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
,
75 struct netlink_ext_ack
*extack
)
77 struct ipvl_dev
*ipvlan
;
78 struct net_device
*mdev
= port
->dev
;
83 if (port
->mode
!= nval
) {
84 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
85 flags
= ipvlan
->dev
->flags
;
86 if (nval
== IPVLAN_MODE_L3
|| nval
== IPVLAN_MODE_L3S
) {
87 err
= dev_change_flags(ipvlan
->dev
,
91 err
= dev_change_flags(ipvlan
->dev
,
98 if (nval
== IPVLAN_MODE_L3S
) {
100 err
= ipvlan_register_nf_hook(read_pnet(&port
->pnet
));
102 mdev
->l3mdev_ops
= &ipvl_l3mdev_ops
;
103 mdev
->priv_flags
|= IFF_L3MDEV_RX_HANDLER
;
106 } else if (port
->mode
== IPVLAN_MODE_L3S
) {
107 /* Old mode was L3S */
108 mdev
->priv_flags
&= ~IFF_L3MDEV_RX_HANDLER
;
109 ipvlan_unregister_nf_hook(read_pnet(&port
->pnet
));
110 mdev
->l3mdev_ops
= NULL
;
117 /* Undo the flags changes that have been done so far. */
118 list_for_each_entry_continue_reverse(ipvlan
, &port
->ipvlans
, pnode
) {
119 flags
= ipvlan
->dev
->flags
;
120 if (port
->mode
== IPVLAN_MODE_L3
||
121 port
->mode
== IPVLAN_MODE_L3S
)
122 dev_change_flags(ipvlan
->dev
, flags
| IFF_NOARP
,
125 dev_change_flags(ipvlan
->dev
, flags
& ~IFF_NOARP
,
132 static int ipvlan_port_create(struct net_device
*dev
)
134 struct ipvl_port
*port
;
137 port
= kzalloc(sizeof(struct ipvl_port
), GFP_KERNEL
);
141 write_pnet(&port
->pnet
, dev_net(dev
));
143 port
->mode
= IPVLAN_MODE_L3
;
144 INIT_LIST_HEAD(&port
->ipvlans
);
145 for (idx
= 0; idx
< IPVLAN_HASH_SIZE
; idx
++)
146 INIT_HLIST_HEAD(&port
->hlhead
[idx
]);
148 skb_queue_head_init(&port
->backlog
);
149 INIT_WORK(&port
->wq
, ipvlan_process_multicast
);
150 ida_init(&port
->ida
);
151 port
->dev_id_start
= 1;
153 err
= netdev_rx_handler_register(dev
, ipvlan_handle_frame
, port
);
164 static void ipvlan_port_destroy(struct net_device
*dev
)
166 struct ipvl_port
*port
= ipvlan_port_get_rtnl(dev
);
169 if (port
->mode
== IPVLAN_MODE_L3S
) {
170 dev
->priv_flags
&= ~IFF_L3MDEV_RX_HANDLER
;
171 ipvlan_unregister_nf_hook(dev_net(dev
));
172 dev
->l3mdev_ops
= NULL
;
174 netdev_rx_handler_unregister(dev
);
175 cancel_work_sync(&port
->wq
);
176 while ((skb
= __skb_dequeue(&port
->backlog
)) != NULL
) {
181 ida_destroy(&port
->ida
);
185 #define IPVLAN_FEATURES \
186 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
187 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_GSO_ROBUST | \
188 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
189 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
191 #define IPVLAN_STATE_MASK \
192 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
194 static int ipvlan_init(struct net_device
*dev
)
196 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
197 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
198 struct ipvl_port
*port
;
201 dev
->state
= (dev
->state
& ~IPVLAN_STATE_MASK
) |
202 (phy_dev
->state
& IPVLAN_STATE_MASK
);
203 dev
->features
= phy_dev
->features
& IPVLAN_FEATURES
;
204 dev
->features
|= NETIF_F_LLTX
| NETIF_F_VLAN_CHALLENGED
;
205 dev
->gso_max_size
= phy_dev
->gso_max_size
;
206 dev
->gso_max_segs
= phy_dev
->gso_max_segs
;
207 dev
->hard_header_len
= phy_dev
->hard_header_len
;
209 netdev_lockdep_set_classes(dev
);
211 ipvlan
