Commit | Line | Data |
---|---|---|
2ad7bf36 MB |
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 | ||
ab5b7013 | 12 | static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev) |
2ad7bf36 MB |
13 | { |
14 | ipvlan->dev->mtu = dev->mtu - ipvlan->mtu_adj; | |
15 | } | |
16 | ||
e93fbc5a | 17 | static void ipvlan_set_port_mode(struct ipvl_port *port, u16 nval) |
2ad7bf36 MB |
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 | ||
32 | static 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 | } | |
764e433b MB |
41 | |
42 | if (netif_is_macvlan_port(dev)) { | |
43 | netdev_err(dev, "Master is a macvlan port.\n"); | |
44 | return -EBUSY; | |
45 | } | |
46 | ||
2ad7bf36 MB |
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 | ||
ba35f858 MB |
57 | skb_queue_head_init(&port->backlog); |
58 | INIT_WORK(&port->wq, ipvlan_process_multicast); | |
59 | ||
2ad7bf36 MB |
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 | ||
67 | err: | |
68 | kfree_rcu(port, rcu); | |
69 | return err; | |
70 | } | |
71 | ||
72 | static 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); | |
ba35f858 MB |
78 | cancel_work_sync(&port->wq); |
79 | __skb_queue_purge(&port->backlog); | |
2ad7bf36 MB |
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 | */ | |
87 | static struct lock_class_key ipvlan_netdev_xmit_lock_key; | |
88 | static struct lock_class_key ipvlan_netdev_addr_lock_key; | |
89 | ||
90 | #define IPVLAN_FEATURES \ | |
a188222b | 91 | (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ |
2ad7bf36 MB |
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 | ||
99 | static 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 | ||
106 | static 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 | ||
112 | static 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; | |
2ad7bf36 MB |
122 | dev->hard_header_len = phy_dev->hard_header_len; |
123 | ||
124 | ipvlan_set_lockdep_class(dev); | |
125 | ||
126 | ipvlan->pcpu_stats = alloc_percpu(struct ipvl_pcpu_stats); | |
127 | if (!ipvlan->pcpu_stats) | |
128 | return -ENOMEM; | |
129 | ||
130 | return 0; | |
131 | } | |
132 | ||
133 | static void ipvlan_uninit(struct net_device *dev) | |
134 | { | |
135 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
136 | struct ipvl_port *port = ipvlan->port; | |
137 | ||
04901cea | 138 | free_percpu(ipvlan->pcpu_stats); |
2ad7bf36 MB |
139 | |
140 | port->count -= 1; | |
141 | if (!port->count) | |
142 | ipvlan_port_destroy(port->dev); | |
143 | } | |
144 | ||
145 | static int ipvlan_open(struct net_device *dev) | |
146 | { | |
147 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
148 | struct net_device *phy_dev = ipvlan->phy_dev; | |
149 | struct ipvl_addr *addr; | |
150 | ||
151 | if (ipvlan->port->mode == IPVLAN_MODE_L3) | |
152 | dev->flags |= IFF_NOARP; | |
153 | else | |
154 | dev->flags &= ~IFF_NOARP; | |
155 | ||
515866f8 KK |
156 | list_for_each_entry(addr, &ipvlan->addrs, anode) |
157 | ipvlan_ht_addr_add(ipvlan, addr); | |
158 | ||
2ad7bf36 MB |
159 | return dev_uc_add(phy_dev, phy_dev->dev_addr); |
160 | } | |
161 | ||
162 | static int ipvlan_stop(struct net_device *dev) | |
163 | { | |
164 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
165 | struct net_device *phy_dev = ipvlan->phy_dev; | |
166 | struct ipvl_addr *addr; | |
167 | ||
168 | dev_uc_unsync(phy_dev, dev); | |
169 | dev_mc_unsync(phy_dev, dev); | |
170 | ||
171 | dev_uc_del(phy_dev, phy_dev->dev_addr); | |
172 | ||
515866f8 | 173 | list_for_each_entry(addr, &ipvlan->addrs, anode) |
6640e673 | 174 | ipvlan_ht_addr_del(addr); |
515866f8 | 175 | |
2ad7bf36 MB |
176 | return 0; |
177 | } | |
178 | ||
92c7b0de MB |
