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 int ipvlan_set_port_mode(struct ipvl_port
*port
, u16 nval
,
13 struct netlink_ext_ack
*extack
)
15 struct ipvl_dev
*ipvlan
;
20 if (port
->mode
!= nval
) {
21 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
22 flags
= ipvlan
->dev
->flags
;
23 if (nval
== IPVLAN_MODE_L3
|| nval
== IPVLAN_MODE_L3S
) {
24 err
= dev_change_flags(ipvlan
->dev
,
28 err
= dev_change_flags(ipvlan
->dev
,
35 if (nval
== IPVLAN_MODE_L3S
) {
37 err
= ipvlan_l3s_register(port
);
40 } else if (port
->mode
== IPVLAN_MODE_L3S
) {
41 /* Old mode was L3S */
42 ipvlan_l3s_unregister(port
);
49 /* Undo the flags changes that have been done so far. */
50 list_for_each_entry_continue_reverse(ipvlan
, &port
->ipvlans
, pnode
) {
51 flags
= ipvlan
->dev
->flags
;
52 if (port
->mode
== IPVLAN_MODE_L3
||
53 port
->mode
== IPVLAN_MODE_L3S
)
54 dev_change_flags(ipvlan
->dev
, flags
| IFF_NOARP
,
57 dev_change_flags(ipvlan
->dev
, flags
& ~IFF_NOARP
,
64 static int ipvlan_port_create(struct net_device
*dev
)
66 struct ipvl_port
*port
;
69 port
= kzalloc(sizeof(struct ipvl_port
), GFP_KERNEL
);
73 write_pnet(&port
->pnet
, dev_net(dev
));
75 port
->mode
= IPVLAN_MODE_L3
;
76 INIT_LIST_HEAD(&port
->ipvlans
);
77 for (idx
= 0; idx
< IPVLAN_HASH_SIZE
; idx
++)
78 INIT_HLIST_HEAD(&port
->hlhead
[idx
]);
80 skb_queue_head_init(&port
->backlog
);
81 INIT_WORK(&port
->wq
, ipvlan_process_multicast
);
83 port
->dev_id_start
= 1;
85 err
= netdev_rx_handler_register(dev
, ipvlan_handle_frame
, port
);
96 static void ipvlan_port_destroy(struct net_device
*dev
)
98 struct ipvl_port
*port
= ipvlan_port_get_rtnl(dev
);
101 if (port
->mode
== IPVLAN_MODE_L3S
)
102 ipvlan_l3s_unregister(port
);
103 netdev_rx_handler_unregister(dev
);
104 cancel_work_sync(&port
->wq
);
105 while ((skb
= __skb_dequeue(&port
->backlog
)) != NULL
) {
110 ida_destroy(&port
->ida
);
114 #define IPVLAN_FEATURES \
115 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
116 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_GSO_ROBUST | \
117 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
118 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
120 #define IPVLAN_STATE_MASK \
121 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
123 static int ipvlan_init(struct net_device
*dev
)
125 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
126 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
127 struct ipvl_port
*port
;
130 dev
->state
= (dev
->state
& ~IPVLAN_STATE_MASK
) |
131 (phy_dev
->state
& IPVLAN_STATE_MASK
);
132 dev
->features
= phy_dev
->features
& IPVLAN_FEATURES
;
133 dev
->features
|= NETIF_F_LLTX
| NETIF_F_VLAN_CHALLENGED
;
134 dev
->gso_max_size
= phy_dev
->gso_max_size
;
135 dev
->gso_max_segs
= phy_dev
->gso_max_segs
;
136 dev
->hard_header_len
= phy_dev
->hard_header_len
;
138 netdev_lockdep_set_classes(dev
);
140 ipvlan
->pcpu_stats
= netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats
);
141 if (!ipvlan
->pcpu_stats
)
144 if (!netif_is_ipvlan_port(phy_dev
)) {
145 err
