ipvlans, pnode) {
48 		flags = ipvlan->dev->flags;
49 		if (port->mode == IPVLAN_MODE_L3 ||
50 		    port->mode == IPVLAN_MODE_L3S)
51 			dev_change_flags(ipvlan->dev, flags | IFF_NOARP,
52 					 NULL);
53 		else
54 			dev_change_flags(ipvlan->dev, flags & ~IFF_NOARP,
55 					 NULL);
56 	}
57 
58 	return err;
59 }
60 
61 static int ipvlan_port_create(struct net_device *dev)
62 {
63 	struct ipvl_port *port;
64 	int err, idx;
65 
66 	port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL);
67 	if (!port)
68 		return -ENOMEM;
69 
70 	write_pnet(&port->pnet, dev_net(dev));
71 	port->dev = dev;
72 	port->mode = IPVLAN_MODE_L3;
73 	INIT_LIST_HEAD(&port->ipvlans);
74 	for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++)
75 		INIT_HLIST_HEAD(&port->hlhead[idx]);
76 
77 	skb_queue_head_init(&port->backlog);
78 	INIT_WORK(&port->wq, ipvlan_process_multicast);
79 	ida_init(&port->ida);
80 	port->dev_id_start = 1;
81 
82 	err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port);
83 	if (err)
84 		goto err;
85 
86 	netdev_hold(dev, &port->dev_tracker, GFP_KERNEL);
87 	return 0;
88 
89 err:
90 	kfree(port);
91 	return err;
92 }
93 
94 static void ipvlan_port_destroy(struct net_device *dev)
95 {
96 	struct ipvl_port *port = ipvlan_port_get_rtnl(dev);
97 	struct sk_buff *skb;
98 
99 	netdev_put(dev, &port->dev_tracker);
100 	if (port->mode == IPVLAN_MODE_L3S)
101 		ipvlan_l3s_unregister(port);
102 	netdev_rx_handler_unregister(dev);
103 	cancel_work_sync(&port->wq);
104 	while ((skb = __skb_dequeue(&port->backlog)) != NULL) {
105 		dev_put(skb->dev);
106 		kfree_skb(skb);
107 	}
108 	ida_destroy(&port->ida);
109 	kfree(port);
110 }
111 
112 #define IPVLAN_ALWAYS_ON_OFLOADS \
113 	(NETIF_F_SG | NETIF_F_HW_CSUM | \
114 	 NETIF_F_GSO_ROBUST | NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL)
115 
116 #define IPVLAN_ALWAYS_ON \
117 	(IPVLAN_ALWAYS_ON_OFLOADS | NETIF_F_LLTX | NETIF_F_VLAN_CHALLENGED)
118 
119 #define IPVLAN_FEATURES \
120 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
121 	 NETIF_F_GSO | NETIF_F_ALL_TSO | NETIF_F_GSO_ROBUST | \
122 	 NETIF_F_GRO | NETIF_F_RXCSUM | \
123 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
124 
125 	/* NETIF_F_GSO_ENCAP_ALL NETIF_F_GSO_SOFTWARE Newly added */
126 
127 #define IPVLAN_STATE_MASK \
128 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
129 
130 static int ipvlan_init(struct net_device *dev)
131 {
132 	struct ipvl_dev *ipvlan = netdev_priv(dev);
133 	struct net_device *phy_dev = ipvlan->phy_dev;
134 	struct ipvl_port *port;
135 	int err;
136 
137 	dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
138 		     (phy_dev->state & IPVLAN_STATE_MASK);
139 	dev->features = phy_dev->features & IPVLAN_FEATURES;
140 	dev->features |= IPVLAN_ALWAYS_ON;
141 	dev->vlan_features = phy_dev->vlan_features & IPVLAN_FEATURES;
142 	dev->vlan_features |= IPVLAN_ALWAYS_ON_OFLOADS;
143 	dev->hw_enc_features |= dev->features;
144 	netif_inherit_tso_max(dev, phy_dev);
145 	dev->hard_header_len = phy_dev->hard_header_len;
146 
147 	netdev_lockdep_set_classes(dev);
148 
149 	ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
150 	if (!ipvlan->pcpu_stats)
151 		return -ENOMEM;
152 
153 	if (!netif_is_ipvlan_port(phy_dev)) {
154 		err = ipvlan_port_create(phy_dev);
155 		if (err < 0) {
156 			free_percpu(ipvlan->pcpu_stats);
157 			return err;
158 		}
159 	}
160 	port = ipvlan_port_get_rtnl(phy_dev);
161 	port->count += 1;
162 	return 0;
163 }
164 