179 | static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb, |
180 | struct net_device *dev) | |
2ad7bf36 MB |
181 | { |
182 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
183 | int skblen = skb->len; | |
184 | int ret; | |
185 | ||
186 | ret = ipvlan_queue_xmit(skb, dev); | |
187 | if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { | |
188 | struct ipvl_pcpu_stats *pcptr; | |
189 | ||
190 | pcptr = this_cpu_ptr(ipvlan->pcpu_stats); | |
191 | ||
192 | u64_stats_update_begin(&pcptr->syncp); | |
193 | pcptr->tx_pkts++; | |
194 | pcptr->tx_bytes += skblen; | |
195 | u64_stats_update_end(&pcptr->syncp); | |
196 | } else { | |
197 | this_cpu_inc(ipvlan->pcpu_stats->tx_drps); | |
198 | } | |
199 | return ret; | |
200 | } | |
201 | ||
202 | static netdev_features_t ipvlan_fix_features(struct net_device *dev, | |
203 | netdev_features_t features) | |
204 | { | |
205 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
206 | ||
207 | return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES); | |
208 | } | |
209 | ||
210 | static void ipvlan_change_rx_flags(struct net_device *dev, int change) | |
211 | { | |
212 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
213 | struct net_device *phy_dev = ipvlan->phy_dev; | |
214 | ||
215 | if (change & IFF_ALLMULTI) | |
216 | dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1); | |
217 | } | |
218 | ||
2ad7bf36 MB |
219 | static void ipvlan_set_multicast_mac_filter(struct net_device *dev) |
220 | { | |
221 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
222 | ||
223 | if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { | |
224 | bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE); | |
225 | } else { | |
226 | struct netdev_hw_addr *ha; | |
227 | DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE); | |
228 | ||
229 | bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE); | |
230 | netdev_for_each_mc_addr(ha, dev) | |
231 | __set_bit(ipvlan_mac_hash(ha->addr), mc_filters); | |
232 | ||
f631c44b MB |
233 | /* Turn-on broadcast bit irrespective of address family, |
234 | * since broadcast is deferred to a work-queue, hence no | |
235 | * impact on fast-path processing. | |
236 | */ | |
237 | __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters); | |
238 | ||
2ad7bf36 MB |
239 | bitmap_copy(ipvlan->mac_filters, mc_filters, |
240 | IPVLAN_MAC_FILTER_SIZE); | |
241 | } | |
242 | dev_uc_sync(ipvlan->phy_dev, dev); | |
243 | dev_mc_sync(ipvlan->phy_dev, dev); | |
244 | } | |
245 | ||
246 | static struct rtnl_link_stats64 *ipvlan_get_stats64(struct net_device *dev, | |
247 | struct rtnl_link_stats64 *s) | |
248 | { | |
249 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
250 | ||
251 | if (ipvlan->pcpu_stats) { | |
252 | struct ipvl_pcpu_stats *pcptr; | |
253 | u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes; | |
254 | u32 rx_errs = 0, tx_drps = 0; | |
255 | u32 strt; | |
256 | int idx; | |
257 | ||
258 | for_each_possible_cpu(idx) { | |
259 | pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx); | |
260 | do { | |
261 | strt= u64_stats_fetch_begin_irq(&pcptr->syncp); | |
262 | rx_pkts = pcptr->rx_pkts; | |
263 | rx_bytes = pcptr->rx_bytes; | |
264 | rx_mcast = pcptr->rx_mcast; | |
265 | tx_pkts = pcptr->tx_pkts; | |
266 | tx_bytes = pcptr->tx_bytes; | |
267 | } while (u64_stats_fetch_retry_irq(&pcptr->syncp, | |
268 | strt)); | |
269 | ||
270 | s->rx_packets += rx_pkts; | |
271 | s->rx_bytes += rx_bytes; | |
272 | s->multicast += rx_mcast; | |
273 | s->tx_packets += tx_pkts; | |
274 | s->tx_bytes += tx_bytes; | |
275 | ||
276 | /* u32 values are updated without syncp protection. */ | |
277 | rx_errs += pcptr->rx_errs; | |
