Loading...
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
9static 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
45fail:
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
61static 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
89err:
90 kfree(port);
91 return err;
92}
93
94static 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_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
130static 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 dev->lltx = true;
145 netif_inherit_tso_max(dev, phy_dev);
146 dev->hard_header_len = phy_dev->hard_header_len;
147
148 netdev_lockdep_set_classes(dev);
149
150 ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
151 if (!ipvlan->pcpu_stats)
152 return -ENOMEM;
153
154 if (!netif_is_ipvlan_port(phy_dev)) {
155 err = ipvlan_port_create(phy_dev);
156 if (err < 0) {
157 free_percpu(ipvlan->pcpu_stats);
158 return err;
159 }
160 }
161 port = ipvlan_port_get_rtnl(phy_dev);
162 port->count += 1;
163 return 0;
164}
165
166static void ipvlan_uninit(struct net_device *dev)
167{
168 struct ipvl_dev *ipvlan = netdev_priv(dev);
169 struct net_device *phy_dev = ipvlan->phy_dev;
170 struct ipvl_port *port;
171
172 free_percpu(ipvlan->pcpu_stats);
173
174 port = ipvlan_port_get_rtnl(phy_dev);
175 port->count -= 1;
176 if (!port->count)
177 ipvlan_port_destroy(port->dev);
178}
179
180static int ipvlan_open(struct net_device *dev)
181{
182 struct ipvl_dev *ipvlan = netdev_priv(dev);
183 struct ipvl_addr *addr;
184
185 if (ipvlan->port->mode == IPVLAN_MODE_L3 ||
186 ipvlan->port->mode == IPVLAN_MODE_L3S)
187 dev->flags |= IFF_NOARP;
188 else
189 dev->flags &= ~IFF_NOARP;
190
191 rcu_read_lock();
192 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
193 ipvlan_ht_addr_add(ipvlan, addr);
194 rcu_read_unlock();
195
196 return 0;
197}
198
199static int ipvlan_stop(struct net_device *dev)
200{
201 struct ipvl_dev *ipvlan = netdev_priv(dev);
202 struct net_device *phy_dev = ipvlan->phy_dev;
203 struct ipvl_addr *addr;
204
205 dev_uc_unsync(phy_dev, dev);
206 dev_mc_unsync(phy_dev, dev);
207
208 rcu_read_lock();
209 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
210 ipvlan_ht_addr_del(addr);
211 rcu_read_unlock();
212
213 return 0;
214}
215
216static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
217 struct net_device *dev)
218{
219 const struct ipvl_dev *ipvlan = netdev_priv(dev);
220 int skblen = skb->len;
221 int ret;
222
223 ret = ipvlan_queue_xmit(skb, dev);
224 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
225 struct ipvl_pcpu_stats *pcptr;
226
227 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
228
229 u64_stats_update_begin(&pcptr->syncp);
230 u64_stats_inc(&pcptr->tx_pkts);
231 u64_stats_add(&pcptr->tx_bytes, skblen);
232 u64_stats_update_end(&pcptr->syncp);
233 } else {
234 this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
235 }
236 return ret;
237}
238
239static netdev_features_t ipvlan_fix_features(struct net_device *dev,
240 netdev_features_t features)
241{
242 struct ipvl_dev *ipvlan = netdev_priv(dev);
243
244 features |= NETIF_F_ALL_FOR_ALL;
245 features &= (ipvlan->sfeatures | ~IPVLAN_FEATURES);
246 features = netdev_increment_features(ipvlan->phy_dev->features,
247 features, features);
248 features |= IPVLAN_ALWAYS_ON;
249 features &= (IPVLAN_FEATURES | IPVLAN_ALWAYS_ON);
250
251 return features;
252}
253
254static void ipvlan_change_rx_flags(struct net_device *dev, int change)
255{
256 struct ipvl_dev *ipvlan = netdev_priv(dev);
257 struct net_device *phy_dev = ipvlan->phy_dev;
258
259 if (change & IFF_ALLMULTI)
260 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
261}
262
263static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
264{
265 struct ipvl_dev *ipvlan = netdev_priv(dev);
266
267 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
268 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
269 } else {
270 struct netdev_hw_addr *ha;
271 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
272
273 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
274 netdev_for_each_mc_addr(ha, dev)
275 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
276
277 /* Turn-on broadcast bit irrespective of address family,
278 * since broadcast is deferred to a work-queue, hence no
279 * impact on fast-path processing.
280 */
281 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
282
283 bitmap_copy(ipvlan->mac_filters, mc_filters,
284 IPVLAN_MAC_FILTER_SIZE);
285 }
286 dev_uc_sync(ipvlan->phy_dev, dev);
287 dev_mc_sync(ipvlan->phy_dev, dev);
288}
289
290static void ipvlan_get_stats64(struct net_device *dev,
291 struct rtnl_link_stats64 *s)
292{
293 struct ipvl_dev *ipvlan = netdev_priv(dev);
294
295 if (ipvlan->pcpu_stats) {
296 struct ipvl_pcpu_stats *pcptr;
297 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
298 u32 rx_errs = 0, tx_drps = 0;
299 u32 strt;
300 int idx;
301
302 for_each_possible_cpu(idx) {
303 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
304 do {
305 strt = u64_stats_fetch_begin(&pcptr->syncp);
306 rx_pkts = u64_stats_read(&pcptr->rx_pkts);
307 rx_bytes = u64_stats_read(&pcptr->rx_bytes);
308 rx_mcast = u64_stats_read(&pcptr->rx_mcast);
309 tx_pkts = u64_stats_read(&pcptr->tx_pkts);
310 tx_bytes = u64_stats_read(&pcptr->tx_bytes);
311 } while (u64_stats_fetch_retry(&pcptr->syncp,
312 strt));
313
314 s->rx_packets += rx_pkts;
315 s->rx_bytes += rx_bytes;
316 s->multicast += rx_mcast;
317 s->tx_packets += tx_pkts;
318 s->tx_bytes += tx_bytes;
319
320 /* u32 values are updated without syncp protection. */
321 rx_errs += READ_ONCE(pcptr->rx_errs);
322 tx_drps += READ_ONCE(pcptr->tx_drps);
323 }
324 s->rx_errors = rx_errs;
325 s->rx_dropped = rx_errs;
326 s->tx_dropped = tx_drps;
327 }
328 s->tx_errors = DEV_STATS_READ(dev, tx_errors);
329}
330
331static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
332{
333 struct ipvl_dev *ipvlan = netdev_priv(dev);
334 struct net_device *phy_dev = ipvlan->phy_dev;
335
336 return vlan_vid_add(phy_dev, proto, vid);
337}
338
339static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
340 u16 vid)
341{
342 struct ipvl_dev *ipvlan = netdev_priv(dev);
343 struct net_device *phy_dev = ipvlan->phy_dev;
344
345 vlan_vid_del(phy_dev, proto, vid);
346 return 0;
347}
348
349static int ipvlan_get_iflink(const struct net_device *dev)
350{
351 struct ipvl_dev *ipvlan = netdev_priv(dev);
352
353 return READ_ONCE(ipvlan->phy_dev->ifindex);
354}
355
356static const struct net_device_ops ipvlan_netdev_ops = {
357 .ndo_init = ipvlan_init,
358 .ndo_uninit = ipvlan_uninit,
359 .ndo_open = ipvlan_open,
360 .ndo_stop = ipvlan_stop,
361 .ndo_start_xmit = ipvlan_start_xmit,
362 .ndo_fix_features = ipvlan_fix_features,
363 .ndo_change_rx_flags = ipvlan_change_rx_flags,
364 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter,
365 .ndo_get_stats64 = ipvlan_get_stats64,
366 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid,
367 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid,
368 .ndo_get_iflink = ipvlan_get_iflink,
369};
370
371static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
372 unsigned short type, const void *daddr,
373 const void *saddr, unsigned len)
374{
375 const struct ipvl_dev *ipvlan = netdev_priv(dev);
376 struct net_device *phy_dev = ipvlan->phy_dev;
377
378 /* TODO Probably use a different field than dev_addr so that the
379 * mac-address on the virtual device is portable and can be carried
380 * while the packets use the mac-addr on the physical device.
