1 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
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.
12 static unsigned int ipvlan_netid __read_mostly;
15 unsigned int ipvl_nf_hook_refcnt;
18 static const struct nf_hook_ops ipvl_nfops[] = {
20 .hook = ipvlan_nf_input,
22 .hooknum = NF_INET_LOCAL_IN,
25 #if IS_ENABLED(CONFIG_IPV6)
27 .hook = ipvlan_nf_input,
29 .hooknum = NF_INET_LOCAL_IN,
35 static const struct l3mdev_ops ipvl_l3mdev_ops = {
36 .l3mdev_l3_rcv = ipvlan_l3_rcv,
39 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
41 ipvlan->dev->mtu = dev->mtu;
44 static int ipvlan_register_nf_hook(struct net *net)
46 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
49 if (!vnet->ipvl_nf_hook_refcnt) {
50 err = nf_register_net_hooks(net, ipvl_nfops,
51 ARRAY_SIZE(ipvl_nfops));
53 vnet->ipvl_nf_hook_refcnt = 1;
55 vnet->ipvl_nf_hook_refcnt++;
61 static void ipvlan_unregister_nf_hook(struct net *net)
63 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
65 if (WARN_ON(!vnet->ipvl_nf_hook_refcnt))
68 vnet->ipvl_nf_hook_refcnt--;
69 if (!vnet->ipvl_nf_hook_refcnt)
70 nf_unregister_net_hooks(net, ipvl_nfops,
71 ARRAY_SIZE(ipvl_nfops));
74 static int ipvlan_set_port_mode(struct ipvl_port *port, u16 nval)
76 struct ipvl_dev *ipvlan;
77 struct net_device *mdev = port->dev;
81 if (port->mode != nval) {
82 if (nval == IPVLAN_MODE_L3S) {
84 err = ipvlan_register_nf_hook(read_pnet(&port->pnet));
86 mdev->l3mdev_ops = &ipvl_l3mdev_ops;
87 mdev->priv_flags |= IFF_L3MDEV_MASTER;
90 } else if (port->mode == IPVLAN_MODE_L3S) {
91 /* Old mode was L3S */
92 mdev->priv_flags &= ~IFF_L3MDEV_MASTER;
93 ipvlan_unregister_nf_hook(read_pnet(&port->pnet));
94 mdev->l3mdev_ops = NULL;
96 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
97 if (nval == IPVLAN_MODE_L3 || nval == IPVLAN_MODE_L3S)
98 ipvlan->dev->flags |= IFF_NOARP;
100 ipvlan->dev->flags &= ~IFF_NOARP;
107 static int ipvlan_port_create(struct net_device *dev)
109 struct ipvl_port *port;
112 port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL);
116 write_pnet(&port->pnet, dev_net(dev));
118 port->mode = IPVLAN_MODE_L3;
119 INIT_LIST_HEAD(&port->ipvlans);
120 for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++)
121 INIT_HLIST_HEAD(&port->hlhead[idx]);
123 skb_queue_head_init(&port->backlog);
124 INIT_WORK(&port->wq, ipvlan_process_multicast);
125 ida_init(&port->ida);
126 port->dev_id_start = 1;
128 err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port);
139 static void ipvlan_port_destroy(struct net_device *dev)
141 struct ipvl_port *port = ipvlan_port_get_rtnl(dev);
144 if (port->mode == IPVLAN_MODE_L3S) {
145 dev->priv_flags &= ~IFF_L3MDEV_MASTER;
146 ipvlan_unregister_nf_hook(dev_net(dev));
147 dev->l3mdev_ops = NULL;
149 netdev_rx_handler_unregister(dev);
150 cancel_work_sync(&port->wq);
151 while ((skb = __skb_dequeue(&port->backlog)) != NULL) {
156 ida_destroy(&port->ida);
160 #define IPVLAN_FEATURES \
161 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
162 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_GSO_ROBUST | \
163 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
164 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
166 #define IPVLAN_STATE_MASK \
167 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
169 static int ipvlan_init(struct net_device *dev)
171 struct ipvl_dev *ipvlan = netdev_priv(dev);
172 struct net_device *phy_dev = ipvlan->phy_dev;
173 struct ipvl_port *port;
