xref: /openbmc/linux/drivers/net/ipvlan/ipvlan_main.c (revision 86e281fc)
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 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 };
391 
392 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
393 {
394 	ipvlan->dev->mtu = dev->mtu;
395 }
396 
397 static bool netif_is_ipvlan(const struct net_device *dev)
398 {
399 	/* both ipvlan and ipvtap devices use the same netdev_ops */
400 	return dev->netdev_ops == &ipvlan_netdev_ops;
401 }
402 
403 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
404 					     struct ethtool_link_ksettings *cmd)
405 {
406 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
407 
408 	return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
409 }
410 
411 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
412 				       struct ethtool_drvinfo *drvinfo)
413 {
414 	strscpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
415 	strscpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
416 }
417 
418 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
419 {
420 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
421 
422 	return ipvlan->msg_enable;
423 }
424 
425 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
426 {
427 	struct ipvl_dev *ipvlan = netdev_priv(dev);
428 
429 	ipvlan->msg_enable = value;
430 }
431 
432 static const struct ethtool_ops ipvlan_ethtool_ops = {
433 	.get_link	= ethtool_op_get_link,
434 	.get_link_ksettings	= ipvlan_ethtool_get_link_ksettings,
435 	.get_drvinfo	= ipvlan_ethtool_get_drvinfo,
436 	.get_msglevel	= ipvlan_ethtool_get_msglevel,
437 	.set_msglevel	= ipvlan_ethtool_set_msglevel,
438 };
439 
440 static int ipvlan_nl_changelink(struct net_device *dev,
441 				struct nlattr *tb[], struct nlattr *data[],
442 				struct netlink_ext_ack *extack)
443 {
444 	struct ipvl_dev *ipvlan = netdev_priv(dev);
445 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
446 	int err = 0;
447 
448 	if (!data)
449 		return 0;
450 	if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
451 		return -EPERM;
452 
453 	if (data[IFLA_IPVLAN_MODE]) {
454 		u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
455 
456 		err = ipvlan_set_port_mode(port, nmode, extack);
457 	}
458 
459 	if (!err && data[IFLA_IPVLAN_FLAGS]) {
460 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
461 
462 		if (flags & IPVLAN_F_PRIVATE)
463 			ipvlan_mark_private(port);
464 		else
465 			ipvlan_clear_private(port);
466 
467 		if (flags & IPVLAN_F_VEPA)
468 			ipvlan_mark_vepa(port);
469 		else
470 			ipvlan_clear_vepa(port);
471 	}
472 
473 	return err;
474 }
475 
476 static size_t ipvlan_nl_getsize(const struct net_device *dev)
477 {
478 	return (0
479 		+ nla_total_size(2) /* IFLA_IPVLAN_MODE */
480 		+ nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
481 		);
482 }
483 
484 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
485 			      struct netlink_ext_ack *extack)
486 {
487 	if (!data)
488 		return 0;
489 
490 	if (data[IFLA_IPVLAN_MODE]) {
491 		u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
492 
493 		if (mode >= IPVLAN_MODE_MAX)
494 			return -EINVAL;
495 	}
496 	if (data[IFLA_IPVLAN_FLAGS]) {
497 		u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
498 
499 		/* Only two bits are used at this moment. */
500 		if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
501 			return -EINVAL;
502 		/* Also both flags can't be active at the same time. */
503 		if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
504 		    (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
505 			return -EINVAL;
506 	}
507 
508 	return 0;
509 }
510 
511 static int ipvlan_nl_fillinfo(struct sk_buff *skb,
512 			      const struct net_device *dev)
513 {
514 	struct ipvl_dev *ipvlan = netdev_priv(dev);
515 	struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
516 	int ret = -EINVAL;
517 
518 	if (!port)
519 		goto err;
520 
521 	ret = -EMSGSIZE;
522 	if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
523 		goto err;
524 	if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
525 		goto err;
526 
527 	return 0;
528 
529 err:
530 	return ret;
531 }
532 
533 int ipvlan_link_new(struct net *src_net, struct net_device *dev,
534 		    struct nlattr *tb[], struct nlattr *data[],
535 		    struct netlink_ext_ack *extack)
536 {
537 	struct ipvl_dev *ipvlan = netdev_priv(dev);
538 	struct ipvl_port *port;
539 	struct net_device *phy_dev;
