blob: c747ab65266559fdb6dff77eb8078a4885c9a7ef [file] [log] [blame]
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001/* 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
Daniel Borkmann207895f2015-01-29 12:15:03 +010012static u32 ipvlan_jhash_secret __read_mostly;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080013
14void ipvlan_init_secret(void)
15{
16 net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
17}
18
19static void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
20 unsigned int len, bool success, bool mcast)
21{
22 if (!ipvlan)
23 return;
24
25 if (likely(success)) {
26 struct ipvl_pcpu_stats *pcptr;
27
28 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
29 u64_stats_update_begin(&pcptr->syncp);
30 pcptr->rx_pkts++;
31 pcptr->rx_bytes += len;
32 if (mcast)
33 pcptr->rx_mcast++;
34 u64_stats_update_end(&pcptr->syncp);
35 } else {
36 this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
37 }
38}
39
40static u8 ipvlan_get_v6_hash(const void *iaddr)
41{
42 const struct in6_addr *ip6_addr = iaddr;
43
44 return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
45 IPVLAN_HASH_MASK;
46}
47
48static u8 ipvlan_get_v4_hash(const void *iaddr)
49{
50 const struct in_addr *ip4_addr = iaddr;
51
52 return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
53 IPVLAN_HASH_MASK;
54}
55
Mahesh Bandewarab5b7012016-02-20 19:31:41 -080056static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
57 const void *iaddr, bool is_v6)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080058{
59 struct ipvl_addr *addr;
60 u8 hash;
61
62 hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
63 ipvlan_get_v4_hash(iaddr);
64 hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) {
65 if (is_v6 && addr->atype == IPVL_IPV6 &&
66 ipv6_addr_equal(&addr->ip6addr, iaddr))
67 return addr;
68 else if (!is_v6 && addr->atype == IPVL_IPV4 &&
69 addr->ip4addr.s_addr ==
70 ((struct in_addr *)iaddr)->s_addr)
71 return addr;
72 }
73 return NULL;
74}
75
76void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
77{
78 struct ipvl_port *port = ipvlan->port;
79 u8 hash;
80
81 hash = (addr->atype == IPVL_IPV6) ?
82 ipvlan_get_v6_hash(&addr->ip6addr) :
83 ipvlan_get_v4_hash(&addr->ip4addr);
Jiri Benc27705f72015-03-28 19:13:22 +010084 if (hlist_unhashed(&addr->hlnode))
85 hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080086}
87
Konstantin Khlebnikov6640e672015-07-14 16:35:53 +030088void ipvlan_ht_addr_del(struct ipvl_addr *addr)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080089{
Jiri Benc27705f72015-03-28 19:13:22 +010090 hlist_del_init_rcu(&addr->hlnode);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080091}
92
Jiri Bence9997c22015-03-28 19:13:25 +010093struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
94 const void *iaddr, bool is_v6)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080095{
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -080096 struct ipvl_addr *addr;
97
98 list_for_each_entry(addr, &ipvlan->addrs, anode) {
99 if ((is_v6 && addr->atype == IPVL_IPV6 &&
100 ipv6_addr_equal(&addr->ip6addr, iaddr)) ||
101 (!is_v6 && addr->atype == IPVL_IPV4 &&
102 addr->ip4addr.s_addr == ((struct in_addr *)iaddr)->s_addr))
Jiri Bence9997c22015-03-28 19:13:25 +0100103 return addr;
104 }
105 return NULL;
106}
107
108bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
109{
110 struct ipvl_dev *ipvlan;
111
112 ASSERT_RTNL();
113
114 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
115 if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800116 return true;
117 }
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800118 return false;
119}
120
121static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
122{
123 void *lyr3h = NULL;
124
125 switch (skb->protocol) {
126 case htons(ETH_P_ARP): {
127 struct arphdr *arph;
128
129 if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
130 return NULL;
131
132 arph = arp_hdr(skb);
133 *type = IPVL_ARP;
134 lyr3h = arph;
135 break;
136 }
137 case htons(ETH_P_IP): {
138 u32 pktlen;
139 struct iphdr *ip4h;
140
141 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
142 return NULL;
143
144 ip4h = ip_hdr(skb);
145 pktlen = ntohs(ip4h->tot_len);
146 if (ip4h->ihl < 5 || ip4h->version != 4)
147 return NULL;
148 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
149 return NULL;
150
151 *type = IPVL_IPV4;
152 lyr3h = ip4h;
153 break;
154 }
155 case htons(ETH_P_IPV6): {
156 struct ipv6hdr *ip6h;
157
158 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
159 return NULL;
160
161 ip6h = ipv6_hdr(skb);
162 if (ip6h->version != 6)
163 return NULL;
164