176 dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
177 (phy_dev->state & IPVLAN_STATE_MASK);
178 dev->features = phy_dev->features & IPVLAN_FEATURES;
179 dev->features |= NETIF_F_LLTX | NETIF_F_VLAN_CHALLENGED;
180 dev->gso_max_size = phy_dev->gso_max_size;
181 dev->gso_max_segs = phy_dev->gso_max_segs;
182 dev->hard_header_len = phy_dev->hard_header_len;
184 netdev_lockdep_set_classes(dev);
186 ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
187 if (!ipvlan->pcpu_stats)
190 if (!netif_is_ipvlan_port(phy_dev)) {
191 err = ipvlan_port_create(phy_dev);
193 free_percpu(ipvlan->pcpu_stats);
197 port = ipvlan_port_get_rtnl(phy_dev);
202 static void ipvlan_uninit(struct net_device *dev)
204 struct ipvl_dev *ipvlan = netdev_priv(dev);
205 struct net_device *phy_dev = ipvlan->phy_dev;
206 struct ipvl_port *port;
208 free_percpu(ipvlan->pcpu_stats);
210 port = ipvlan_port_get_rtnl(phy_dev);
213 ipvlan_port_destroy(port->dev);
216 static int ipvlan_open(struct net_device *dev)
218 struct ipvl_dev *ipvlan = netdev_priv(dev);
219 struct net_device *phy_dev = ipvlan->phy_dev;
220 struct ipvl_addr *addr;
222 if (ipvlan->port->mode == IPVLAN_MODE_L3 ||
223 ipvlan->port->mode == IPVLAN_MODE_L3S)
224 dev->flags |= IFF_NOARP;
226 dev->flags &= ~IFF_NOARP;
229 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
230 ipvlan_ht_addr_add(ipvlan, addr);
233 return dev_uc_add(phy_dev, phy_dev->dev_addr);
236 static int ipvlan_stop(struct net_device *dev)
238 struct ipvl_dev *ipvlan = netdev_priv(dev);
239 struct net_device *phy_dev = ipvlan->phy_dev;
240 struct ipvl_addr *addr;
242 dev_uc_unsync(phy_dev, dev);
243 dev_mc_unsync(phy_dev, dev);
245 dev_uc_del(phy_dev, phy_dev->dev_addr);
248 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode)
249 ipvlan_ht_addr_del(addr);
255 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
256 struct net_device *dev)
258 const struct ipvl_dev *ipvlan = netdev_priv(dev);
259 int skblen = skb->len;
262 ret = ipvlan_queue_xmit(skb, dev);
263 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
264 struct ipvl_pcpu_stats *pcptr;
266 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
268 u64_stats_update_begin(&pcptr->syncp);
270 pcptr->tx_bytes += skblen;
271 u64_stats_update_end(&pcptr->syncp);
273 this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
278 static netdev_features_t ipvlan_fix_features(struct net_device *dev,
279 netdev_features_t features)
281 struct ipvl_dev *ipvlan = netdev_priv(dev);
283 return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES);
286 static void ipvlan_change_rx_flags(struct net_device *dev, int change)
288 struct ipvl_dev *ipvlan = netdev_priv(dev);
289 struct net_device *phy_dev = ipvlan->phy_dev;
291 if (change & IFF_ALLMULTI)
292 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
295 static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
297 struct ipvl_dev *ipvlan = netdev_priv(dev);
299 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
300 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
302 struct netdev_hw_addr *ha;
303 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
305 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
306 netdev_for_each_mc_addr(ha, dev)
307 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
309 /* Turn-on broadcast bit irrespective of address family,
310 * since broadcast is deferred to a work-queue, hence no
311 * impact on fast-path processing.