381 */
382 return dev_hard_header(skb, phy_dev, type, daddr,
383 saddr ? : phy_dev->dev_addr, len);
384}
385
386static const struct header_ops ipvlan_header_ops = {
387 .create = ipvlan_hard_header,
388 .parse = eth_header_parse,
389 .cache = eth_header_cache,
390 .cache_update = eth_header_cache_update,
391 .parse_protocol = eth_header_parse_protocol,
392};
393
394static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
395{
396 ipvlan->dev->mtu = dev->mtu;
397}
398
399static bool netif_is_ipvlan(const struct net_device *dev)
400{
401 /* both ipvlan and ipvtap devices use the same netdev_ops */
402 return dev->netdev_ops == &ipvlan_netdev_ops;
403}
404
405static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
406 struct ethtool_link_ksettings *cmd)
407{
408 const struct ipvl_dev *ipvlan = netdev_priv(dev);
409
410 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
411}
412
413static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
414 struct ethtool_drvinfo *drvinfo)
415{
416 strscpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
417 strscpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
418}
419
420static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
421{
422 const struct ipvl_dev *ipvlan = netdev_priv(dev);
423
424 return ipvlan->msg_enable;
425}
426
427static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
428{
429 struct ipvl_dev *ipvlan = netdev_priv(dev);
430
431 ipvlan->msg_enable = value;
432}
433
434static const struct ethtool_ops ipvlan_ethtool_ops = {
435 .get_link = ethtool_op_get_link,
436 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings,
437 .get_drvinfo = ipvlan_ethtool_get_drvinfo,
438 .get_msglevel = ipvlan_ethtool_get_msglevel,
439 .set_msglevel = ipvlan_ethtool_set_msglevel,
440};
441
442static int ipvlan_nl_changelink(struct net_device *dev,
443 struct nlattr *tb[], struct nlattr *data[],
444 struct netlink_ext_ack *extack)
445{
446 struct ipvl_dev *ipvlan = netdev_priv(dev);
447 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
448 int err = 0;
449
450 if (!data)
451 return 0;
452 if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
453 return -EPERM;
454
455 if (data[IFLA_IPVLAN_MODE]) {
456 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
457
458 err = ipvlan_set_port_mode(port, nmode, extack);
459 }
460
461 if (!err && data[IFLA_IPVLAN_FLAGS]) {
462 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
463
464 if (flags & IPVLAN_F_PRIVATE)
465 ipvlan_mark_private(port);
466 else
467 ipvlan_clear_private(port);
468
469 if (flags & IPVLAN_F_VEPA)
470 ipvlan_mark_vepa(port);
471 else
472 ipvlan_clear_vepa(port);
473 }
474
475 return err;
476}
477
478static size_t ipvlan_nl_getsize(const struct net_device *dev)
479{
480 return (0
481 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
482 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
483 );
484}
485
486static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
487 struct netlink_ext_ack *extack)
488{
489 if (!data)
490 return 0;
491
492 if (data[IFLA_IPVLAN_MODE]) {
493 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
494
495 if (mode >= IPVLAN_MODE_MAX)
496 return -EINVAL;
497 }
498 if (data[IFLA_IPVLAN_FLAGS]) {
499 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
500
501 /* Only two bits are used at this moment. */
502 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
503 return -EINVAL;
504 /* Also both flags can't be active at the same time. */
505 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
506 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
507 return -EINVAL;
508 }
509
510 return 0;
511}
512
513static int ipvlan_nl_fillinfo(struct sk_buff *skb,
514 const struct net_device *dev)
515{
516 struct ipvl_dev *ipvlan = netdev_priv(dev);
517 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
518 int ret = -EINVAL;
519
520 if (!port)
521 goto err;
522
523 ret = -EMSGSIZE;
524 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
525 goto err;
526 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
527 goto err;
528
529 return 0;
530
531err:
532 return ret;
533}
534
535int ipvlan_link_new(struct net *src_net, struct net_device *dev,
536 struct nlattr *tb[], struct nlattr *data[],
537 struct netlink_ext_ack *extack)
538{
539 struct ipvl_dev *ipvlan = netdev_priv(dev);
540 struct ipvl_port *port;
541 struct net_device *phy_dev;
542 int err;
543 u16 mode = IPVLAN_MODE_L3;
544
545 if (!tb[IFLA_LINK])
546 return -EINVAL;
547
548 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
549 if (!phy_dev)
550 return -ENODEV;
551
552 if (netif_is_ipvlan(phy_dev)) {
553 struct ipvl_dev *tmp = netdev_priv(phy_dev);
554
555 phy_dev = tmp->phy_dev;
556 if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
557 return -EPERM;
558 } else if (!netif_is_ipvlan_port(phy_dev)) {
559 /* Exit early if the underlying link is invalid or busy */
560 if (phy_dev->type != ARPHRD_ETHER ||
561 phy_dev->flags & IFF_LOOPBACK) {
562 netdev_err(phy_dev,
563 "Master is either lo or non-ether device\n");
564 return -EINVAL;
565 }
566
567 if (netdev_is_rx_handler_busy(phy_dev)) {
568 netdev_err(phy_dev, "Device is already in use.\n");
569 return -EBUSY;
570 }
571 }
572
573 ipvlan->phy_dev = phy_dev;
574 ipvlan->dev = dev;
575 ipvlan->sfeatures = IPVLAN_FEATURES;
576 if (!tb[IFLA_MTU])
577 ipvlan_adjust_mtu(ipvlan, phy_dev);
578 INIT_LIST_HEAD(&ipvlan->addrs);
579 spin_lock_init(&ipvlan->addrs_lock);
580
581 /* TODO Probably put random address here to be presented to the
582 * world but keep using the physical-dev address for the outgoing
583 * packets.
584 */
585 eth_hw_addr_set(dev, phy_dev->dev_addr);
586
587 dev->priv_flags |= IFF_NO_RX_HANDLER;
588
589 err = register_netdevice(dev);
590 if (err < 0)
591 return err;
592
593 /* ipvlan_init() would have created the port, if required */
594 port = ipvlan_port_get_rtnl(phy_dev);
595 ipvlan->port = port;
596
597 /* If the port-id base is at the MAX value, then wrap it around and
598 * begin from 0x1 again. This may be due to a busy system where lots
599 * of slaves are getting created and deleted.
600 */
601 if (port->dev_id_start == 0xFFFE)
602 port->dev_id_start = 0x1;
603
604 /* Since L2 address is shared among all IPvlan slaves including
605 * master, use unique 16 bit dev-ids to differentiate among them.
606 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
607 * slave link [see addrconf_ifid_eui48()].