->pcpu_stats
= netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats
);
212 if (!ipvlan
->pcpu_stats
)
215 if (!netif_is_ipvlan_port(phy_dev
)) {
216 err
= ipvlan_port_create(phy_dev
);
218 free_percpu(ipvlan
->pcpu_stats
);
222 port
= ipvlan_port_get_rtnl(phy_dev
);
227 static void ipvlan_uninit(struct net_device
*dev
)
229 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
230 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
231 struct ipvl_port
*port
;
233 free_percpu(ipvlan
->pcpu_stats
);
235 port
= ipvlan_port_get_rtnl(phy_dev
);
238 ipvlan_port_destroy(port
->dev
);
241 static int ipvlan_open(struct net_device
*dev
)
243 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
244 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
245 struct ipvl_addr
*addr
;
247 if (ipvlan
->port
->mode
== IPVLAN_MODE_L3
||
248 ipvlan
->port
->mode
== IPVLAN_MODE_L3S
)
249 dev
->flags
|= IFF_NOARP
;
251 dev
->flags
&= ~IFF_NOARP
;
254 list_for_each_entry_rcu(addr
, &ipvlan
->addrs
, anode
)
255 ipvlan_ht_addr_add(ipvlan
, addr
);
258 return dev_uc_add(phy_dev
, phy_dev
->dev_addr
);
261 static int ipvlan_stop(struct net_device
*dev
)
263 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
264 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
265 struct ipvl_addr
*addr
;
267 dev_uc_unsync(phy_dev
, dev
);
268 dev_mc_unsync(phy_dev
, dev
);
270 dev_uc_del(phy_dev
, phy_dev
->dev_addr
);
273 list_for_each_entry_rcu(addr
, &ipvlan
->addrs
, anode
)
274 ipvlan_ht_addr_del(addr
);
280 static netdev_tx_t
ipvlan_start_xmit(struct sk_buff
*skb
,
281 struct net_device
*dev
)
283 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
284 int skblen
= skb
->len
;
287 ret
= ipvlan_queue_xmit(skb
, dev
);
288 if (likely(ret
== NET_XMIT_SUCCESS
|| ret
== NET_XMIT_CN
)) {
289 struct ipvl_pcpu_stats
*pcptr
;
291 pcptr
= this_cpu_ptr(ipvlan
->pcpu_stats
);
293 u64_stats_update_begin(&pcptr
->syncp
);
295 pcptr
->tx_bytes
+= skblen
;
296 u64_stats_update_end(&pcptr
->syncp
);
298 this_cpu_inc(ipvlan
->pcpu_stats
->tx_drps
);
303 static netdev_features_t
ipvlan_fix_features(struct net_device
*dev
,
304 netdev_features_t features
)
306 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
308 return features
& (ipvlan
->sfeatures
| ~IPVLAN_FEATURES
);
311 static void ipvlan_change_rx_flags(struct net_device
*dev
, int change
)
313 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
314 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
316 if (change
& IFF_ALLMULTI
)
317 dev_set_allmulti(phy_dev
, dev
->flags
& IFF_ALLMULTI
? 1 : -1);
320 static void ipvlan_set_multicast_mac_filter(struct net_device
*dev
)
322 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
324 if (dev
->flags
& (IFF_PROMISC
| IFF_ALLMULTI
)) {
325 bitmap_fill(ipvlan
->mac_filters
, IPVLAN_MAC_FILTER_SIZE
);
327 struct netdev_hw_addr
*ha
;
328 DECLARE_BITMAP(mc_filters
, IPVLAN_MAC_FILTER_SIZE
);
330 bitmap_zero(mc_filters
, IPVLAN_MAC_FILTER_SIZE
);
331 netdev_for_each_mc_addr(ha
, dev
)
332 __set_bit(ipvlan_mac_hash(ha
->addr
), mc_filters
);
334 /* Turn-on broadcast bit irrespective of address family,
335 * since broadcast is deferred to a work-queue, hence no
336 * impact on fast-path processing.