313 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters);
315 bitmap_copy(ipvlan->mac_filters, mc_filters,
316 IPVLAN_MAC_FILTER_SIZE);
318 dev_uc_sync(ipvlan->phy_dev, dev);
319 dev_mc_sync(ipvlan->phy_dev, dev);
322 static void ipvlan_get_stats64(struct net_device *dev,
323 struct rtnl_link_stats64 *s)
325 struct ipvl_dev *ipvlan = netdev_priv(dev);
327 if (ipvlan->pcpu_stats) {
328 struct ipvl_pcpu_stats *pcptr;
329 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
330 u32 rx_errs = 0, tx_drps = 0;
334 for_each_possible_cpu(idx) {
335 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
337 strt= u64_stats_fetch_begin_irq(&pcptr->syncp);
338 rx_pkts = pcptr->rx_pkts;
339 rx_bytes = pcptr->rx_bytes;
340 rx_mcast = pcptr->rx_mcast;
341 tx_pkts = pcptr->tx_pkts;
342 tx_bytes = pcptr->tx_bytes;
343 } while (u64_stats_fetch_retry_irq(&pcptr->syncp,
346 s->rx_packets += rx_pkts;
347 s->rx_bytes += rx_bytes;
348 s->multicast += rx_mcast;
349 s->tx_packets += tx_pkts;
350 s->tx_bytes += tx_bytes;
352 /* u32 values are updated without syncp protection. */
353 rx_errs += pcptr->rx_errs;
354 tx_drps += pcptr->tx_drps;
356 s->rx_errors = rx_errs;
357 s->rx_dropped = rx_errs;
358 s->tx_dropped = tx_drps;
362 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
364 struct ipvl_dev *ipvlan = netdev_priv(dev);
365 struct net_device *phy_dev = ipvlan->phy_dev;
367 return vlan_vid_add(phy_dev, proto, vid);
370 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
373 struct ipvl_dev *ipvlan = netdev_priv(dev);
374 struct net_device *phy_dev = ipvlan->phy_dev;
376 vlan_vid_del(phy_dev, proto, vid);
380 static int ipvlan_get_iflink(const struct net_device *dev)
382 struct ipvl_dev *ipvlan = netdev_priv(dev);
384 return ipvlan->phy_dev->ifindex;
387 static const struct net_device_ops ipvlan_netdev_ops = {
388 .ndo_init = ipvlan_init,
389 .ndo_uninit = ipvlan_uninit,
390 .ndo_open = ipvlan_open,
391 .ndo_stop = ipvlan_stop,
392 .ndo_start_xmit = ipvlan_start_xmit,
393 .ndo_fix_features = ipvlan_fix_features,
394 .ndo_change_rx_flags = ipvlan_change_rx_flags,
395 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter,
396 .ndo_get_stats64 = ipvlan_get_stats64,
397 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid,
398 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid,
399 .ndo_get_iflink = ipvlan_get_iflink,
402 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
403 unsigned short type, const void *daddr,
404 const void *saddr, unsigned len)
406 const struct ipvl_dev *ipvlan = netdev_priv(dev);
407 struct net_device *phy_dev = ipvlan->phy_dev;
409 /* TODO Probably use a different field than dev_addr so that the
410 * mac-address on the virtual device is portable and can be carried
411 * while the packets use the mac-addr on the physical device.