= ipvlan_port_create(phy_dev
);
147 free_percpu(ipvlan
->pcpu_stats
);
151 port
= ipvlan_port_get_rtnl(phy_dev
);
156 static void ipvlan_uninit(struct net_device
*dev
)
158 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
159 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
160 struct ipvl_port
*port
;
162 free_percpu(ipvlan
->pcpu_stats
);
164 port
= ipvlan_port_get_rtnl(phy_dev
);
167 ipvlan_port_destroy(port
->dev
);
170 static int ipvlan_open(struct net_device
*dev
)
172 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
173 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
174 struct ipvl_addr
*addr
;
176 if (ipvlan
->port
->mode
== IPVLAN_MODE_L3
||
177 ipvlan
->port
->mode
== IPVLAN_MODE_L3S
)
178 dev
->flags
|= IFF_NOARP
;
180 dev
->flags
&= ~IFF_NOARP
;
183 list_for_each_entry_rcu(addr
, &ipvlan
->addrs
, anode
)
184 ipvlan_ht_addr_add(ipvlan
, addr
);
187 return dev_uc_add(phy_dev
, phy_dev
->dev_addr
);
190 static int ipvlan_stop(struct net_device
*dev
)
192 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
193 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
194 struct ipvl_addr
*addr
;
196 dev_uc_unsync(phy_dev
, dev
);
197 dev_mc_unsync(phy_dev
, dev
);
199 dev_uc_del(phy_dev
, phy_dev
->dev_addr
);
202 list_for_each_entry_rcu(addr
, &ipvlan
->addrs
, anode
)
203 ipvlan_ht_addr_del(addr
);
209 static netdev_tx_t
ipvlan_start_xmit(struct sk_buff
*skb
,
210 struct net_device
*dev
)
212 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
213 int skblen
= skb
->len
;
216 ret
= ipvlan_queue_xmit(skb
, dev
);
217 if (likely(ret
== NET_XMIT_SUCCESS
|| ret
== NET_XMIT_CN
)) {
218 struct ipvl_pcpu_stats
*pcptr
;
220 pcptr
= this_cpu_ptr(ipvlan
->pcpu_stats
);
222 u64_stats_update_begin(&pcptr
->syncp
);
224 pcptr
->tx_bytes
+= skblen
;
225 u64_stats_update_end(&pcptr
->syncp
);
227 this_cpu_inc(ipvlan
->pcpu_stats
->tx_drps
);
232 static netdev_features_t
ipvlan_fix_features(struct net_device
*dev
,
233 netdev_features_t features
)
235 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
237 return features
& (ipvlan
->sfeatures
| ~IPVLAN_FEATURES
);
240 static void ipvlan_change_rx_flags(struct net_device
*dev
, int change
)
242 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
243 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
245 if (change
& IFF_ALLMULTI
)
246 dev_set_allmulti(phy_dev
, dev
->flags
& IFF_ALLMULTI
? 1 : -1);
249 static void ipvlan_set_multicast_mac_filter(struct net_device
*dev
)
251 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
253 if (dev
->flags
& (IFF_PROMISC
| IFF_ALLMULTI
)) {
254 bitmap_fill(ipvlan
->mac_filters
, IPVLAN_MAC_FILTER_SIZE
);
256 struct netdev_hw_addr
*ha
;
257 DECLARE_BITMAP(mc_filters
, IPVLAN_MAC_FILTER_SIZE
);
259 bitmap_zero(mc_filters
, IPVLAN_MAC_FILTER_SIZE
);
260 netdev_for_each_mc_addr(ha
, dev
)
261 __set_bit(ipvlan_mac_hash(ha
->addr
), mc_filters
);
263 /* Turn-on broadcast bit irrespective of address family,
264 * since broadcast is deferred to a work-queue, hence no
265 * impact on fast-path processing.
