xref: /openbmc/linux/drivers/net/ipvlan/ipvlan_main.c (revision 15e3ae36)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
3  */
4 
5 #include "ipvlan.h"
6 
7 static int ipvlan_set_port_mode(struct ipvl_port *port, u16 nval,
8 				struct netlink_ext_ack *extack)
9 {
10 	struct ipvl_dev *ipvlan;
11 	unsigned int flags;
12 	int err;
13 
14 	ASSERT_RTNL();
15 	if (port->mode != nval) {
16 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
17 			flags = ipvlan->dev->flags;
18 			if (nval == IPVLAN_MODE_L3 || nval == IPVLAN_MODE_L3S) {
19 				err = dev_change_flags(ipvlan->dev,
20 						       flags | IFF_NOARP,
21 						       extack);
22 			} else {
23 				err = dev_change_flags(ipvlan->dev,
24 						       flags & ~IFF_NOARP,
25 						       extack);
26 			}
27 			if (unlikely(err))
28 				goto fail;
29 		}
30 		if (nval == IPVLAN_MODE_L3S) {
31 			/* New mode is L3S */
32 			err = ipvlan_l3s_register(port);
33 			if (err)
34 				goto fail;
35 		} else if (port->mode == IPVLAN_MODE_L3S) {
36 			/* Old mode was L3S */
37 			ipvlan_l3s_unregister(port);
38 		}
39 		port->mode = nval;
40 	}
41 	return 0;
42 
43 fail:
44 	/* Undo the flags changes that have been done so far. */
45 	list_for_each_entry_continue_reverse(ipvlan, &port->ipvlans, pnode) {
46 		flags = ipvlan->dev->flags;
47 		if (port->mode == IPVLAN_MODE_L3 ||
48 		    port->mode == IPVLAN_MODE_L3S)
49 			dev_change_flags(ipvlan->dev, flags | IFF_NOARP,
50 					 NULL);
51 		else
52 			dev_change_flags(ipvlan->dev, flags & ~IFF_NOARP,
53 					 NULL);
54 	}
55 
56 	return err;
57 }
58 
59 static int ipvlan_port_create(struct net_device *dev)
60 {
61 	struct ipvl_port *port;
62 	int err, idx;
63 
64 	port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL);
65 	if (!port)
66 		return -ENOMEM;
67 
68 	write_pnet(&port->pnet, dev_net(dev));
69 	port->dev = dev;
70 	port->mode = IPVLAN_MODE_L3;
71 	INIT_LIST_HEAD(&port->ipvlans);
72 	for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++)
73 		INIT_HLIST_HEAD(&port->hlhead[idx]);
74 
75 	skb_queue_head_init(&port->backlog);
76 	INIT_WORK(&port->wq, ipvlan_process_multicast);
77 	ida_init(&port->ida);
78 	port->dev_id_start = 1;
79 
80 	err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port);
81 	if (err)
82 		goto err;
83 
84 	return 0;
85 
86 err:
87 	kfree(port);
88 	return err;
89 }
90 
91 static void ipvlan_port_destroy(struct net_device *dev)
92 {
93 	struct ipvl_port *port = ipvlan_port_get_rtnl(dev);
94 	struct sk_buff *skb;
95 
96 	if (port->mode == IPVLAN_MODE_L3S)
97 		ipvlan_l3s_unregister(port);
98 	netdev_rx_handler_unregister(dev);
99 	cancel_work_sync(&port->wq);
100 	while ((skb = __skb_dequeue(&port->backlog)) != NULL) {
101 		if (skb->dev)
102 			dev_put(skb->dev);
103 		kfree_skb(skb);
104 	}
105 	ida_destroy(&port->ida);
106 	kfree(port);
107 }
108 
109 #define IPVLAN_FEATURES \
110 	(NETIF_F_SG | NETIF_F_CSUM_MASK | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
111 	 NETIF_F_GSO | NETIF_F_ALL_TSO | NETIF_F_GSO_ROBUST | \
112 	 NETIF_F_GRO | NETIF_F_RXCSUM | \
113 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
114 
115 #define IPVLAN_STATE_MASK \
116 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
117 
118 static int ipvlan_init(struct net_device *dev)
119 {
120 	struct ipvl_dev *ipvlan = netdev_priv(dev);
121 	struct net_device *phy_dev = ipvlan->phy_dev;
122 	struct ipvl_port *port;
123 	int err;
124 
125 	dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
126 		     (phy_dev->state & IPVLAN_STATE_MASK);
127 	dev->features = phy_dev->features & IPVLAN_FEATURES;
128 	dev->features |= NETIF_F_LLTX | NETIF_F_VLAN_CHALLENGED;
129 	dev->hw_enc_features |= dev->features;
130 	dev->gso_max_size = phy_dev->gso_max_size;
