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_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
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 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
165static 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
179static 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
198static 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
215static 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
238static 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
253static 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
262static 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
289static 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
330static 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
338static 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
348static 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
355static 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
370static 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
385static const struct header_ops ipvlan_header_ops = {
386 .create = ipvlan_hard_header,
387 .parse = eth_header_parse,
388 .cache = eth_header_cache,
389 .cache_update = eth_header_cache_update,
390 .parse_protocol = eth_header_parse_protocol,
391};
392
393static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
394{
395 ipvlan->dev->mtu = dev->mtu;
396}
397
398static bool netif_is_ipvlan(const struct net_device *dev)
399{
400 /* both ipvlan and ipvtap devices use the same netdev_ops */
401 return dev->netdev_ops == &ipvlan_netdev_ops;
402}
403
404static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
405 struct ethtool_link_ksettings *cmd)
406{
407 const struct ipvl_dev *ipvlan = netdev_priv(dev);
408
409 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
410}
411
412static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
413 struct ethtool_drvinfo *drvinfo)
414{
415 strscpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
416 strscpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
417}
418
419static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
420{
421 const struct ipvl_dev *ipvlan = netdev_priv(dev);
422
423 return ipvlan->msg_enable;
424}
425
426static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
427{
428 struct ipvl_dev *ipvlan = netdev_priv(dev);
429
430 ipvlan->msg_enable = value;
431}
432
433static const struct ethtool_ops ipvlan_ethtool_ops = {
434 .get_link = ethtool_op_get_link,
435 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings,
436 .get_drvinfo = ipvlan_ethtool_get_drvinfo,
437 .get_msglevel = ipvlan_ethtool_get_msglevel,
438 .set_msglevel = ipvlan_ethtool_set_msglevel,
439};
440
441static int ipvlan_nl_changelink(struct net_device *dev,
442 struct nlattr *tb[], struct nlattr *data[],
443 struct netlink_ext_ack *extack)
444{
445 struct ipvl_dev *ipvlan = netdev_priv(dev);
446 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
447 int err = 0;
448
449 if (!data)
450 return 0;
451 if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN))
452 return -EPERM;
453
454 if (data[IFLA_IPVLAN_MODE]) {
455 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
456
457 err = ipvlan_set_port_mode(port, nmode, extack);
458 }
459
460 if (!err && data[IFLA_IPVLAN_FLAGS]) {
461 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
462
463 if (flags & IPVLAN_F_PRIVATE)
464 ipvlan_mark_private(port);
465 else
466 ipvlan_clear_private(port);
467
468 if (flags & IPVLAN_F_VEPA)
469 ipvlan_mark_vepa(port);
470 else
471 ipvlan_clear_vepa(port);
472 }
473
474 return err;
475}
476
477static size_t ipvlan_nl_getsize(const struct net_device *dev)
478{
479 return (0
480 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
481 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
482 );
483}
484
485static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
486 struct netlink_ext_ack *extack)
487{
488 if (!data)
489 return 0;
490
491 if (data[IFLA_IPVLAN_MODE]) {
492 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
493
494 if (mode >= IPVLAN_MODE_MAX)
495 return -EINVAL;
496 }
497 if (data[IFLA_IPVLAN_FLAGS]) {
498 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
499
500 /* Only two bits are used at this moment. */
501 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
502 return -EINVAL;
503 /* Also both flags can't be active at the same time. */
504 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
505 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
506 return -EINVAL;
507 }
508
509 return 0;
510}
511
512static int ipvlan_nl_fillinfo(struct sk_buff *skb,
513 const struct net_device *dev)
514{
515 struct ipvl_dev *ipvlan = netdev_priv(dev);
516 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
517 int ret = -EINVAL;
518
519 if (!port)
520 goto err;
521
522 ret = -EMSGSIZE;
523 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
524 goto err;
525 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
526 goto err;
527
528 return 0;
529
530err:
531 return ret;
532}
533
534int ipvlan_link_new(struct net *src_net, struct net_device *dev,
535 struct nlattr *tb[], struct nlattr *data[],
536 struct netlink_ext_ack *extack)
537{
538 struct ipvl_dev *ipvlan = netdev_priv(dev);
539 struct ipvl_port *port;
540 struct net_device *phy_dev;
541 int err;
542 u16 mode = IPVLAN_MODE_L3;
543
544 if (!tb[IFLA_LINK])
545 return -EINVAL;
546
547 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
548 if (!phy_dev)
549 return -ENODEV;
550
551 if (netif_is_ipvlan(phy_dev)) {
552 struct ipvl_dev *tmp = netdev_priv(phy_dev);
553
554 phy_dev = tmp->phy_dev;
555 if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN))
556 return -EPERM;
557 } else if (!netif_is_ipvlan_port(phy_dev)) {
558 /* Exit early if the underlying link is invalid or busy */
559 if (phy_dev->type != ARPHRD_ETHER ||
560 phy_dev->flags & IFF_LOOPBACK) {
561 netdev_err(phy_dev,
562 "Master is either lo or non-ether device\n");
563 return -EINVAL;
564 }
565
566 if (netdev_is_rx_handler_busy(phy_dev)) {
567 netdev_err(phy_dev, "Device is already in use.\n");
568 return -EBUSY;
569 }
570 }
571
572 ipvlan->phy_dev = phy_dev;
573 ipvlan->dev = dev;
574 ipvlan->sfeatures = IPVLAN_FEATURES;
575 if (!tb[IFLA_MTU])
576 ipvlan_adjust_mtu(ipvlan, phy_dev);
577 INIT_LIST_HEAD(&ipvlan->addrs);
578 spin_lock_init(&ipvlan->addrs_lock);
579
580 /* TODO Probably put random address here to be presented to the
581 * world but keep using the physical-dev address for the outgoing
582 * packets.