278 | tx_drps += pcptr->tx_drps; | |
279 | } | |
280 | s->rx_errors = rx_errs; | |
281 | s->rx_dropped = rx_errs; | |
282 | s->tx_dropped = tx_drps; | |
283 | } | |
284 | return s; | |
285 | } | |
286 | ||
287 | static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid) | |
288 | { | |
289 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
290 | struct net_device *phy_dev = ipvlan->phy_dev; | |
291 | ||
292 | return vlan_vid_add(phy_dev, proto, vid); | |
293 | } | |
294 | ||
295 | static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto, | |
296 | u16 vid) | |
297 | { | |
298 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
299 | struct net_device *phy_dev = ipvlan->phy_dev; | |
300 | ||
301 | vlan_vid_del(phy_dev, proto, vid); | |
302 | return 0; | |
303 | } | |
304 | ||
7c411658 ND |
305 | static int ipvlan_get_iflink(const struct net_device *dev) |
306 | { | |
307 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
308 | ||
309 | return ipvlan->phy_dev->ifindex; | |
310 | } | |
311 | ||
2ad7bf36 MB |
312 | static const struct net_device_ops ipvlan_netdev_ops = { |
313 | .ndo_init = ipvlan_init, | |
314 | .ndo_uninit = ipvlan_uninit, | |
315 | .ndo_open = ipvlan_open, | |
316 | .ndo_stop = ipvlan_stop, | |
317 | .ndo_start_xmit = ipvlan_start_xmit, | |
318 | .ndo_fix_features = ipvlan_fix_features, | |
319 | .ndo_change_rx_flags = ipvlan_change_rx_flags, | |
320 | .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter, | |
321 | .ndo_get_stats64 = ipvlan_get_stats64, | |
322 | .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid, | |
323 | .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid, | |
7c411658 | 324 | .ndo_get_iflink = ipvlan_get_iflink, |
2ad7bf36 MB |
325 | }; |
326 | ||
327 | static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev, | |
328 | unsigned short type, const void *daddr, | |
329 | const void *saddr, unsigned len) | |
330 | { | |
331 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
332 | struct net_device *phy_dev = ipvlan->phy_dev; | |
333 | ||
334 | /* TODO Probably use a different field than dev_addr so that the | |
335 | * mac-address on the virtual device is portable and can be carried | |
336 | * while the packets use the mac-addr on the physical device. | |
337 | */ | |
338 | return dev_hard_header(skb, phy_dev, type, daddr, | |
339 | saddr ? : dev->dev_addr, len); | |
340 | } | |
341 | ||
342 | static const struct header_ops ipvlan_header_ops = { | |
343 | .create = ipvlan_hard_header, | |
2ad7bf36 MB |
344 | .parse = eth_header_parse, |
345 | .cache = eth_header_cache, | |
346 | .cache_update = eth_header_cache_update, | |
347 | }; | |
348 | ||
314d10d7 DD |
349 | static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev, |
350 | struct ethtool_link_ksettings *cmd) | |
2ad7bf36 MB |
351 | { |
352 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
353 | ||
314d10d7 | 354 | return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd); |
2ad7bf36 MB |
355 | } |
356 | ||
357 | static void ipvlan_ethtool_get_drvinfo(struct net_device *dev, | |
358 | struct ethtool_drvinfo *drvinfo) | |
359 | { | |
360 | strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver)); | |
361 | strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version)); | |
362 | } | |
363 | ||
364 | static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev) | |
365 | { | |
366 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
367 | ||
368 | return ipvlan->msg_enable; | |
369 | } | |
370 | ||
371 | static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value) | |
372 | { | |
373 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
374 | ||
375 | ipvlan->msg_enable = value; | |