165 static void ipvlan_uninit(struct net_device *dev)
166 {
167 	struct ipvl_dev *ipvlan = netdev_priv(dev);
168 	struct net_device *phy_dev = ipvlan->phy_dev;
169 	struct ipvl_port *port;
170 
171 	free_percpu(ipvlan->pcpu_stats);
172 
173 	port = ipvlan_port_get_rtnl(phy_dev);
174 	port->count -= 1;
175 	if (!port->count)
176 		ipvlan_port_destroy(port->dev);
177 }
178 
179 static int ipvlan_open(struct net_device *dev)
180 {
181 	struct ipvl_dev *ipvlan = netdev_priv(dev);
182 	struct ipvl_addr *addr;
183 
184 	if (ipvlan->port->mode == IPVLAN_MODE_L3 ||
185 	    ipvlan->port->mode == IPVLAN_MODE_L3S)
186 		dev->flags |= IFF_NOARP;
187 	else
188 		dev->flags &= ~IFF_NOARP;
189 
190 	rcu_read_lock();
191 	list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
192 		ipvlan_ht_addr_add(ipvlan, addr);
193 	rcu_read_unlock();
194 
195 	return 0;
196 }
197 
198 static int ipvlan_stop(struct net_device *dev)
199 {
200 	struct ipvl_dev *ipvlan = netdev_priv(dev);
201 	struct net_device *phy_dev = ipvlan->phy_dev;
202 	struct ipvl_addr *addr;
203 
204 	dev_uc_unsync(phy_dev, dev);
205 	dev_mc_unsync(phy_dev, dev);
206 
207 	rcu_read_lock();
208 	list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
209 		ipvlan_ht_addr_del(addr);
210 	rcu_read_unlock();
211 
212 	return 0;
213 }
214 
215 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
216 				     struct net_device *dev)
217 {
218 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
219 	int skblen = skb->len;
220 	int ret;
221 
222 	ret = ipvlan_queue_xmit(skb, dev);
223 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
224 		struct ipvl_pcpu_stats *pcptr;
225 
226 		pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
227 
228 		u64_stats_update_begin(&pcptr->syncp);
229 		u64_stats_inc(&pcptr->tx_pkts);
230 		u64_stats_add(&pcptr->tx_bytes, skblen);
231 		u64_stats_update_end(&pcptr->syncp);
232 	} else {
233 		this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
234 	}
235 	return ret;
236 }
237 
238 static netdev_features_t ipvlan_fix_features(struct net_device *dev,
239 					     netdev_features_t features)
240 {
241 	struct ipvl_dev *ipvlan = netdev_priv(dev);
242 
243 	features |= NETIF_F_ALL_FOR_ALL;
244 	features &= (ipvlan->sfeatures | ~IPVLAN_FEATURES);
245 	features = netdev_increment_features(ipvlan->phy_dev->features,
246 					     features, features);
247 	features |= IPVLAN_ALWAYS_ON;
248 	features &= (IPVLAN_FEATURES | IPVLAN_ALWAYS_ON);
249 
250 	return features;
251 }
252 
253 static void ipvlan_change_rx_flags(struct net_device *dev, int change)
254 {
255 	struct ipvl_dev *ipvlan = netdev_priv(dev);
256 	struct net_device *phy_dev = ipvlan->phy_dev;
257 
258 	if (change & IFF_ALLMULTI)
259 		dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
260 }
261 
262 static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
263 {
264 	struct ipvl_dev *ipvlan = netdev_priv(dev);
265 
266 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
267 		bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
268 	} else {
269 		struct netdev_hw_addr *ha;
270 		DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
271 
272 		bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
273 		netdev_for_each_mc_addr(ha, dev)
274 			__set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
275 
276 		/* Turn-on broadcast bit irrespective of address family,
277 		 * since broadcast is deferred to a work-queue, hence no
278 		 * impact on fast-path processing.