131 	dev->gso_max_segs = phy_dev->gso_max_segs;
132 	dev->hard_header_len = phy_dev->hard_header_len;
133 
134 	ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
135 	if (!ipvlan->pcpu_stats)
136 		return -ENOMEM;
137 
138 	if (!netif_is_ipvlan_port(phy_dev)) {
139 		err = ipvlan_port_create(phy_dev);
140 		if (err < 0) {
141 			free_percpu(ipvlan->pcpu_stats);
142 			return err;
143 		}
144 	}
145 	port = ipvlan_port_get_rtnl(phy_dev);
146 	port->count += 1;
147 	return 0;
148 }
149 
150 static void ipvlan_uninit(struct net_device *dev)
151 {
152 	struct ipvl_dev *ipvlan = netdev_priv(dev);
153 	struct net_device *phy_dev = ipvlan->phy_dev;
154 	struct ipvl_port *port;
155 
156 	free_percpu(ipvlan->pcpu_stats);
157 
158 	port = ipvlan_port_get_rtnl(phy_dev);
159 	port->count -= 1;
160 	if (!port->count)
161 		ipvlan_port_destroy(port->dev);
162 }
163 
164 static int ipvlan_open(struct net_device *dev)
165 {
166 	struct ipvl_dev *ipvlan = netdev_priv(dev);
167 	struct ipvl_addr *addr;
168 
169 	if (ipvlan->port->mode == IPVLAN_MODE_L3 ||
170 	    ipvlan->port->mode == IPVLAN_MODE_L3S)
171 		dev->flags |= IFF_NOARP;
172 	else
173 		dev->flags &= ~IFF_NOARP;
174 
175 	rcu_read_lock();
176 	list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
177 		ipvlan_ht_addr_add(ipvlan, addr);
178 	rcu_read_unlock();
179 
180 	return 0;
181 }
182 
183 static int ipvlan_stop(struct net_device *dev)
184 {
185 	struct ipvl_dev *ipvlan = netdev_priv(dev);
186 	struct net_device *phy_dev = ipvlan->phy_dev;
187 	struct ipvl_addr *addr;
188 
189 	dev_uc_unsync(phy_dev, dev);
190 	dev_mc_unsync(phy_dev, dev);
191 
192 	rcu_read_lock();
193 	list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
194 		ipvlan_ht_addr_del(addr);
195 	rcu_read_unlock();
196 
197 	return 0;
198 }
199 
200 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
201 				     struct net_device *dev)
202 {
203 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
204 	int skblen = skb->len;
205 	int ret;
206 
207 	ret = ipvlan_queue_xmit(skb, dev);
208 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
209 		struct ipvl_pcpu_stats *pcptr;
210 
211 		pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
212 
213 		u64_stats_update_begin(&pcptr->syncp);
214 		pcptr->tx_pkts++;
215 		pcptr->tx_bytes += skblen;
216 		u64_stats_update_end(&pcptr->syncp);
217 	} else {
218 		this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
219 	}
220 	return ret;
221 }
222 
223 static netdev_features_t ipvlan_fix_features(struct net_device *dev,
224 					     netdev_features_t features)
225 {
226 	struct ipvl_dev *ipvlan = netdev_priv(dev);
227 
228 	return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES);
229 }
230 
231 static void ipvlan_change_rx_flags(struct net_device *dev, int change)
232 {
233 	struct ipvl_dev *ipvlan = netdev_priv(dev);
234 	struct net_device *phy_dev = ipvlan->phy_dev;
235 
236 	if (change & IFF_ALLMULTI)
237 		dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
238 }
239 
240 static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
241 {
242 	struct ipvl_dev *ipvlan = netdev_priv(dev);
243 
244 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
245 		bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
246 	} else {
247 		struct netdev_hw_addr *ha;
248 		DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
249 
250 		bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
251 		netdev_for_each_mc_addr(ha, dev)
252 			__set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
253 
254 		/* Turn-on broadcast bit irrespective of address family,
255 		 * since broadcast is deferred to a work-queue, hence no
256 		 * impact on fast-path processing.