376 | } | |
377 | ||
378 | static const struct ethtool_ops ipvlan_ethtool_ops = { | |
379 | .get_link = ethtool_op_get_link, | |
314d10d7 | 380 | .get_link_ksettings = ipvlan_ethtool_get_link_ksettings, |
2ad7bf36 MB |
381 | .get_drvinfo = ipvlan_ethtool_get_drvinfo, |
382 | .get_msglevel = ipvlan_ethtool_get_msglevel, | |
383 | .set_msglevel = ipvlan_ethtool_set_msglevel, | |
384 | }; | |
385 | ||
386 | static int ipvlan_nl_changelink(struct net_device *dev, | |
387 | struct nlattr *tb[], struct nlattr *data[]) | |
388 | { | |
389 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
390 | struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); | |
391 | ||
392 | if (data && data[IFLA_IPVLAN_MODE]) { | |
393 | u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]); | |
394 | ||
395 | ipvlan_set_port_mode(port, nmode); | |
396 | } | |
397 | return 0; | |
398 | } | |
399 | ||
400 | static size_t ipvlan_nl_getsize(const struct net_device *dev) | |
401 | { | |
402 | return (0 | |
403 | + nla_total_size(2) /* IFLA_IPVLAN_MODE */ | |
404 | ); | |
405 | } | |
406 | ||
407 | static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[]) | |
408 | { | |
409 | if (data && data[IFLA_IPVLAN_MODE]) { | |
410 | u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); | |
411 | ||
412 | if (mode < IPVLAN_MODE_L2 || mode >= IPVLAN_MODE_MAX) | |
413 | return -EINVAL; | |
414 | } | |
415 | return 0; | |
416 | } | |
417 | ||
418 | static int ipvlan_nl_fillinfo(struct sk_buff *skb, | |
419 | const struct net_device *dev) | |
420 | { | |
421 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
422 | struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); | |
423 | int ret = -EINVAL; | |
424 | ||
425 | if (!port) | |
426 | goto err; | |
427 | ||
428 | ret = -EMSGSIZE; | |
429 | if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode)) | |
430 | goto err; | |
431 | ||
432 | return 0; | |
433 | ||
434 | err: | |
435 | return ret; | |
436 | } | |
437 | ||
438 | static int ipvlan_link_new(struct net *src_net, struct net_device *dev, | |
439 | struct nlattr *tb[], struct nlattr *data[]) | |
440 | { | |
441 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
442 | struct ipvl_port *port; | |
443 | struct net_device *phy_dev; | |
444 | int err; | |
e93fbc5a | 445 | u16 mode = IPVLAN_MODE_L3; |
2ad7bf36 MB |
446 | |
447 | if (!tb[IFLA_LINK]) | |
448 | return -EINVAL; | |
449 | ||
450 | phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); | |
451 | if (!phy_dev) | |
452 | return -ENODEV; | |
453 | ||
5933fea7 | 454 | if (netif_is_ipvlan(phy_dev)) { |
2ad7bf36 MB |
455 | struct ipvl_dev *tmp = netdev_priv(phy_dev); |
456 | ||
457 | phy_dev = tmp->phy_dev; | |
5933fea7 | 458 | } else if (!netif_is_ipvlan_port(phy_dev)) { |
2ad7bf36 MB |
459 | err = ipvlan_port_create(phy_dev); |
460 | if (err < 0) | |
461 | return err; | |
462 | } | |
463 | ||
2ad7bf36 | 464 | if (data && data[IFLA_IPVLAN_MODE]) |
e93fbc5a | 465 | mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); |
2ad7bf36 | 466 | |
e93fbc5a | 467 | port = ipvlan_port_get_rtnl(phy_dev); |
2ad7bf36 MB |
468 | ipvlan->phy_dev = phy_dev; |
469 | ipvlan->dev = dev; | |
470 | ipvlan->port = port; | |
471 | ipvlan->sfeatures = IPVLAN_FEATURES; | |
296d4856 | 472 | ipvlan_adjust_mtu(ipvlan, phy_dev); |
2ad7bf36 | 473 | INIT_LIST_HEAD(&ipvlan->addrs); |
2ad7bf36 MB |
474 | |
475 | /* TODO Probably put random address here to be presented to the | |
476 | * world but keep using the physical-dev address for the outgoing | |
477 | * packets. | |
478 | */ | |
479 | memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN); | |
480 | ||
481 | dev->priv_flags |= IFF_IPVLAN_SLAVE; | |