338 __set_bit(ipvlan_mac_hash(dev
->broadcast
), mc_filters
);
340 bitmap_copy(ipvlan
->mac_filters
, mc_filters
,
341 IPVLAN_MAC_FILTER_SIZE
);
343 dev_uc_sync(ipvlan
->phy_dev
, dev
);
344 dev_mc_sync(ipvlan
->phy_dev
, dev
);
347 static void ipvlan_get_stats64(struct net_device
*dev
,
348 struct rtnl_link_stats64
*s
)
350 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
352 if (ipvlan
->pcpu_stats
) {
353 struct ipvl_pcpu_stats
*pcptr
;
354 u64 rx_pkts
, rx_bytes
, rx_mcast
, tx_pkts
, tx_bytes
;
355 u32 rx_errs
= 0, tx_drps
= 0;
359 for_each_possible_cpu(idx
) {
360 pcptr
= per_cpu_ptr(ipvlan
->pcpu_stats
, idx
);
362 strt
= u64_stats_fetch_begin_irq(&pcptr
->syncp
);
363 rx_pkts
= pcptr
->rx_pkts
;
364 rx_bytes
= pcptr
->rx_bytes
;
365 rx_mcast
= pcptr
->rx_mcast
;
366 tx_pkts
= pcptr
->tx_pkts
;
367 tx_bytes
= pcptr
->tx_bytes
;
368 } while (u64_stats_fetch_retry_irq(&pcptr
->syncp
,
371 s
->rx_packets
+= rx_pkts
;
372 s
->rx_bytes
+= rx_bytes
;
373 s
->multicast
+= rx_mcast
;
374 s
->tx_packets
+= tx_pkts
;
375 s
->tx_bytes
+= tx_bytes
;
377 /* u32 values are updated without syncp protection. */
378 rx_errs
+= pcptr
->rx_errs
;
379 tx_drps
+= pcptr
->tx_drps
;
381 s
->rx_errors
= rx_errs
;
382 s
->rx_dropped
= rx_errs
;
383 s
->tx_dropped
= tx_drps
;
387 static int ipvlan_vlan_rx_add_vid(struct net_device
*dev
, __be16 proto
, u16 vid
)
389 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
390 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
392 return vlan_vid_add(phy_dev
, proto
, vid
);
395 static int ipvlan_vlan_rx_kill_vid(struct net_device
*dev
, __be16 proto
,
398 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
399 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
401 vlan_vid_del(phy_dev
, proto
, vid
);
405 static int ipvlan_get_iflink(const struct net_device
*dev
)
407 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
409 return ipvlan
->phy_dev
->ifindex
;
412 static const struct net_device_ops ipvlan_netdev_ops
= {
413 .ndo_init
= ipvlan_init
,
414 .ndo_uninit
= ipvlan_uninit
,
415 .ndo_open
= ipvlan_open
,
416 .ndo_stop
= ipvlan_stop
,
417 .ndo_start_xmit
= ipvlan_start_xmit
,
418 .ndo_fix_features
= ipvlan_fix_features
,
419 .ndo_change_rx_flags
= ipvlan_change_rx_flags
,
420 .ndo_set_rx_mode
= ipvlan_set_multicast_mac_filter
,
421 .ndo_get_stats64
= ipvlan_get_stats64
,
422 .ndo_vlan_rx_add_vid
= ipvlan_vlan_rx_add_vid
,
423 .ndo_vlan_rx_kill_vid
= ipvlan_vlan_rx_kill_vid
,
424 .ndo_get_iflink
= ipvlan_get_iflink
,
427 static int ipvlan_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
428 unsigned short type
, const void *daddr
,
429 const void *saddr
, unsigned len
)
431 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
432 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
434 /* TODO Probably use a different field than dev_addr so that the
435 * mac-address on the virtual device is portable and can be carried
436 * while the packets use the mac-addr on the physical device.