279 		 */
280 		__set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
281 
282 		bitmap_copy(ipvlan->mac_filters, mc_filters,
283 			    IPVLAN_MAC_FILTER_SIZE);
284 	}
285 	dev_uc_sync(ipvlan->phy_dev, dev);
286 	dev_mc_sync(ipvlan->phy_dev, dev);
287 }
288 
289 static void ipvlan_get_stats64(struct net_device *dev,
290 			       struct rtnl_link_stats64 *s)
291 {
292 	struct ipvl_dev *ipvlan = netdev_priv(dev);
293 
294 	if (ipvlan->pcpu_stats) {
295 		struct ipvl_pcpu_stats *pcptr;
296 		u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
297 		u32 rx_errs = 0, tx_drps = 0;
298 		u32 strt;
299 		int idx;
300 
301 		for_each_possible_cpu(idx) {
302 			pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
303 			do {
304 				strt = u64_stats_fetch_begin(&pcptr->syncp);
305 				rx_pkts = u64_stats_read(&pcptr->rx_pkts);
306 				rx_bytes = u64_stats_read(&pcptr->rx_bytes);
307 				rx_mcast = u64_stats_read(&pcptr->rx_mcast);
308 				tx_pkts = u64_stats_read(&pcptr->tx_pkts);
309 				tx_bytes = u64_stats_read(&pcptr->tx_bytes);
310 			} while (u64_stats_fetch_retry(&pcptr->syncp,
311 							   strt));
312 
313 			s->rx_packets += rx_pkts;
314 			s->rx_bytes += rx_bytes;
315 			s->multicast += rx_mcast;
316 			s->tx_packets += tx_pkts;
317 			s->tx_bytes += tx_bytes;
318 
319 			/* u32 values are updated without syncp protection. */
320 			rx_errs += READ_ONCE(pcptr->rx_errs);
321 			tx_drps += READ_ONCE(pcptr->tx_drps);
322 		}
323 		s->rx_errors = rx_errs;
324 		s->rx_dropped = rx_errs;
325 		s->tx_dropped = tx_drps;
326 	}
327 	s->tx_errors = DEV_STATS_READ(dev, tx_errors);
328 }
329 
330 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
331 {
332 	struct ipvl_dev *ipvlan = netdev_priv(dev);
333 	struct net_device *phy_dev = ipvlan->phy_dev;
334 
335 	return vlan_vid_add(phy_dev, proto, vid);
336 }
337 
338 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
339 				   u16 vid)
340 {
341 	struct ipvl_dev *ipvlan = netdev_priv(dev);
342 	struct net_device *phy_dev = ipvlan->phy_dev;
343 
344 	vlan_vid_del(phy_dev, proto, vid);
345 	return 0;
346 }
347 
348 static int ipvlan_get_iflink(const struct net_device *dev)
349 {
350 	struct ipvl_dev *ipvlan = netdev_priv(dev);
351 
352 	return READ_ONCE(ipvlan->phy_dev->ifindex);
353 }
354 
355 static const struct net_device_ops ipvlan_netdev_ops = {
356 	.ndo_init		= ipvlan_init,
357 	.ndo_uninit		= ipvlan_uninit,
358 	.ndo_open		= ipvlan_open,
359 	.ndo_stop		= ipvlan_stop,
360 	.ndo_start_xmit		= ipvlan_start_xmit,
361 	.ndo_fix_features	= ipvlan_fix_features,
362 	.ndo_change_rx_flags	= ipvlan_change_rx_flags,
363 	.ndo_set_rx_mode	= ipvlan_set_multicast_mac_filter,
364 	.ndo_get_stats64	= ipvlan_get_stats64,
365 	.ndo_vlan_rx_add_vid	= ipvlan_vlan_rx_add_vid,
366 	.ndo_vlan_rx_kill_vid	= ipvlan_vlan_rx_kill_vid,
367 	.ndo_get_iflink		= ipvlan_get_iflink,
368 };
369 
370 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
371 			      unsigned short type, const void *daddr,
372 			      const void *saddr, unsigned len)
373 {
374 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
375 	struct net_device *phy_dev = ipvlan->phy_dev;
376 
377 	/* TODO Probably use a different field than dev_addr so that the
378 	 * mac-address on the virtual device is portable and can be carried
379 	 * while the packets use the mac-addr on the physical device.