482 | ||
483 | port->count += 1; | |
484 | err = register_netdevice(dev); | |
485 | if (err < 0) | |
486 | goto ipvlan_destroy_port; | |
487 | ||
488 | err = netdev_upper_dev_link(phy_dev, dev); | |
489 | if (err) | |
490 | goto ipvlan_destroy_port; | |
491 | ||
492 | list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans); | |
e93fbc5a MB |
493 | ipvlan_set_port_mode(port, mode); |
494 | ||
2ad7bf36 MB |
495 | netif_stacked_transfer_operstate(phy_dev, dev); |
496 | return 0; | |
497 | ||
498 | ipvlan_destroy_port: | |
499 | port->count -= 1; | |
500 | if (!port->count) | |
501 | ipvlan_port_destroy(phy_dev); | |
502 | ||
503 | return err; | |
504 | } | |
505 | ||
506 | static void ipvlan_link_delete(struct net_device *dev, struct list_head *head) | |
507 | { | |
508 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
509 | struct ipvl_addr *addr, *next; | |
510 | ||
515866f8 | 511 | list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) { |
6640e673 | 512 | ipvlan_ht_addr_del(addr); |
515866f8 | 513 | list_del(&addr->anode); |
6a725497 | 514 | kfree_rcu(addr, rcu); |
2ad7bf36 | 515 | } |
515866f8 | 516 | |
2ad7bf36 MB |
517 | list_del_rcu(&ipvlan->pnode); |
518 | unregister_netdevice_queue(dev, head); | |
519 | netdev_upper_dev_unlink(ipvlan->phy_dev, dev); | |
520 | } | |
521 | ||
522 | static void ipvlan_link_setup(struct net_device *dev) | |
523 | { | |
524 | ether_setup(dev); | |
525 | ||
526 | dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); | |
bf485bcf | 527 | dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE; |
2ad7bf36 MB |
528 | dev->netdev_ops = &ipvlan_netdev_ops; |
529 | dev->destructor = free_netdev; | |
530 | dev->header_ops = &ipvlan_header_ops; | |
531 | dev->ethtool_ops = &ipvlan_ethtool_ops; | |
2ad7bf36 MB |
532 | } |
533 | ||
534 | static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] = | |
535 | { | |
536 | [IFLA_IPVLAN_MODE] = { .type = NLA_U16 }, | |
537 | }; | |
538 | ||
539 | static struct rtnl_link_ops ipvlan_link_ops = { | |
540 | .kind = "ipvlan", | |
541 | .priv_size = sizeof(struct ipvl_dev), | |
542 | ||
543 | .get_size = ipvlan_nl_getsize, | |
544 | .policy = ipvlan_nl_policy, | |
545 | .validate = ipvlan_nl_validate, | |
546 | .fill_info = ipvlan_nl_fillinfo, | |
547 | .changelink = ipvlan_nl_changelink, | |
548 | .maxtype = IFLA_IPVLAN_MAX, | |
549 | ||
550 | .setup = ipvlan_link_setup, | |
551 | .newlink = ipvlan_link_new, | |
552 | .dellink = ipvlan_link_delete, | |
553 | }; | |
554 | ||
92c7b0de | 555 | static int ipvlan_link_register(struct rtnl_link_ops *ops) |
2ad7bf36 MB |
556 | { |
557 | return rtnl_link_register(ops); | |
558 | } | |
559 | ||
560 | static int ipvlan_device_event(struct notifier_block *unused, | |
561 | unsigned long event, void *ptr) | |
562 | { | |
563 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
564 | struct ipvl_dev *ipvlan, *next; | |
565 | struct ipvl_port *port; | |
566 | LIST_HEAD(lst_kill); | |
567 | ||
5933fea7 | 568 | if (!netif_is_ipvlan_port(dev)) |
2ad7bf36 MB |
569 | return NOTIFY_DONE; |
570 | ||
571 | port = ipvlan_port_get_rtnl(dev); | |
572 | ||
573 | switch (event) { | |
574 | case NETDEV_CHANGE: | |
575 | list_for_each_entry(ipvlan, &port->ipvlans, pnode) | |
576 | netif_stacked_transfer_operstate(ipvlan->phy_dev, | |
577 | ipvlan->dev); | |
578 | break; | |
579 | ||
580 | case NETDEV_UNREGISTER: | |
581 | if (dev->reg_state != NETREG_UNREGISTERING) | |
582 | break; | |
583 | ||
584 | list_for_each_entry_safe(ipvlan, next, &port->ipvlans, | |
585 | pnode) | |
586 | ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev, | |
587 | &lst_kill); | |
588 | unregister_netdevice_many(&lst_kill); | |