413 return dev_hard_header(skb, phy_dev, type, daddr,
414 saddr ? : phy_dev->dev_addr, len);
417 static const struct header_ops ipvlan_header_ops = {
418 .create = ipvlan_hard_header,
419 .parse = eth_header_parse,
420 .cache = eth_header_cache,
421 .cache_update = eth_header_cache_update,
424 static bool netif_is_ipvlan(const struct net_device *dev)
426 /* both ipvlan and ipvtap devices use the same netdev_ops */
427 return dev->netdev_ops == &ipvlan_netdev_ops;
430 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev,
431 struct ethtool_link_ksettings *cmd)
433 const struct ipvl_dev *ipvlan = netdev_priv(dev);
435 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd);
438 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
439 struct ethtool_drvinfo *drvinfo)
441 strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
442 strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
445 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
447 const struct ipvl_dev *ipvlan = netdev_priv(dev);
449 return ipvlan->msg_enable;
452 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
454 struct ipvl_dev *ipvlan = netdev_priv(dev);
456 ipvlan->msg_enable = value;
459 static const struct ethtool_ops ipvlan_ethtool_ops = {
460 .get_link = ethtool_op_get_link,
461 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings,
462 .get_drvinfo = ipvlan_ethtool_get_drvinfo,
463 .get_msglevel = ipvlan_ethtool_get_msglevel,
464 .set_msglevel = ipvlan_ethtool_set_msglevel,
467 static int ipvlan_nl_changelink(struct net_device *dev,
468 struct nlattr *tb[], struct nlattr *data[],
469 struct netlink_ext_ack *extack)
471 struct ipvl_dev *ipvlan = netdev_priv(dev);
472 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
478 if (data[IFLA_IPVLAN_MODE]) {
479 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
481 err = ipvlan_set_port_mode(port, nmode);
484 if (!err && data[IFLA_IPVLAN_FLAGS]) {
485 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
487 if (flags & IPVLAN_F_PRIVATE)
488 ipvlan_mark_private(port);
490 ipvlan_clear_private(port);
492 if (flags & IPVLAN_F_VEPA)
493 ipvlan_mark_vepa(port);
495 ipvlan_clear_vepa(port);
501 static size_t ipvlan_nl_getsize(const struct net_device *dev)
504 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
505 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
509 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[],
510 struct netlink_ext_ack *extack)
515 if (data[IFLA_IPVLAN_MODE]) {
516 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
518 if (mode < IPVLAN_MODE_L2 || mode >= IPVLAN_MODE_MAX)
521 if (data[IFLA_IPVLAN_FLAGS]) {
522 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
524 /* Only two bits are used at this moment. */
525 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
527 /* Also both flags can't be active at the same time. */
528 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) ==
529 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA))
536 static int ipvlan_nl_fillinfo(struct sk_buff *skb,
537 const struct net_device *dev)
539 struct ipvl_dev *ipvlan = netdev_priv(dev);
540 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
547 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
549 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags))
558 int ipvlan_link_new(struct net *src_net, struct net_device *dev,
559 struct nlattr *tb[], struct nlattr *data[],
560 struct netlink_ext_ack *extack)
562 struct ipvl_dev *ipvlan = netdev_priv(dev);
563 struct ipvl_port *port;
564 struct net_device *phy_dev;
566 u16 mode = IPVLAN_MODE_L3;
571 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
575 if (netif_is_ipvlan(phy_dev)) {
576 struct ipvl_dev *tmp = netdev_priv(phy_dev);
578 phy_dev = tmp->phy_dev;
579 } else if (!netif_is_ipvlan_port(phy_dev)) {
580 /* Exit early if the underlying link is invalid or busy */
581 if (phy_dev->type != ARPHRD_ETHER ||
582 phy_dev->flags & IFF_LOOPBACK) {
584 "Master is either lo or non-ether device\n");
588 if (netdev_is_rx_handler_busy(phy_dev)) {
589 netdev_err(phy_dev, "Device is already in use.\n");
594 ipvlan->phy_dev = phy_dev;
596 ipvlan->sfeatures = IPVLAN_FEATURES;
598 ipvlan_adjust_mtu(ipvlan, phy_dev);
599 INIT_LIST_HEAD(&ipvlan->addrs);
600 spin_lock_init(&ipvlan->addrs_lock);
602 /* TODO Probably put random address here to be presented to the
603 * world but keep using the physical-dev address for the outgoing
606 memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN);
608 dev->priv_flags |= IFF_NO_RX_HANDLER;
610 err = register_netdevice(dev);
614 /* ipvlan_init() would have created the port, if required */
615 port = ipvlan_port_get_rtnl(phy_dev);
618 /* If the port-id base is at the MAX value, then wrap it around and
619 * begin from 0x1 again. This may be due to a busy system where lots
620 * of slaves are getting created and deleted.
622 if (port->dev_id_start == 0xFFFE)
623 port->dev_id_start = 0x1;
625 /* Since L2 address is shared among all IPvlan slaves including
626 * master, use unique 16 bit dev-ids to diffentiate among them.