583 */
584 eth_hw_addr_set(dev, phy_dev->dev_addr);
585
586 dev->priv_flags |= IFF_NO_RX_HANDLER;
587
588 err = register_netdevice(dev);
589 if (err < 0)
590 return err;
591
592 /* ipvlan_init() would have created the port, if required */
593 port = ipvlan_port_get_rtnl(phy_dev);
594 ipvlan->port = port;
595
596 /* If the port-id base is at the MAX value, then wrap it around and
597 * begin from 0x1 again. This may be due to a busy system where lots
598 * of slaves are getting created and deleted.
599 */
600 if (port->dev_id_start == 0xFFFE)
601 port->dev_id_start = 0x1;
602
603 /* Since L2 address is shared among all IPvlan slaves including
604 * master, use unique 16 bit dev-ids to differentiate among them.
605 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
606 * slave link [see addrconf_ifid_eui48()].
607 */
608 err = ida_alloc_range(&port->ida, port->dev_id_start, 0xFFFD,
609 GFP_KERNEL);
610 if (err < 0)
611 err = ida_alloc_range(&port->ida, 0x1, port->dev_id_start - 1,
612 GFP_KERNEL);
613 if (err < 0)
614 goto unregister_netdev;
615 dev->dev_id = err;
616
617 /* Increment id-base to the next slot for the future assignment */
618 port->dev_id_start = err + 1;
619
620 err = netdev_upper_dev_link(phy_dev, dev, extack);
621 if (err)
622 goto remove_ida;
623
624 /* Flags are per port and latest update overrides. User has
625 * to be consistent in setting it just like the mode attribute.
626 */
627 if (data && data[IFLA_IPVLAN_FLAGS])
628 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
629
630 if (data && data[IFLA_IPVLAN_MODE])
631 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
632
633 err = ipvlan_set_port_mode(port, mode, extack);
634 if (err)
635 goto unlink_netdev;
636
637 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
638 netif_stacked_transfer_operstate(phy_dev, dev);
639 return 0;
640
641unlink_netdev:
642 netdev_upper_dev_unlink(phy_dev, dev);
643remove_ida:
644 ida_free(&port->ida, dev->dev_id);
645unregister_netdev:
646 unregister_netdevice(dev);
647 return err;
648}
649EXPORT_SYMBOL_GPL(ipvlan_link_new);
650
651void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
652{
653 struct ipvl_dev *ipvlan = netdev_priv(dev);
654 struct ipvl_addr *addr, *next;
655
656 spin_lock_bh(&ipvlan->addrs_lock);
657 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
658 ipvlan_ht_addr_del(addr);
659 list_del_rcu(&addr->anode);
660 kfree_rcu(addr, rcu);
661 }
662 spin_unlock_bh(&ipvlan->addrs_lock);
663
664 ida_free(&ipvlan->port->ida, dev->dev_id);
665 list_del_rcu(&ipvlan->pnode);
666 unregister_netdevice_queue(dev, head);
667 netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
668}
669EXPORT_SYMBOL_GPL(ipvlan_link_delete);
670
671void ipvlan_link_setup(struct net_device *dev)
672{
673 ether_setup(dev);
674
675 dev->max_mtu = ETH_MAX_MTU;
676 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
677 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
678 dev->netdev_ops = &ipvlan_netdev_ops;
679 dev->needs_free_netdev = true;
680 dev->header_ops = &ipvlan_header_ops;
681 dev->ethtool_ops = &ipvlan_ethtool_ops;
682}
683EXPORT_SYMBOL_GPL(ipvlan_link_setup);
684
685static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
686{
687 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
688 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
689};
690
691static struct net *ipvlan_get_link_net(const struct net_device *dev)
692{
693 struct ipvl_dev *ipvlan = netdev_priv(dev);
694
695 return dev_net(ipvlan->phy_dev);
696}
697
698static struct rtnl_link_ops ipvlan_link_ops = {
699 .kind = "ipvlan",
700 .priv_size = sizeof(struct ipvl_dev),
701
702 .setup = ipvlan_link_setup,
703 .newlink = ipvlan_link_new,
704 .dellink = ipvlan_link_delete,
705 .get_link_net = ipvlan_get_link_net,
706};
707
708int ipvlan_link_register(struct rtnl_link_ops *ops)
709{
710 ops->get_size = ipvlan_nl_getsize;
711 ops->policy = ipvlan_nl_policy;
712 ops->validate = ipvlan_nl_validate;
713 ops->fill_info = ipvlan_nl_fillinfo;
714 ops->changelink = ipvlan_nl_changelink;
715 ops->maxtype = IFLA_IPVLAN_MAX;
716 return rtnl_link_register(ops);
717}
718EXPORT_SYMBOL_GPL(ipvlan_link_register);
719
720static int ipvlan_device_event(struct notifier_block *unused,
721 unsigned long event, void *ptr)
722{
723 struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr);
724 struct netdev_notifier_pre_changeaddr_info *prechaddr_info;
725 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
726 struct ipvl_dev *ipvlan, *next;
727 struct ipvl_port *port;
728 LIST_HEAD(lst_kill);
729 int err;
730
731 if (!netif_is_ipvlan_port(dev))
732 return NOTIFY_DONE;
733
734 port = ipvlan_port_get_rtnl(dev);
735
736 switch (event) {
737 case NETDEV_UP:
738 case NETDEV_CHANGE:
739 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
740 netif_stacked_transfer_operstate(ipvlan->phy_dev,
741 ipvlan->dev);
742 break;
743
744 case NETDEV_REGISTER: {
745 struct net *oldnet, *newnet = dev_net(dev);
746
747 oldnet = read_pnet(&port->pnet);
748 if (net_eq(newnet, oldnet))
749 break;
750
751 write_pnet(&port->pnet, newnet);
752
753 if (port->mode == IPVLAN_MODE_L3S)
754 ipvlan_migrate_l3s_hook(oldnet, newnet);
755 break;
756 }
757 case NETDEV_UNREGISTER:
758 if (dev->reg_state != NETREG_UNREGISTERING)
759 break;
760
761 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
762 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
763 &lst_kill);
764 unregister_netdevice_many(&lst_kill);
765 break;
766
767 case NETDEV_FEAT_CHANGE:
768 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
769 netif_inherit_tso_max(ipvlan->dev, dev);
770 netdev_update_features(ipvlan->dev);
771 }
772 break;
773
774 case NETDEV_CHANGEMTU:
775 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
776 ipvlan_adjust_mtu(ipvlan, dev);
777 break;
778
779 case NETDEV_PRE_CHANGEADDR:
780 prechaddr_info = ptr;
781 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
782 err = dev_pre_changeaddr_notify(ipvlan->dev,
783 prechaddr_info->dev_addr,
784 extack);
785 if (err)
786 return notifier_from_errno(err);
787 }
788 break;
789
790 case NETDEV_CHANGEADDR:
791 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
792 eth_hw_addr_set(ipvlan->dev, dev->dev_addr);
793 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
794 }
795 break;
796
797 case NETDEV_PRE_TYPE_CHANGE:
798 /* Forbid underlying device to change its type. */
799 return NOTIFY_BAD;
800 }
801 return NOTIFY_DONE;
802}
803
804/* the caller must held the addrs lock */
805static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
806{
807 struct ipvl_addr *addr;
808
809 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
810 if (!addr)
811 return -ENOMEM;
812
813 addr->master = ipvlan;
814 if (!is_v6) {
815 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
816 addr->atype = IPVL_IPV4;
817#if IS_ENABLED(CONFIG_IPV6)
818 } else {
819 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
820 addr->atype = IPVL_IPV6;
821#endif
822 }
823
824 list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
825
826 /* If the interface is not up, the address will be added to the hash
827 * list by ipvlan_open.
828 */
829 if (netif_running(ipvlan->dev))
830 ipvlan_ht_addr_add(ipvlan, addr);
831
832 return 0;
833}
834
835static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
836{
837 struct ipvl_addr *addr;
838
839 spin_lock_bh(&ipvlan->addrs_lock);
840 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
841 if (!addr) {
842 spin_unlock_bh(&ipvlan->addrs_lock);
843 return;
844 }
845
846 ipvlan_ht_addr_del(addr);
847 list_del_rcu(&addr->anode);
848 spin_unlock_bh(&ipvlan->addrs_lock);
849 kfree_rcu(addr, rcu);
850}
851
852static bool ipvlan_is_valid_dev(const struct net_device *dev)
853{
854 struct ipvl_dev *ipvlan = netdev_priv(dev);
855
856 if (!netif_is_ipvlan(dev))
857 return false;
858
859 if (!ipvlan || !ipvlan->port)
860 return false;
861
862 return true;
863}
864
865#if IS_ENABLED(CONFIG_IPV6)
866static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
867{
868 int ret = -EINVAL;
869
870 spin_lock_bh(&ipvlan->addrs_lock);
871 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
872 netif_err(ipvlan, ifup, ipvlan->dev,
873 "Failed to add IPv6=%pI6c addr for %s intf\n",
874 ip6_addr, ipvlan->dev->name);
875 else
876 ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
877 spin_unlock_bh(&ipvlan->addrs_lock);
878 return ret;
879}
880
881static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
882{
883 return ipvlan_del_addr(ipvlan, ip6_addr, true);
884}
885
886static int ipvlan_addr6_event(struct notifier_block *unused,
887 unsigned long event, void *ptr)
888{
889 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
890 struct net_device *dev = (struct net_device *)if6->idev->dev;
891 struct ipvl_dev *ipvlan = netdev_priv(dev);
892
893 if (!ipvlan_is_valid_dev(dev))
894 return NOTIFY_DONE;
895
896 switch (event) {
897 case NETDEV_UP:
898 if (ipvlan_add_addr6(ipvlan, &if6->addr))
899 return NOTIFY_BAD;
900 break;
901
902 case NETDEV_DOWN:
903 ipvlan_del_addr6(ipvlan, &if6->addr);
904 break;
905 }
906
907 return NOTIFY_OK;
908}
909
910static int ipvlan_addr6_validator_event(struct notifier_block *unused,
911 unsigned long event, void *ptr)
912{
913 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
914 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
915 struct ipvl_dev *ipvlan = netdev_priv(dev);
916
917 if (!ipvlan_is_valid_dev(dev))
918 return NOTIFY_DONE;
919
920 switch (event) {
921 case NETDEV_UP:
922 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
923 NL_SET_ERR_MSG(i6vi->extack,
924 "Address already assigned to an ipvlan device");
925 return notifier_from_errno(-EADDRINUSE);
926 }
927 break;
928 }
929
930 return NOTIFY_OK;
931}
932#endif
933
934static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
935{
936 int ret = -EINVAL;
937
938 spin_lock_bh(&ipvlan->addrs_lock);
939 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
940 netif_err(ipvlan, ifup, ipvlan->dev,
941 "Failed to add IPv4=%pI4 on %s intf.\n",
942 ip4_addr, ipvlan->dev->name);
943 else
944 ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
945 spin_unlock_bh(&ipvlan->addrs_lock);
946 return ret;
947}
948
949static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
950{
951 return ipvlan_del_addr(ipvlan, ip4_addr, false);
952}
953
954static int ipvlan_addr4_event(struct notifier_block *unused,
955 unsigned long event, void *ptr)
956{
957 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
958 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
959 struct ipvl_dev *ipvlan = netdev_priv(dev);
960 struct in_addr ip4_addr;
961
962 if (!ipvlan_is_valid_dev(dev))
963 return NOTIFY_DONE;
964
965 switch (event) {
966 case NETDEV_UP:
967 ip4_addr.s_addr = if4->ifa_address;
968 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
969 return NOTIFY_BAD;
970 break;
971
972 case NETDEV_DOWN:
973 ip4_addr.s_addr = if4->ifa_address;
974 ipvlan_del_addr4(ipvlan, &ip4_addr);
975 break;
976 }
977
978 return NOTIFY_OK;
979}