165 *type = IPVL_IPV6;
166 lyr3h = ip6h;
167 /* Only Neighbour Solicitation pkts need different treatment */
168 if (ipv6_addr_any(&ip6h->saddr) &&
169 ip6h->nexthdr == NEXTHDR_ICMP) {
170 *type = IPVL_ICMPV6;
171 lyr3h = ip6h + 1;
172 }
173 break;
174 }
175 default:
176 return NULL;
177 }
178
179 return lyr3h;
180}
181
182unsigned int ipvlan_mac_hash(const unsigned char *addr)
183{
184 u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
185 ipvlan_jhash_secret);
186
187 return hash & IPVLAN_MAC_FILTER_MASK;
188}
189
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700190void ipvlan_process_multicast(struct work_struct *work)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800191{
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700192 struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
193 struct ethhdr *ethh;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800194 struct ipvl_dev *ipvlan;
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700195 struct sk_buff *skb, *nskb;
196 struct sk_buff_head list;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800197 unsigned int len;
198 unsigned int mac_hash;
199 int ret;
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700200 u8 pkt_type;
201 bool hlocal, dlocal;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800202
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700203 __skb_queue_head_init(&list);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800204
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700205 spin_lock_bh(&port->backlog.lock);
206 skb_queue_splice_tail_init(&port->backlog, &list);
207 spin_unlock_bh(&port->backlog.lock);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800208
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700209 while ((skb = __skb_dequeue(&list)) != NULL) {
210 ethh = eth_hdr(skb);
211 hlocal = ether_addr_equal(ethh->h_source, port->dev->dev_addr);
212 mac_hash = ipvlan_mac_hash(ethh->h_dest);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800213
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700214 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
215 pkt_type = PACKET_BROADCAST;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800216 else
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700217 pkt_type = PACKET_MULTICAST;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800218
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700219 dlocal = false;
220 rcu_read_lock();
221 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
222 if (hlocal && (ipvlan->dev == skb->dev)) {
223 dlocal = true;
224 continue;
225 }
226 if (!test_bit(mac_hash, ipvlan->mac_filters))
227 continue;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800228
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700229 ret = NET_RX_DROP;
230 len = skb->len + ETH_HLEN;
231 nskb = skb_clone(skb, GFP_ATOMIC);
232 if (!nskb)
233 goto acct;
234
235 nskb->pkt_type = pkt_type;
236 nskb->dev = ipvlan->dev;
237 if (hlocal)
238 ret = dev_forward_skb(ipvlan->dev, nskb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800239 else
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700240 ret = netif_rx(nskb);
241acct:
242 ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
243 }
244 rcu_read_unlock();
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800245
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700246 if (dlocal) {
247 /* If the packet originated here, send it out. */
248 skb->dev = port->dev;
249 skb->pkt_type = pkt_type;
250 dev_queue_xmit(skb);
251 } else {
252 kfree_skb(skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800253 }
254 }
255}
256
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700257static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
258{
259 bool xnet = true;
260
261 if (dev)
262 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
263
264 skb_scrub_packet(skb, xnet);
265 if (dev)
266 skb->dev = dev;
267}
268
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100269static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800270 bool local)
271{
272 struct ipvl_dev *ipvlan = addr->master;
273 struct net_device *dev = ipvlan->dev;
274 unsigned int len;
275 rx_handler_result_t ret = RX_HANDLER_CONSUMED;
276 bool success = false;
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100277 struct sk_buff *skb = *pskb;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800278
279 len = skb->len + ETH_HLEN;
Mahesh Bandewarab5b7012016-02-20 19:31:41 -0800280 /* Only packets exchanged between two local slaves need to have
281 * device-up check as well as skb-share check.