608 */
609 err = ida_alloc_range(&port->ida, port->dev_id_start, 0xFFFD,
610 GFP_KERNEL);
611 if (err < 0)
612 err = ida_alloc_range(&port->ida, 0x1, port->dev_id_start - 1,
613 GFP_KERNEL);
614 if (err < 0)
615 goto unregister_netdev;
616 dev->dev_id = err;
617
618 /* Increment id-base to the next slot for the future assignment */
619 port->dev_id_start = err + 1;
620
621 err = netdev_upper_dev_link(phy_dev, dev, extack);
622 if (err)
623 goto remove_ida;
624
625 /* Flags are per port and latest update overrides. User has
626 * to be consistent in setting it just like the mode attribute.
627 */
628 if (data && data[IFLA_IPVLAN_FLAGS])
629 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
630
631 if (data && data[IFLA_IPVLAN_MODE])
632 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
633
634 err = ipvlan_set_port_mode(port, mode, extack);
635 if (err)
636 goto unlink_netdev;
637
638 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
639 netif_stacked_transfer_operstate(phy_dev, dev);
640 return 0;
641
642unlink_netdev:
643 netdev_upper_dev_unlink(phy_dev, dev);
644remove_ida:
645 ida_free(&port->ida, dev->dev_id);
646unregister_netdev:
647 unregister_netdevice(dev);
648 return err;
649}
650EXPORT_SYMBOL_GPL(ipvlan_link_new);
651
652void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
653{
654 struct ipvl_dev *ipvlan = netdev_priv(dev);
655 struct ipvl_addr *addr, *next;
656
657 spin_lock_bh(&ipvlan->addrs_lock);
658 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
659 ipvlan_ht_addr_del(addr);
660 list_del_rcu(&addr->anode);
661 kfree_rcu(addr, rcu);
662 }
663 spin_unlock_bh(&ipvlan->addrs_lock);
664
665 ida_free(&ipvlan->port->ida, dev->dev_id);
666 list_del_rcu(&ipvlan->pnode);
667 unregister_netdevice_queue(dev, head);
668 netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
669}
670EXPORT_SYMBOL_GPL(ipvlan_link_delete);
671
672void ipvlan_link_setup(struct net_device *dev)
673{
674 ether_setup(dev);
675
676 dev->max_mtu = ETH_MAX_MTU;
677 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
678 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
679 dev->netdev_ops = &ipvlan_netdev_ops;
680 dev->needs_free_netdev = true;
681 dev->header_ops = &ipvlan_header_ops;
682 dev->ethtool_ops = &ipvlan_ethtool_ops;
683}
684EXPORT_SYMBOL_GPL(ipvlan_link_setup);
685
686static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
687{
688 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
689 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
690};
691
692static struct net *ipvlan_get_link_net(const struct net_device *dev)
693{
694 struct ipvl_dev *ipvlan = netdev_priv(dev);
695
696 return dev_net(ipvlan->phy_dev);
697}
698
699static struct rtnl_link_ops ipvlan_link_ops = {
700 .kind = "ipvlan",
701 .priv_size = sizeof(struct ipvl_dev),
702
703 .setup = ipvlan_link_setup,
704 .newlink = ipvlan_link_new,
705 .dellink = ipvlan_link_delete,
706 .get_link_net = ipvlan_get_link_net,
707};
708
709int ipvlan_link_register(struct rtnl_link_ops *ops)
710{
711 ops->get_size = ipvlan_nl_getsize;
712 ops->policy = ipvlan_nl_policy;
713 ops->validate = ipvlan_nl_validate;
714 ops->fill_info = ipvlan_nl_fillinfo;
715 ops->changelink = ipvlan_nl_changelink;
716 ops->maxtype = IFLA_IPVLAN_MAX;
717 return rtnl_link_register(ops);
718}
719EXPORT_SYMBOL_GPL(ipvlan_link_register);
720
721static int ipvlan_device_event(struct notifier_block *unused,
722 unsigned long event, void *ptr)
723{
724 struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
725 struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
726 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
727 struct ipvl_dev *ipvlan, *next;
728 struct ipvl_port *port;
729 LIST_HEAD(lst_kill);
730 int err;
731
732 if (!netif_is_ipvlan_port(dev))
733 return NOTIFY_DONE;
734
735 port = ipvlan_port_get_rtnl(dev);
736
737 switch (event) {
738 case NETDEV_UP:
739 case NETDEV_DOWN:
740 case NETDEV_CHANGE:
741 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
742 netif_stacked_transfer_operstate(ipvlan->phy_dev,
743 ipvlan->dev);
744 break;
745
746 case NETDEV_REGISTER: {
747 struct net *oldnet, *newnet = dev_net(dev);
748
749 oldnet = read_pnet(&port->pnet);
750 if (net_eq(newnet, oldnet))
751 break;
752
753 write_pnet(&port->pnet, newnet);
754
755 if (port->mode == IPVLAN_MODE_L3S)
756 ipvlan_migrate_l3s_hook(oldnet, newnet);
757 break;
758 }
759 case NETDEV_UNREGISTER:
760 if (dev->reg_state != NETREG_UNREGISTERING)
761 break;
762
763 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
764 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
765 &lst_kill);
766 unregister_netdevice_many(&lst_kill);
767 break;
768
769 case NETDEV_FEAT_CHANGE:
770 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
771 netif_inherit_tso_max(ipvlan->dev, dev);
772 netdev_update_features(ipvlan->dev);
773 }
774 break;
775
776 case NETDEV_CHANGEMTU:
777 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
778 ipvlan_adjust_mtu(ipvlan, dev);
779 break;
780
781 case NETDEV_PRE_CHANGEADDR:
782 prechaddr_info = ptr;
783 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
784 err = dev_pre_changeaddr_notify(ipvlan->dev,
785 prechaddr_info->dev_addr,
786 extack);
787 if (err)
788 return notifier_from_errno(err);
789 }
790 break;
791
792 case NETDEV_CHANGEADDR:
793 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
794 eth_hw_addr_set(ipvlan->dev, dev->dev_addr);
795 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
796 }
797 break;
798
799 case NETDEV_PRE_TYPE_CHANGE:
800 /* Forbid underlying device to change its type. */
801 return NOTIFY_BAD;
802 }
803 return NOTIFY_DONE;
804}
805
806/* the caller must held the addrs lock */
807static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
808{
809 struct ipvl_addr *addr;
810
811 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
812 if (!addr)
813 return -ENOMEM;
814
815 addr->master = ipvlan;
816 if (!is_v6) {
817 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
818 addr->atype = IPVL_IPV4;
819#if IS_ENABLED(CONFIG_IPV6)
820 } else {
821 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
822 addr->atype = IPVL_IPV6;
823#endif
824 }
825
826 list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
827
828 /* If the interface is not up, the address will be added to the hash
829 * list by ipvlan_open.