438 return dev_hard_header(skb
, phy_dev
, type
, daddr
,
439 saddr
? : phy_dev
->dev_addr
, len
);
442 static const struct header_ops ipvlan_header_ops
= {
443 .create
= ipvlan_hard_header
,
444 .parse
= eth_header_parse
,
445 .cache
= eth_header_cache
,
446 .cache_update
= eth_header_cache_update
,
449 static bool netif_is_ipvlan(const struct net_device
*dev
)
451 /* both ipvlan and ipvtap devices use the same netdev_ops */
452 return dev
->netdev_ops
== &ipvlan_netdev_ops
;
455 static int ipvlan_ethtool_get_link_ksettings(struct net_device
*dev
,
456 struct ethtool_link_ksettings
*cmd
)
458 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
460 return __ethtool_get_link_ksettings(ipvlan
->phy_dev
, cmd
);
463 static void ipvlan_ethtool_get_drvinfo(struct net_device
*dev
,
464 struct ethtool_drvinfo
*drvinfo
)
466 strlcpy(drvinfo
->driver
, IPVLAN_DRV
, sizeof(drvinfo
->driver
));
467 strlcpy(drvinfo
->version
, IPV_DRV_VER
, sizeof(drvinfo
->version
));
470 static u32
ipvlan_ethtool_get_msglevel(struct net_device
*dev
)
472 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
474 return ipvlan
->msg_enable
;
477 static void ipvlan_ethtool_set_msglevel(struct net_device
*dev
, u32 value
)
479 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
481 ipvlan
->msg_enable
= value
;
484 static const struct ethtool_ops ipvlan_ethtool_ops
= {
485 .get_link
= ethtool_op_get_link
,
486 .get_link_ksettings
= ipvlan_ethtool_get_link_ksettings
,
487 .get_drvinfo
= ipvlan_ethtool_get_drvinfo
,
488 .get_msglevel
= ipvlan_ethtool_get_msglevel
,
489 .set_msglevel
= ipvlan_ethtool_set_msglevel
,
492 static int ipvlan_nl_changelink(struct net_device
*dev
,
493 struct nlattr
*tb
[], struct nlattr
*data
[],
494 struct netlink_ext_ack
*extack
)
496 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
497 struct ipvl_port
*port
= ipvlan_port_get_rtnl(ipvlan
->phy_dev
);
503 if (data
[IFLA_IPVLAN_MODE
]) {
504 u16 nmode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
506 err
= ipvlan_set_port_mode(port
, nmode
, extack
);
509 if (!err
&& data
[IFLA_IPVLAN_FLAGS
]) {
510 u16 flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
512 if (flags
& IPVLAN_F_PRIVATE
)
513 ipvlan_mark_private(port
);
515 ipvlan_clear_private(port
);
517 if (flags
& IPVLAN_F_VEPA
)
518 ipvlan_mark_vepa(port
);
520 ipvlan_clear_vepa(port
);
526 static size_t ipvlan_nl_getsize(const struct net_device
*dev
)
529 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
530 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
534 static int ipvlan_nl_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
535 struct netlink_ext_ack
*extack
)
540 if (data
[IFLA_IPVLAN_MODE
]) {
541 u16 mode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
543 if (mode
>= IPVLAN_MODE_MAX
)
546 if (data
[IFLA_IPVLAN_FLAGS
]) {
547 u16 flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
549 /* Only two bits are used at this moment. */
550 if (flags
& ~(IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
))
552 /* Also both flags can't be active at the same time. */
553 if ((flags
& (IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
)) ==
554 (IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
))
561 static int ipvlan_nl_fillinfo(struct sk_buff
*skb
,
562 const struct net_device
*dev
)
564 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
565 struct ipvl_port
*port
= ipvlan_port_get_rtnl(ipvlan
->phy_dev
);
572 if (nla_put_u16(skb
, IFLA_IPVLAN_MODE
, port
->mode
))
574 if (nla_put_u16(skb
, IFLA_IPVLAN_FLAGS
, port
->flags
))
583 int ipvlan_link_new(struct net
*src_net
, struct net_device
*dev
,
584 struct nlattr
*tb
[], struct nlattr
*data
[],
585 struct netlink_ext_ack
*extack
)
587 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
588 struct ipvl_port
*port
;
589 struct net_device
*phy_dev
;
591 u16 mode
= IPVLAN_MODE_L3
;
596 phy_dev
= __dev_get_by_index(src_net
, nla_get_u32(tb
[IFLA_LINK
]));
600 if (netif_is_ipvlan(phy_dev
)) {
601 struct ipvl_dev
*tmp
= netdev_priv(phy_dev
);
603 phy_dev
= tmp
->phy_dev
;
604 } else if (!netif_is_ipvlan_port(phy_dev
)) {
605 /* Exit early if the underlying link is invalid or busy */
606 if (phy_dev
->type
!= ARPHRD_ETHER
||
607 phy_dev
->flags
& IFF_LOOPBACK
) {
609 "Master is either lo or non-ether device\n");
613 if (netdev_is_rx_handler_busy(phy_dev
)) {
614 netdev_err(phy_dev
, "Device is already in use.\n");
619 ipvlan
->phy_dev
= phy_dev
;
621 ipvlan
->sfeatures
= IPVLAN_FEATURES
;
623 ipvlan_adjust_mtu(ipvlan
, phy_dev
);
624 INIT_LIST_HEAD(&ipvlan
->addrs
);
625 spin_lock_init(&ipvlan
->addrs_lock
);
627 /* TODO Probably put random address here to be presented to the
628 * world but keep using the physical-dev address for the outgoing
631 memcpy(dev
->dev_addr
, phy_dev
->dev_addr
, ETH_ALEN
);
633 dev
->priv_flags
|= IFF_NO_RX_HANDLER
;
635 err
= register_netdevice(dev
);
639 /* ipvlan_init() would have created the port, if required */
640 port
= ipvlan_port_get_rtnl(phy_dev
);
643 /* If the port-id base is at the MAX value, then wrap it around and
644 * begin from 0x1 again. This may be due to a busy system where lots
645 * of slaves are getting created and deleted.