589 | break; | |
590 | ||
591 | case NETDEV_FEAT_CHANGE: | |
592 | list_for_each_entry(ipvlan, &port->ipvlans, pnode) { | |
593 | ipvlan->dev->features = dev->features & IPVLAN_FEATURES; | |
594 | ipvlan->dev->gso_max_size = dev->gso_max_size; | |
595 | netdev_features_change(ipvlan->dev); | |
596 | } | |
597 | break; | |
598 | ||
599 | case NETDEV_CHANGEMTU: | |
600 | list_for_each_entry(ipvlan, &port->ipvlans, pnode) | |
601 | ipvlan_adjust_mtu(ipvlan, dev); | |
602 | break; | |
603 | ||
604 | case NETDEV_PRE_TYPE_CHANGE: | |
605 | /* Forbid underlying device to change its type. */ | |
606 | return NOTIFY_BAD; | |
607 | } | |
608 | return NOTIFY_DONE; | |
609 | } | |
610 | ||
611 | static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) | |
612 | { | |
613 | struct ipvl_addr *addr; | |
614 | ||
e9997c29 | 615 | if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true)) { |
2ad7bf36 MB |
616 | netif_err(ipvlan, ifup, ipvlan->dev, |
617 | "Failed to add IPv6=%pI6c addr for %s intf\n", | |
618 | ip6_addr, ipvlan->dev->name); | |
619 | return -EINVAL; | |
620 | } | |
621 | addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC); | |
622 | if (!addr) | |
623 | return -ENOMEM; | |
624 | ||
625 | addr->master = ipvlan; | |
626 | memcpy(&addr->ip6addr, ip6_addr, sizeof(struct in6_addr)); | |
627 | addr->atype = IPVL_IPV6; | |
40891e8a | 628 | list_add_tail(&addr->anode, &ipvlan->addrs); |
515866f8 | 629 | |
27705f70 JB |
630 | /* If the interface is not up, the address will be added to the hash |
631 | * list by ipvlan_open. | |
632 | */ | |
633 | if (netif_running(ipvlan->dev)) | |
634 | ipvlan_ht_addr_add(ipvlan, addr); | |
2ad7bf36 MB |
635 | |
636 | return 0; | |
637 | } | |
638 | ||
639 | static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) | |
640 | { | |
641 | struct ipvl_addr *addr; | |
642 | ||
e9997c29 | 643 | addr = ipvlan_find_addr(ipvlan, ip6_addr, true); |
2ad7bf36 MB |
644 | if (!addr) |
645 | return; | |
646 | ||
6640e673 | 647 | ipvlan_ht_addr_del(addr); |
40891e8a | 648 | list_del(&addr->anode); |
2ad7bf36 MB |
649 | kfree_rcu(addr, rcu); |
650 | ||
651 | return; | |
652 | } | |
653 | ||
654 | static int ipvlan_addr6_event(struct notifier_block *unused, | |
655 | unsigned long event, void *ptr) | |
656 | { | |
657 | struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr; | |
658 | struct net_device *dev = (struct net_device *)if6->idev->dev; | |
659 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
660 | ||
23a5a49c KK |
661 | /* FIXME IPv6 autoconf calls us from bh without RTNL */ |
662 | if (in_softirq()) | |
663 | return NOTIFY_DONE; | |
664 | ||
5933fea7 | 665 | if (!netif_is_ipvlan(dev)) |
2ad7bf36 MB |
666 | return NOTIFY_DONE; |
667 | ||
668 | if (!ipvlan || !ipvlan->port) | |
669 | return NOTIFY_DONE; | |
670 | ||
671 | switch (event) { | |
672 | case NETDEV_UP: | |
673 | if (ipvlan_add_addr6(ipvlan, &if6->addr)) | |
674 | return NOTIFY_BAD; | |
675 | break; | |
676 | ||
677 | case NETDEV_DOWN: | |
678 | ipvlan_del_addr6(ipvlan, &if6->addr); | |
679 | break; | |
680 | } | |
681 | ||
682 | return NOTIFY_OK; | |
683 | } | |
684 | ||
685 | static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) | |
686 | { | |
687 | struct ipvl_addr *addr; | |
688 | ||
e9997c29 | 689 | if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false)) { |
2ad7bf36 MB |
690 | netif_err(ipvlan, ifup, ipvlan->dev, |
691 | "Failed to add IPv4=%pI4 on %s intf.\n", | |
692 | ip4_addr, ipvlan->dev->name); | |
693 | return -EINVAL; | |
694 | } | |
695 | addr = kzalloc(sizeof(struct ipvl_addr), GFP_KERNEL); | |
696 | if (!addr) | |
697 | return -ENOMEM; | |
698 | ||