830 */
831 if (netif_running(ipvlan->dev))
832 ipvlan_ht_addr_add(ipvlan, addr);
833
834 return 0;
835}
836
837static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
838{
839 struct ipvl_addr *addr;
840
841 spin_lock_bh(&ipvlan->addrs_lock);
842 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
843 if (!addr) {
844 spin_unlock_bh(&ipvlan->addrs_lock);
845 return;
846 }
847
848 ipvlan_ht_addr_del(addr);
849 list_del_rcu(&addr->anode);
850 spin_unlock_bh(&ipvlan->addrs_lock);
851 kfree_rcu(addr, rcu);
852}
853
854static bool ipvlan_is_valid_dev(const struct net_device *dev)
855{
856 struct ipvl_dev *ipvlan = netdev_priv(dev);
857
858 if (!netif_is_ipvlan(dev))
859 return false;
860
861 if (!ipvlan || !ipvlan->port)
862 return false;
863
864 return true;
865}
866
867#if IS_ENABLED(CONFIG_IPV6)
868static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
869{
870 int ret = -EINVAL;
871
872 spin_lock_bh(&ipvlan->addrs_lock);
873 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
874 netif_err(ipvlan, ifup, ipvlan->dev,
875 "Failed to add IPv6=%pI6c addr for %s intf\n",
876 ip6_addr, ipvlan->dev->name);
877 else
878 ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
879 spin_unlock_bh(&ipvlan->addrs_lock);
880 return ret;
881}
882
883static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
884{
885 return ipvlan_del_addr(ipvlan, ip6_addr, true);
886}
887
888static int ipvlan_addr6_event(struct notifier_block *unused,
889 unsigned long event, void *ptr)
890{
891 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
892 struct net_device *dev = (struct net_device *)if6->idev->dev;
893 struct ipvl_dev *ipvlan = netdev_priv(dev);
894
895 if (!ipvlan_is_valid_dev(dev))
896 return NOTIFY_DONE;
897
898 switch (event) {
899 case NETDEV_UP:
900 if (ipvlan_add_addr6(ipvlan, &if6->addr))
901 return NOTIFY_BAD;
902 break;
903
904 case NETDEV_DOWN:
905 ipvlan_del_addr6(ipvlan, &if6->addr);
906 break;
907 }
908
909 return NOTIFY_OK;
910}
911
912static int ipvlan_addr6_validator_event(struct notifier_block *unused,
913 unsigned long event, void *ptr)
914{
915 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
916 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
917 struct ipvl_dev *ipvlan = netdev_priv(dev);
918
919 if (!ipvlan_is_valid_dev(dev))
920 return NOTIFY_DONE;
921
922 switch (event) {
923 case NETDEV_UP:
924 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
925 NL_SET_ERR_MSG(i6vi->extack,
926 "Address already assigned to an ipvlan device");
927 return notifier_from_errno(-EADDRINUSE);
928 }
929 break;
930 }
931
932 return NOTIFY_OK;
933}
934#endif
935
936static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
937{
938 int ret = -EINVAL;
939
940 spin_lock_bh(&ipvlan->addrs_lock);
941 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
942 netif_err(ipvlan, ifup, ipvlan->dev,
943 "Failed to add IPv4=%pI4 on %s intf.\n",
944 ip4_addr, ipvlan->dev->name);
945 else
946 ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
947 spin_unlock_bh(&ipvlan->addrs_lock);
948 return ret;
949}
950
951static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
952{
953 return ipvlan_del_addr(ipvlan, ip4_addr, false);
954}
955
956static int ipvlan_addr4_event(struct notifier_block *unused,
957 unsigned long event, void *ptr)
958{
959 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
960 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
961 struct ipvl_dev *ipvlan = netdev_priv(dev);
962 struct in_addr ip4_addr;
963
964 if (!ipvlan_is_valid_dev(dev))
965 return NOTIFY_DONE;
966
967 switch (event) {
968 case NETDEV_UP:
969 ip4_addr.s_addr = if4->ifa_address;
970 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
971 return NOTIFY_BAD;
972 break;
973
974 case NETDEV_DOWN:
975 ip4_addr.s_addr = if4->ifa_address;
976 ipvlan_del_addr4(ipvlan, &ip4_addr);
977 break;
978 }
979
980 return NOTIFY_OK;
981}
982
983static int ipvlan_addr4_validator_event(struct notifier_block *unused,
984 unsigned long event, void *ptr)
985{
986 struct in_validator_info *ivi = (struct in_validator_info *)ptr;
987 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
988 struct ipvl_dev *ipvlan = netdev_priv(dev);
989
990 if (!ipvlan_is_valid_dev(dev))
991 return NOTIFY_DONE;
992
993 switch (event) {
994 case NETDEV_UP:
995 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
996 NL_SET_ERR_MSG(ivi->extack,
997 "Address already assigned to an ipvlan device");
998 return notifier_from_errno(-EADDRINUSE);
999 }
1000 break;
1001 }
1002
1003 return NOTIFY_OK;
1004}
1005
1006static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
1007 .notifier_call = ipvlan_addr4_event,
1008};
1009
1010static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
1011 .notifier_call = ipvlan_addr4_validator_event,
1012};
1013
1014static struct notifier_block ipvlan_notifier_block __read_mostly = {
1015 .notifier_call = ipvlan_device_event,
1016};
1017
1018#if IS_ENABLED(CONFIG_IPV6)
1019static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1020 .notifier_call = ipvlan_addr6_event,
1021};
1022
1023static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1024 .notifier_call = ipvlan_addr6_validator_event,
1025};
1026#endif
1027
1028static int __init ipvlan_init_module(void)
1029{
1030 int err;
1031
1032 ipvlan_init_secret();
1033 register_netdevice_notifier(&ipvlan_notifier_block);
1034#if IS_ENABLED(CONFIG_IPV6)
1035 register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1036 register_inet6addr_validator_notifier(
1037 &ipvlan_addr6_vtor_notifier_block);
1038#endif
1039 register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1040 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1041
1042 err = ipvlan_l3s_init();
1043 if (err < 0)
1044 goto error;
1045
1046 err = ipvlan_link_register(&ipvlan_link_ops);
1047 if (err < 0) {
1048 ipvlan_l3s_cleanup();
1049 goto error;
1050 }
1051
1052 return 0;
1053error:
1054 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1055 unregister_inetaddr_validator_notifier(
1056 &ipvlan_addr4_vtor_notifier_block);
1057#if IS_ENABLED(CONFIG_IPV6)
1058 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1059 unregister_inet6addr_validator_notifier(
1060 &ipvlan_addr6_vtor_notifier_block);
1061#endif
1062 unregister_netdevice_notifier(&ipvlan_notifier_block);
1063 return err;
1064}
1065
1066static void __exit ipvlan_cleanup_module(void)
1067{
1068 rtnl_link_unregister(&ipvlan_link_ops);
1069 ipvlan_l3s_cleanup();
1070 unregister_netdevice_notifier(&ipvlan_notifier_block);
1071 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1072 unregister_inetaddr_validator_notifier(
1073 &ipvlan_addr4_vtor_notifier_block);
1074#if IS_ENABLED(CONFIG_IPV6)
1075 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1076 unregister_inet6addr_validator_notifier(
1077 &ipvlan_addr6_vtor_notifier_block);
1078#endif
1079}
1080
1081module_init(ipvlan_init_module);
1082module_exit(ipvlan_cleanup_module);
1083
1084MODULE_LICENSE("GPL");
1085MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1086MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1087MODULE_ALIAS_RTNL_LINK("ipvlan");
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
7static 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
43fail:
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
59static 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
86err:
87 kfree(port);
88 return err;
89}
90
91static 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_ALWAYS_ON_OFLOADS \
110 (NETIF_F_SG | NETIF_F_HW_CSUM | \
111 NETIF_F_GSO_ROBUST | NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL)
112
113#define IPVLAN_ALWAYS_ON \
114 (IPVLAN_ALWAYS_ON_OFLOADS | NETIF_F_LLTX | NETIF_F_VLAN_CHALLENGED)
115
116#define IPVLAN_FEATURES \