380 	 */
381 	return dev_hard_header(skb, phy_dev, type, daddr,
382 			       saddr ? : phy_dev->dev_addr, len);
383 }
384 
385 static const struct header_ops ipvlan_header_ops = {
386 	.create  	= ipvlan_hard_header,
387 	.parse		= eth_header_parse,
388 	.cache		= eth_header_cache,
389 	.cache_update	= eth_header_cache_update,
390 	.parse_protocol	= eth_header_parse_protocol,
391 };
392 
393 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
394 {
395 	ipvlan->dev->mtu = dev->mtu;
396 }
397 
398 static bool netif_is_ipvlan(const struct net_device *dev)
399 {
400 	/* both ipvlan and ipvtap devices use the same netdev_ops */
401 	return dev->netdev_ops == &ipvlan_netdev_ops;
402 }
403 
404 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
405 					     struct ethtool_link_ksettings *cmd)
406 {
407 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
408 
409 	return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
410 }
411 
412 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
413 				       struct ethtool_drvinfo *drvinfo)
414 {
415 	strscpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
416 	strscpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
417 }
418 
419 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
420 {
421 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
422 
423 	return ipvlan->msg_enable;
424 }
425 
426 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
427 {
428 	struct ipvl_dev *ipvlan = netdev_priv(dev);
429 
430 	ipvlan->msg_enable = value;
431 }
432 
433 static const struct ethtool_ops ipvlan_ethtool_ops = {
434 	.get_link	= ethtool_op_get_link,
435 	.get_link_ksettings	= ipvlan_ethtool_get_link_ksettings,
436 	.get_drvinfo	= ipvlan_ethtool_get_drvinfo,
437 	.get_msglevel	= ipvlan_ethtool_get_msglevel,
438 	.set_msglevel	= ipvlan_ethtool_set_msglevel,
439 };
440 
441 static int ipvlan_nl_changelink(struct net_device *dev,
442 				struct nlattr *tb[], struct nlattr *data[],
443 				struct netlink_ext_ack *extack)
444 {
445 	struct ipvl_dev *ipvlan = netdev_priv(dev);
446 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
447 	int err = 0;
448 
449 	if (!data)
450 		return 0;
451 	if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
452 		return -EPERM;
453 
454 	if (data[IFLA_IPVLAN_MODE]) {
455 		u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
456 
457 		err = ipvlan_set_port_mode(port, nmode, extack);
458 	}
459 
460 	if (!err && data[IFLA_IPVLAN_FLAGS]) {
461 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
462 
463 		if (flags & IPVLAN_F_PRIVATE)
464 			ipvlan_mark_private(port);
465 		else
466 			ipvlan_clear_private(port);
467 
468 		if (flags & IPVLAN_F_VEPA)
469 			ipvlan_mark_vepa(port);
470 		else
471 			ipvlan_clear_vepa(port);
472 	}
473 
474 	return err;
475 }
476 
477 static size_t ipvlan_nl_getsize(const struct net_device *dev)
478 {
479 	return (0
480 		+ nla_total_size(2) /* IFLA_IPVLAN_MODE */
481 		+ nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
482 		);
483 }
484 
485 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
486 			      struct netlink_ext_ack *extack)
487 {
488 	if (!data)
489 		return 0;
490 
491 	if (data[IFLA_IPVLAN_MODE]) {
492 		u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
493 
494 		if (mode >= IPVLAN_MODE_MAX)
495 			return -EINVAL;
496 	}
497 	if (data[IFLA_IPVLAN_FLAGS]) {
498 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
499 
500 		/* Only two bits are used at this moment. */
501 		if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
502 			return -EINVAL;
503 		/* Also both flags can't be active at the same time. */
504 		if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
505 		    (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
506 			return -EINVAL;
507 	}
508 
509 	return 0;
510 }
511 
512 static int ipvlan_nl_fillinfo(struct sk_buff *skb,
513 			      const struct net_device *dev)
514 {
515 	struct ipvl_dev *ipvlan = netdev_priv(dev);
516 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
517 	int ret = -EINVAL;
518 
519 	if (!port)
520 		goto err;
521 
522 	ret = -EMSGSIZE;
523 	if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
524 		goto err;
525 	if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
526 		goto err;
527 
528 	return 0;
529 
530 err:
531 	return ret;
532 }
533 
534 int ipvlan_link_new(struct net *src_net, struct net_device *dev,
535 		    struct nlattr *tb[], struct nlattr *data[],
536 		    struct netlink_ext_ack *extack)
537 {
538 	struct ipvl_dev *ipvlan = netdev_priv(dev);
539 	struct ipvl_port *port;
540 	struct net_device *phy_dev;
541 	int err;
542 	u16 mode = IPVLAN_MODE_L3;
543 
544 	if (!tb[IFLA_LINK])
545 		return -EINVAL;
546 
547 	phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
548 	if (!phy_dev)
549 		return -ENODEV;
550 
551 	if (netif_is_ipvlan(phy_dev)) {
552 		struct ipvl_dev *tmp = netdev_priv(phy_dev);
553 
554 		phy_dev = tmp->phy_dev;
555 		if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
556 			return -EPERM;
557 	} else if (!netif_is_ipvlan_port(phy_dev)) {
558 		/* Exit early if the underlying link is invalid or busy */
559 		if (phy_dev->type != ARPHRD_ETHER ||
560 		    phy_dev->flags & IFF_LOOPBACK) {
561 			netdev_err(phy_dev,
562 				   "Master is either lo or non-ether device\n");
563 			return -EINVAL;
564 		}
565 
566 		if (netdev_is_rx_handler_busy(phy_dev)) {
567 			netdev_err(phy_dev, "Device is already in use.\n");
568 			return -EBUSY;
569 		}
570 	}
571 
572 	ipvlan->phy_dev = phy_dev;
573 	ipvlan->dev = dev;
574 	ipvlan->sfeatures = IPVLAN_FEATURES;
575 	if (!tb[IFLA_MTU])
576 		ipvlan_adjust_mtu(ipvlan, phy_dev);
577 	INIT_LIST_HEAD(&ipvlan->addrs);
578 	spin_lock_init(&ipvlan->addrs_lock);
579 
580 	/* TODO Probably put random address here to be presented to the
581 	 * world but keep using the physical-dev address for the outgoing
582 	 * packets.