267 __set_bit(ipvlan_mac_hash(dev
->broadcast
), mc_filters
);
269 bitmap_copy(ipvlan
->mac_filters
, mc_filters
,
270 IPVLAN_MAC_FILTER_SIZE
);
272 dev_uc_sync(ipvlan
->phy_dev
, dev
);
273 dev_mc_sync(ipvlan
->phy_dev
, dev
);
276 static void ipvlan_get_stats64(struct net_device
*dev
,
277 struct rtnl_link_stats64
*s
)
279 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
281 if (ipvlan
->pcpu_stats
) {
282 struct ipvl_pcpu_stats
*pcptr
;
283 u64 rx_pkts
, rx_bytes
, rx_mcast
, tx_pkts
, tx_bytes
;
284 u32 rx_errs
= 0, tx_drps
= 0;
288 for_each_possible_cpu(idx
) {
289 pcptr
= per_cpu_ptr(ipvlan
->pcpu_stats
, idx
);
291 strt
= u64_stats_fetch_begin_irq(&pcptr
->syncp
);
292 rx_pkts
= pcptr
->rx_pkts
;
293 rx_bytes
= pcptr
->rx_bytes
;
294 rx_mcast
= pcptr
->rx_mcast
;
295 tx_pkts
= pcptr
->tx_pkts
;
296 tx_bytes
= pcptr
->tx_bytes
;
297 } while (u64_stats_fetch_retry_irq(&pcptr
->syncp
,
300 s
->rx_packets
+= rx_pkts
;
301 s
->rx_bytes
+= rx_bytes
;
302 s
->multicast
+= rx_mcast
;
303 s
->tx_packets
+= tx_pkts
;
304 s
->tx_bytes
+= tx_bytes
;
306 /* u32 values are updated without syncp protection. */
307 rx_errs
+= pcptr
->rx_errs
;
308 tx_drps
+= pcptr
->tx_drps
;
310 s
->rx_errors
= rx_errs
;
311 s
->rx_dropped
= rx_errs
;
312 s
->tx_dropped
= tx_drps
;
316 static int ipvlan_vlan_rx_add_vid(struct net_device
*dev
, __be16 proto
, u16 vid
)
318 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
319 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
321 return vlan_vid_add(phy_dev
, proto
, vid
);
324 static int ipvlan_vlan_rx_kill_vid(struct net_device
*dev
, __be16 proto
,
327 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
328 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
330 vlan_vid_del(phy_dev
, proto
, vid
);
334 static int ipvlan_get_iflink(const struct net_device
*dev
)
336 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
338 return ipvlan
->phy_dev
->ifindex
;
341 static const struct net_device_ops ipvlan_netdev_ops
= {
342 .ndo_init
= ipvlan_init
,
343 .ndo_uninit
= ipvlan_uninit
,
344 .ndo_open
= ipvlan_open
,
345 .ndo_stop
= ipvlan_stop
,
346 .ndo_start_xmit
= ipvlan_start_xmit
,
347 .ndo_fix_features
= ipvlan_fix_features
,
348 .ndo_change_rx_flags
= ipvlan_change_rx_flags
,
349 .ndo_set_rx_mode
= ipvlan_set_multicast_mac_filter
,
350 .ndo_get_stats64
= ipvlan_get_stats64
,
351 .ndo_vlan_rx_add_vid
= ipvlan_vlan_rx_add_vid
,
352 .ndo_vlan_rx_kill_vid
= ipvlan_vlan_rx_kill_vid
,
353 .ndo_get_iflink
= ipvlan_get_iflink
,
356 static int ipvlan_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
357 unsigned short type
, const void *daddr
,
358 const void *saddr
, unsigned len
)
360 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
361 struct net_device
*phy_dev
= ipvlan
->phy_dev
;
363 /* TODO Probably use a different field than dev_addr so that the
364 * mac-address on the virtual device is portable and can be carried
365 * while the packets use the mac-addr on the physical device.