282 */
283 if (local) {
284 if (unlikely(!(dev->flags & IFF_UP))) {
285 kfree_skb(skb);
286 goto out;
287 }
288
289 skb = skb_share_check(skb, GFP_ATOMIC);
290 if (!skb)
291 goto out;
292
293 *pskb = skb;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800294 }
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700295 ipvlan_skb_crossing_ns(skb, dev);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800296
297 if (local) {
Mahesh Bandewarab5b7012016-02-20 19:31:41 -0800298 skb->pkt_type = PACKET_HOST;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800299 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
300 success = true;
301 } else {
Mahesh Bandewar283b46f2017-12-07 15:15:43 -0800302 if (!ether_addr_equal_64bits(eth_hdr(skb)->h_dest,
303 ipvlan->phy_dev->dev_addr))
304 skb->pkt_type = PACKET_OTHERHOST;
305
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800306 ret = RX_HANDLER_ANOTHER;
307 success = true;
308 }
309
310out:
311 ipvlan_count_rx(ipvlan, len, success, false);
312 return ret;
313}
314
315static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
316 void *lyr3h, int addr_type,
317 bool use_dest)
318{
319 struct ipvl_addr *addr = NULL;
320
321 if (addr_type == IPVL_IPV6) {
322 struct ipv6hdr *ip6h;
323 struct in6_addr *i6addr;
324
325 ip6h = (struct ipv6hdr *)lyr3h;
326 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
327 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
328 } else if (addr_type == IPVL_ICMPV6) {
329 struct nd_msg *ndmh;
330 struct in6_addr *i6addr;
331
332 /* Make sure that the NeighborSolicitation ICMPv6 packets
333 * are handled to avoid DAD issue.
334 */
335 ndmh = (struct nd_msg *)lyr3h;
336 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
337 i6addr = &ndmh->target;
338 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
339 }
340 } else if (addr_type == IPVL_IPV4) {
341 struct iphdr *ip4h;
342 __be32 *i4addr;
343
344 ip4h = (struct iphdr *)lyr3h;
345 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
346 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
347 } else if (addr_type == IPVL_ARP) {
348 struct arphdr *arph;
349 unsigned char *arp_ptr;
350 __be32 dip;
351
352 arph = (struct arphdr *)lyr3h;
353 arp_ptr = (unsigned char *)(arph + 1);
354 if (use_dest)
355 arp_ptr += (2 * port->dev->addr_len) + 4;
356 else
357 arp_ptr += port->dev->addr_len;
358
359 memcpy(&dip, arp_ptr, 4);
360 addr = ipvlan_ht_addr_lookup(port, &dip, false);
361 }
362
363 return addr;
364}
365
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700366static int ipvlan_process_v4_outbound(struct sk_buff *skb)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800367{
368 const struct iphdr *ip4h = ip_hdr(skb);
369 struct net_device *dev = skb->dev;
Eric W. Biederman57c4bf82015-10-07 16:48:44 -0500370 struct net *net = dev_net(dev);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800371 struct rtable *rt;
372 int err, ret = NET_XMIT_DROP;
373 struct flowi4 fl4 = {
Brenden Blanco63b11e72015-10-20 16:47:33 -0700374 .flowi4_oif = dev->ifindex,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800375 .flowi4_tos = RT_TOS(ip4h->tos),
376 .flowi4_flags = FLOWI_FLAG_ANYSRC,