627 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
628 * slave link [see addrconf_ifid_eui48()].
630 err = ida_simple_get(&port->ida, port->dev_id_start, 0xFFFE,
633 err = ida_simple_get(&port->ida, 0x1, port->dev_id_start,
636 goto unregister_netdev;
639 /* Increment id-base to the next slot for the future assignment */
640 port->dev_id_start = err + 1;
642 err = netdev_upper_dev_link(phy_dev, dev, extack);
646 /* Flags are per port and latest update overrides. User has
647 * to be consistent in setting it just like the mode attribute.
649 if (data && data[IFLA_IPVLAN_FLAGS])
650 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
652 if (data && data[IFLA_IPVLAN_MODE])
653 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
655 err = ipvlan_set_port_mode(port, mode);
659 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
660 netif_stacked_transfer_operstate(phy_dev, dev);
664 netdev_upper_dev_unlink(phy_dev, dev);
666 ida_simple_remove(&port->ida, dev->dev_id);
668 unregister_netdevice(dev);
671 EXPORT_SYMBOL_GPL(ipvlan_link_new);
673 void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
675 struct ipvl_dev *ipvlan = netdev_priv(dev);
676 struct ipvl_addr *addr, *next;
678 spin_lock_bh(&ipvlan->addrs_lock);
679 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
680 ipvlan_ht_addr_del(addr);
681 list_del_rcu(&addr->anode);
682 kfree_rcu(addr, rcu);
684 spin_unlock_bh(&ipvlan->addrs_lock);
686 ida_simple_remove(&ipvlan->port->ida, dev->dev_id);
687 list_del_rcu(&ipvlan->pnode);
688 unregister_netdevice_queue(dev, head);
689 netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
691 EXPORT_SYMBOL_GPL(ipvlan_link_delete);
693 void ipvlan_link_setup(struct net_device *dev)
697 dev->max_mtu = ETH_MAX_MTU;
698 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
699 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE;
700 dev->netdev_ops = &ipvlan_netdev_ops;
701 dev->needs_free_netdev = true;
702 dev->header_ops = &ipvlan_header_ops;
703 dev->ethtool_ops = &ipvlan_ethtool_ops;
705 EXPORT_SYMBOL_GPL(ipvlan_link_setup);
707 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
709 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
710 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 },
713 static struct rtnl_link_ops ipvlan_link_ops = {
715 .priv_size = sizeof(struct ipvl_dev),
717 .setup = ipvlan_link_setup,
718 .newlink = ipvlan_link_new,
719 .dellink = ipvlan_link_delete,
722 int ipvlan_link_register(struct rtnl_link_ops *ops)
724 ops->get_size = ipvlan_nl_getsize;
725 ops->policy = ipvlan_nl_policy;
726 ops->validate = ipvlan_nl_validate;
727 ops->fill_info = ipvlan_nl_fillinfo;
728 ops->changelink = ipvlan_nl_changelink;
729 ops->maxtype = IFLA_IPVLAN_MAX;
730 return rtnl_link_register(ops);
732 EXPORT_SYMBOL_GPL(ipvlan_link_register);
734 static int ipvlan_device_event(struct notifier_block *unused,
735 unsigned long event, void *ptr)
737 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
738 struct ipvl_dev *ipvlan, *next;
739 struct ipvl_port *port;
742 if (!netif_is_ipvlan_port(dev))
745 port = ipvlan_port_get_rtnl(dev);
749 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
750 netif_stacked_transfer_operstate(ipvlan->phy_dev,
754 case NETDEV_REGISTER: {
755 struct net *oldnet, *newnet = dev_net(dev);
756 struct ipvlan_netns *old_vnet;
758 oldnet = read_pnet(&port->pnet);
759 if (net_eq(newnet, oldnet))
762 write_pnet(&port->pnet, newnet);
764 old_vnet = net_generic(oldnet, ipvlan_netid);
765 if (!old_vnet->ipvl_nf_hook_refcnt)
768 ipvlan_register_nf_hook(newnet);
769 ipvlan_unregister_nf_hook(oldnet);
772 case NETDEV_UNREGISTER:
773 if (dev->reg_state != NETREG_UNREGISTERING)
776 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode)
777 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