117 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
118 NETIF_F_GSO | NETIF_F_ALL_TSO | NETIF_F_GSO_ROBUST | \
119 NETIF_F_GRO | NETIF_F_RXCSUM | \
120 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
121
122 /* NETIF_F_GSO_ENCAP_ALL NETIF_F_GSO_SOFTWARE Newly added */
123
124#define IPVLAN_STATE_MASK \
125 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
126
127static int ipvlan_init(struct net_device *dev)
128{
129 struct ipvl_dev *ipvlan = netdev_priv(dev);
130 struct net_device *phy_dev = ipvlan->phy_dev;
131 struct ipvl_port *port;
132 int err;
133
134 dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
135 (phy_dev->state & IPVLAN_STATE_MASK);
136 dev->features = phy_dev->features & IPVLAN_FEATURES;
137 dev->features |= IPVLAN_ALWAYS_ON;
138 dev->vlan_features = phy_dev->vlan_features & IPVLAN_FEATURES;
139 dev->vlan_features |= IPVLAN_ALWAYS_ON_OFLOADS;
140 dev->hw_enc_features |= dev->features;
141 dev->gso_max_size = phy_dev->gso_max_size;
142 dev->gso_max_segs = phy_dev->gso_max_segs;
143 dev->hard_header_len = phy_dev->hard_header_len;
144
145 netdev_lockdep_set_classes(dev);
146
147 ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
148 if (!ipvlan->pcpu_stats)
149 return -ENOMEM;
150
151 if (!netif_is_ipvlan_port(phy_dev)) {
152 err = ipvlan_port_create(phy_dev);
153 if (err < 0) {
154 free_percpu(ipvlan->pcpu_stats);
155 return err;
156 }
157 }
158 port = ipvlan_port_get_rtnl(phy_dev);
159 port->count += 1;
160 return 0;
161}
162
163static void ipvlan_uninit(struct net_device *dev)
164{
165 struct ipvl_dev *ipvlan = netdev_priv(dev);
166 struct net_device *phy_dev = ipvlan->phy_dev;
167 struct ipvl_port *port;
168
169 free_percpu(ipvlan->pcpu_stats);
170
171 port = ipvlan_port_get_rtnl(phy_dev);
172 port->count -= 1;
173 if (!port->count)
174 ipvlan_port_destroy(port->dev);
175}
176
177static int ipvlan_open(struct net_device *dev)
178{
179 struct ipvl_dev *ipvlan = netdev_priv(dev);
180 struct ipvl_addr *addr;
181
182 if (ipvlan->port->mode == IPVLAN_MODE_L3 ||
183 ipvlan->port->mode == IPVLAN_MODE_L3S)
184 dev->flags |= IFF_NOARP;
185 else
186 dev->flags &= ~IFF_NOARP;
187
188 rcu_read_lock();
189 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
190 ipvlan_ht_addr_add(ipvlan, addr);
191 rcu_read_unlock();
192
193 return 0;
194}
195
196static int ipvlan_stop(struct net_device *dev)
197{
198 struct ipvl_dev *ipvlan = netdev_priv(dev);
199 struct net_device *phy_dev = ipvlan->phy_dev;
200 struct ipvl_addr *addr;
201
202 dev_uc_unsync(phy_dev, dev);
203 dev_mc_unsync(phy_dev, dev);
204
205 rcu_read_lock();
206 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
207 ipvlan_ht_addr_del(addr);
208 rcu_read_unlock();
209
210 return 0;
211}
212
213static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
214 struct net_device *dev)
215{
216 const struct ipvl_dev *ipvlan = netdev_priv(dev);
217 int skblen = skb->len;
218 int ret;
219
220 ret = ipvlan_queue_xmit(skb, dev);
221 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
222 struct ipvl_pcpu_stats *pcptr;
223
224 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
225
226 u64_stats_update_begin(&pcptr->syncp);
227 pcptr->tx_pkts++;
228 pcptr->tx_bytes += skblen;
229 u64_stats_update_end(&pcptr->syncp);
230 } else {
231 this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
232 }
233 return ret;
234}
235
236static netdev_features_t ipvlan_fix_features(struct net_device *dev,
237 netdev_features_t features)
238{
239 struct ipvl_dev *ipvlan = netdev_priv(dev);
240
241 features |= NETIF_F_ALL_FOR_ALL;
242 features &= (ipvlan->sfeatures | ~IPVLAN_FEATURES);
243 features = netdev_increment_features(ipvlan->phy_dev->features,
244 features, features);
245 features |= IPVLAN_ALWAYS_ON;
246 features &= (IPVLAN_FEATURES | IPVLAN_ALWAYS_ON);
247
248 return features;
249}
250
251static void ipvlan_change_rx_flags(struct net_device *dev, int change)
252{
253 struct ipvl_dev *ipvlan = netdev_priv(dev);
254 struct net_device *phy_dev = ipvlan->phy_dev;
255
256 if (change & IFF_ALLMULTI)
257 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
258}
259
260static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
261{
262 struct ipvl_dev *ipvlan = netdev_priv(dev);
263
264 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
265 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
266 } else {
267 struct netdev_hw_addr *ha;
268 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
269
270 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
271 netdev_for_each_mc_addr(ha, dev)
272 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
273
274 /* Turn-on broadcast bit irrespective of address family,
275 * since broadcast is deferred to a work-queue, hence no
276 * impact on fast-path processing.
277 */
278 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
279
280 bitmap_copy(ipvlan->mac_filters, mc_filters,
281 IPVLAN_MAC_FILTER_SIZE);
282 }
283 dev_uc_sync(ipvlan->phy_dev, dev);
284 dev_mc_sync(ipvlan->phy_dev, dev);
285}
286
287static void ipvlan_get_stats64(struct net_device *dev,
288 struct rtnl_link_stats64 *s)
289{
290 struct ipvl_dev *ipvlan = netdev_priv(dev);
291
292 if (ipvlan->pcpu_stats) {
293 struct ipvl_pcpu_stats *pcptr;
294 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
295 u32 rx_errs = 0, tx_drps = 0;
296 u32 strt;
297 int idx;
298
299 for_each_possible_cpu(idx) {
300 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
301 do {
302 strt= u64_stats_fetch_begin_irq(&pcptr->syncp);
303 rx_pkts = pcptr->rx_pkts;
304 rx_bytes = pcptr->rx_bytes;
305 rx_mcast = pcptr->rx_mcast;
306 tx_pkts = pcptr->tx_pkts;
307 tx_bytes = pcptr->tx_bytes;
308 } while (u64_stats_fetch_retry_irq(&pcptr->syncp,
309 strt));
310
311 s->rx_packets += rx_pkts;
312 s->rx_bytes += rx_bytes;
313 s->multicast += rx_mcast;
314 s->tx_packets += tx_pkts;
315 s->tx_bytes += tx_bytes;
316
317 /* u32 values are updated without syncp protection. */
318 rx_errs += pcptr->rx_errs;
319 tx_drps += pcptr->tx_drps;
320 }
321 s->rx_errors = rx_errs;
322 s->rx_dropped = rx_errs;
323 s->tx_dropped = tx_drps;
324 }
325}
326
327static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
328{
329 struct ipvl_dev *ipvlan = netdev_priv(dev);
330 struct net_device *phy_dev = ipvlan->phy_dev;
331
332 return vlan_vid_add(phy_dev, proto, vid);
333}
334
335static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
336 u16 vid)
337{
338 struct ipvl_dev *ipvlan = netdev_priv(dev);
339 struct net_device *phy_dev = ipvlan->phy_dev;
340
341 vlan_vid_del(phy_dev, proto, vid);
342 return 0;
343}
344
345static int ipvlan_get_iflink(const struct net_device *dev)
346{
347 struct ipvl_dev *ipvlan = netdev_priv(dev);
348
349 return ipvlan->phy_dev->ifindex;
350}
351
352static const struct net_device_ops ipvlan_netdev_ops = {
353 .ndo_init = ipvlan_init,
354 .ndo_uninit = ipvlan_uninit,
355 .ndo_open = ipvlan_open,
356 .ndo_stop = ipvlan_stop,
357 .ndo_start_xmit = ipvlan_start_xmit,
358 .ndo_fix_features = ipvlan_fix_features,
359 .ndo_change_rx_flags = ipvlan_change_rx_flags,
360 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter,
361 .ndo_get_stats64 = ipvlan_get_stats64,
362 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid,
363 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid,
364 .ndo_get_iflink = ipvlan_get_iflink,
365};
366
367static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
368 unsigned short type, const void *daddr,
369 const void *saddr, unsigned len)
370{
371 const struct ipvl_dev *ipvlan = netdev_priv(dev);
372 struct net_device *phy_dev = ipvlan->phy_dev;
373
374 /* TODO Probably use a different field than dev_addr so that the
375 * mac-address on the virtual device is portable and can be carried
376 * while the packets use the mac-addr on the physical device.