367 return dev_hard_header(skb
, phy_dev
, type
, daddr
,
368 saddr
? : phy_dev
->dev_addr
, len
);
371 static const struct header_ops ipvlan_header_ops
= {
372 .create
= ipvlan_hard_header
,
373 .parse
= eth_header_parse
,
374 .cache
= eth_header_cache
,
375 .cache_update
= eth_header_cache_update
,
378 static void ipvlan_adjust_mtu(struct ipvl_dev
*ipvlan
, struct net_device
*dev
)
380 ipvlan
->dev
->mtu
= dev
->mtu
;
383 static bool netif_is_ipvlan(const struct net_device
*dev
)
385 /* both ipvlan and ipvtap devices use the same netdev_ops */
386 return dev
->netdev_ops
== &ipvlan_netdev_ops
;
389 static int ipvlan_ethtool_get_link_ksettings(struct net_device
*dev
,
390 struct ethtool_link_ksettings
*cmd
)
392 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
394 return __ethtool_get_link_ksettings(ipvlan
->phy_dev
, cmd
);
397 static void ipvlan_ethtool_get_drvinfo(struct net_device
*dev
,
398 struct ethtool_drvinfo
*drvinfo
)
400 strlcpy(drvinfo
->driver
, IPVLAN_DRV
, sizeof(drvinfo
->driver
));
401 strlcpy(drvinfo
->version
, IPV_DRV_VER
, sizeof(drvinfo
->version
));
404 static u32
ipvlan_ethtool_get_msglevel(struct net_device
*dev
)
406 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
408 return ipvlan
->msg_enable
;
411 static void ipvlan_ethtool_set_msglevel(struct net_device
*dev
, u32 value
)
413 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
415 ipvlan
->msg_enable
= value
;
418 static const struct ethtool_ops ipvlan_ethtool_ops
= {
419 .get_link
= ethtool_op_get_link
,
420 .get_link_ksettings
= ipvlan_ethtool_get_link_ksettings
,
421 .get_drvinfo
= ipvlan_ethtool_get_drvinfo
,
422 .get_msglevel
= ipvlan_ethtool_get_msglevel
,
423 .set_msglevel
= ipvlan_ethtool_set_msglevel
,
426 static int ipvlan_nl_changelink(struct net_device
*dev
,
427 struct nlattr
*tb
[], struct nlattr
*data
[],
428 struct netlink_ext_ack
*extack
)
430 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
431 struct ipvl_port
*port
= ipvlan_port_get_rtnl(ipvlan
->phy_dev
);
437 if (data
[IFLA_IPVLAN_MODE
]) {
438 u16 nmode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
440 err
= ipvlan_set_port_mode(port
, nmode
, extack
);
443 if (!err
&& data
[IFLA_IPVLAN_FLAGS
]) {
444 u16 flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
446 if (flags
& IPVLAN_F_PRIVATE
)
447 ipvlan_mark_private(port
);
449 ipvlan_clear_private(port
);
451 if (flags
& IPVLAN_F_VEPA
)
452 ipvlan_mark_vepa(port
);
454 ipvlan_clear_vepa(port
);
460 static size_t ipvlan_nl_getsize(const struct net_device
*dev
)
463 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
464 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */
468 static int ipvlan_nl_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
469 struct netlink_ext_ack
*extack
)
474 if (data
[IFLA_IPVLAN_MODE
]) {
475 u16 mode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
477 if (mode
>= IPVLAN_MODE_MAX
)
480 if (data
[IFLA_IPVLAN_FLAGS
]) {
481 u16 flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
483 /* Only two bits are used at this moment. */
484 if (flags
& ~(IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
))
486 /* Also both flags can't be active at the same time. */
487 if ((flags
& (IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
)) ==
488 (IPVLAN_F_PRIVATE
| IPVLAN_F_VEPA
))
495 static int ipvlan_nl_fillinfo(struct sk_buff
*skb
,
496 const struct net_device
*dev
)
498 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
499 struct ipvl_port
*port
= ipvlan_port_get_rtnl(ipvlan
->phy_dev
);
506 if (nla_put_u16(skb
, IFLA_IPVLAN_MODE
, port
->mode
))
508 if (nla_put_u16(skb
, IFLA_IPVLAN_FLAGS
, port
->flags
))
517 int ipvlan_link_new(struct net
*src_net
, struct net_device
*dev
,
518 struct nlattr
*tb
[], struct nlattr
*data
[],
519 struct netlink_ext_ack
*extack
)
521 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
522 struct ipvl_port
*port
;
523 struct net_device
*phy_dev
;
525 u16 mode
= IPVLAN_MODE_L3
;
530 phy_dev
= __dev_get_by_index(src_net
, nla_get_u32(tb
[IFLA_LINK
]));
534 if (netif_is_ipvlan(phy_dev
)) {
535 struct ipvl_dev
*tmp
= netdev_priv(phy_dev
);
537 phy_dev
= tmp
->phy_dev
;
538 } else if (!netif_is_ipvlan_port(phy_dev
)) {
539 /* Exit early if the underlying link is invalid or busy */
540 if (phy_dev
->type
!= ARPHRD_ETHER
||
541 phy_dev
->flags
& IFF_LOOPBACK
) {
543 "Master is either lo or non-ether device\n");
547 if (netdev_is_rx_handler_busy(phy_dev
)) {
548 netdev_err(phy_dev
, "Device is already in use.\n");
553 ipvlan
->phy_dev
= phy_dev
;
555 ipvlan
->sfeatures
= IPVLAN_FEATURES
;
557 ipvlan_adjust_mtu(ipvlan
, phy_dev
);
558 INIT_LIST_HEAD(&ipvlan
->addrs
);
559 spin_lock_init(&ipvlan
->addrs_lock
);
561 /* TODO Probably put random address here to be presented to the
562 * world but keep using the physical-dev address for the outgoing
565 memcpy(dev
->dev_addr
, phy_dev
->dev_addr
, ETH_ALEN
);
567 dev
->priv_flags
|= IFF_NO_RX_HANDLER
;
569 err
= register_netdevice(dev
);
573 /* ipvlan_init() would have created the port, if required */
574 port
= ipvlan_port_get_rtnl(phy_dev
);
577 /* If the port-id base is at the MAX value, then wrap it around and
578 * begin from 0x1 again. This may be due to a busy system where lots
579 * of slaves are getting created and deleted.