257 		 */
258 		__set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
259 
260 		bitmap_copy(ipvlan->mac_filters, mc_filters,
261 			    IPVLAN_MAC_FILTER_SIZE);
262 	}
263 	dev_uc_sync(ipvlan->phy_dev, dev);
264 	dev_mc_sync(ipvlan->phy_dev, dev);
265 }
266 
267 static void ipvlan_get_stats64(struct net_device *dev,
268 			       struct rtnl_link_stats64 *s)
269 {
270 	struct ipvl_dev *ipvlan = netdev_priv(dev);
271 
272 	if (ipvlan->pcpu_stats) {
273 		struct ipvl_pcpu_stats *pcptr;
274 		u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
275 		u32 rx_errs = 0, tx_drps = 0;
276 		u32 strt;
277 		int idx;
278 
279 		for_each_possible_cpu(idx) {
280 			pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
281 			do {
282 				strt= u64_stats_fetch_begin_irq(&pcptr->syncp);
283 				rx_pkts = pcptr->rx_pkts;
284 				rx_bytes = pcptr->rx_bytes;
285 				rx_mcast = pcptr->rx_mcast;
286 				tx_pkts = pcptr->tx_pkts;
287 				tx_bytes = pcptr->tx_bytes;
288 			} while (u64_stats_fetch_retry_irq(&pcptr->syncp,
289 							   strt));
290 
291 			s->rx_packets += rx_pkts;
292 			s->rx_bytes += rx_bytes;
293 			s->multicast += rx_mcast;
294 			s->tx_packets += tx_pkts;
295 			s->tx_bytes += tx_bytes;
296 
297 			/* u32 values are updated without syncp protection. */
298 			rx_errs += pcptr->rx_errs;
299 			tx_drps += pcptr->tx_drps;
300 		}
301 		s->rx_errors = rx_errs;
302 		s->rx_dropped = rx_errs;
303 		s->tx_dropped = tx_drps;
304 	}
305 }
306 
307 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
308 {
309 	struct ipvl_dev *ipvlan = netdev_priv(dev);
310 	struct net_device *phy_dev = ipvlan->phy_dev;
311 
312 	return vlan_vid_add(phy_dev, proto, vid);
313 }
314 
315 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
316 				   u16 vid)
317 {
318 	struct ipvl_dev *ipvlan = netdev_priv(dev);
319 	struct net_device *phy_dev = ipvlan->phy_dev;
320 
321 	vlan_vid_del(phy_dev, proto, vid);
322 	return 0;
323 }
324 
325 static int ipvlan_get_iflink(const struct net_device *dev)
326 {
327 	struct ipvl_dev *ipvlan = netdev_priv(dev);
328 
329 	return ipvlan->phy_dev->ifindex;
330 }
331 
332 static const struct net_device_ops ipvlan_netdev_ops = {
333 	.ndo_init		= ipvlan_init,
334 	.ndo_uninit		= ipvlan_uninit,
335 	.ndo_open		= ipvlan_open,
336 	.ndo_stop		= ipvlan_stop,
337 	.ndo_start_xmit		= ipvlan_start_xmit,
338 	.ndo_fix_features	= ipvlan_fix_features,
339 	.ndo_change_rx_flags	= ipvlan_change_rx_flags,
340 	.ndo_set_rx_mode	= ipvlan_set_multicast_mac_filter,
341 	.ndo_get_stats64	= ipvlan_get_stats64,
342 	.ndo_vlan_rx_add_vid	= ipvlan_vlan_rx_add_vid,
343 	.ndo_vlan_rx_kill_vid	= ipvlan_vlan_rx_kill_vid,
344 	.ndo_get_iflink		= ipvlan_get_iflink,
345 };
346 
347 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
348 			      unsigned short type, const void *daddr,
349 			      const void *saddr, unsigned len)
350 {
351 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
352 	struct net_device *phy_dev = ipvlan->phy_dev;
353 
354 	/* TODO Probably use a different field than dev_addr so that the
355 	 * mac-address on the virtual device is portable and can be carried
356 	 * while the packets use the mac-addr on the physical device.