540 	int err;
541 	u16 mode = IPVLAN_MODE_L3;
542 
543 	if (!tb[IFLA_LINK])
544 		return -EINVAL;
545 
546 	phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
547 	if (!phy_dev)
548 		return -ENODEV;
549 
550 	if (netif_is_ipvlan(phy_dev)) {
551 		struct ipvl_dev *tmp = netdev_priv(phy_dev);
552 
553 		phy_dev = tmp->phy_dev;
554 		if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
555 			return -EPERM;
556 	} else if (!netif_is_ipvlan_port(phy_dev)) {
557 		/* Exit early if the underlying link is invalid or busy */
558 		if (phy_dev->type != ARPHRD_ETHER ||
559 		    phy_dev->flags & IFF_LOOPBACK) {
560 			netdev_err(phy_dev,
561 				   "Master is either lo or non-ether device\n");
562 			return -EINVAL;
563 		}
564 
565 		if (netdev_is_rx_handler_busy(phy_dev)) {
566 			netdev_err(phy_dev, "Device is already in use.\n");
567 			return -EBUSY;
568 		}
569 	}
570 
571 	ipvlan->phy_dev = phy_dev;
572 	ipvlan->dev = dev;
573 	ipvlan->sfeatures = IPVLAN_FEATURES;
574 	if (!tb[IFLA_MTU])
575 		ipvlan_adjust_mtu(ipvlan, phy_dev);
576 	INIT_LIST_HEAD(&ipvlan->addrs);
577 	spin_lock_init(&ipvlan->addrs_lock);
578 
579 	/* TODO Probably put random address here to be presented to the
580 	 * world but keep using the physical-dev address for the outgoing
581 	 * packets.
582 	 */
583 	eth_hw_addr_set(dev, phy_dev->dev_addr);
584 
585 	dev->priv_flags |= IFF_NO_RX_HANDLER;
586 
587 	err = register_netdevice(dev);
588 	if (err < 0)
589 		return err;
590 
591 	/* ipvlan_init() would have created the port, if required */
592 	port = ipvlan_port_get_rtnl(phy_dev);
593 	ipvlan->port = port;
594 
595 	/* If the port-id base is at the MAX value, then wrap it around and
596 	 * begin from 0x1 again. This may be due to a busy system where lots
597 	 * of slaves are getting created and deleted.
598 	 */
599 	if (port->dev_id_start == 0xFFFE)
600 		port->dev_id_start = 0x1;
601 
602 	/* Since L2 address is shared among all IPvlan slaves including
603 	 * master, use unique 16 bit dev-ids to diffentiate among them.
604 	 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
605 	 * slave link [see addrconf_ifid_eui48()].
606 	 */
607 	err = ida_simple_get(&port->ida, port->dev_id_start, 0xFFFE,
608 			     GFP_KERNEL);
609 	if (err < 0)
610 		err = ida_simple_get(&port->ida, 0x1, port->dev_id_start,
611 				     GFP_KERNEL);
612 	if (err < 0)
613 		goto unregister_netdev;
614 	dev->dev_id = err;
615 
616 	/* Increment id-base to the next slot for the future assignment */
617 	port->dev_id_start = err + 1;
618 
619 	err = netdev_upper_dev_link(phy_dev, dev, extack);
620 	if (err)
621 		goto remove_ida;
622 
623 	/* Flags are per port and latest update overrides. User has
624 	 * to be consistent in setting it just like the mode attribute.
625 	 */
626 	if (data && data[IFLA_IPVLAN_FLAGS])
627 		port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
628 
629 	if (data && data[IFLA_IPVLAN_MODE])
630 		mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
631 
632 	err = ipvlan_set_port_mode(port, mode, extack);
633 	if (err)
634 		goto unlink_netdev;
635 
636 	list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
637 	netif_stacked_transfer_operstate(phy_dev, dev);
638 	return 0;
639 
640 unlink_netdev:
641 	netdev_upper_dev_unlink(phy_dev, dev);
642 remove_ida:
643 	ida_simple_remove(&port->ida, dev->dev_id);
644 unregister_netdev:
645 	unregister_netdevice(dev);
646 	return err;
647 }
648 EXPORT_SYMBOL_GPL(ipvlan_link_new);
649 
650 void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
651 {
652 	struct ipvl_dev *ipvlan = netdev_priv(dev);
653 	struct ipvl_addr *addr, *next;
654 
655 	spin_lock_bh(&ipvlan->addrs_lock);
656 	list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
657 		ipvlan_ht_addr_del(addr);
658 		list_del_rcu(&addr->anode);
659 		kfree_rcu(addr, rcu);
660 	}
661 	spin_unlock_bh(&ipvlan->addrs_lock);
662 
663 	ida_simple_remove(&ipvlan->port->ida, dev->dev_id);
664 	list_del_rcu(&ipvlan->pnode);
665 	unregister_netdevice_queue(dev, head);
666 	netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
667 }
668 EXPORT_SYMBOL_GPL(ipvlan_link_delete);
669 
670 void ipvlan_link_setup(struct net_device *dev)
671 {
672 	ether_setup(dev);
673 
674 	dev->max_mtu = ETH_MAX_MTU;
675 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
676 	dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
677 	dev->netdev_ops = &ipvlan_netdev_ops;