583 	 */
584 	eth_hw_addr_set(dev, phy_dev->dev_addr);
585 
586 	dev->priv_flags |= IFF_NO_RX_HANDLER;
587 
588 	err = register_netdevice(dev);
589 	if (err < 0)
590 		return err;
591 
592 	/* ipvlan_init() would have created the port, if required */
593 	port = ipvlan_port_get_rtnl(phy_dev);
594 	ipvlan->port = port;
595 
596 	/* If the port-id base is at the MAX value, then wrap it around and
597 	 * begin from 0x1 again. This may be due to a busy system where lots
598 	 * of slaves are getting created and deleted.
599 	 */
600 	if (port->dev_id_start == 0xFFFE)
601 		port->dev_id_start = 0x1;
602 
603 	/* Since L2 address is shared among all IPvlan slaves including
604 	 * master, use unique 16 bit dev-ids to differentiate among them.
605 	 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
606 	 * slave link [see addrconf_ifid_eui48()].
607 	 */
608 	err = ida_alloc_range(&port->ida, port->dev_id_start, 0xFFFD,
609 			      GFP_KERNEL);
610 	if (err < 0)
611 		err = ida_alloc_range(&port->ida, 0x1, port->dev_id_start - 1,
612 				      GFP_KERNEL);
613 	if (err < 0)
614 		goto unregister_netdev;
615 	dev->dev_id = err;
616 
617 	/* Increment id-base to the next slot for the future assignment */
618 	port->dev_id_start = err + 1;
619 
620 	err = netdev_upper_dev_link(phy_dev, dev, extack);
621 	if (err)
622 		goto remove_ida;
623 
624 	/* Flags are per port and latest update overrides. User has
625 	 * to be consistent in setting it just like the mode attribute.
626 	 */
627 	if (data && data[IFLA_IPVLAN_FLAGS])
628 		port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
629 
630 	if (data && data[IFLA_IPVLAN_MODE])
631 		mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
632 
633 	err = ipvlan_set_port_mode(port, mode, extack);
634 	if (err)
635 		goto unlink_netdev;
636 
637 	list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
638 	netif_stacked_transfer_operstate(phy_dev, dev);
639 	return 0;
640 
641 unlink_netdev:
642 	netdev_upper_dev_unlink(phy_dev, dev);
643 remove_ida:
644 	ida_free(&port->ida, dev->dev_id);
645 unregister_netdev:
646 	unregister_netdevice(dev);
647 	return err;
648 }
649 EXPORT_SYMBOL_GPL(ipvlan_link_new);
650 
651 void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
652 {
653 	struct ipvl_dev *ipvlan = netdev_priv(dev);
654 	struct ipvl_addr *addr, *next;
655 
656 	spin_lock_bh(&ipvlan->addrs_lock);
657 	list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
658 		ipvlan_ht_addr_del(addr);
659 		list_del_rcu(&addr->anode);
660 		kfree_rcu(addr, rcu);
661 	}
662 	spin_unlock_bh(&ipvlan->addrs_lock);
663 
664 	ida_free(&ipvlan->port->ida, dev->dev_id);
665 	list_del_rcu(&ipvlan->pnode);
666 	unregister_netdevice_queue(dev, head);
667 	netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
668 }
669 EXPORT_SYMBOL_GPL(ipvlan_link_delete);
670 
671 void ipvlan_link_setup(struct net_device *dev)
672 {
673 	ether_setup(dev);
674 
675 	dev->max_mtu = ETH_MAX_MTU;
676 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
677 	dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
678 	dev->netdev_ops = &ipvlan_netdev_ops;
679 	dev->needs_free_netdev = true;
680 	dev->header_ops = &ipvlan_header_ops;
681 	dev->ethtool_ops = &ipvlan_ethtool_ops;
682 }
683 EXPORT_SYMBOL_GPL(ipvlan_link_setup);
684 
685 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
686 {
687 	[IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
688 	[IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
689 };
690 
691 static struct net *ipvlan_get_link_net(const struct net_device *dev)
692 {
693 	struct ipvl_dev *ipvlan = netdev_priv(dev);
694 
695 	return dev_net(ipvlan->phy_dev);
696 }
697 
698 static struct rtnl_link_ops ipvlan_link_ops = {
699 	.kind		= "ipvlan",