Gao Feng671d9012017-12-01 09:58:42 +0800377 .flowi4_mark = skb->mark,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800378 .daddr = ip4h->daddr,
379 .saddr = ip4h->saddr,
380 };
381
Eric W. Biederman57c4bf82015-10-07 16:48:44 -0500382 rt = ip_route_output_flow(net, &fl4, NULL);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800383 if (IS_ERR(rt))
384 goto err;
385
386 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
387 ip_rt_put(rt);
388 goto err;
389 }
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800390 skb_dst_set(skb, &rt->dst);
Eric W. Biederman33224b12015-10-07 16:48:46 -0500391 err = ip_local_out(net, skb->sk, skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800392 if (unlikely(net_xmit_eval(err)))
393 dev->stats.tx_errors++;
394 else
395 ret = NET_XMIT_SUCCESS;
396 goto out;
397err:
398 dev->stats.tx_errors++;
399 kfree_skb(skb);
400out:
401 return ret;
402}
403
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700404static int ipvlan_process_v6_outbound(struct sk_buff *skb)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800405{
406 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
407 struct net_device *dev = skb->dev;
Eric W. Biederman57c4bf82015-10-07 16:48:44 -0500408 struct net *net = dev_net(dev);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800409 struct dst_entry *dst;
410 int err, ret = NET_XMIT_DROP;
411 struct flowi6 fl6 = {
Keefe Liua625a162017-11-09 20:09:31 +0800412 .flowi6_oif = dev->ifindex,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800413 .daddr = ip6h->daddr,
414 .saddr = ip6h->saddr,
415 .flowi6_flags = FLOWI_FLAG_ANYSRC,
416 .flowlabel = ip6_flowinfo(ip6h),
417 .flowi6_mark = skb->mark,
418 .flowi6_proto = ip6h->nexthdr,
419 };
420
Eric W. Biederman57c4bf82015-10-07 16:48:44 -0500421 dst = ip6_route_output(net, NULL, &fl6);
Mahesh Bandewar2aab9522015-01-24 21:53:43 -0800422 if (dst->error) {
423 ret = dst->error;
424 dst_release(dst);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800425 goto err;
Mahesh Bandewar2aab9522015-01-24 21:53:43 -0800426 }
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800427 skb_dst_set(skb, dst);
Eric W. Biederman33224b12015-10-07 16:48:46 -0500428 err = ip6_local_out(net, skb->sk, skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800429 if (unlikely(net_xmit_eval(err)))
430 dev->stats.tx_errors++;
431 else
432 ret = NET_XMIT_SUCCESS;
433 goto out;
434err:
435 dev->stats.tx_errors++;
436 kfree_skb(skb);
437out:
438 return ret;
439}
440
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700441static int ipvlan_process_outbound(struct sk_buff *skb)
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800442{
443 struct ethhdr *ethh = eth_hdr(skb);
444 int ret = NET_XMIT_DROP;
445
446 /* In this mode we dont care about multicast and broadcast traffic */
447 if (is_multicast_ether_addr(ethh->h_dest)) {
448 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
449 ntohs(skb->protocol));
450 kfree_skb(skb);
451 goto out;
452 }
453
454 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
455 * will have L2; which need to discarded and processed further
456 * in the net-ns of the main-device.