357 	 */
358 	return dev_hard_header(skb, phy_dev, type, daddr,
359 			       saddr ? : phy_dev->dev_addr, len);
360 }
361 
362 static const struct header_ops ipvlan_header_ops = {
363 	.create  	= ipvlan_hard_header,
364 	.parse		= eth_header_parse,
365 	.cache		= eth_header_cache,
366 	.cache_update	= eth_header_cache_update,
367 };
368 
369 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
370 {
371 	ipvlan->dev->mtu = dev->mtu;
372 }
373 
374 static bool netif_is_ipvlan(const struct net_device *dev)
375 {
376 	/* both ipvlan and ipvtap devices use the same netdev_ops */
377 	return dev->netdev_ops == &ipvlan_netdev_ops;
378 }
379 
380 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
381 					     struct ethtool_link_ksettings *cmd)
382 {
383 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
384 
385 	return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
386 }
387 
388 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
389 				       struct ethtool_drvinfo *drvinfo)
390 {
391 	strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
392 	strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
393 }
394 
395 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
396 {
397 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
398 
399 	return ipvlan->msg_enable;
400 }
401 
402 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
403 {
404 	struct ipvl_dev *ipvlan = netdev_priv(dev);
405 
406 	ipvlan->msg_enable = value;
407 }
408 
409 static const struct ethtool_ops ipvlan_ethtool_ops = {
410 	.get_link	= ethtool_op_get_link,
411 	.get_link_ksettings	= ipvlan_ethtool_get_link_ksettings,
412 	.get_drvinfo	= ipvlan_ethtool_get_drvinfo,
413 	.get_msglevel	= ipvlan_ethtool_get_msglevel,
414 	.set_msglevel	= ipvlan_ethtool_set_msglevel,
415 };
416 
417 static int ipvlan_nl_changelink(struct net_device *dev,
418 				struct nlattr *tb[], struct nlattr *data[],
419 				struct netlink_ext_ack *extack)
420 {
421 	struct ipvl_dev *ipvlan = netdev_priv(dev);
422 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
423 	int err = 0;
424 
425 	if (!data)
426 		return 0;
427 	if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
428 		return -EPERM;
429 
430 	if (data[IFLA_IPVLAN_MODE]) {
431 		u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
432 
433 		err = ipvlan_set_port_mode(port, nmode, extack);
434 	}
435 
436 	if (!err && data[IFLA_IPVLAN_FLAGS]) {
437 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
438 
439 		if (flags & IPVLAN_F_PRIVATE)
440 			ipvlan_mark_private(port);
441 		else
442 			ipvlan_clear_private(port);
443 
444 		if (flags & IPVLAN_F_VEPA)
445 			ipvlan_mark_vepa(port);
446 		else
447 			ipvlan_clear_vepa(port);
448 	}
449 
450 	return err;
451 }
452 
453 static size_t ipvlan_nl_getsize(const struct net_device *dev)
454 {
455 	return (0
456 		+ nla_total_size(2) /* IFLA_IPVLAN_MODE */
457 		+ nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
458 		);
459 }
460 
461 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
462 			      struct netlink_ext_ack *extack)
463 {
464 	if (!data)
465 		return 0;
466 
467 	if (data[IFLA_IPVLAN_MODE]) {
468 		u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
469 
470 		if (mode >= IPVLAN_MODE_MAX)
471 			return -EINVAL;
472 	}
473 	if (data[IFLA_IPVLAN_FLAGS]) {
474 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
475 
476 		/* Only two bits are used at this moment. */
477 		if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
478 			return -EINVAL;
479 		/* Also both flags can't be active at the same time. */
480 		if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
481 		    (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