700 	.priv_size	= sizeof(struct ipvl_dev),
701 
702 	.setup		= ipvlan_link_setup,
703 	.newlink	= ipvlan_link_new,
704 	.dellink	= ipvlan_link_delete,
705 	.get_link_net   = ipvlan_get_link_net,
706 };
707 
708 int ipvlan_link_register(struct rtnl_link_ops *ops)
709 {
710 	ops->get_size	= ipvlan_nl_getsize;
711 	ops->policy	= ipvlan_nl_policy;
712 	ops->validate	= ipvlan_nl_validate;
713 	ops->fill_info	= ipvlan_nl_fillinfo;
714 	ops->changelink = ipvlan_nl_changelink;
715 	ops->maxtype	= IFLA_IPVLAN_MAX;
716 	return rtnl_link_register(ops);
717 }
718 EXPORT_SYMBOL_GPL(ipvlan_link_register);
719 
720 static int ipvlan_device_event(struct notifier_block *unused,
721 			       unsigned long event, void *ptr)
722 {
723 	struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
724 	struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
725 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
726 	struct ipvl_dev *ipvlan, *next;
727 	struct ipvl_port *port;
728 	LIST_HEAD(lst_kill);
729 	int err;
730 
731 	if (!netif_is_ipvlan_port(dev))
732 		return NOTIFY_DONE;
733 
734 	port = ipvlan_port_get_rtnl(dev);
735 
736 	switch (event) {
737 	case NETDEV_UP:
738 	case NETDEV_CHANGE:
739 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
740 			netif_stacked_transfer_operstate(ipvlan->phy_dev,
741 							 ipvlan->dev);
742 		break;
743 
744 	case NETDEV_REGISTER: {
745 		struct net *oldnet, *newnet = dev_net(dev);
746 
747 		oldnet = read_pnet(&port->pnet);
748 		if (net_eq(newnet, oldnet))
749 			break;
750 
751 		write_pnet(&port->pnet, newnet);
752 
753 		if (port->mode == IPVLAN_MODE_L3S)
754 			ipvlan_migrate_l3s_hook(oldnet, newnet);
755 		break;
756 	}
757 	case NETDEV_UNREGISTER:
758 		if (dev->reg_state != NETREG_UNREGISTERING)
759 			break;
760 
761 		list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
762 			ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
763 							    &lst_kill);
764 		unregister_netdevice_many(&lst_kill);
765 		break;
766 
767 	case NETDEV_FEAT_CHANGE:
768 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
769 			netif_inherit_tso_max(ipvlan->dev, dev);
770 			netdev_update_features(ipvlan->dev);
771 		}
772 		break;
773 
774 	case NETDEV_CHANGEMTU:
775 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
776 			ipvlan_adjust_mtu(ipvlan, dev);
777 		break;
778 
779 	case NETDEV_PRE_CHANGEADDR:
780 		prechaddr_info = ptr;
781 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
782 			err = dev_pre_changeaddr_notify(ipvlan->dev,
783 						    prechaddr_info->dev_addr,
784 						    extack);
785 			if (err)
786 				return notifier_from_errno(err);
787 		}
788 		break;
789 
790 	case NETDEV_CHANGEADDR:
791 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
792 			eth_hw_addr_set(ipvlan->dev, dev->dev_addr);
793 			call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
794 		}
795 		break;
796 
797 	case NETDEV_PRE_TYPE_CHANGE:
798 		/* Forbid underlying device to change its type. */
799 		return NOTIFY_BAD;
800 	}
801 	return NOTIFY_DONE;
802 }
803 
804 /* the caller must held the addrs lock */
805 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
806 {
807 	struct ipvl_addr *addr;
808 
809 	addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
810 	if (!addr)
811 		return -ENOMEM;
812 
813 	addr->master = ipvlan;
814 	if (!is_v6) {
815 		memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
816 		addr->atype = IPVL_IPV4;
817 #if IS_ENABLED(CONFIG_IPV6)
818 	} else {
819 		memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
820 		addr->atype = IPVL_IPV6;
821 #endif
822 	}
823 
824 	list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
825 
826 	/* If the interface is not up, the address will be added to the hash
827 	 * list by ipvlan_open.