457 */
458 if (skb_mac_header_was_set(skb)) {
459 skb_pull(skb, sizeof(*ethh));
460 skb->mac_header = (typeof(skb->mac_header))~0U;
461 skb_reset_network_header(skb);
462 }
463
464 if (skb->protocol == htons(ETH_P_IPV6))
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700465 ret = ipvlan_process_v6_outbound(skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800466 else if (skb->protocol == htons(ETH_P_IP))
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700467 ret = ipvlan_process_v4_outbound(skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800468 else {
469 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
470 ntohs(skb->protocol));
471 kfree_skb(skb);
472 }
473out:
474 return ret;
475}
476
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700477static void ipvlan_multicast_enqueue(struct ipvl_port *port,
478 struct sk_buff *skb)
479{
480 if (skb->protocol == htons(ETH_P_PAUSE)) {
481 kfree_skb(skb);
482 return;
483 }
484
485 spin_lock(&port->backlog.lock);
486 if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
487 __skb_queue_tail(&port->backlog, skb);
488 spin_unlock(&port->backlog.lock);
489 schedule_work(&port->wq);
490 } else {
491 spin_unlock(&port->backlog.lock);
492 atomic_long_inc(&skb->dev->rx_dropped);
493 kfree_skb(skb);
494 }
495}
496
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800497static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
498{
499 const struct ipvl_dev *ipvlan = netdev_priv(dev);
500 void *lyr3h;
501 struct ipvl_addr *addr;
502 int addr_type;
503
504 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
505 if (!lyr3h)
506 goto out;
507
508 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
509 if (addr)
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100510 return ipvlan_rcv_frame(addr, &skb, true);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800511
512out:
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700513 ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
514 return ipvlan_process_outbound(skb);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800515}
516
517static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
518{
519 const struct ipvl_dev *ipvlan = netdev_priv(dev);
520 struct ethhdr *eth = eth_hdr(skb);
521 struct ipvl_addr *addr;
522 void *lyr3h;
523 int addr_type;
524
525 if (ether_addr_equal(eth->h_dest, eth->h_source)) {
526 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
527 if (lyr3h) {
528 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
529 if (addr)
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100530 return ipvlan_rcv_frame(addr, &skb, true);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800531 }
532 skb = skb_share_check(skb, GFP_ATOMIC);
533 if (!skb)
534 return NET_XMIT_DROP;
535
536 /* Packet definitely does not belong to any of the
537 * virtual devices, but the dest is local. So forward
538 * the skb for the main-dev. At the RX side we just return
539 * RX_PASS for it to be processed further on the stack.
540 */
541 return dev_forward_skb(ipvlan->phy_dev, skb);
542
543 } else if (is_multicast_ether_addr(eth->h_dest)) {
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700544 ipvlan_skb_crossing_ns(skb, NULL);
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700545 ipvlan_multicast_enqueue(ipvlan->port, skb);
546 return NET_XMIT_SUCCESS;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800547 }
548
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700549 ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800550 return dev_queue_xmit(skb);
551}
552
553int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
554{
555 struct ipvl_dev *ipvlan = netdev_priv(dev);
WANG Cong0fba37a2015-07-14 16:35:54 +0300556 struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800557
558 if (!port)
559 goto out;
560
561 if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
562 goto out;
563
564 switch(port->mode) {
565 case IPVLAN_MODE_L2:
566 return ipvlan_xmit_mode_l2(skb, dev);
567 case IPVLAN_MODE_L3:
Mahesh Bandewar4fbae7d2016-09-16 12:59:19 -0700568 case IPVLAN_MODE_L3S:
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800569 return ipvlan_xmit_mode_l3(skb, dev);
570 }
571
572 /* Should not reach here */
573 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
574 port->mode);
575out:
576 kfree_skb(skb);
577 return NET_XMIT_DROP;
578}
579
580static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
581{
582 struct ethhdr *eth = eth_hdr(skb);
583 struct ipvl_addr *addr;
584 void *lyr3h;
585 int addr_type;
586
587 if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
588 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
589 if (!lyr3h)
590 return true;
591
592 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
593 if (addr)
594 return false;
595 }
596
597 return true;
598}
599
600static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
601 struct ipvl_port *port)
602{
603 void *lyr3h;
604 int addr_type;
605 struct ipvl_addr *addr;
606 struct sk_buff *skb = *pskb;
607 rx_handler_result_t ret = RX_HANDLER_PASS;
608
609 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
610 if (!lyr3h)
611 goto out;
612
613 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
614 if (addr)
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100615 ret = ipvlan_rcv_frame(addr, pskb, false);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800616
617out:
618 return ret;
619}
620
621static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
622 struct ipvl_port *port)
623{
624 struct sk_buff *skb = *pskb;
625 struct ethhdr *eth = eth_hdr(skb);
626 rx_handler_result_t ret = RX_HANDLER_PASS;
627 void *lyr3h;
628 int addr_type;
629
630 if (is_multicast_ether_addr(eth->h_dest)) {
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700631 if (ipvlan_external_frame(skb, port)) {
632 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
633
634 /* External frames are queued for device local
635 * distribution, but a copy is given to master
636 * straight away to avoid sending duplicates later
637 * when work-queue processes this frame. This is
638 * achieved by returning RX_HANDLER_PASS.