647 if (port
->dev_id_start
== 0xFFFE)
648 port
->dev_id_start
= 0x1;
650 /* Since L2 address is shared among all IPvlan slaves including
651 * master, use unique 16 bit dev-ids to diffentiate among them.
652 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
653 * slave link [see addrconf_ifid_eui48()].
655 err
= ida_simple_get(&port
->ida
, port
->dev_id_start
, 0xFFFE,
658 err
= ida_simple_get(&port
->ida
, 0x1, port
->dev_id_start
,
661 goto unregister_netdev
;
664 /* Increment id-base to the next slot for the future assignment */
665 port
->dev_id_start
= err
+ 1;
667 err
= netdev_upper_dev_link(phy_dev
, dev
, extack
);
671 /* Flags are per port and latest update overrides. User has
672 * to be consistent in setting it just like the mode attribute.
674 if (data
&& data
[IFLA_IPVLAN_FLAGS
])
675 port
->flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
677 if (data
&& data
[IFLA_IPVLAN_MODE
])
678 mode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
680 err
= ipvlan_set_port_mode(port
, mode
, extack
);
684 list_add_tail_rcu(&ipvlan
->pnode
, &port
->ipvlans
);
685 netif_stacked_transfer_operstate(phy_dev
, dev
);
689 netdev_upper_dev_unlink(phy_dev
, dev
);
691 ida_simple_remove(&port
->ida
, dev
->dev_id
);
693 unregister_netdevice(dev
);
696 EXPORT_SYMBOL_GPL(ipvlan_link_new
);
698 void ipvlan_link_delete(struct net_device
*dev
, struct list_head
*head
)
700 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
701 struct ipvl_addr
*addr
, *next
;
703 spin_lock_bh(&ipvlan
->addrs_lock
);
704 list_for_each_entry_safe(addr
, next
, &ipvlan
->addrs
, anode
) {
705 ipvlan_ht_addr_del(addr
);
706 list_del_rcu(&addr
->anode
);
707 kfree_rcu(addr
, rcu
);
709 spin_unlock_bh(&ipvlan
->addrs_lock
);
711 ida_simple_remove(&ipvlan
->port
->ida
, dev
->dev_id
);
712 list_del_rcu(&ipvlan
->pnode
);
713 unregister_netdevice_queue(dev
, head
);
714 netdev_upper_dev_unlink(ipvlan
->phy_dev
, dev
);
716 EXPORT_SYMBOL_GPL(ipvlan_link_delete
);
718 void ipvlan_link_setup(struct net_device
*dev
)
722 dev
->max_mtu
= ETH_MAX_MTU
;
723 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_TX_SKB_SHARING
);
724 dev
->priv_flags
|= IFF_UNICAST_FLT
| IFF_NO_QUEUE
;
725 dev
->netdev_ops
= &ipvlan_netdev_ops
;
726 dev
->needs_free_netdev
= true;
727 dev
->header_ops
= &ipvlan_header_ops
;
728 dev
->ethtool_ops
= &ipvlan_ethtool_ops
;
730 EXPORT_SYMBOL_GPL(ipvlan_link_setup
);
732 static const struct nla_policy ipvlan_nl_policy
[IFLA_IPVLAN_MAX
+ 1] =
734 [IFLA_IPVLAN_MODE
] = { .type
= NLA_U16
},
735 [IFLA_IPVLAN_FLAGS
] = { .type
= NLA_U16
},
738 static struct rtnl_link_ops ipvlan_link_ops
= {
740 .priv_size
= sizeof(struct ipvl_dev
),
742 .setup
= ipvlan_link_setup
,
743 .newlink
= ipvlan_link_new
,
744 .dellink
= ipvlan_link_delete
,
747 int ipvlan_link_register(struct rtnl_link_ops
*ops
)
749 ops
->get_size
= ipvlan_nl_getsize
;
750 ops
->policy
= ipvlan_nl_policy
;
751 ops
->validate
= ipvlan_nl_validate
;
752 ops
->fill_info
= ipvlan_nl_fillinfo