699 | addr->master = ipvlan; | |
700 | memcpy(&addr->ip4addr, ip4_addr, sizeof(struct in_addr)); | |
701 | addr->atype = IPVL_IPV4; | |
40891e8a | 702 | list_add_tail(&addr->anode, &ipvlan->addrs); |
515866f8 | 703 | |
27705f70 JB |
704 | /* If the interface is not up, the address will be added to the hash |
705 | * list by ipvlan_open. | |
706 | */ | |
707 | if (netif_running(ipvlan->dev)) | |
708 | ipvlan_ht_addr_add(ipvlan, addr); | |
2ad7bf36 MB |
709 | |
710 | return 0; | |
711 | } | |
712 | ||
713 | static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) | |
714 | { | |
715 | struct ipvl_addr *addr; | |
716 | ||
e9997c29 | 717 | addr = ipvlan_find_addr(ipvlan, ip4_addr, false); |
2ad7bf36 MB |
718 | if (!addr) |
719 | return; | |
720 | ||
6640e673 | 721 | ipvlan_ht_addr_del(addr); |
40891e8a | 722 | list_del(&addr->anode); |
2ad7bf36 MB |
723 | kfree_rcu(addr, rcu); |
724 | ||
725 | return; | |
726 | } | |
727 | ||
728 | static int ipvlan_addr4_event(struct notifier_block *unused, | |
729 | unsigned long event, void *ptr) | |
730 | { | |
731 | struct in_ifaddr *if4 = (struct in_ifaddr *)ptr; | |
732 | struct net_device *dev = (struct net_device *)if4->ifa_dev->dev; | |
733 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
734 | struct in_addr ip4_addr; | |
735 | ||
5933fea7 | 736 | if (!netif_is_ipvlan(dev)) |
2ad7bf36 MB |
737 | return NOTIFY_DONE; |
738 | ||
739 | if (!ipvlan || !ipvlan->port) | |
740 | return NOTIFY_DONE; | |
741 | ||
742 | switch (event) { | |
743 | case NETDEV_UP: | |
744 | ip4_addr.s_addr = if4->ifa_address; | |
745 | if (ipvlan_add_addr4(ipvlan, &ip4_addr)) | |
746 | return NOTIFY_BAD; | |
747 | break; | |
748 | ||
749 | case NETDEV_DOWN: | |
750 | ip4_addr.s_addr = if4->ifa_address; | |
751 | ipvlan_del_addr4(ipvlan, &ip4_addr); | |
752 | break; | |
753 | } | |
754 | ||
755 | return NOTIFY_OK; | |
756 | } | |
757 | ||
758 | static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = { | |
759 | .notifier_call = ipvlan_addr4_event, | |
760 | }; | |
761 | ||
762 | static struct notifier_block ipvlan_notifier_block __read_mostly = { | |
763 | .notifier_call = ipvlan_device_event, | |
764 | }; | |
765 | ||
766 | static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = { | |
767 | .notifier_call = ipvlan_addr6_event, | |
768 | }; | |
769 | ||
770 | static int __init ipvlan_init_module(void) | |
771 | { | |
772 | int err; | |
773 | ||
774 | ipvlan_init_secret(); | |
775 | register_netdevice_notifier(&ipvlan_notifier_block); | |
776 | register_inet6addr_notifier(&ipvlan_addr6_notifier_block); | |
777 | register_inetaddr_notifier(&ipvlan_addr4_notifier_block); | |
778 | ||
779 | err = ipvlan_link_register(&ipvlan_link_ops); | |
780 | if (err < 0) | |
781 | goto error; | |
782 | ||
783 | return 0; | |
784 | error: | |
785 | unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); | |
786 | unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); | |
787 | unregister_netdevice_notifier(&ipvlan_notifier_block); | |
788 | return err; | |
789 | } | |
790 | ||
791 | static void __exit ipvlan_cleanup_module(void) | |
792 | { | |
793 | rtnl_link_unregister(&ipvlan_link_ops); | |
794 | unregister_netdevice_notifier(&ipvlan_notifier_block); | |
795 | unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); | |
796 | unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); | |
797 | } | |
798 | ||
799 | module_init(ipvlan_init_module); | |
800 | module_exit(ipvlan_cleanup_module); | |
801 | ||
802 | MODULE_LICENSE("GPL"); | |
803 | MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>"); | |
804 | MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs"); | |
805 | MODULE_ALIAS_RTNL_LINK("ipvlan"); |