639 */
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700640 if (nskb) {
641 ipvlan_skb_crossing_ns(nskb, NULL);
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700642 ipvlan_multicast_enqueue(port, nskb);
Mahesh Bandewarb93dd492016-07-25 14:38:16 -0700643 }
Mahesh Bandewarba35f852015-05-04 17:06:03 -0700644 }
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800645 } else {
646 struct ipvl_addr *addr;
647
648 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
649 if (!lyr3h)
650 return ret;
651
652 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
653 if (addr)
Sabrina Dubrocacf554ad2015-11-16 22:34:26 +0100654 ret = ipvlan_rcv_frame(addr, pskb, false);
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800655 }
656
657 return ret;
658}
659
660rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
661{
662 struct sk_buff *skb = *pskb;
663 struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
664
665 if (!port)
666 return RX_HANDLER_PASS;
667
668 switch (port->mode) {
669 case IPVLAN_MODE_L2:
670 return ipvlan_handle_mode_l2(pskb, port);
671 case IPVLAN_MODE_L3:
672 return ipvlan_handle_mode_l3(pskb, port);
Mahesh Bandewar4fbae7d2016-09-16 12:59:19 -0700673 case IPVLAN_MODE_L3S:
674 return RX_HANDLER_PASS;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800675 }
676
677 /* Should not reach here */
678 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
679 port->mode);
680 kfree_skb(skb);
Sabrina Dubrocaa534dc52015-11-16 22:44:53 +0100681 return RX_HANDLER_CONSUMED;
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -0800682}
Mahesh Bandewar4fbae7d2016-09-16 12:59:19 -0700683
684static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
685 struct net_device *dev)
686{
687 struct ipvl_addr *addr = NULL;
688 struct ipvl_port *port;
689 void *lyr3h;
690 int addr_type;
691
692 if (!dev || !netif_is_ipvlan_port(dev))
693 goto out;
694
695 port = ipvlan_port_get_rcu(dev);
696 if (!port || port->mode != IPVLAN_MODE_L3S)
697 goto out;
698
699 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
700 if (!lyr3h)
701 goto out;
702
703 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
704out:
705 return addr;
706}
707
708struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
709 u16 proto)
710{
711 struct ipvl_addr *addr;
712 struct net_device *sdev;
713
714 addr = ipvlan_skb_to_addr(skb, dev);
715 if (!addr)
716 goto out;
717
718 sdev = addr->master->dev;
719 switch (proto) {
720 case AF_INET:
721 {
722 int err;
723 struct iphdr *ip4h = ip_hdr(skb);
724
725 err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
726 ip4h->tos, sdev);
727 if (unlikely(err))
728 goto out;
729 break;
730 }
731 case AF_INET6:
732 {
733 struct dst_entry *dst;
734 struct ipv6hdr *ip6h = ipv6_hdr(skb);
735 int flags = RT6_LOOKUP_F_HAS_SADDR;
736 struct flowi6 fl6 = {
737 .flowi6_iif = sdev->ifindex,
738 .daddr = ip6h->daddr,
739 .saddr = ip6h->saddr,
740 .flowlabel = ip6_flowinfo(ip6h),
741 .flowi6_mark = skb->mark,
742 .flowi6_proto = ip6h->nexthdr,
743 };
744
745 skb_dst_drop(skb);
746 dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
747 skb_dst_set(skb, dst);
748 break;
749 }
750 default:
751 break;
752 }
753
754out:
755 return skb;
756}
757
758unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
759 const struct nf_hook_state *state)
760{
761 struct ipvl_addr *addr;
762 unsigned int len;
763
764 addr = ipvlan_skb_to_addr(skb, skb->dev);
765 if (!addr)
766 goto out;
767
768 skb->dev = addr->master->dev;
769 len = skb->len + ETH_HLEN;
770 ipvlan_count_rx(addr->master, len, true, false);
771out:
772 return NF_ACCEPT;
773}