482 			return -EINVAL;
483 	}
484 
485 	return 0;
486 }
487 
488 static int ipvlan_nl_fillinfo(struct sk_buff *skb,
489 			      const struct net_device *dev)
490 {
491 	struct ipvl_dev *ipvlan = netdev_priv(dev);
492 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
493 	int ret = -EINVAL;
494 
495 	if (!port)
496 		goto err;
497 
498 	ret = -EMSGSIZE;
499 	if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
500 		goto err;
501 	if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
502 		goto err;
503 
504 	return 0;
505 
506 err:
507 	return ret;
508 }
509 
510 int ipvlan_link_new(struct net *src_net, struct net_device *dev,
511 		    struct nlattr *tb[], struct nlattr *data[],
512 		    struct netlink_ext_ack *extack)
513 {
514 	struct ipvl_dev *ipvlan = netdev_priv(dev);
515 	struct ipvl_port *port;
516 	struct net_device *phy_dev;
517 	int err;
518 	u16 mode = IPVLAN_MODE_L3;
519 
520 	if (!tb[IFLA_LINK])
521 		return -EINVAL;
522 
523 	phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
524 	if (!phy_dev)
525 		return -ENODEV;
526 
527 	if (netif_is_ipvlan(phy_dev)) {
528 		struct ipvl_dev *tmp = netdev_priv(phy_dev);
529 
530 		phy_dev = tmp->phy_dev;
531 		if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
532 			return -EPERM;
533 	} else if (!netif_is_ipvlan_port(phy_dev)) {
534 		/* Exit early if the underlying link is invalid or busy */
535 		if (phy_dev->type != ARPHRD_ETHER ||
536 		    phy_dev->flags & IFF_LOOPBACK) {
537 			netdev_err(phy_dev,
538 				   "Master is either lo or non-ether device\n");
539 			return -EINVAL;
540 		}
541 
542 		if (netdev_is_rx_handler_busy(phy_dev)) {
543 			netdev_err(phy_dev, "Device is already in use.\n");
544 			return -EBUSY;
545 		}
546 	}
547 
548 	ipvlan->phy_dev = phy_dev;
549 	ipvlan->dev = dev;
550 	ipvlan->sfeatures = IPVLAN_FEATURES;
551 	if (!tb[IFLA_MTU])
552 		ipvlan_adjust_mtu(ipvlan, phy_dev);
553 	INIT_LIST_HEAD(&ipvlan->addrs);
554 	spin_lock_init(&ipvlan->addrs_lock);
555 
556 	/* TODO Probably put random address here to be presented to the
557 	 * world but keep using the physical-dev address for the outgoing
558 	 * packets.
559 	 */
560 	memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN);
561 
562 	dev->priv_flags |= IFF_NO_RX_HANDLER;
563 
564 	err = register_netdevice(dev);
565 	if (err < 0)
566 		return err;
567 
568 	/* ipvlan_init() would have created the port, if required */
569 	port = ipvlan_port_get_rtnl(phy_dev);
570 	ipvlan->port = port;
571 
572 	/* If the port-id base is at the MAX value, then wrap it around and
573 	 * begin from 0x1 again. This may be due to a busy system where lots
574 	 * of slaves are getting created and deleted.
575 	 */
576 	if (port->dev_id_start == 0xFFFE)
577 		port->dev_id_start = 0x1;
578 
579 	/* Since L2 address is shared among all IPvlan slaves including
580 	 * master, use unique 16 bit dev-ids to diffentiate among them.
581 	 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
582 	 * slave link [see addrconf_ifid_eui48()].
583 	 */
584 	err = ida_simple_get(&port->ida, port->dev_id_start, 0xFFFE,
585 			     GFP_KERNEL);
586 	if (err < 0)
587 		err = ida_simple_get(&port->ida, 0x1, port->dev_id_start,
588 				     GFP_KERNEL);
589 	if (err < 0)
590 		goto unregister_netdev;
591 	dev->dev_id = err;
592 
593 	/* Increment id-base to the next slot for the future assignment */
594 	port->dev_id_start = err + 1;
595 
596 	err = netdev_upper_dev_link(phy_dev, dev, extack);
597 	if (err)
598 		goto remove_ida;
599 
600 	/* Flags are per port and latest update overrides. User has
601 	 * to be consistent in setting it just like the mode attribute.