678 	dev->needs_free_netdev = true;
679 	dev->header_ops = &ipvlan_header_ops;
680 	dev->ethtool_ops = &ipvlan_ethtool_ops;
681 }
682 EXPORT_SYMBOL_GPL(ipvlan_link_setup);
683 
684 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
685 {
686 	[IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
687 	[IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
688 };
689 
690 static struct net *ipvlan_get_link_net(const struct net_device *dev)
691 {
692 	struct ipvl_dev *ipvlan = netdev_priv(dev);
693 
694 	return dev_net(ipvlan->phy_dev);
695 }
696 
697 static struct rtnl_link_ops ipvlan_link_ops = {
698 	.kind		= "ipvlan",
699 	.priv_size	= sizeof(struct ipvl_dev),
700 
701 	.setup		= ipvlan_link_setup,
702 	.newlink	= ipvlan_link_new,
703 	.dellink	= ipvlan_link_delete,
704 	.get_link_net   = ipvlan_get_link_net,
705 };
706 
707 int ipvlan_link_register(struct rtnl_link_ops *ops)
708 {
709 	ops->get_size	= ipvlan_nl_getsize;
710 	ops->policy	= ipvlan_nl_policy;
711 	ops->validate	= ipvlan_nl_validate;
712 	ops->fill_info	= ipvlan_nl_fillinfo;
713 	ops->changelink = ipvlan_nl_changelink;
714 	ops->maxtype	= IFLA_IPVLAN_MAX;
715 	return rtnl_link_register(ops);
716 }
717 EXPORT_SYMBOL_GPL(ipvlan_link_register);
718 
719 static int ipvlan_device_event(struct notifier_block *unused,
720 			       unsigned long event, void *ptr)
721 {
722 	struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
723 	struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
724 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
725 	struct ipvl_dev *ipvlan, *next;
726 	struct ipvl_port *port;
727 	LIST_HEAD(lst_kill);
728 	int err;
729 
730 	if (!netif_is_ipvlan_port(dev))
731 		return NOTIFY_DONE;
732 
733 	port = ipvlan_port_get_rtnl(dev);
734 
735 	switch (event) {
736 	case NETDEV_UP:
737 	case NETDEV_CHANGE:
738 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
739 			netif_stacked_transfer_operstate(ipvlan->phy_dev,
740 							 ipvlan->dev);
741 		break;
742 
743 	case NETDEV_REGISTER: {
744 		struct net *oldnet, *newnet = dev_net(dev);
745 
746 		oldnet = read_pnet(&port->pnet);
747 		if (net_eq(newnet, oldnet))
748 			break;
749 
750 		write_pnet(&port->pnet, newnet);
751 
752 		if (port->mode == IPVLAN_MODE_L3S)
753 			ipvlan_migrate_l3s_hook(oldnet, newnet);
754 		break;
755 	}
756 	case NETDEV_UNREGISTER:
757 		if (dev->reg_state != NETREG_UNREGISTERING)
758 			break;
759 
760 		list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
761 			ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
762 							    &lst_kill);
763 		unregister_netdevice_many(&lst_kill);
764 		break;
765 
766 	case NETDEV_FEAT_CHANGE:
767 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
768 			netif_inherit_tso_max(ipvlan->dev, dev);
769 			netdev_update_features(ipvlan->dev);
770 		}
771 		break;
772 
773 	case NETDEV_CHANGEMTU:
774 		list_for_each_entry(ipvlan, &port->ipvlans, pnode)
775 			ipvlan_adjust_mtu(ipvlan, dev);
776 		break;
777 
778 	case NETDEV_PRE_CHANGEADDR:
779 		prechaddr_info = ptr;
780 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
781 			err = dev_pre_changeaddr_notify(ipvlan->dev,
782 						    prechaddr_info->dev_addr,
783 						    extack);
784 			if (err)
785 				return notifier_from_errno(err);
786 		}
787 		break;
788 
789 	case NETDEV_CHANGEADDR:
790 		list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
791 			eth_hw_addr_set(ipvlan->dev, dev->dev_addr);
792 			call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
793 		}
794 		break;
795 
796 	case NETDEV_PRE_TYPE_CHANGE:
797 		/* Forbid underlying device to change its type. */
798 		return NOTIFY_BAD;
799 	}
800 	return NOTIFY_DONE;
801 }
802 
803 /* the caller must held the addrs lock */
804 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
805 {
806 	struct ipvl_addr *addr;
807 
808 	addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
809 	if (!addr)
810 		return -ENOMEM;
811 
812 	addr->master = ipvlan;
813 	if (!is_v6) {
814 		memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
815 		addr->atype = IPVL_IPV4;
816 #if IS_ENABLED(CONFIG_IPV6)
817 	} else {
818 		memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
819 		addr->atype = IPVL_IPV6;
820 #endif
821 	}
822 
823 	list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
824 
825 	/* If the interface is not up, the address will be added to the hash
826 	 * list by ipvlan_open.