980
981static int ipvlan_addr4_validator_event(struct notifier_block *unused,
982 unsigned long event, void *ptr)
983{
984 struct in_validator_info *ivi = (struct in_validator_info *)ptr;
985 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
986 struct ipvl_dev *ipvlan = netdev_priv(dev);
987
988 if (!ipvlan_is_valid_dev(dev))
989 return NOTIFY_DONE;
990
991 switch (event) {
992 case NETDEV_UP:
993 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
994 NL_SET_ERR_MSG(ivi->extack,
995 "Address already assigned to an ipvlan device");
996 return notifier_from_errno(-EADDRINUSE);
997 }
998 break;
999 }
1000
1001 return NOTIFY_OK;
1002}
1003
1004static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
1005 .notifier_call = ipvlan_addr4_event,
1006};
1007
1008static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
1009 .notifier_call = ipvlan_addr4_validator_event,
1010};
1011
1012static struct notifier_block ipvlan_notifier_block __read_mostly = {
1013 .notifier_call = ipvlan_device_event,
1014};
1015
1016#if IS_ENABLED(CONFIG_IPV6)
1017static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1018 .notifier_call = ipvlan_addr6_event,
1019};
1020
1021static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1022 .notifier_call = ipvlan_addr6_validator_event,
1023};
1024#endif
1025
1026static int __init ipvlan_init_module(void)
1027{
1028 int err;
1029
1030 ipvlan_init_secret();
1031 register_netdevice_notifier(&ipvlan_notifier_block);
1032#if IS_ENABLED(CONFIG_IPV6)
1033 register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1034 register_inet6addr_validator_notifier(
1035 &ipvlan_addr6_vtor_notifier_block);
1036#endif
1037 register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1038 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1039
1040 err = ipvlan_l3s_init();
1041 if (err < 0)
1042 goto error;
1043
1044 err = ipvlan_link_register(&ipvlan_link_ops);
1045 if (err < 0) {
1046 ipvlan_l3s_cleanup();
1047 goto error;
1048 }
1049
1050 return 0;
1051error:
1052 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1053 unregister_inetaddr_validator_notifier(
1054 &ipvlan_addr4_vtor_notifier_block);
1055#if IS_ENABLED(CONFIG_IPV6)
1056 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1057 unregister_inet6addr_validator_notifier(
1058 &ipvlan_addr6_vtor_notifier_block);
1059#endif
1060 unregister_netdevice_notifier(&ipvlan_notifier_block);
1061 return err;
1062}
1063
1064static void __exit ipvlan_cleanup_module(void)
1065{
1066 rtnl_link_unregister(&ipvlan_link_ops);
1067 ipvlan_l3s_cleanup();
1068 unregister_netdevice_notifier(&ipvlan_notifier_block);
1069 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1070 unregister_inetaddr_validator_notifier(
1071 &ipvlan_addr4_vtor_notifier_block);
1072#if IS_ENABLED(CONFIG_IPV6)
1073 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1074 unregister_inet6addr_validator_notifier(
1075 &ipvlan_addr6_vtor_notifier_block);
1076#endif
1077}
1078
1079module_init(ipvlan_init_module);
1080module_exit(ipvlan_cleanup_module);
1081
1082MODULE_LICENSE("GPL");
1083MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1084MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1085MODULE_ALIAS_RTNL_LINK("ipvlan");
1/* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
2 *
3 * This program is free software; you can redistribute it and/or
4 * modify it under the terms of the GNU General Public License as
5 * published by the Free Software Foundation; either version 2 of
6 * the License, or (at your option) any later version.
7 *
8 */
9
10#include "ipvlan.h"
11
12static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
13{
14 ipvlan->dev->mtu = dev->mtu - ipvlan->mtu_adj;
15}
16
17static void ipvlan_set_port_mode(struct ipvl_port *port, u16 nval)
18{
19 struct ipvl_dev *ipvlan;
20
21 if (port->mode != nval) {
22 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
23 if (nval == IPVLAN_MODE_L3)
24 ipvlan->dev->flags |= IFF_NOARP;
25 else
26 ipvlan->dev->flags &= ~IFF_NOARP;
27 }
28 port->mode = nval;
29 }
30}
31
32static int ipvlan_port_create(struct net_device *dev)
33{
34 struct ipvl_port *port;
35 int err, idx;
36
37 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) {
38 netdev_err(dev, "Master is either lo or non-ether device\n");
39 return -EINVAL;
40 }
41
42 if (netif_is_macvlan_port(dev)) {
43 netdev_err(dev, "Master is a macvlan port.\n");
44 return -EBUSY;
45 }
46
47 port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL);
48 if (!port)
49 return -ENOMEM;
50
51 port->dev = dev;
52 port->mode = IPVLAN_MODE_L3;
53 INIT_LIST_HEAD(&port->ipvlans);
54 for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++)
55 INIT_HLIST_HEAD(&port->hlhead[idx]);
56
57 skb_queue_head_init(&port->backlog);
58 INIT_WORK(&port->wq, ipvlan_process_multicast);
59
60 err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port);
61 if (err)
62 goto err;
63
64 dev->priv_flags |= IFF_IPVLAN_MASTER;
65 return 0;
66
67err:
68 kfree_rcu(port, rcu);
69 return err;
70}
71
72static void ipvlan_port_destroy(struct net_device *dev)
73{
74 struct ipvl_port *port = ipvlan_port_get_rtnl(dev);
75
76 dev->priv_flags &= ~IFF_IPVLAN_MASTER;
77 netdev_rx_handler_unregister(dev);
78 cancel_work_sync(&port->wq);
79 __skb_queue_purge(&port->backlog);
80 kfree_rcu(port, rcu);
81}
82
83/* ipvlan network devices have devices nesting below it and are a special
84 * "super class" of normal network devices; split their locks off into a
85 * separate class since they always nest.