377 */
378 return dev_hard_header(skb, phy_dev, type, daddr,
379 saddr ? : phy_dev->dev_addr, len);
380}
381
382static const struct header_ops ipvlan_header_ops = {
383 .create = ipvlan_hard_header,
384 .parse = eth_header_parse,
385 .cache = eth_header_cache,
386 .cache_update = eth_header_cache_update,
387};
388
389static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
390{
391 ipvlan->dev->mtu = dev->mtu;
392}
393
394static bool netif_is_ipvlan(const struct net_device *dev)
395{
396 /* both ipvlan and ipvtap devices use the same netdev_ops */
397 return dev->netdev_ops == &ipvlan_netdev_ops;
398}
399
400static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
401 struct ethtool_link_ksettings *cmd)
402{
403 const struct ipvl_dev *ipvlan = netdev_priv(dev);
404
405 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
406}
407
408static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
409 struct ethtool_drvinfo *drvinfo)
410{
411 strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
412 strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
413}
414
415static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
416{
417 const struct ipvl_dev *ipvlan = netdev_priv(dev);
418
419 return ipvlan->msg_enable;
420}
421
422static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
423{
424 struct ipvl_dev *ipvlan = netdev_priv(dev);
425
426 ipvlan->msg_enable = value;
427}
428
429static const struct ethtool_ops ipvlan_ethtool_ops = {
430 .get_link = ethtool_op_get_link,
431 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings,
432 .get_drvinfo = ipvlan_ethtool_get_drvinfo,
433 .get_msglevel = ipvlan_ethtool_get_msglevel,
434 .set_msglevel = ipvlan_ethtool_set_msglevel,
435};
436
437static int ipvlan_nl_changelink(struct net_device *dev,
438 struct nlattr *tb[], struct nlattr *data[],
439 struct netlink_ext_ack *extack)
440{
441 struct ipvl_dev *ipvlan = netdev_priv(dev);
442 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
443 int err = 0;
444
445 if (!data)
446 return 0;
447 if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
448 return -EPERM;
449
450 if (data[IFLA_IPVLAN_MODE]) {
451 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
452
453 err = ipvlan_set_port_mode(port, nmode, extack);
454 }
455
456 if (!err && data[IFLA_IPVLAN_FLAGS]) {
457 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
458
459 if (flags & IPVLAN_F_PRIVATE)
460 ipvlan_mark_private(port);
461 else
462 ipvlan_clear_private(port);
463
464 if (flags & IPVLAN_F_VEPA)
465 ipvlan_mark_vepa(port);
466 else
467 ipvlan_clear_vepa(port);
468 }
469
470 return err;
471}
472
473static size_t ipvlan_nl_getsize(const struct net_device *dev)
474{
475 return (0
476 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
477 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
478 );
479}
480
481static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
482 struct netlink_ext_ack *extack)
483{
484 if (!data)
485 return 0;
486
487 if (data[IFLA_IPVLAN_MODE]) {
488 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
489
490 if (mode >= IPVLAN_MODE_MAX)
491 return -EINVAL;
492 }
493 if (data[IFLA_IPVLAN_FLAGS]) {
494 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
495
496 /* Only two bits are used at this moment. */
497 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
498 return -EINVAL;
499 /* Also both flags can't be active at the same time. */
500 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
501 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
502 return -EINVAL;
503 }
504
505 return 0;
506}
507
508static int ipvlan_nl_fillinfo(struct sk_buff *skb,
509 const struct net_device *dev)
510{
511 struct ipvl_dev *ipvlan = netdev_priv(dev);
512 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
513 int ret = -EINVAL;
514
515 if (!port)
516 goto err;
517
518 ret = -EMSGSIZE;
519 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
520 goto err;
521 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
522 goto err;
523
524 return 0;
525
526err:
527 return ret;
528}
529
530int ipvlan_link_new(struct net *src_net, struct net_device *dev,
531 struct nlattr *tb[], struct nlattr *data[],
532 struct netlink_ext_ack *extack)
533{
534 struct ipvl_dev *ipvlan = netdev_priv(dev);
535 struct ipvl_port *port;
536 struct net_device *phy_dev;
537 int err;
538 u16 mode = IPVLAN_MODE_L3;
539
540 if (!tb[IFLA_LINK])
541 return -EINVAL;
542
543 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
544 if (!phy_dev)
545 return -ENODEV;
546
547 if (netif_is_ipvlan(phy_dev)) {
548 struct ipvl_dev *tmp = netdev_priv(phy_dev);
549
550 phy_dev = tmp->phy_dev;
551 if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
552 return -EPERM;
553 } else if (!netif_is_ipvlan_port(phy_dev)) {
554 /* Exit early if the underlying link is invalid or busy */
555 if (phy_dev->type != ARPHRD_ETHER ||
556 phy_dev->flags & IFF_LOOPBACK) {
557 netdev_err(phy_dev,
558 "Master is either lo or non-ether device\n");
559 return -EINVAL;
560 }
561
562 if (netdev_is_rx_handler_busy(phy_dev)) {
563 netdev_err(phy_dev, "Device is already in use.\n");
564 return -EBUSY;
565 }
566 }
567
568 ipvlan->phy_dev = phy_dev;
569 ipvlan->dev = dev;
570 ipvlan->sfeatures = IPVLAN_FEATURES;
571 if (!tb[IFLA_MTU])
572 ipvlan_adjust_mtu(ipvlan, phy_dev);
573 INIT_LIST_HEAD(&ipvlan->addrs);
574 spin_lock_init(&ipvlan->addrs_lock);
575
576 /* TODO Probably put random address here to be presented to the
577 * world but keep using the physical-dev address for the outgoing
578 * packets.