827 	 */
828 	if (netif_running(ipvlan->dev))
829 		ipvlan_ht_addr_add(ipvlan, addr);
830 
831 	return 0;
832 }
833 
834 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
835 {
836 	struct ipvl_addr *addr;
837 
838 	spin_lock_bh(&ipvlan->addrs_lock);
839 	addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
840 	if (!addr) {
841 		spin_unlock_bh(&ipvlan->addrs_lock);
842 		return;
843 	}
844 
845 	ipvlan_ht_addr_del(addr);
846 	list_del_rcu(&addr->anode);
847 	spin_unlock_bh(&ipvlan->addrs_lock);
848 	kfree_rcu(addr, rcu);
849 }
850 
851 static bool ipvlan_is_valid_dev(const struct net_device *dev)
852 {
853 	struct ipvl_dev *ipvlan = netdev_priv(dev);
854 
855 	if (!netif_is_ipvlan(dev))
856 		return false;
857 
858 	if (!ipvlan || !ipvlan->port)
859 		return false;
860 
861 	return true;
862 }
863 
864 #if IS_ENABLED(CONFIG_IPV6)
865 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
866 {
867 	int ret = -EINVAL;
868 
869 	spin_lock_bh(&ipvlan->addrs_lock);
870 	if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
871 		netif_err(ipvlan, ifup, ipvlan->dev,
872 			  "Failed to add IPv6=%pI6c addr for %s intf\n",
873 			  ip6_addr, ipvlan->dev->name);
874 	else
875 		ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
876 	spin_unlock_bh(&ipvlan->addrs_lock);
877 	return ret;
878 }
879 
880 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
881 {
882 	return ipvlan_del_addr(ipvlan, ip6_addr, true);
883 }
884 
885 static int ipvlan_addr6_event(struct notifier_block *unused,
886 			      unsigned long event, void *ptr)
887 {
888 	struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
889 	struct net_device *dev = (struct net_device *)if6->idev->dev;
890 	struct ipvl_dev *ipvlan = netdev_priv(dev);
891 
892 	if (!ipvlan_is_valid_dev(dev))
893 		return NOTIFY_DONE;
894 
895 	switch (event) {
896 	case NETDEV_UP:
897 		if (ipvlan_add_addr6(ipvlan, &if6->addr))
898 			return NOTIFY_BAD;
899 		break;
900 
901 	case NETDEV_DOWN:
902 		ipvlan_del_addr6(ipvlan, &if6->addr);
903 		break;
904 	}
905 
906 	return NOTIFY_OK;
907 }
908 
909 static int ipvlan_addr6_validator_event(struct notifier_block *unused,
910 					unsigned long event, void *ptr)
911 {
912 	struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
913 	struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
914 	struct ipvl_dev *ipvlan = netdev_priv(dev);
915 
916 	if (!ipvlan_is_valid_dev(dev))
917 		return NOTIFY_DONE;
918 
919 	switch (event) {
920 	case NETDEV_UP:
921 		if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
922 			NL_SET_ERR_MSG(i6vi->extack,
923 				       "Address already assigned to an ipvlan device");
924 			return notifier_from_errno(-EADDRINUSE);
925 		}
926 		break;
927 	}
928 
929 	return NOTIFY_OK;
930 }
931 #endif
932 
933 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
934 {
935 	int ret = -EINVAL;
936 
937 	spin_lock_bh(&ipvlan->addrs_lock);
938 	if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
939 		netif_err(ipvlan, ifup, ipvlan->dev,
940 			  "Failed to add IPv4=%pI4 on %s intf.\n",
941 			  ip4_addr, ipvlan->dev->name);
942 	else
943 		ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
944 	spin_unlock_bh(&ipvlan->addrs_lock);
945 	return ret;
946 }
947 
948 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
949 {
950 	return ipvlan_del_addr(ipvlan, ip4_addr, false);
951 }
952 
953 static int ipvlan_addr4_event(struct notifier_block *unused,
954 			      unsigned long event, void *ptr)
955 {
956 	struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
957 	struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
958 	struct ipvl_dev *ipvlan = netdev_priv(dev);
959 	struct in_addr ip4_addr;
960 
961 	if (!ipvlan_is_valid_dev(dev))