602 	 */
603 	if (data && data[IFLA_IPVLAN_FLAGS])
604 		port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
605 
606 	if (data && data[IFLA_IPVLAN_MODE])
607 		mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
608 
609 	err = ipvlan_set_port_mode(port, mode, extack);
610 	if (err)
611 		goto unlink_netdev;
612 
613 	list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
614 	netif_stacked_transfer_operstate(phy_dev, dev);
615 	return 0;
616 
617 unlink_netdev:
618 	netdev_upper_dev_unlink(phy_dev, dev);
619 remove_ida:
620 	ida_simple_remove(&port->ida, dev->dev_id);
621 unregister_netdev:
622 	unregister_netdevice(dev);
623 	return err;
624 }
625 EXPORT_SYMBOL_GPL(ipvlan_link_new);
626 
627 void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
628 {
629 	struct ipvl_dev *ipvlan = netdev_priv(dev);
630 	struct ipvl_addr *addr, *next;
631 
632 	spin_lock_bh(&ipvlan->addrs_lock);
633 	list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
634 		ipvlan_ht_addr_del(addr);
635 		list_del_rcu(&addr->anode);
636 		kfree_rcu(addr, rcu);
637 	}
638 	spin_unlock_bh(&ipvlan->addrs_lock);
639 
640 	ida_simple_remove(&ipvlan->port->ida, dev->dev_id);
641 	list_del_rcu(&ipvlan->pnode);
642 	unregister_netdevice_queue(dev, head);
643 	netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
644 }
645 EXPORT_SYMBOL_GPL(ipvlan_link_delete);
646 
647 void ipvlan_link_setup(struct net_device *dev)
648 {
649 	ether_setup(dev);
650 
651 	dev->max_mtu = ETH_MAX_MTU;
652 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
653 	dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
654 	dev->netdev_ops = &ipvlan_netdev_ops;
655 	dev->needs_free_netdev = true;
656 	dev->header_ops = &ipvlan_header_ops;
657 	dev->ethtool_ops = &ipvlan_ethtool_ops;
658 }
659 EXPORT_SYMBOL_GPL(ipvlan_link_setup);
660 
661 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
662 {
663 	[IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
664 	[IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
665 };
666 
667 static struct rtnl_link_ops ipvlan_link_ops = {
668 	.kind		= "ipvlan",
669 	.priv_size	= sizeof(struct ipvl_dev),
670 
671 	.setup		= ipvlan_link_setup,
672 	.newlink	= ipvlan_link_new,
673 	.dellink	= ipvlan_link_delete,
674 };
675 
676 int ipvlan_link_register(struct rtnl_link_ops *ops)
677 {
678 	ops->get_size	= ipvlan_nl_getsize;
679 	ops->policy	= ipvlan_nl_policy;
680 	ops->validate	= ipvlan_nl_validate;
681 	ops->fill_info	= ipvlan_nl_fillinfo;
682 	ops->changelink = ipvlan_nl_changelink;
683 	ops->maxtype	= IFLA_IPVLAN_MAX;
684 	return rtnl_link_register(ops);
685 }
686 EXPORT_SYMBOL_GPL(ipvlan_link_register);
687 
688 static int ipvlan_device_event(struct notifier_block *unused,
689 			       unsigned long event, void *ptr)
690 {
691 	struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
692 	struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
693 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
694 	struct ipvl_dev *ipvlan, *next;
695 	struct ipvl_port *port;
696 	LIST_HEAD(lst_kill);
697 	int err;
698 
699 	if (!netif_is_ipvlan_port(dev))
700 		return NOTIFY_DONE;
701 
702 	port = ipvlan_port_get_rtnl(dev);
703 
704 	switch (event) {
705 	case NETDEV_CHANGE:
706 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
707 			netif_stacked_transfer_operstate(ipvlan->phy_dev,
708 							 ipvlan->dev);
709 		break;
710 
711 	case NETDEV_REGISTER: {
712 		struct net *oldnet, *newnet = dev_net(dev);
713 
714 		oldnet = read_pnet(&port->pnet);
715 		if (net_eq(newnet, oldnet))
716 			break;
717 
718 		write_pnet(&port->pnet, newnet);
719 
720 		ipvlan_migrate_l3s_hook(oldnet, newnet);
721 		break;
722 	}
723 	case NETDEV_UNREGISTER:
724 		if (dev->reg_state != NETREG_UNREGISTERING)
725 			break;
726 
727 		list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
728 			ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
729 							    &lst_kill);
730 		unregister_netdevice_many(&lst_kill);
731 		break;
732 
733 	case NETDEV_FEAT_CHANGE:
734 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
735 			ipvlan->dev->features = dev->features & IPVLAN_FEATURES;
736 			ipvlan->dev->gso_max_size = dev->gso_max_size;
737 			ipvlan->dev->gso_max_segs = dev->gso_max_segs;
738 			netdev_features_change(ipvlan->dev);
739 		}
740 		break;
741 
742 	case NETDEV_CHANGEMTU:
743 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
744 			ipvlan_adjust_mtu(ipvlan, dev);
745 		break;
746 
747 	case NETDEV_PRE_CHANGEADDR:
748 		prechaddr_info = ptr;
749 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
750 			err = dev_pre_changeaddr_notify(ipvlan->dev,
751 						    prechaddr_info->dev_addr,
752 						    extack);
753 			if (err)
754 				return notifier_from_errno(err);
755 		}
756 		break;
757 
758 	case NETDEV_CHANGEADDR:
759 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
760 			ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr);
761 			call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
762 		}
763 		break;
764 
765 	case NETDEV_PRE_TYPE_CHANGE:
766 		/* Forbid underlying device to change its type. */
767 		return NOTIFY_BAD;
768 	}
769 	return NOTIFY_DONE;
770 }
771 
772 /* the caller must held the addrs lock */
773 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
774 {
775 	struct ipvl_addr *addr;
776 
777 	addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
778 	if (!addr)
779 		return -ENOMEM;
780 
781 	addr->master = ipvlan;
782 	if (!is_v6) {
783 		memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
784 		addr->atype = IPVL_IPV4;
785 #if IS_ENABLED(CONFIG_IPV6)
786 	} else {
787 		memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
788 		addr->atype = IPVL_IPV6;
789 #endif
790 	}
791 
792 	list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
793 
794 	/* If the interface is not up, the address will be added to the hash
795 	 * list by ipvlan_open.