579 */
580 memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN);
581
582 dev->priv_flags |= IFF_NO_RX_HANDLER;
583
584 err = register_netdevice(dev);
585 if (err < 0)
586 return err;
587
588 /* ipvlan_init() would have created the port, if required */
589 port = ipvlan_port_get_rtnl(phy_dev);
590 ipvlan->port = port;
591
592 /* If the port-id base is at the MAX value, then wrap it around and
593 * begin from 0x1 again. This may be due to a busy system where lots
594 * of slaves are getting created and deleted.
595 */
596 if (port->dev_id_start == 0xFFFE)
597 port->dev_id_start = 0x1;
598
599 /* Since L2 address is shared among all IPvlan slaves including
600 * master, use unique 16 bit dev-ids to diffentiate among them.
601 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
602 * slave link [see addrconf_ifid_eui48()].
603 */
604 err = ida_simple_get(&port->ida, port->dev_id_start, 0xFFFE,
605 GFP_KERNEL);
606 if (err < 0)
607 err = ida_simple_get(&port->ida, 0x1, port->dev_id_start,
608 GFP_KERNEL);
609 if (err < 0)
610 goto unregister_netdev;
611 dev->dev_id = err;
612
613 /* Increment id-base to the next slot for the future assignment */
614 port->dev_id_start = err + 1;
615
616 err = netdev_upper_dev_link(phy_dev, dev, extack);
617 if (err)
618 goto remove_ida;
619
620 /* Flags are per port and latest update overrides. User has
621 * to be consistent in setting it just like the mode attribute.
622 */
623 if (data && data[IFLA_IPVLAN_FLAGS])
624 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
625
626 if (data && data[IFLA_IPVLAN_MODE])
627 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
628
629 err = ipvlan_set_port_mode(port, mode, extack);
630 if (err)
631 goto unlink_netdev;
632
633 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
634 netif_stacked_transfer_operstate(phy_dev, dev);
635 return 0;
636
637unlink_netdev:
638 netdev_upper_dev_unlink(phy_dev, dev);
639remove_ida:
640 ida_simple_remove(&port->ida, dev->dev_id);
641unregister_netdev:
642 unregister_netdevice(dev);
643 return err;
644}
645EXPORT_SYMBOL_GPL(ipvlan_link_new);
646
647void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
648{
649 struct ipvl_dev *ipvlan = netdev_priv(dev);
650 struct ipvl_addr *addr, *next;
651
652 spin_lock_bh(&ipvlan->addrs_lock);
653 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
654 ipvlan_ht_addr_del(addr);
655 list_del_rcu(&addr->anode);
656 kfree_rcu(addr, rcu);
657 }
658 spin_unlock_bh(&ipvlan->addrs_lock);
659
660 ida_simple_remove(&ipvlan->port->ida, dev->dev_id);
661 list_del_rcu(&ipvlan->pnode);
662 unregister_netdevice_queue(dev, head);
663 netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
664}
665EXPORT_SYMBOL_GPL(ipvlan_link_delete);
666
667void ipvlan_link_setup(struct net_device *dev)
668{
669 ether_setup(dev);
670
671 dev->max_mtu = ETH_MAX_MTU;
672 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
673 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
674 dev->netdev_ops = &ipvlan_netdev_ops;
675 dev->needs_free_netdev = true;
676 dev->header_ops = &ipvlan_header_ops;
677 dev->ethtool_ops = &ipvlan_ethtool_ops;
678}
679EXPORT_SYMBOL_GPL(ipvlan_link_setup);
680
681static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
682{
683 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
684 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
685};
686
687static struct rtnl_link_ops ipvlan_link_ops = {
688 .kind = "ipvlan",
689 .priv_size = sizeof(struct ipvl_dev),
690
691 .setup = ipvlan_link_setup,
692 .newlink = ipvlan_link_new,
693 .dellink = ipvlan_link_delete,
694};
695
696int ipvlan_link_register(struct rtnl_link_ops *ops)
697{
698 ops->get_size = ipvlan_nl_getsize;
699 ops->policy = ipvlan_nl_policy;
700 ops->validate = ipvlan_nl_validate;
701 ops->fill_info = ipvlan_nl_fillinfo;
702 ops->changelink = ipvlan_nl_changelink;
703 ops->maxtype = IFLA_IPVLAN_MAX;
704 return rtnl_link_register(ops);
705}
706EXPORT_SYMBOL_GPL(ipvlan_link_register);
707
708static int ipvlan_device_event(struct notifier_block *unused,
709 unsigned long event, void *ptr)
710{
711 struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
712 struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
713 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
714 struct ipvl_dev *ipvlan, *next;
715 struct ipvl_port *port;
716 LIST_HEAD(lst_kill);
717 int err;
718
719 if (!netif_is_ipvlan_port(dev))
720 return NOTIFY_DONE;
721
722 port = ipvlan_port_get_rtnl(dev);
723
724 switch (event) {
725 case NETDEV_CHANGE:
726 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
727 netif_stacked_transfer_operstate(ipvlan->phy_dev,
728 ipvlan->dev);
729 break;
730
731 case NETDEV_REGISTER: {
732 struct net *oldnet, *newnet = dev_net(dev);
733
734 oldnet = read_pnet(&port->pnet);
735 if (net_eq(newnet, oldnet))
736 break;
737
738 write_pnet(&port->pnet, newnet);
739
740 ipvlan_migrate_l3s_hook(oldnet, newnet);
741 break;
742 }
743 case NETDEV_UNREGISTER:
744 if (dev->reg_state != NETREG_UNREGISTERING)
745 break;
746
747 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
748 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
749 &lst_kill);
750 unregister_netdevice_many(&lst_kill);
751 break;
752
753 case NETDEV_FEAT_CHANGE:
754 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
755 ipvlan->dev->gso_max_size = dev->gso_max_size;
756 ipvlan->dev->gso_max_segs = dev->gso_max_segs;
757 netdev_update_features(ipvlan->dev);
758 }
759 break;
760
761 case NETDEV_CHANGEMTU:
762 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
763 ipvlan_adjust_mtu(ipvlan, dev);
764 break;
765
766 case NETDEV_PRE_CHANGEADDR:
767 prechaddr_info = ptr;
768 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
769 err = dev_pre_changeaddr_notify(ipvlan->dev,
770 prechaddr_info->dev_addr,
771 extack);
772 if (err)
773 return notifier_from_errno(err);
774 }
775 break;
776
777 case NETDEV_CHANGEADDR:
778 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
779 ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr);
780 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
781 }
782 break;
783
784 case NETDEV_PRE_TYPE_CHANGE:
785 /* Forbid underlying device to change its type. */
786 return NOTIFY_BAD;
787 }
788 return NOTIFY_DONE;
789}
790
791/* the caller must held the addrs lock */
792static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
793{
794 struct ipvl_addr *addr;
795
796 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
797 if (!addr)
798 return -ENOMEM;
799
800 addr->master = ipvlan;
801 if (!is_v6) {
802 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
803 addr->atype = IPVL_IPV4;
804#if IS_ENABLED(CONFIG_IPV6)
805 } else {
806 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
807 addr->atype = IPVL_IPV6;
808#endif
809 }
810
811 list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
812
813 /* If the interface is not up, the address will be added to the hash
814 * list by ipvlan_open.