828 	 */
829 	if (netif_running(ipvlan->dev))
830 		ipvlan_ht_addr_add(ipvlan, addr);
831 
832 	return 0;
833 }
834 
835 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
836 {
837 	struct ipvl_addr *addr;
838 
839 	spin_lock_bh(&ipvlan->addrs_lock);
840 	addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
841 	if (!addr) {
842 		spin_unlock_bh(&ipvlan->addrs_lock);
843 		return;
844 	}
845 
846 	ipvlan_ht_addr_del(addr);
847 	list_del_rcu(&addr->anode);
848 	spin_unlock_bh(&ipvlan->addrs_lock);
849 	kfree_rcu(addr, rcu);
850 }
851 
852 static bool ipvlan_is_valid_dev(const struct net_device *dev)
853 {
854 	struct ipvl_dev *ipvlan = netdev_priv(dev);
855 
856 	if (!netif_is_ipvlan(dev))
857 		return false;
858 
859 	if (!ipvlan || !ipvlan->port)
860 		return false;
861 
862 	return true;
863 }
864 
865 #if IS_ENABLED(CONFIG_IPV6)
866 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
867 {
868 	int ret = -EINVAL;
869 
870 	spin_lock_bh(&ipvlan->addrs_lock);
871 	if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
872 		netif_err(ipvlan, ifup, ipvlan->dev,
873 			  "Failed to add IPv6=%pI6c addr for %s intf\n",
874 			  ip6_addr, ipvlan->dev->name);
875 	else
876 		ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
877 	spin_unlock_bh(&ipvlan->addrs_lock);
878 	return ret;
879 }
880 
881 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
882 {
883 	return ipvlan_del_addr(ipvlan, ip6_addr, true);
884 }
885 
886 static int ipvlan_addr6_event(struct notifier_block *unused,
887 			      unsigned long event, void *ptr)
888 {
889 	struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
890 	struct net_device *dev = (struct net_device *)if6->idev->dev;
891 	struct ipvl_dev *ipvlan = netdev_priv(dev);
892 
893 	if (!ipvlan_is_valid_dev(dev))
894 		return NOTIFY_DONE;
895 
896 	switch (event) {
897 	case NETDEV_UP:
898 		if (ipvlan_add_addr6(ipvlan, &if6->addr))
899 			return NOTIFY_BAD;
900 		break;
901 
902 	case NETDEV_DOWN:
903 		ipvlan_del_addr6(ipvlan, &if6->addr);
904 		break;
905 	}
906 
907 	return NOTIFY_OK;
908 }
909 
910 static int ipvlan_addr6_validator_event(struct notifier_block *unused,
911 					unsigned long event, void *ptr)
912 {
913 	struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
914 	struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
915 	struct ipvl_dev *ipvlan = netdev_priv(dev);
916 
917 	if (!ipvlan_is_valid_dev(dev))
918 		return NOTIFY_DONE;
919 
920 	switch (event) {
921 	case NETDEV_UP:
922 		if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
923 			NL_SET_ERR_MSG(i6vi->extack,
924 				       "Address already assigned to an ipvlan device");
925 			return notifier_from_errno(-EADDRINUSE);
926 		}
927 		break;
928 	}
929 
930 	return NOTIFY_OK;
931 }
932 #endif
933 
934 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
935 {
936 	int ret = -EINVAL;
937 
938 	spin_lock_bh(&ipvlan->addrs_lock);
939 	if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
940 		netif_err(ipvlan, ifup, ipvlan->dev,
941 			  "Failed to add IPv4=%pI4 on %s intf.\n",
942 			  ip4_addr, ipvlan->dev->name);
943 	else
944 		ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
945 	spin_unlock_bh(&ipvlan->addrs_lock);
946 	return ret;
947 }
948 
949 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
950 {
951 	return ipvlan_del_addr(ipvlan, ip4_addr, false);
952 }
953 
954 static int ipvlan_addr4_event(struct notifier_block *unused,
955 			      unsigned long event, void *ptr)
956 {
957 	struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
958 	struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
959 	struct ipvl_dev *ipvlan = netdev_priv(dev);
960 	struct in_addr ip4_addr;
961 
962 	if (!ipvlan_is_valid_dev(dev))