86 */
87static struct lock_class_key ipvlan_netdev_xmit_lock_key;
88static struct lock_class_key ipvlan_netdev_addr_lock_key;
89
90#define IPVLAN_FEATURES \
91 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
92 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
93 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
94 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
95
96#define IPVLAN_STATE_MASK \
97 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
98
99static void ipvlan_set_lockdep_class_one(struct net_device *dev,
100 struct netdev_queue *txq,
101 void *_unused)
102{
103 lockdep_set_class(&txq->_xmit_lock, &ipvlan_netdev_xmit_lock_key);
104}
105
106static void ipvlan_set_lockdep_class(struct net_device *dev)
107{
108 lockdep_set_class(&dev->addr_list_lock, &ipvlan_netdev_addr_lock_key);
109 netdev_for_each_tx_queue(dev, ipvlan_set_lockdep_class_one, NULL);
110}
111
112static int ipvlan_init(struct net_device *dev)
113{
114 struct ipvl_dev *ipvlan = netdev_priv(dev);
115 const struct net_device *phy_dev = ipvlan->phy_dev;
116
117 dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
118 (phy_dev->state & IPVLAN_STATE_MASK);
119 dev->features = phy_dev->features & IPVLAN_FEATURES;
120 dev->features |= NETIF_F_LLTX;
121 dev->gso_max_size = phy_dev->gso_max_size;
122 dev->gso_max_segs = phy_dev->gso_max_segs;
123 dev->hard_header_len = phy_dev->hard_header_len;
124
125 ipvlan_set_lockdep_class(dev);
126
127 ipvlan->pcpu_stats = alloc_percpu(struct ipvl_pcpu_stats);
128 if (!ipvlan->pcpu_stats)
129 return -ENOMEM;
130
131 return 0;
132}
133
134static void ipvlan_uninit(struct net_device *dev)
135{
136 struct ipvl_dev *ipvlan = netdev_priv(dev);
137 struct ipvl_port *port = ipvlan->port;
138
139 free_percpu(ipvlan->pcpu_stats);
140
141 port->count -= 1;
142 if (!port->count)
143 ipvlan_port_destroy(port->dev);
144}
145
146static int ipvlan_open(struct net_device *dev)
147{
148 struct ipvl_dev *ipvlan = netdev_priv(dev);
149 struct net_device *phy_dev = ipvlan->phy_dev;
150 struct ipvl_addr *addr;
151
152 if (ipvlan->port->mode == IPVLAN_MODE_L3)
153 dev->flags |= IFF_NOARP;
154 else
155 dev->flags &= ~IFF_NOARP;
156
157 list_for_each_entry(addr, &ipvlan->addrs, anode)
158 ipvlan_ht_addr_add(ipvlan, addr);
159
160 return dev_uc_add(phy_dev, phy_dev->dev_addr);
161}
162
163static int ipvlan_stop(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_addr *addr;
168
169 dev_uc_unsync(phy_dev, dev);
170 dev_mc_unsync(phy_dev, dev);
171
172 dev_uc_del(phy_dev, phy_dev->dev_addr);
173
174 list_for_each_entry(addr, &ipvlan->addrs, anode)
175 ipvlan_ht_addr_del(addr);
176
177 return 0;
178}
179
180static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
181 struct net_device *dev)
182{
183 const struct ipvl_dev *ipvlan = netdev_priv(dev);
184 int skblen = skb->len;
185 int ret;
186
187 ret = ipvlan_queue_xmit(skb, dev);
188 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
189 struct ipvl_pcpu_stats *pcptr;
190
191 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
192
193 u64_stats_update_begin(&pcptr->syncp);
194 pcptr->tx_pkts++;
195 pcptr->tx_bytes += skblen;
196 u64_stats_update_end(&pcptr->syncp);
197 } else {
198 this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
199 }
200 return ret;
201}
202
203static netdev_features_t ipvlan_fix_features(struct net_device *dev,
204 netdev_features_t features)
205{
206 struct ipvl_dev *ipvlan = netdev_priv(dev);
207
208 return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES);
209}
210
211static void ipvlan_change_rx_flags(struct net_device *dev, int change)
212{
213 struct ipvl_dev *ipvlan = netdev_priv(dev);
214 struct net_device *phy_dev = ipvlan->phy_dev;
215
216 if (change & IFF_ALLMULTI)
217 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
218}
219
220static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
221{
222 struct ipvl_dev *ipvlan = netdev_priv(dev);
223
224 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
225 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
226 } else {
227 struct netdev_hw_addr *ha;
228 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
229
230 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
231 netdev_for_each_mc_addr(ha, dev)
232 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
233
234 /* Turn-on broadcast bit irrespective of address family,
235 * since broadcast is deferred to a work-queue, hence no
236 * impact on fast-path processing.