;
753 ops
->changelink
= ipvlan_nl_changelink
;
754 ops
->maxtype
= IFLA_IPVLAN_MAX
;
755 return rtnl_link_register(ops
);
757 EXPORT_SYMBOL_GPL(ipvlan_link_register
);
759 static int ipvlan_device_event(struct notifier_block
*unused
,
760 unsigned long event
, void *ptr
)
762 struct netlink_ext_ack
*extack
= netdev_notifier_info_to_extack(ptr
);
763 struct netdev_notifier_pre_changeaddr_info
*prechaddr_info
;
764 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
765 struct ipvl_dev
*ipvlan
, *next
;
766 struct ipvl_port
*port
;
770 if (!netif_is_ipvlan_port(dev
))
773 port
= ipvlan_port_get_rtnl(dev
);
777 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
)
778 netif_stacked_transfer_operstate(ipvlan
->phy_dev
,
782 case NETDEV_REGISTER
: {
783 struct net
*oldnet
, *newnet
= dev_net(dev
);
784 struct ipvlan_netns
*old_vnet
;
786 oldnet
= read_pnet(&port
->pnet
);
787 if (net_eq(newnet
, oldnet
))
790 write_pnet(&port
->pnet
, newnet
);
792 old_vnet
= net_generic(oldnet
, ipvlan_netid
);
793 if (!old_vnet
->ipvl_nf_hook_refcnt
)
796 ipvlan_register_nf_hook(newnet
);
797 ipvlan_unregister_nf_hook(oldnet
);
800 case NETDEV_UNREGISTER
:
801 if (dev
->reg_state
!= NETREG_UNREGISTERING
)
804 list_for_each_entry_safe(ipvlan
, next
, &port
->ipvlans
, pnode
)
805 ipvlan
->dev
->rtnl_link_ops
->dellink(ipvlan
->dev
,
807 unregister_netdevice_many(&lst_kill
);
810 case NETDEV_FEAT_CHANGE
:
811 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
812 ipvlan
->dev
->features
= dev
->features
& IPVLAN_FEATURES
;
813 ipvlan
->dev
->gso_max_size
= dev
->gso_max_size
;
814 ipvlan
->dev
->gso_max_segs
= dev
->gso_max_segs
;
815 netdev_features_change(ipvlan
->dev
);
819 case NETDEV_CHANGEMTU
:
820 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
)
821 ipvlan_adjust_mtu(ipvlan
, dev
);
824 case NETDEV_PRE_CHANGEADDR
:
825 prechaddr_info
= ptr
;
826 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
827 err
= dev_pre_changeaddr_notify(ipvlan
->dev
,
828 prechaddr_info
->dev_addr
,
831 return notifier_from_errno(err
);
835 case NETDEV_CHANGEADDR
:
836 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
837 ether_addr_copy(ipvlan
->dev
->dev_addr
, dev
->dev_addr
);
838 call_netdevice_notifiers(NETDEV_CHANGEADDR
, ipvlan
->dev
);
842 case NETDEV_PRE_TYPE_CHANGE
:
843 /* Forbid underlying device to change its type. */
849 /* the caller must held the addrs lock */
850 static int ipvlan_add_addr(struct ipvl_dev
*ipvlan
, void *iaddr
, bool is_v6
)
852 struct ipvl_addr
*addr
;
854 addr
= kzalloc(sizeof(struct ipvl_addr
), GFP_ATOMIC
);
858 addr
->master
= ipvlan
;
860 memcpy(&addr
->ip4addr
, iaddr
, sizeof(struct in_addr
));
861 addr
->atype
= IPVL_IPV4
;
862 #if IS_ENABLED(CONFIG_IPV6)
864 memcpy(&addr
->ip6addr
, iaddr
, sizeof(struct in6_addr
));
865 addr
->atype
= IPVL_IPV6
;
869 list_add_tail_rcu(&addr
->anode
, &ipvlan
->addrs
);
871 /* If the interface is not up, the address will be added to the hash
872 * list by ipvlan_open.