963 		return NOTIFY_DONE;
964 
965 	switch (event) {
966 	case NETDEV_UP:
967 		ip4_addr.s_addr = if4->ifa_address;
968 		if (ipvlan_add_addr4(ipvlan, &ip4_addr))
969 			return NOTIFY_BAD;
970 		break;
971 
972 	case NETDEV_DOWN:
973 		ip4_addr.s_addr = if4->ifa_address;
974 		ipvlan_del_addr4(ipvlan, &ip4_addr);
975 		break;
976 	}
977 
978 	return NOTIFY_OK;
979 }
980 
981 static int ipvlan_addr4_validator_event(struct notifier_block *unused,
982 					unsigned long event, void *ptr)
983 {
984 	struct in_validator_info *ivi = (struct in_validator_info *)ptr;
985 	struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
986 	struct ipvl_dev *ipvlan = netdev_priv(dev);
987 
988 	if (!ipvlan_is_valid_dev(dev))
989 		return NOTIFY_DONE;
990 
991 	switch (event) {
992 	case NETDEV_UP:
993 		if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
994 			NL_SET_ERR_MSG(ivi->extack,
995 				       "Address already assigned to an ipvlan device");
996 			return notifier_from_errno(-EADDRINUSE);
997 		}
998 		break;
999 	}
1000 
1001 	return NOTIFY_OK;
1002 }
1003 
1004 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
1005 	.notifier_call = ipvlan_addr4_event,
1006 };
1007 
1008 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
1009 	.notifier_call = ipvlan_addr4_validator_event,
1010 };
1011 
1012 static struct notifier_block ipvlan_notifier_block __read_mostly = {
1013 	.notifier_call = ipvlan_device_event,
1014 };
1015 
1016 #if IS_ENABLED(CONFIG_IPV6)
1017 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1018 	.notifier_call = ipvlan_addr6_event,
1019 };
1020 
1021 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1022 	.notifier_call = ipvlan_addr6_validator_event,
1023 };
1024 #endif
1025 
1026 static int __init ipvlan_init_module(void)
1027 {
1028 	int err;
1029 
1030 	ipvlan_init_secret();
1031 	register_netdevice_notifier(&ipvlan_notifier_block);
1032 #if IS_ENABLED(CONFIG_IPV6)
1033 	register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1034 	register_inet6addr_validator_notifier(
1035 	    &ipvlan_addr6_vtor_notifier_block);
1036 #endif
1037 	register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1038 	register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1039 
1040 	err = ipvlan_l3s_init();
1041 	if (err < 0)
1042 		goto error;
1043 
1044 	err = ipvlan_link_register(&ipvlan_link_ops);
1045 	if (err < 0) {
1046 		ipvlan_l3s_cleanup();
1047 		goto error;
1048 	}
1049 
1050 	return 0;
1051 error:
1052 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1053 	unregister_inetaddr_validator_notifier(
1054 	    &ipvlan_addr4_vtor_notifier_block);
1055 #if IS_ENABLED(CONFIG_IPV6)
1056 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1057 	unregister_inet6addr_validator_notifier(
1058 	    &ipvlan_addr6_vtor_notifier_block);
1059 #endif
1060 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1061 	return err;
1062 }
1063 
1064 static void __exit ipvlan_cleanup_module(void)
1065 {
1066 	rtnl_link_unregister(&ipvlan_link_ops);
1067 	ipvlan_l3s_cleanup();
1068 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1069 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1070 	unregister_inetaddr_validator_notifier(
1071 	    &ipvlan_addr4_vtor_notifier_block);
1072 #if IS_ENABLED(CONFIG_IPV6)
1073 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1074 	unregister_inet6addr_validator_notifier(
1075 	    &ipvlan_addr6_vtor_notifier_block);
1076 #endif
1077 }
1078 
1079 module_init(ipvlan_init_module);
1080 module_exit(ipvlan_cleanup_module);
1081 
1082 MODULE_LICENSE("GPL");
1083 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1084 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1085 MODULE_ALIAS_RTNL_LINK("ipvlan");
1086