796 	 */
797 	if (netif_running(ipvlan->dev))
798 		ipvlan_ht_addr_add(ipvlan, addr);
799 
800 	return 0;
801 }
802 
803 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
804 {
805 	struct ipvl_addr *addr;
806 
807 	spin_lock_bh(&ipvlan->addrs_lock);
808 	addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
809 	if (!addr) {
810 		spin_unlock_bh(&ipvlan->addrs_lock);
811 		return;
812 	}
813 
814 	ipvlan_ht_addr_del(addr);
815 	list_del_rcu(&addr->anode);
816 	spin_unlock_bh(&ipvlan->addrs_lock);
817 	kfree_rcu(addr, rcu);
818 }
819 
820 static bool ipvlan_is_valid_dev(const struct net_device *dev)
821 {
822 	struct ipvl_dev *ipvlan = netdev_priv(dev);
823 
824 	if (!netif_is_ipvlan(dev))
825 		return false;
826 
827 	if (!ipvlan || !ipvlan->port)
828 		return false;
829 
830 	return true;
831 }
832 
833 #if IS_ENABLED(CONFIG_IPV6)
834 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
835 {
836 	int ret = -EINVAL;
837 
838 	spin_lock_bh(&ipvlan->addrs_lock);
839 	if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
840 		netif_err(ipvlan, ifup, ipvlan->dev,
841 			  "Failed to add IPv6=%pI6c addr for %s intf\n",
842 			  ip6_addr, ipvlan->dev->name);
843 	else
844 		ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
845 	spin_unlock_bh(&ipvlan->addrs_lock);
846 	return ret;
847 }
848 
849 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
850 {
851 	return ipvlan_del_addr(ipvlan, ip6_addr, true);
852 }
853 
854 static int ipvlan_addr6_event(struct notifier_block *unused,
855 			      unsigned long event, void *ptr)
856 {
857 	struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
858 	struct net_device *dev = (struct net_device *)if6->idev->dev;
859 	struct ipvl_dev *ipvlan = netdev_priv(dev);
860 
861 	if (!ipvlan_is_valid_dev(dev))
862 		return NOTIFY_DONE;
863 
864 	switch (event) {
865 	case NETDEV_UP:
866 		if (ipvlan_add_addr6(ipvlan, &if6->addr))
867 			return NOTIFY_BAD;
868 		break;
869 
870 	case NETDEV_DOWN:
871 		ipvlan_del_addr6(ipvlan, &if6->addr);
872 		break;
873 	}
874 
875 	return NOTIFY_OK;
876 }
877 
878 static int ipvlan_addr6_validator_event(struct notifier_block *unused,
879 					unsigned long event, void *ptr)
880 {
881 	struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
882 	struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
883 	struct ipvl_dev *ipvlan = netdev_priv(dev);
884 
885 	if (!ipvlan_is_valid_dev(dev))
886 		return NOTIFY_DONE;
887 
888 	switch (event) {
889 	case NETDEV_UP:
890 		if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
891 			NL_SET_ERR_MSG(i6vi->extack,
892 				       "Address already assigned to an ipvlan device");
893 			return notifier_from_errno(-EADDRINUSE);
894 		}
895 		break;
896 	}
897 
898 	return NOTIFY_OK;
899 }
900 #endif
901 
902 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
903 {
904 	int ret = -EINVAL;
905 
906 	spin_lock_bh(&ipvlan->addrs_lock);
907 	if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
908 		netif_err(ipvlan, ifup, ipvlan->dev,
909 			  "Failed to add IPv4=%pI4 on %s intf.\n",
910 			  ip4_addr, ipvlan->dev->name);
911 	else
912 		ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
913 	spin_unlock_bh(&ipvlan->addrs_lock);
914 	return ret;
915 }
916 
917 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
918 {
919 	return ipvlan_del_addr(ipvlan, ip4_addr, false);
920 }
921 
922 static int ipvlan_addr4_event(struct notifier_block *unused,
923 			      unsigned long event, void *ptr)
924 {
925 	struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
926 	struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
927 	struct ipvl_dev *ipvlan = netdev_priv(dev);
928 	struct in_addr ip4_addr;
929 
930 	if (!ipvlan_is_valid_dev(dev))