779 unregister_netdevice_many(&lst_kill);
782 case NETDEV_FEAT_CHANGE:
783 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
784 ipvlan->dev->features = dev->features & IPVLAN_FEATURES;
785 ipvlan->dev->gso_max_size = dev->gso_max_size;
786 ipvlan->dev->gso_max_segs = dev->gso_max_segs;
787 netdev_features_change(ipvlan->dev);
791 case NETDEV_CHANGEMTU:
792 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
793 ipvlan_adjust_mtu(ipvlan, dev);
796 case NETDEV_CHANGEADDR:
797 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
798 ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr);
799 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
803 case NETDEV_PRE_TYPE_CHANGE:
804 /* Forbid underlying device to change its type. */
810 /* the caller must held the addrs lock */
811 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
813 struct ipvl_addr *addr;
815 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
819 addr->master = ipvlan;
821 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr));
822 addr->atype = IPVL_IPV4;
823 #if IS_ENABLED(CONFIG_IPV6)
825 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr));
826 addr->atype = IPVL_IPV6;
830 list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
832 /* If the interface is not up, the address will be added to the hash
833 * list by ipvlan_open.
835 if (netif_running(ipvlan->dev))
836 ipvlan_ht_addr_add(ipvlan, addr);
841 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6)
843 struct ipvl_addr *addr;
845 spin_lock_bh(&ipvlan->addrs_lock);
846 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6);
848 spin_unlock_bh(&ipvlan->addrs_lock);
852 ipvlan_ht_addr_del(addr);
853 list_del_rcu(&addr->anode);
854 spin_unlock_bh(&ipvlan->addrs_lock);
855 kfree_rcu(addr, rcu);
858 static bool ipvlan_is_valid_dev(const struct net_device *dev)
860 struct ipvl_dev *ipvlan = netdev_priv(dev);
862 if (!netif_is_ipvlan(dev))
865 if (!ipvlan || !ipvlan->port)
871 #if IS_ENABLED(CONFIG_IPV6)
872 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
876 spin_lock_bh(&ipvlan->addrs_lock);
877 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true))
878 netif_err(ipvlan, ifup, ipvlan->dev,
879 "Failed to add IPv6=%pI6c addr for %s intf\n",
880 ip6_addr, ipvlan->dev->name);
882 ret = ipvlan_add_addr(ipvlan, ip6_addr, true);
883 spin_unlock_bh(&ipvlan->addrs_lock);
887 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
889 return ipvlan_del_addr(ipvlan, ip6_addr, true);
892 static int ipvlan_addr6_event(struct notifier_block *unused,
893 unsigned long event, void *ptr)
895 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
896 struct net_device *dev = (struct net_device *)if6->idev->dev;
897 struct ipvl_dev *ipvlan = netdev_priv(dev);
899 if (!ipvlan_is_valid_dev(dev))
904 if (ipvlan_add_addr6(ipvlan, &if6->addr))
909 ipvlan_del_addr6(ipvlan, &if6->addr);
916 static int ipvlan_addr6_validator_event(struct notifier_block *unused,
917 unsigned long event, void *ptr)
919 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr;
920 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev;
921 struct ipvl_dev *ipvlan = netdev_priv(dev);
923 if (!ipvlan_is_valid_dev(dev))
928 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) {
929 NL_SET_ERR_MSG(i6vi->extack,
930 "Address already assigned to an ipvlan device");
931 return notifier_from_errno(-EADDRINUSE);
940 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
944 spin_lock_bh(&ipvlan->addrs_lock);
945 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false))
946 netif_err(ipvlan, ifup, ipvlan->dev,
947 "Failed to add IPv4=%pI4 on %s intf.\n",
948 ip4_addr, ipvlan->dev->name);
950 ret = ipvlan_add_addr(ipvlan, ip4_addr, false);
951 spin_unlock_bh(&ipvlan->addrs_lock);
955 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
957 return ipvlan_del_addr(ipvlan, ip4_addr, false);