874 if (netif_running(ipvlan
->dev
))
875 ipvlan_ht_addr_add(ipvlan
, addr
);
880 static void ipvlan_del_addr(struct ipvl_dev
*ipvlan
, void *iaddr
, bool is_v6
)
882 struct ipvl_addr
*addr
;
884 spin_lock_bh(&ipvlan
->addrs_lock
);
885 addr
= ipvlan_find_addr(ipvlan
, iaddr
, is_v6
);
887 spin_unlock_bh(&ipvlan
->addrs_lock
);
891 ipvlan_ht_addr_del(addr
);
892 list_del_rcu(&addr
->anode
);
893 spin_unlock_bh(&ipvlan
->addrs_lock
);
894 kfree_rcu(addr
, rcu
);
897 static bool ipvlan_is_valid_dev(const struct net_device
*dev
)
899 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
901 if (!netif_is_ipvlan(dev
))
904 if (!ipvlan
|| !ipvlan
->port
)
910 #if IS_ENABLED(CONFIG_IPV6)
911 static int ipvlan_add_addr6(struct ipvl_dev
*ipvlan
, struct in6_addr
*ip6_addr
)
915 spin_lock_bh(&ipvlan
->addrs_lock
);
916 if (ipvlan_addr_busy(ipvlan
->port
, ip6_addr
, true))
917 netif_err(ipvlan
, ifup
, ipvlan
->dev
,
918 "Failed to add IPv6=%pI6c addr for %s intf\n",
919 ip6_addr
, ipvlan
->dev
->name
);
921 ret
= ipvlan_add_addr(ipvlan
, ip6_addr
, true);
922 spin_unlock_bh(&ipvlan
->addrs_lock
);
926 static void ipvlan_del_addr6(struct ipvl_dev
*ipvlan
, struct in6_addr
*ip6_addr
)
928 return ipvlan_del_addr(ipvlan
, ip6_addr
, true);
931 static int ipvlan_addr6_event(struct notifier_block
*unused
,
932 unsigned long event
, void *ptr
)
934 struct inet6_ifaddr
*if6
= (struct inet6_ifaddr
*)ptr
;
935 struct net_device
*dev
= (struct net_device
*)if6
->idev
->dev
;
936 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
938 if (!ipvlan_is_valid_dev(dev
))
943 if (ipvlan_add_addr6(ipvlan
, &if6
->addr
))
948 ipvlan_del_addr6(ipvlan
, &if6
->addr
);
955 static int ipvlan_addr6_validator_event(struct notifier_block
*unused
,
956 unsigned long event
, void *ptr
)
958 struct in6_validator_info
*i6vi
= (struct in6_validator_info
*)ptr
;
959 struct net_device
*dev
= (struct net_device
*)i6vi
->i6vi_dev
->dev
;
960 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
962 if (!ipvlan_is_valid_dev(dev
))
967 if (ipvlan_addr_busy(ipvlan
->port
, &i6vi
->i6vi_addr
, true)) {
968 NL_SET_ERR_MSG(i6vi
->extack
,
969 "Address already assigned to an ipvlan device");
970 return notifier_from_errno(-EADDRINUSE
);
979 static int ipvlan_add_addr4(struct ipvl_dev
*ipvlan
, struct in_addr
*ip4_addr
)
983 spin_lock_bh(&ipvlan
->addrs_lock
);
984 if (ipvlan_addr_busy(ipvlan
->port
, ip4_addr
, false))
985 netif_err(ipvlan
, ifup
, ipvlan
->dev
,
986 "Failed to add IPv4=%pI4 on %s intf.\n",
987 ip4_addr
, ipvlan
->dev
->name
);
989 ret
= ipvlan_add_addr(ipvlan
, ip4_addr
, false);
990 spin_unlock_bh(&ipvlan
->addrs_lock
);
994 static void ipvlan_del_addr4(struct ipvl_dev
*ipvlan
, struct in_addr
*ip4_addr
)
996 return ipvlan_del_addr(ipvlan
, ip4_addr
, false);
999 static int ipvlan_addr4_event(struct notifier_block
*unused
,
1000 unsigned long event
, void *ptr
)
1002 struct in_ifaddr
*if4
= (struct in_ifaddr
*)ptr
;
1003 struct net_device
*dev
= (struct net_device
*)if4
->ifa_dev
->dev
;
1004 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
1005 struct in_addr ip4_addr
;
1007 if (!ipvlan_is_valid_dev(dev
))
1012 ip4_addr
.s_addr
= if4
->ifa_address
;
1013 if (ipvlan_add_addr4(ipvlan
, &ip4_addr
))
1018 ip4_addr
.s_addr
= if4
->ifa_address
;
1019 ipvlan_del_addr4(ipvlan
, &ip4_addr
);
1026 static int ipvlan_addr4_validator_event(struct notifier_block