962 		return NOTIFY_DONE;
963 
964 	switch (event) {
965 	case NETDEV_UP:
966 		ip4_addr.s_addr = if4->ifa_address;
967 		if (ipvlan_add_addr4(ipvlan, &ip4_addr))
968 			return NOTIFY_BAD;
969 		break;
970 
971 	case NETDEV_DOWN:
972 		ip4_addr.s_addr = if4->ifa_address;
973 		ipvlan_del_addr4(ipvlan, &ip4_addr);
974 		break;
975 	}
976 
977 	return NOTIFY_OK;
978 }
979 
980 static int ipvlan_addr4_validator_event(struct notifier_block *unused,
981 					unsigned long event, void *ptr)
982 {
983 	struct in_validator_info *ivi = (struct in_validator_info *)ptr;
984 	struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
985 	struct ipvl_dev *ipvlan = netdev_priv(dev);
986 
987 	if (!ipvlan_is_valid_dev(dev))
988 		return NOTIFY_DONE;
989 
990 	switch (event) {
991 	case NETDEV_UP:
992 		if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
993 			NL_SET_ERR_MSG(ivi->extack,
994 				       "Address already assigned to an ipvlan device");
995 			return notifier_from_errno(-EADDRINUSE);
996 		}
997 		break;
998 	}
999 
1000 	return NOTIFY_OK;
1001 }
1002 
1003 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
1004 	.notifier_call = ipvlan_addr4_event,
1005 };
1006 
1007 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
1008 	.notifier_call = ipvlan_addr4_validator_event,
1009 };
1010 
1011 static struct notifier_block ipvlan_notifier_block __read_mostly = {
1012 	.notifier_call = ipvlan_device_event,
1013 };
1014 
1015 #if IS_ENABLED(CONFIG_IPV6)
1016 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1017 	.notifier_call = ipvlan_addr6_event,
1018 };
1019 
1020 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1021 	.notifier_call = ipvlan_addr6_validator_event,
1022 };
1023 #endif
1024 
1025 static int __init ipvlan_init_module(void)
1026 {
1027 	int err;
1028 
1029 	ipvlan_init_secret();
1030 	register_netdevice_notifier(&ipvlan_notifier_block);
1031 #if IS_ENABLED(CONFIG_IPV6)
1032 	register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1033 	register_inet6addr_validator_notifier(
1034 	    &ipvlan_addr6_vtor_notifier_block);
1035 #endif
1036 	register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1037 	register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1038 
1039 	err = ipvlan_l3s_init();
1040 	if (err < 0)
1041 		goto error;
1042 
1043 	err = ipvlan_link_register(&ipvlan_link_ops);
1044 	if (err < 0) {
1045 		ipvlan_l3s_cleanup();
1046 		goto error;
1047 	}
1048 
1049 	return 0;
1050 error:
1051 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1052 	unregister_inetaddr_validator_notifier(
1053 	    &ipvlan_addr4_vtor_notifier_block);
1054 #if IS_ENABLED(CONFIG_IPV6)
1055 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1056 	unregister_inet6addr_validator_notifier(
1057 	    &ipvlan_addr6_vtor_notifier_block);
1058 #endif
1059 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1060 	return err;
1061 }
1062 
1063 static void __exit ipvlan_cleanup_module(void)
1064 {
1065 	rtnl_link_unregister(&ipvlan_link_ops);
1066 	ipvlan_l3s_cleanup();
1067 	unregister_netdevice_notifier(&ipvlan_notifier_block);
1068 	unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1069 	unregister_inetaddr_validator_notifier(
1070 	    &ipvlan_addr4_vtor_notifier_block);
1071 #if IS_ENABLED(CONFIG_IPV6)
1072 	unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1073 	unregister_inet6addr_validator_notifier(
1074 	    &ipvlan_addr6_vtor_notifier_block);
1075 #endif
1076 }
1077 
1078 module_init(ipvlan_init_module);
1079 module_exit(ipvlan_cleanup_module);
1080 
1081 MODULE_LICENSE("GPL");
1082 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1083 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1084 MODULE_ALIAS_RTNL_LINK("ipvlan");
1085