237 */
238 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
239
240 bitmap_copy(ipvlan->mac_filters, mc_filters,
241 IPVLAN_MAC_FILTER_SIZE);
242 }
243 dev_uc_sync(ipvlan->phy_dev, dev);
244 dev_mc_sync(ipvlan->phy_dev, dev);
245}
246
247static struct rtnl_link_stats64 *ipvlan_get_stats64(struct net_device *dev,
248 struct rtnl_link_stats64 *s)
249{
250 struct ipvl_dev *ipvlan = netdev_priv(dev);
251
252 if (ipvlan->pcpu_stats) {
253 struct ipvl_pcpu_stats *pcptr;
254 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
255 u32 rx_errs = 0, tx_drps = 0;
256 u32 strt;
257 int idx;
258
259 for_each_possible_cpu(idx) {
260 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
261 do {
262 strt= u64_stats_fetch_begin_irq(&pcptr->syncp);
263 rx_pkts = pcptr->rx_pkts;
264 rx_bytes = pcptr->rx_bytes;
265 rx_mcast = pcptr->rx_mcast;
266 tx_pkts = pcptr->tx_pkts;
267 tx_bytes = pcptr->tx_bytes;
268 } while (u64_stats_fetch_retry_irq(&pcptr->syncp,
269 strt));
270
271 s->rx_packets += rx_pkts;
272 s->rx_bytes += rx_bytes;
273 s->multicast += rx_mcast;
274 s->tx_packets += tx_pkts;
275 s->tx_bytes += tx_bytes;
276
277 /* u32 values are updated without syncp protection. */
278 rx_errs += pcptr->rx_errs;
279 tx_drps += pcptr->tx_drps;
280 }
281 s->rx_errors = rx_errs;
282 s->rx_dropped = rx_errs;
283 s->tx_dropped = tx_drps;
284 }
285 return s;
286}
287
288static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
289{
290 struct ipvl_dev *ipvlan = netdev_priv(dev);
291 struct net_device *phy_dev = ipvlan->phy_dev;
292
293 return vlan_vid_add(phy_dev, proto, vid);
294}
295
296static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
297 u16 vid)
298{
299 struct ipvl_dev *ipvlan = netdev_priv(dev);
300 struct net_device *phy_dev = ipvlan->phy_dev;
301
302 vlan_vid_del(phy_dev, proto, vid);
303 return 0;
304}
305
306static int ipvlan_get_iflink(const struct net_device *dev)
307{
308 struct ipvl_dev *ipvlan = netdev_priv(dev);
309
310 return ipvlan->phy_dev->ifindex;
311}
312
313static const struct net_device_ops ipvlan_netdev_ops = {
314 .ndo_init = ipvlan_init,
315 .ndo_uninit = ipvlan_uninit,
316 .ndo_open = ipvlan_open,
317 .ndo_stop = ipvlan_stop,
318 .ndo_start_xmit = ipvlan_start_xmit,
319 .ndo_fix_features = ipvlan_fix_features,
320 .ndo_change_rx_flags = ipvlan_change_rx_flags,
321 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter,
322 .ndo_get_stats64 = ipvlan_get_stats64,
323 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid,
324 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid,
325 .ndo_get_iflink = ipvlan_get_iflink,
326};
327
328static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
329 unsigned short type, const void *daddr,
330 const void *saddr, unsigned len)
331{
332 const struct ipvl_dev *ipvlan = netdev_priv(dev);
333 struct net_device *phy_dev = ipvlan->phy_dev;
334
335 /* TODO Probably use a different field than dev_addr so that the
336 * mac-address on the virtual device is portable and can be carried
337 * while the packets use the mac-addr on the physical device.
338 */
339 return dev_hard_header(skb, phy_dev, type, daddr,
340 saddr ? : dev->dev_addr, len);
341}
342
343static const struct header_ops ipvlan_header_ops = {
344 .create = ipvlan_hard_header,
345 .parse = eth_header_parse,
346 .cache = eth_header_cache,
347 .cache_update = eth_header_cache_update,
348};
349
350static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
351 struct ethtool_link_ksettings *cmd)
352{
353 const struct ipvl_dev *ipvlan = netdev_priv(dev);
354
355 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
356}
357
358static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
359 struct ethtool_drvinfo *drvinfo)
360{
361 strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
362 strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
363}
364
365static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
366{
367 const struct ipvl_dev *ipvlan = netdev_priv(dev);
368
369 return ipvlan->msg_enable;
370}
371
372static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
373{
374 struct ipvl_dev *ipvlan = netdev_priv(dev);
375
376 ipvlan->msg_enable = value;
377}
378
379static const struct ethtool_ops ipvlan_ethtool_ops = {
380 .get_link = ethtool_op_get_link,
381 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings,
382 .get_drvinfo = ipvlan_ethtool_get_drvinfo,
383 .get_msglevel = ipvlan_ethtool_get_msglevel,
384 .set_msglevel = ipvlan_ethtool_set_msglevel,
385};
386
387static int ipvlan_nl_changelink(struct net_device *dev,
388 struct nlattr *tb[], struct nlattr *data[])
389{
390 struct ipvl_dev *ipvlan = netdev_priv(dev);
391 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
392
393 if (data && data[IFLA_IPVLAN_MODE]) {
394 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
395
396 ipvlan_set_port_mode(port, nmode);
397 }
398 return 0;
399}
400
401static size_t ipvlan_nl_getsize(const struct net_device *dev)
402{
403 return (0
404 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
405 );
406}
407
408static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[])
409{
410 if (data && data[IFLA_IPVLAN_MODE]) {
411 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
412
413 if (mode < IPVLAN_MODE_L2 || mode >= IPVLAN_MODE_MAX)
414 return -EINVAL;
415 }
416 return 0;
417}
418
419static int ipvlan_nl_fillinfo(struct sk_buff *skb,
420 const struct net_device *dev)
421{
422 struct ipvl_dev *ipvlan = netdev_priv(dev);
423 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
424 int ret = -EINVAL;
425
426 if (!port)
427 goto err;
428
429 ret = -EMSGSIZE;