*unused
,
1027 unsigned long event
, void *ptr
)
1029 struct in_validator_info
*ivi
= (struct in_validator_info
*)ptr
;
1030 struct net_device
*dev
= (struct net_device
*)ivi
->ivi_dev
->dev
;
1031 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
1033 if (!ipvlan_is_valid_dev(dev
))
1038 if (ipvlan_addr_busy(ipvlan
->port
, &ivi
->ivi_addr
, false)) {
1039 NL_SET_ERR_MSG(ivi
->extack
,
1040 "Address already assigned to an ipvlan device");
1041 return notifier_from_errno(-EADDRINUSE
);
1049 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly
= {
1050 .notifier_call
= ipvlan_addr4_event
,
1053 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly
= {
1054 .notifier_call
= ipvlan_addr4_validator_event
,
1057 static struct notifier_block ipvlan_notifier_block __read_mostly
= {
1058 .notifier_call
= ipvlan_device_event
,
1061 #if IS_ENABLED(CONFIG_IPV6)
1062 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly
= {
1063 .notifier_call
= ipvlan_addr6_event
,
1066 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly
= {
1067 .notifier_call
= ipvlan_addr6_validator_event
,
1071 static void ipvlan_ns_exit(struct net
*net
)
1073 struct ipvlan_netns
*vnet
= net_generic(net
, ipvlan_netid
);
1075 if (WARN_ON_ONCE(vnet
->ipvl_nf_hook_refcnt
)) {
1076 vnet
->ipvl_nf_hook_refcnt
= 0;
1077 nf_unregister_net_hooks(net
, ipvl_nfops
,
1078 ARRAY_SIZE(ipvl_nfops
));
1082 static struct pernet_operations ipvlan_net_ops
= {
1083 .id
= &ipvlan_netid
,
1084 .size
= sizeof(struct ipvlan_netns
),
1085 .exit
= ipvlan_ns_exit
,
1088 static int __init
ipvlan_init_module(void)
1092 ipvlan_init_secret();
1093 register_netdevice_notifier(&ipvlan_notifier_block
);
1094 #if IS_ENABLED(CONFIG_IPV6)
1095 register_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1096 register_inet6addr_validator_notifier(
1097 &ipvlan_addr6_vtor_notifier_block
);
1099 register_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1100 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block
);
1102 err
= register_pernet_subsys(&ipvlan_net_ops
);
1106 err
= ipvlan_link_register(&ipvlan_link_ops
);
1108 unregister_pernet_subsys(&ipvlan_net_ops
);
1114 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1115 unregister_inetaddr_validator_notifier(
1116 &ipvlan_addr4_vtor_notifier_block
);
1117 #if IS_ENABLED(CONFIG_IPV6)
1118 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1119 unregister_inet6addr_validator_notifier(
1120 &ipvlan_addr6_vtor_notifier_block
);
1122 unregister_netdevice_notifier(&ipvlan_notifier_block
);
1126 static void __exit
ipvlan_cleanup_module(void)
1128 rtnl_link_unregister(&ipvlan_link_ops
);
1129 unregister_pernet_subsys(&ipvlan_net_ops
);
1130 unregister_netdevice_notifier(&ipvlan_notifier_block
);
1131 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1132 unregister_inetaddr_validator_notifier(
1133 &ipvlan_addr4_vtor_notifier_block
);
1134 #if IS_ENABLED(CONFIG_IPV6)
1135 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1136 unregister_inet6addr_validator_notifier(
1137 &ipvlan_addr6_vtor_notifier_block
);
1141 module_init(ipvlan_init_module
);
1142 module_exit(ipvlan_cleanup_module
);
1144 MODULE_LICENSE("GPL");
1145 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1146 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1147 MODULE_ALIAS_RTNL_LINK("ipvlan");