931 		return NOTIFY_DONE;
932 
933 	switch (event) {
934 	case NETDEV_UP:
935 		ip4_addr.s_addr = if4->ifa_address;
936 		if (ipvlan_add_addr4(ipvlan, &ip4_addr))
937 			return NOTIFY_BAD;
938 		break;
939 
940 	case NETDEV_DOWN:
941 		ip4_addr.s_addr = if4->ifa_address;
942 		ipvlan_del_addr4(ipvlan, &ip4_addr);
943 		break;
944 	}
945 
946 	return NOTIFY_OK;
947 }
948 
949 static int ipvlan_addr4_validator_event(struct notifier_block *unused,
950 					unsigned long event, void *ptr)
951 {
952 	struct in_validator_info *ivi = (struct in_validator_info *)ptr;
953 	struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
954 	struct ipvl_dev *ipvlan = netdev_priv(dev);
955 
956 	if (!ipvlan_is_valid_dev(dev))
957 		return NOTIFY_DONE;
958 
959 	switch (event) {
960 	case NETDEV_UP:
961 		if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
962 			NL_SET_ERR_MSG(ivi->extack,
963 				       "Address already assigned to an ipvlan device");
964 			return notifier_from_errno(-EADDRINUSE);
965 		}
966 		break;
967 	}
968 
969 	return NOTIFY_OK;
970 }
971 
972 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
973 	.notifier_call = ipvlan_addr4_event,
974 };
975 
976 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
977 	.notifier_call = ipvlan_addr4_validator_event,
978 };
979 
980 static struct notifier_block ipvlan_notifier_block __read_mostly = {
981 	.notifier_call = ipvlan_device_event,
982 };
983 
984 #if IS_ENABLED(CONFIG_IPV6)
985 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
986 	.notifier_call = ipvlan_addr6_event,
987 };
988 
989 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
990 	.notifier_call = ipvlan_addr6_validator_event,
991 };
992 #endif
993 
994 static int __init ipvlan_init_module(void)
995 {
996 	int err;
997 
998 	ipvlan_init_secret();
999 	register_netdevice_notifier(&ipvlan_notifier_block);
1000 #if IS_ENABLED(CONFIG_IPV6)
1001 	register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1002 	register_inet6addr_validator_notifier(
1003 	    &ipvlan_addr6_vtor_notifier_block);
1004 #endif
1005 	register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1006 	register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1007 
1008 	err = ipvlan_l3s_init();
1009 	if (err < 0)
1010 		goto error;
1011 
1012 	err = ipvlan_link_register(&ipvlan_link_ops);
1013 	if (err < 0) {
1014 		ipvlan_l3s_cleanup();
1015 		goto error;
1016 	}
1017 
1018 	return 0;
1019 error:
1020 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1021 	unregister_inetaddr_validator_notifier(
1022 	    &ipvlan_addr4_vtor_notifier_block);
1023 #if IS_ENABLED(CONFIG_IPV6)
1024 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1025 	unregister_inet6addr_validator_notifier(
1026 	    &ipvlan_addr6_vtor_notifier_block);
1027 #endif
1028 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1029 	return err;
1030 }
1031 
1032 static void __exit ipvlan_cleanup_module(void)
1033 {
1034 	rtnl_link_unregister(&ipvlan_link_ops);
1035 	ipvlan_l3s_cleanup();
1036 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1037 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1038 	unregister_inetaddr_validator_notifier(
1039 	    &ipvlan_addr4_vtor_notifier_block);
1040 #if IS_ENABLED(CONFIG_IPV6)
1041 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1042 	unregister_inet6addr_validator_notifier(
1043 	    &ipvlan_addr6_vtor_notifier_block);
1044 #endif
1045 }
1046 
1047 module_init(ipvlan_init_module);
1048 module_exit(ipvlan_cleanup_module);
1049 
1050 MODULE_LICENSE("GPL");
1051 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1052 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1053 MODULE_ALIAS_RTNL_LINK("ipvlan");
1054