581 if (port
->dev_id_start
== 0xFFFE)
582 port
->dev_id_start
= 0x1;
584 /* Since L2 address is shared among all IPvlan slaves including
585 * master, use unique 16 bit dev-ids to diffentiate among them.
586 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each
587 * slave link [see addrconf_ifid_eui48()].
589 err
= ida_simple_get(&port
->ida
, port
->dev_id_start
, 0xFFFE,
592 err
= ida_simple_get(&port
->ida
, 0x1, port
->dev_id_start
,
595 goto unregister_netdev
;
598 /* Increment id-base to the next slot for the future assignment */
599 port
->dev_id_start
= err
+ 1;
601 err
= netdev_upper_dev_link(phy_dev
, dev
, extack
);
605 /* Flags are per port and latest update overrides. User has
606 * to be consistent in setting it just like the mode attribute.
608 if (data
&& data
[IFLA_IPVLAN_FLAGS
])
609 port
->flags
= nla_get_u16(data
[IFLA_IPVLAN_FLAGS
]);
611 if (data
&& data
[IFLA_IPVLAN_MODE
])
612 mode
= nla_get_u16(data
[IFLA_IPVLAN_MODE
]);
614 err
= ipvlan_set_port_mode(port
, mode
, extack
);
618 list_add_tail_rcu(&ipvlan
->pnode
, &port
->ipvlans
);
619 netif_stacked_transfer_operstate(phy_dev
, dev
);
623 netdev_upper_dev_unlink(phy_dev
, dev
);
625 ida_simple_remove(&port
->ida
, dev
->dev_id
);
627 unregister_netdevice(dev
);
630 EXPORT_SYMBOL_GPL(ipvlan_link_new
);
632 void ipvlan_link_delete(struct net_device
*dev
, struct list_head
*head
)
634 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
635 struct ipvl_addr
*addr
, *next
;
637 spin_lock_bh(&ipvlan
->addrs_lock
);
638 list_for_each_entry_safe(addr
, next
, &ipvlan
->addrs
, anode
) {
639 ipvlan_ht_addr_del(addr
);
640 list_del_rcu(&addr
->anode
);
641 kfree_rcu(addr
, rcu
);
643 spin_unlock_bh(&ipvlan
->addrs_lock
);
645 ida_simple_remove(&ipvlan
->port
->ida
, dev
->dev_id
);
646 list_del_rcu(&ipvlan
->pnode
);
647 unregister_netdevice_queue(dev
, head
);
648 netdev_upper_dev_unlink(ipvlan
->phy_dev
, dev
);
650 EXPORT_SYMBOL_GPL(ipvlan_link_delete
);
652 void ipvlan_link_setup(struct net_device
*dev
)
656 dev
->max_mtu
= ETH_MAX_MTU
;
657 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_TX_SKB_SHARING
);
658 dev
->priv_flags
|= IFF_UNICAST_FLT
| IFF_NO_QUEUE
;
659 dev
->netdev_ops
= &ipvlan_netdev_ops
;
660 dev
->needs_free_netdev
= true;
661 dev
->header_ops
= &ipvlan_header_ops
;
662 dev
->ethtool_ops
= &ipvlan_ethtool_ops
;
664 EXPORT_SYMBOL_GPL(ipvlan_link_setup
);
666 static const struct nla_policy ipvlan_nl_policy
[IFLA_IPVLAN_MAX
+ 1] =
668 [IFLA_IPVLAN_MODE
] = { .type
= NLA_U16
},
669 [IFLA_IPVLAN_FLAGS
] = { .type
= NLA_U16
},
672 static struct rtnl_link_ops ipvlan_link_ops
= {
674 .priv_size
= sizeof(struct ipvl_dev
),
676 .setup
= ipvlan_link_setup
,
677 .newlink
= ipvlan_link_new
,
678 .dellink
= ipvlan_link_delete
,
681 int ipvlan_link_register(struct rtnl_link_ops
*ops
)
683 ops
->get_size
= ipvlan_nl_getsize
;
684 ops
->policy
= ipvlan_nl_policy
;