815 */
816 if (netif_running(ipvlan->dev))
817 ipvlan_ht_addr_add(ipvlan, addr);
818
819 return 0;
820}
821
822static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
823{
824 struct ipvl_addr *addr;
825
826 spin_lock_bh(&ipvlan->addrs_lock);
827 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
828 if (!addr) {
829 spin_unlock_bh(&ipvlan->addrs_lock);
830 return;
831 }
832
833 ipvlan_ht_addr_del(addr);
834 list_del_rcu(&addr->anode);
835 spin_unlock_bh(&ipvlan->addrs_lock);
836 kfree_rcu(addr, rcu);
837}
838
839static bool ipvlan_is_valid_dev(const struct net_device *dev)
840{
841 struct ipvl_dev *ipvlan = netdev_priv(dev);
842
843 if (!netif_is_ipvlan(dev))
844 return false;
845
846 if (!ipvlan || !ipvlan->port)
847 return false;
848
849 return true;
850}
851
852#if IS_ENABLED(CONFIG_IPV6)
853static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
854{
855 int ret = -EINVAL;
856
857 spin_lock_bh(&ipvlan->addrs_lock);
858 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
859 netif_err(ipvlan, ifup, ipvlan->dev,
860 "Failed to add IPv6=%pI6c addr for %s intf\n",
861 ip6_addr, ipvlan->dev->name);
862 else
863 ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
864 spin_unlock_bh(&ipvlan->addrs_lock);
865 return ret;
866}
867
868static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
869{
870 return ipvlan_del_addr(ipvlan, ip6_addr, true);
871}
872
873static int ipvlan_addr6_event(struct notifier_block *unused,
874 unsigned long event, void *ptr)
875{
876 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
877 struct net_device *dev = (struct net_device *)if6->idev->dev;
878 struct ipvl_dev *ipvlan = netdev_priv(dev);
879
880 if (!ipvlan_is_valid_dev(dev))
881 return NOTIFY_DONE;
882
883 switch (event) {
884 case NETDEV_UP:
885 if (ipvlan_add_addr6(ipvlan, &if6->addr))
886 return NOTIFY_BAD;
887 break;
888
889 case NETDEV_DOWN:
890 ipvlan_del_addr6(ipvlan, &if6->addr);
891 break;
892 }
893
894 return NOTIFY_OK;
895}
896
897static int ipvlan_addr6_validator_event(struct notifier_block *unused,
898 unsigned long event, void *ptr)
899{
900 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
901 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
902 struct ipvl_dev *ipvlan = netdev_priv(dev);
903
904 if (!ipvlan_is_valid_dev(dev))
905 return NOTIFY_DONE;
906
907 switch (event) {
908 case NETDEV_UP:
909 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
910 NL_SET_ERR_MSG(i6vi->extack,
911 "Address already assigned to an ipvlan device");
912 return notifier_from_errno(-EADDRINUSE);
913 }
914 break;
915 }
916
917 return NOTIFY_OK;
918}
919#endif
920
921static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
922{
923 int ret = -EINVAL;
924
925 spin_lock_bh(&ipvlan->addrs_lock);
926 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
927 netif_err(ipvlan, ifup, ipvlan->dev,
928 "Failed to add IPv4=%pI4 on %s intf.\n",
929 ip4_addr, ipvlan->dev->name);
930 else
931 ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
932 spin_unlock_bh(&ipvlan->addrs_lock);
933 return ret;
934}
935
936static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
937{
938 return ipvlan_del_addr(ipvlan, ip4_addr, false);
939}
940
941static int ipvlan_addr4_event(struct notifier_block *unused,
942 unsigned long event, void *ptr)
943{
944 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
945 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
946 struct ipvl_dev *ipvlan = netdev_priv(dev);
947 struct in_addr ip4_addr;
948
949 if (!ipvlan_is_valid_dev(dev))
950 return NOTIFY_DONE;
951
952 switch (event) {
953 case NETDEV_UP:
954 ip4_addr.s_addr = if4->ifa_address;
955 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
956 return NOTIFY_BAD;
957 break;
958
959 case NETDEV_DOWN:
960 ip4_addr.s_addr = if4->ifa_address;
961 ipvlan_del_addr4(ipvlan, &ip4_addr);
962 break;
963 }
964
965 return NOTIFY_OK;
966}
967
968static int ipvlan_addr4_validator_event(struct notifier_block *unused,
969 unsigned long event, void *ptr)
970{
971 struct in_validator_info *ivi = (struct in_validator_info *)ptr;
972 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
973 struct ipvl_dev *ipvlan = netdev_priv(dev);
974
975 if (!ipvlan_is_valid_dev(dev))
976 return NOTIFY_DONE;
977
978 switch (event) {
979 case NETDEV_UP:
980 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
981 NL_SET_ERR_MSG(ivi->extack,
982 "Address already assigned to an ipvlan device");
983 return notifier_from_errno(-EADDRINUSE);
984 }
985 break;
986 }
987
988 return NOTIFY_OK;
989}
990
991static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
992 .notifier_call = ipvlan_addr4_event,
993};
994
995static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
996 .notifier_call = ipvlan_addr4_validator_event,
997};
998
999static struct notifier_block ipvlan_notifier_block __read_mostly = {
1000 .notifier_call = ipvlan_device_event,
1001};
1002
1003#if IS_ENABLED(CONFIG_IPV6)
1004static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1005 .notifier_call = ipvlan_addr6_event,
1006};
1007
1008static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1009 .notifier_call = ipvlan_addr6_validator_event,
1010};
1011#endif
1012
1013static int __init ipvlan_init_module(void)
1014{
1015 int err;
1016
1017 ipvlan_init_secret();
1018 register_netdevice_notifier(&ipvlan_notifier_block);
1019#if IS_ENABLED(CONFIG_IPV6)
1020 register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1021 register_inet6addr_validator_notifier(
1022 &ipvlan_addr6_vtor_notifier_block);
1023#endif
1024 register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1025 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1026
1027 err = ipvlan_l3s_init();
1028 if (err < 0)
1029 goto error;
1030
1031 err = ipvlan_link_register(&ipvlan_link_ops);
1032 if (err < 0) {
1033 ipvlan_l3s_cleanup();
1034 goto error;
1035 }
1036
1037 return 0;
1038error:
1039 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1040 unregister_inetaddr_validator_notifier(
1041 &ipvlan_addr4_vtor_notifier_block);
1042#if IS_ENABLED(CONFIG_IPV6)
1043 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1044 unregister_inet6addr_validator_notifier(
1045 &ipvlan_addr6_vtor_notifier_block);
1046#endif
1047 unregister_netdevice_notifier(&ipvlan_notifier_block);
1048 return err;
1049}
1050
1051static void __exit ipvlan_cleanup_module(void)
1052{
1053 rtnl_link_unregister(&ipvlan_link_ops);
1054 ipvlan_l3s_cleanup();
1055 unregister_netdevice_notifier(&ipvlan_notifier_block);
1056 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1057 unregister_inetaddr_validator_notifier(
1058 &ipvlan_addr4_vtor_notifier_block);
1059#if IS_ENABLED(CONFIG_IPV6)
1060 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1061 unregister_inet6addr_validator_notifier(
1062 &ipvlan_addr6_vtor_notifier_block);
1063#endif
1064}
1065
1066module_init(ipvlan_init_module);
1067module_exit(ipvlan_cleanup_module);
1068
1069MODULE_LICENSE("GPL");
1070MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1071MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1072MODULE_ALIAS_RTNL_LINK("ipvlan");