430 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
431 goto err;
432
433 return 0;
434
435err:
436 return ret;
437}
438
439static int ipvlan_link_new(struct net *src_net, struct net_device *dev,
440 struct nlattr *tb[], struct nlattr *data[])
441{
442 struct ipvl_dev *ipvlan = netdev_priv(dev);
443 struct ipvl_port *port;
444 struct net_device *phy_dev;
445 int err;
446 u16 mode = IPVLAN_MODE_L3;
447
448 if (!tb[IFLA_LINK])
449 return -EINVAL;
450
451 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
452 if (!phy_dev)
453 return -ENODEV;
454
455 if (netif_is_ipvlan(phy_dev)) {
456 struct ipvl_dev *tmp = netdev_priv(phy_dev);
457
458 phy_dev = tmp->phy_dev;
459 } else if (!netif_is_ipvlan_port(phy_dev)) {
460 err = ipvlan_port_create(phy_dev);
461 if (err < 0)
462 return err;
463 }
464
465 if (data && data[IFLA_IPVLAN_MODE])
466 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
467
468 port = ipvlan_port_get_rtnl(phy_dev);
469 ipvlan->phy_dev = phy_dev;
470 ipvlan->dev = dev;
471 ipvlan->port = port;
472 ipvlan->sfeatures = IPVLAN_FEATURES;
473 ipvlan_adjust_mtu(ipvlan, phy_dev);
474 INIT_LIST_HEAD(&ipvlan->addrs);
475
476 /* TODO Probably put random address here to be presented to the
477 * world but keep using the physical-dev address for the outgoing
478 * packets.
479 */
480 memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN);
481
482 dev->priv_flags |= IFF_IPVLAN_SLAVE;
483
484 port->count += 1;
485 err = register_netdevice(dev);
486 if (err < 0)
487 goto ipvlan_destroy_port;
488
489 err = netdev_upper_dev_link(phy_dev, dev);
490 if (err)
491 goto ipvlan_destroy_port;
492
493 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
494 ipvlan_set_port_mode(port, mode);
495
496 netif_stacked_transfer_operstate(phy_dev, dev);
497 return 0;
498
499ipvlan_destroy_port:
500 port->count -= 1;
501 if (!port->count)
502 ipvlan_port_destroy(phy_dev);
503
504 return err;
505}
506
507static void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
508{
509 struct ipvl_dev *ipvlan = netdev_priv(dev);
510 struct ipvl_addr *addr, *next;
511
512 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
513 ipvlan_ht_addr_del(addr);
514 list_del(&addr->anode);
515 kfree_rcu(addr, rcu);
516 }
517
518 list_del_rcu(&ipvlan->pnode);
519 unregister_netdevice_queue(dev, head);
520 netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
521}
522
523static void ipvlan_link_setup(struct net_device *dev)
524{
525 ether_setup(dev);
526
527 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
528 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
529 dev->netdev_ops = &ipvlan_netdev_ops;
530 dev->destructor = free_netdev;
531 dev->header_ops = &ipvlan_header_ops;
532 dev->ethtool_ops = &ipvlan_ethtool_ops;
533}
534
535static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
536{
537 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
538};
539
540static struct rtnl_link_ops ipvlan_link_ops = {
541 .kind = "ipvlan",
542 .priv_size = sizeof(struct ipvl_dev),
543
544 .get_size = ipvlan_nl_getsize,
545 .policy = ipvlan_nl_policy,
546 .validate = ipvlan_nl_validate,
547 .fill_info = ipvlan_nl_fillinfo,
548 .changelink = ipvlan_nl_changelink,
549 .maxtype = IFLA_IPVLAN_MAX,
550
551 .setup = ipvlan_link_setup,
552 .newlink = ipvlan_link_new,
553 .dellink = ipvlan_link_delete,
554};
555
556static int ipvlan_link_register(struct rtnl_link_ops *ops)
557{
558 return rtnl_link_register(ops);
559}
560
561static int ipvlan_device_event(struct notifier_block *unused,
562 unsigned long event, void *ptr)
563{
564 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
565 struct ipvl_dev *ipvlan, *next;
566 struct ipvl_port *port;
567 LIST_HEAD(lst_kill);
568
569 if (!netif_is_ipvlan_port(dev))
570 return NOTIFY_DONE;
571
572 port = ipvlan_port_get_rtnl(dev);
573
574 switch (event) {
575 case NETDEV_CHANGE:
576 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
577 netif_stacked_transfer_operstate(ipvlan->phy_dev,
578 ipvlan->dev);
579 break;
580
581 case NETDEV_UNREGISTER:
582 if (dev->reg_state != NETREG_UNREGISTERING)
583 break;
584
585 list_for_each_entry_safe(ipvlan, next, &port->ipvlans,
586 pnode)
587 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
588 &lst_kill);
589 unregister_netdevice_many(&lst_kill);
590 break;
591
592 case NETDEV_FEAT_CHANGE:
593 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
594 ipvlan->dev->features = dev->features & IPVLAN_FEATURES;
595 ipvlan->dev->gso_max_size = dev->gso_max_size;
596 ipvlan->dev->gso_max_segs = dev->gso_max_segs;
597 netdev_features_change(ipvlan->dev);
598 }
599 break;
600
601 case NETDEV_CHANGEMTU:
602 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
603 ipvlan_adjust_mtu(ipvlan, dev);
604 break;
605
606 case NETDEV_PRE_TYPE_CHANGE:
607 /* Forbid underlying device to change its type. */
608 return NOTIFY_BAD;
609 }
610 return NOTIFY_DONE;
611}
612
613static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
614{
615 struct ipvl_addr *addr;
616
617 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true)) {
618 netif_err(ipvlan, ifup, ipvlan->dev,
619 "Failed to add IPv6=%pI6c addr for %s intf\n",
620 ip6_addr, ipvlan->dev->name);
621 return -EINVAL;
622 }
623 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
624 if (!addr)
625 return -ENOMEM;
626
627 addr->master = ipvlan;
628 memcpy(&addr->ip6addr, ip6_addr, sizeof(struct in6_addr));
629 addr->atype = IPVL_IPV6;
630 list_add_tail(&addr->anode, &ipvlan->addrs);
631
632 /* If the interface is not up, the address will be added to the hash
633 * list by ipvlan_open.