685 ops
->validate
= ipvlan_nl_validate
;
686 ops
->fill_info
= ipvlan_nl_fillinfo
;
687 ops
->changelink
= ipvlan_nl_changelink
;
688 ops
->maxtype
= IFLA_IPVLAN_MAX
;
689 return rtnl_link_register(ops
);
691 EXPORT_SYMBOL_GPL(ipvlan_link_register
);
693 static int ipvlan_device_event(struct notifier_block
*unused
,
694 unsigned long event
, void *ptr
)
696 struct netlink_ext_ack
*extack
= netdev_notifier_info_to_extack(ptr
);
697 struct netdev_notifier_pre_changeaddr_info
*prechaddr_info
;
698 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
699 struct ipvl_dev
*ipvlan
, *next
;
700 struct ipvl_port
*port
;
704 if (!netif_is_ipvlan_port(dev
))
707 port
= ipvlan_port_get_rtnl(dev
);
711 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
)
712 netif_stacked_transfer_operstate(ipvlan
->phy_dev
,
716 case NETDEV_REGISTER
: {
717 struct net
*oldnet
, *newnet
= dev_net(dev
);
719 oldnet
= read_pnet(&port
->pnet
);
720 if (net_eq(newnet
, oldnet
))
723 write_pnet(&port
->pnet
, newnet
);
725 ipvlan_migrate_l3s_hook(oldnet
, newnet
);
728 case NETDEV_UNREGISTER
:
729 if (dev
->reg_state
!= NETREG_UNREGISTERING
)
732 list_for_each_entry_safe(ipvlan
, next
, &port
->ipvlans
, pnode
)
733 ipvlan
->dev
->rtnl_link_ops
->dellink(ipvlan
->dev
,
735 unregister_netdevice_many(&lst_kill
);
738 case NETDEV_FEAT_CHANGE
:
739 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
740 ipvlan
->dev
->features
= dev
->features
& IPVLAN_FEATURES
;
741 ipvlan
->dev
->gso_max_size
= dev
->gso_max_size
;
742 ipvlan
->dev
->gso_max_segs
= dev
->gso_max_segs
;
743 netdev_features_change(ipvlan
->dev
);
747 case NETDEV_CHANGEMTU
:
748 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
)
749 ipvlan_adjust_mtu(ipvlan
, dev
);
752 case NETDEV_PRE_CHANGEADDR
:
753 prechaddr_info
= ptr
;
754 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
755 err
= dev_pre_changeaddr_notify(ipvlan
->dev
,
756 prechaddr_info
->dev_addr
,
759 return notifier_from_errno(err
);
763 case NETDEV_CHANGEADDR
:
764 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
765 ether_addr_copy(ipvlan
->dev
->dev_addr
, dev
->dev_addr
);
766 call_netdevice_notifiers(NETDEV_CHANGEADDR
, ipvlan
->dev
);
770 case NETDEV_PRE_TYPE_CHANGE
:
771 /* Forbid underlying device to change its type. */
777 /* the caller must held the addrs lock */
778 static int ipvlan_add_addr(struct ipvl_dev
*ipvlan
, void *iaddr
, bool is_v6
)
780 struct ipvl_addr
*addr
;
782 addr
= kzalloc(sizeof(struct ipvl_addr
), GFP_ATOMIC
);
786 addr
->master
= ipvlan
;
788 memcpy(&addr
->ip4addr
, iaddr
, sizeof(struct in_addr
));
789 addr
->atype
= IPVL_IPV4
;
790 #if IS_ENABLED(CONFIG_IPV6)
792 memcpy(&addr
->ip6addr
, iaddr
, sizeof(struct in6_addr
));
793 addr
->atype
= IPVL_IPV6
;
797 list_add_tail_rcu(&addr
->anode
, &ipvlan
->addrs
);
799 /* If the interface is not up, the address will be added to the hash
800 * list by ipvlan_open.