960 static int ipvlan_addr4_event(struct notifier_block *unused,
961 unsigned long event, void *ptr)
963 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
964 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
965 struct ipvl_dev *ipvlan = netdev_priv(dev);
966 struct in_addr ip4_addr;
968 if (!ipvlan_is_valid_dev(dev))
973 ip4_addr.s_addr = if4->ifa_address;
974 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
979 ip4_addr.s_addr = if4->ifa_address;
980 ipvlan_del_addr4(ipvlan, &ip4_addr);
987 static int ipvlan_addr4_validator_event(struct notifier_block *unused,
988 unsigned long event, void *ptr)
990 struct in_validator_info *ivi = (struct in_validator_info *)ptr;
991 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev;
992 struct ipvl_dev *ipvlan = netdev_priv(dev);
994 if (!ipvlan_is_valid_dev(dev))
999 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) {
1000 NL_SET_ERR_MSG(ivi->extack,
1001 "Address already assigned to an ipvlan device");
1002 return notifier_from_errno(-EADDRINUSE);
1010 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
1011 .notifier_call = ipvlan_addr4_event,
1014 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = {
1015 .notifier_call = ipvlan_addr4_validator_event,
1018 static struct notifier_block ipvlan_notifier_block __read_mostly = {
1019 .notifier_call = ipvlan_device_event,
1022 #if IS_ENABLED(CONFIG_IPV6)
1023 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
1024 .notifier_call = ipvlan_addr6_event,
1027 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = {
1028 .notifier_call = ipvlan_addr6_validator_event,
1032 static void ipvlan_ns_exit(struct net *net)
1034 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
1036 if (WARN_ON_ONCE(vnet->ipvl_nf_hook_refcnt)) {
1037 vnet->ipvl_nf_hook_refcnt = 0;
1038 nf_unregister_net_hooks(net, ipvl_nfops,
1039 ARRAY_SIZE(ipvl_nfops));
1043 static struct pernet_operations ipvlan_net_ops = {
1044 .id = &ipvlan_netid,
1045 .size = sizeof(struct ipvlan_netns),
1046 .exit = ipvlan_ns_exit,
1049 static int __init ipvlan_init_module(void)
1053 ipvlan_init_secret();
1054 register_netdevice_notifier(&ipvlan_notifier_block);
1055 #if IS_ENABLED(CONFIG_IPV6)
1056 register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1057 register_inet6addr_validator_notifier(
1058 &ipvlan_addr6_vtor_notifier_block);
1060 register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1061 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block);
1063 err = register_pernet_subsys(&ipvlan_net_ops);
1067 err = ipvlan_link_register(&ipvlan_link_ops);
1069 unregister_pernet_subsys(&ipvlan_net_ops);
1075 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1076 unregister_inetaddr_validator_notifier(
1077 &ipvlan_addr4_vtor_notifier_block);
1078 #if IS_ENABLED(CONFIG_IPV6)
1079 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1080 unregister_inet6addr_validator_notifier(
1081 &ipvlan_addr6_vtor_notifier_block);
1083 unregister_netdevice_notifier(&ipvlan_notifier_block);
1087 static void __exit ipvlan_cleanup_module(void)
1089 rtnl_link_unregister(&ipvlan_link_ops);
1090 unregister_pernet_subsys(&ipvlan_net_ops);
1091 unregister_netdevice_notifier(&ipvlan_notifier_block);
1092 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
1093 unregister_inetaddr_validator_notifier(
1094 &ipvlan_addr4_vtor_notifier_block);
1095 #if IS_ENABLED(CONFIG_IPV6)
1096 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
1097 unregister_inet6addr_validator_notifier(
1098 &ipvlan_addr6_vtor_notifier_block);
1102 module_init(ipvlan_init_module);
1103 module_exit(ipvlan_cleanup_module);
1105 MODULE_LICENSE("GPL");
1106 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1107 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1108 MODULE_ALIAS_RTNL_LINK("ipvlan");