634 */
635 if (netif_running(ipvlan->dev))
636 ipvlan_ht_addr_add(ipvlan, addr);
637
638 return 0;
639}
640
641static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
642{
643 struct ipvl_addr *addr;
644
645 addr = ipvlan_find_addr(ipvlan, ip6_addr, true);
646 if (!addr)
647 return;
648
649 ipvlan_ht_addr_del(addr);
650 list_del(&addr->anode);
651 kfree_rcu(addr, rcu);
652
653 return;
654}
655
656static int ipvlan_addr6_event(struct notifier_block *unused,
657 unsigned long event, void *ptr)
658{
659 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
660 struct net_device *dev = (struct net_device *)if6->idev->dev;
661 struct ipvl_dev *ipvlan = netdev_priv(dev);
662
663 /* FIXME IPv6 autoconf calls us from bh without RTNL */
664 if (in_softirq())
665 return NOTIFY_DONE;
666
667 if (!netif_is_ipvlan(dev))
668 return NOTIFY_DONE;
669
670 if (!ipvlan || !ipvlan->port)
671 return NOTIFY_DONE;
672
673 switch (event) {
674 case NETDEV_UP:
675 if (ipvlan_add_addr6(ipvlan, &if6->addr))
676 return NOTIFY_BAD;
677 break;
678
679 case NETDEV_DOWN:
680 ipvlan_del_addr6(ipvlan, &if6->addr);
681 break;
682 }
683
684 return NOTIFY_OK;
685}
686
687static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
688{
689 struct ipvl_addr *addr;
690
691 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false)) {
692 netif_err(ipvlan, ifup, ipvlan->dev,
693 "Failed to add IPv4=%pI4 on %s intf.\n",
694 ip4_addr, ipvlan->dev->name);
695 return -EINVAL;
696 }
697 addr = kzalloc(sizeof(struct ipvl_addr), GFP_KERNEL);
698 if (!addr)
699 return -ENOMEM;
700
701 addr->master = ipvlan;
702 memcpy(&addr->ip4addr, ip4_addr, sizeof(struct in_addr));
703 addr->atype = IPVL_IPV4;
704 list_add_tail(&addr->anode, &ipvlan->addrs);
705
706 /* If the interface is not up, the address will be added to the hash
707 * list by ipvlan_open.
708 */
709 if (netif_running(ipvlan->dev))
710 ipvlan_ht_addr_add(ipvlan, addr);
711
712 return 0;
713}
714
715static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
716{
717 struct ipvl_addr *addr;
718
719 addr = ipvlan_find_addr(ipvlan, ip4_addr, false);
720 if (!addr)
721 return;
722
723 ipvlan_ht_addr_del(addr);
724 list_del(&addr->anode);
725 kfree_rcu(addr, rcu);
726
727 return;
728}
729
730static int ipvlan_addr4_event(struct notifier_block *unused,
731 unsigned long event, void *ptr)
732{
733 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
734 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
735 struct ipvl_dev *ipvlan = netdev_priv(dev);
736 struct in_addr ip4_addr;
737
738 if (!netif_is_ipvlan(dev))
739 return NOTIFY_DONE;
740
741 if (!ipvlan || !ipvlan->port)
742 return NOTIFY_DONE;
743
744 switch (event) {
745 case NETDEV_UP:
746 ip4_addr.s_addr = if4->ifa_address;
747 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
748 return NOTIFY_BAD;
749 break;
750
751 case NETDEV_DOWN:
752 ip4_addr.s_addr = if4->ifa_address;
753 ipvlan_del_addr4(ipvlan, &ip4_addr);
754 break;
755 }
756
757 return NOTIFY_OK;
758}
759
760static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
761 .notifier_call = ipvlan_addr4_event,
762};
763
764static struct notifier_block ipvlan_notifier_block __read_mostly = {
765 .notifier_call = ipvlan_device_event,
766};
767
768static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
769 .notifier_call = ipvlan_addr6_event,
770};
771
772static int __init ipvlan_init_module(void)
773{
774 int err;
775
776 ipvlan_init_secret();
777 register_netdevice_notifier(&ipvlan_notifier_block);
778 register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
779 register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
780
781 err = ipvlan_link_register(&ipvlan_link_ops);
782 if (err < 0)
783 goto error;
784
785 return 0;
786error:
787 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
788 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
789 unregister_netdevice_notifier(&ipvlan_notifier_block);
790 return err;
791}
792
793static void __exit ipvlan_cleanup_module(void)
794{
795 rtnl_link_unregister(&ipvlan_link_ops);
796 unregister_netdevice_notifier(&ipvlan_notifier_block);
797 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
798 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
799}
800
801module_init(ipvlan_init_module);
802module_exit(ipvlan_cleanup_module);
803
804MODULE_LICENSE("GPL");
805MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
806MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
807MODULE_ALIAS_RTNL_LINK("ipvlan");