802 if (netif_running(ipvlan
->dev
))
803 ipvlan_ht_addr_add(ipvlan
, addr
);
808 static void ipvlan_del_addr(struct ipvl_dev
*ipvlan
, void *iaddr
, bool is_v6
)
810 struct ipvl_addr
*addr
;
812 spin_lock_bh(&ipvlan
->addrs_lock
);
813 addr
= ipvlan_find_addr(ipvlan
, iaddr
, is_v6
);
815 spin_unlock_bh(&ipvlan
->addrs_lock
);
819 ipvlan_ht_addr_del(addr
);
820 list_del_rcu(&addr
->anode
);
821 spin_unlock_bh(&ipvlan
->addrs_lock
);
822 kfree_rcu(addr
, rcu
);
825 static bool ipvlan_is_valid_dev(const struct net_device
*dev
)
827 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
829 if (!netif_is_ipvlan(dev
))
832 if (!ipvlan
|| !ipvlan
->port
)
838 #if IS_ENABLED(CONFIG_IPV6)
839 static int ipvlan_add_addr6(struct ipvl_dev
*ipvlan
, struct in6_addr
*ip6_addr
)
843 spin_lock_bh(&ipvlan
->addrs_lock
);
844 if (ipvlan_addr_busy(ipvlan
->port
, ip6_addr
, true))
845 netif_err(ipvlan
, ifup
, ipvlan
->dev
,
846 "Failed to add IPv6=%pI6c addr for %s intf\n",
847 ip6_addr
, ipvlan
->dev
->name
);
849 ret
= ipvlan_add_addr(ipvlan
, ip6_addr
, true);
850 spin_unlock_bh(&ipvlan
->addrs_lock
);
854 static void ipvlan_del_addr6(struct ipvl_dev
*ipvlan
, struct in6_addr
*ip6_addr
)
856 return ipvlan_del_addr(ipvlan
, ip6_addr
, true);
859 static int ipvlan_addr6_event(struct notifier_block
*unused
,
860 unsigned long event
, void *ptr
)
862 struct inet6_ifaddr
*if6
= (struct inet6_ifaddr
*)ptr
;
863 struct net_device
*dev
= (struct net_device
*)if6
->idev
->dev
;
864 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
866 if (!ipvlan_is_valid_dev(dev
))
871 if (ipvlan_add_addr6(ipvlan
, &if6
->addr
))
876 ipvlan_del_addr6(ipvlan
, &if6
->addr
);
883 static int ipvlan_addr6_validator_event(struct notifier_block
*unused
,
884 unsigned long event
, void *ptr
)
886 struct in6_validator_info
*i6vi
= (struct in6_validator_info
*)ptr
;
887 struct net_device
*dev
= (struct net_device
*)i6vi
->i6vi_dev
->dev
;
888 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
890 if (!ipvlan_is_valid_dev(dev
))
895 if (ipvlan_addr_busy(ipvlan
->port
, &i6vi
->i6vi_addr
, true)) {
896 NL_SET_ERR_MSG(i6vi
->extack
,
897 "Address already assigned to an ipvlan device");
898 return notifier_from_errno(-EADDRINUSE
);
907 static int ipvlan_add_addr4(struct ipvl_dev
*ipvlan
, struct in_addr
*ip4_addr
)
911 spin_lock_bh(&ipvlan
->addrs_lock
);
912 if (ipvlan_addr_busy(ipvlan
->port
, ip4_addr
, false))
913 netif_err(ipvlan
, ifup
, ipvlan
->dev
,
914 "Failed to add IPv4=%pI4 on %s intf.\n",
915 ip4_addr
, ipvlan
->dev
->name
);
917 ret
= ipvlan_add_addr(ipvlan
, ip4_addr
, false);
918 spin_unlock_bh(&ipvlan
->addrs_lock
);
922 static void ipvlan_del_addr4(struct ipvl_dev
*ipvlan
, struct in_addr
*ip4_addr
)
924 return ipvlan_del_addr(ipvlan
, ip4_addr
, false);
927 static int ipvlan_addr4_event(struct notifier_block
*unused
,
928 unsigned long event
, void *ptr
)
930 struct in_ifaddr
*if4
= (struct in_ifaddr
*)ptr
;
931 struct net_device
*dev
= (struct net_device
*)if4
->ifa_dev
->dev
;
932 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
933 struct in_addr ip4_addr
;
935 if (!ipvlan_is_valid_dev(dev
))
940 ip4_addr
.s_addr
= if4
->ifa_address
;
941 if (ipvlan_add_addr4(ipvlan
, &ip4_addr
))
946 ip4_addr
.s_addr
= if4
->ifa_address
;
947 ipvlan_del_addr4(ipvlan
, &ip4_addr
);
954 static int ipvlan_addr4_validator_event(struct notifier_block
*unused
,
955 unsigned long event
, void *ptr
)
957 struct in_validator_info
*ivi
= (struct in_validator_info
*)ptr
;
958 struct net_device
*dev
= (struct net_device
*)ivi
->ivi_dev
->dev
;
959 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
961 if (!ipvlan_is_valid_dev(dev
))
966 if (ipvlan_addr_busy(ipvlan
->port
, &ivi
->ivi_addr
, false)) {
967 NL_SET_ERR_MSG(ivi
->extack
,
968 "Address already assigned to an ipvlan device");
969 return notifier_from_errno(-EADDRINUSE
);
977 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly
= {
978 .notifier_call
= ipvlan_addr4_event
,
981 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly
= {
982 .notifier_call
= ipvlan_addr4_validator_event
,
985 static struct notifier_block ipvlan_notifier_block __read_mostly
= {
986 .notifier_call
= ipvlan_device_event
,
989 #if IS_ENABLED(CONFIG_IPV6)
990 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly
= {
991 .notifier_call
= ipvlan_addr6_event
,
994 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly
= {
995 .notifier_call
= ipvlan_addr6_validator_event
,
999 static int __init
ipvlan_init_module(void)
1003 ipvlan_init_secret();
1004 register_netdevice_notifier(&ipvlan_notifier_block
);
1005 #if IS_ENABLED(CONFIG_IPV6)
1006 register_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1007 register_inet6addr_validator_notifier(
1008 &ipvlan_addr6_vtor_notifier_block
);
1010 register_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1011 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block
);
1013 err
= ipvlan_l3s_init();
1017 err
= ipvlan_link_register(&ipvlan_link_ops
);
1019 ipvlan_l3s_cleanup();
1025 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1026 unregister_inetaddr_validator_notifier(
1027 &ipvlan_addr4_vtor_notifier_block
);
1028 #if IS_ENABLED(CONFIG_IPV6)
1029 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1030 unregister_inet6addr_validator_notifier(
1031 &ipvlan_addr6_vtor_notifier_block
);
1033 unregister_netdevice_notifier(&ipvlan_notifier_block
);
1037 static void __exit
ipvlan_cleanup_module(void)
1039 rtnl_link_unregister(&ipvlan_link_ops
);
1040 ipvlan_l3s_cleanup();
1041 unregister_netdevice_notifier(&ipvlan_notifier_block
);
1042 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block
);
1043 unregister_inetaddr_validator_notifier(
1044 &ipvlan_addr4_vtor_notifier_block
);
1045 #if IS_ENABLED(CONFIG_IPV6)
1046 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block
);
1047 unregister_inet6addr_validator_notifier(
1048 &ipvlan_addr6_vtor_notifier_block
);
1052 module_init(ipvlan_init_module
);
1053 module_exit(ipvlan_cleanup_module
);
1055 MODULE_LICENSE("GPL");
1056 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
1057 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
1058 MODULE_ALIAS_RTNL_LINK("ipvlan");