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 u32 ipvlan_jhash_secret __read_mostly
;
14 void ipvlan_init_secret(void)
16 net_get_random_once(&ipvlan_jhash_secret
, sizeof(ipvlan_jhash_secret
));
19 static void ipvlan_count_rx(const struct ipvl_dev
*ipvlan
,
20 unsigned int len
, bool success
, bool mcast
)
25 if (likely(success
)) {
26 struct ipvl_pcpu_stats
*pcptr
;
28 pcptr
= this_cpu_ptr(ipvlan
->pcpu_stats
);
29 u64_stats_update_begin(&pcptr
->syncp
);
31 pcptr
->rx_bytes
+= len
;
34 u64_stats_update_end(&pcptr
->syncp
);
36 this_cpu_inc(ipvlan
->pcpu_stats
->rx_errs
);
40 static u8
ipvlan_get_v6_hash(const void *iaddr
)
42 const struct in6_addr
*ip6_addr
= iaddr
;
44 return __ipv6_addr_jhash(ip6_addr
, ipvlan_jhash_secret
) &
48 static u8
ipvlan_get_v4_hash(const void *iaddr
)
50 const struct in_addr
*ip4_addr
= iaddr
;
52 return jhash_1word(ip4_addr
->s_addr
, ipvlan_jhash_secret
) &
56 struct ipvl_addr
*ipvlan_ht_addr_lookup(const struct ipvl_port
*port
,
57 const void *iaddr
, bool is_v6
)
59 struct ipvl_addr
*addr
;
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
))
68 else if (!is_v6
&& addr
->atype
== IPVL_IPV4
&&
69 addr
->ip4addr
.s_addr
==
70 ((struct in_addr
*)iaddr
)->s_addr
)
76 void ipvlan_ht_addr_add(struct ipvl_dev
*ipvlan
, struct ipvl_addr
*addr
)
78 struct ipvl_port
*port
= ipvlan
->port
;
81 hash
= (addr
->atype
== IPVL_IPV6
) ?
82 ipvlan_get_v6_hash(&addr
->ip6addr
) :
83 ipvlan_get_v4_hash(&addr
->ip4addr
);
84 if (hlist_unhashed(&addr
->hlnode
))
85 hlist_add_head_rcu(&addr
->hlnode
, &port
->hlhead
[hash
]);
88 void ipvlan_ht_addr_del(struct ipvl_addr
*addr
)
90 hlist_del_init_rcu(&addr
->hlnode
);
93 struct ipvl_addr
*ipvlan_find_addr(const struct ipvl_dev
*ipvlan
,
94 const void *iaddr
, bool is_v6
)
96 struct ipvl_addr
*addr
;
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
))
108 bool ipvlan_addr_busy(struct ipvl_port
*port
, void *iaddr
, bool is_v6
)
110 struct ipvl_dev
*ipvlan
;
114 list_for_each_entry(ipvlan
, &port
->ipvlans
, pnode
) {
115 if (ipvlan_find_addr(ipvlan
, iaddr
, is_v6
))
121 static void *ipvlan_get_L3_hdr(struct sk_buff
*skb
, int *type
)
125 switch (skb
->protocol
) {
126 case htons(ETH_P_ARP
): {
129 if (unlikely(!pskb_may_pull(skb
, sizeof(*arph
))))
137 case htons(ETH_P_IP
): {
141 if (unlikely(!pskb_may_pull(skb
, sizeof(*ip4h
))))
145 pktlen
= ntohs(ip4h
->tot_len
);
146 if (ip4h
->ihl
< 5 || ip4h
->version
!= 4)
148 if (skb
->len
< pktlen
|| pktlen
< (ip4h
->ihl
* 4))
155 case htons(ETH_P_IPV6
): {
156 struct ipv6hdr
*ip6h
;
158 if (unlikely(!pskb_may_pull(skb
, sizeof(*ip6h
))))
161 ip6h
= ipv6_hdr(skb
);
162 if (ip6h
->version
!= 6)
167 /* Only Neighbour Solicitation pkts need different treatment */
168 if (ipv6_addr_any(&ip6h
->saddr
) &&
169 ip6h
->nexthdr
== NEXTHDR_ICMP
) {
182 unsigned int ipvlan_mac_hash(const unsigned char *addr
)
184 u32 hash
= jhash_1word(__get_unaligned_cpu32(addr
+2),
185 ipvlan_jhash_secret
);
187 return hash
& IPVLAN_MAC_FILTER_MASK
;
190 void ipvlan_process_multicast(struct work_struct
*work
)
192 struct ipvl_port
*port
= container_of(work
, struct ipvl_port
, wq
);
194 struct ipvl_dev
*ipvlan
;
195 struct sk_buff
*skb
, *nskb
;
196 struct sk_buff_head list
;
198 unsigned int mac_hash
;
203 __skb_queue_head_init(&list
);
205 spin_lock_bh(&port
->backlog
.lock
);
206 skb_queue_splice_tail_init(&port
->backlog
, &list
);
207 spin_unlock_bh(&port
->backlog
.lock
);
209 while ((skb
= __skb_dequeue(&list
)) != NULL
) {
211 hlocal
= ether_addr_equal(ethh
->h_source
, port
->dev
->dev_addr
);
212 mac_hash
= ipvlan_mac_hash(ethh
->h_dest
);
214 if (ether_addr_equal(ethh
->h_dest
, port
->dev
->broadcast
))
215 pkt_type
= PACKET_BROADCAST
;
217 pkt_type
= PACKET_MULTICAST
;
221 list_for_each_entry_rcu(ipvlan
, &port
->ipvlans
, pnode
) {
222 if (hlocal
&& (ipvlan
->dev
== skb
->dev
)) {
226 if (!test_bit(mac_hash
, ipvlan
->mac_filters
))
230 len
= skb
->len
+ ETH_HLEN
;
231 nskb
= skb_clone(skb
, GFP_ATOMIC
);
235 nskb
->pkt_type
= pkt_type
;
236 nskb
->dev
= ipvlan
->dev
;
238 ret
= dev_forward_skb(ipvlan
->dev
, nskb
);
240 ret
= netif_rx(nskb
);
242 ipvlan_count_rx(ipvlan
, len
, ret
== NET_RX_SUCCESS
, true);
247 /* If the packet originated here, send it out. */
248 skb
->dev
= port
->dev
;
249 skb
->pkt_type
= pkt_type
;
257 static int ipvlan_rcv_frame(struct ipvl_addr
*addr
, struct sk_buff
**pskb
,
260 struct ipvl_dev
*ipvlan
= addr
->master
;
261 struct net_device
*dev
= ipvlan
->dev
;
263 rx_handler_result_t ret
= RX_HANDLER_CONSUMED
;
264 bool success
= false;
265 struct sk_buff
*skb
= *pskb
;
267 len
= skb
->len
+ ETH_HLEN
;
268 if (unlikely(!(dev
->flags
& IFF_UP
))) {
273 skb
= skb_share_check(skb
, GFP_ATOMIC
);
279 skb
->pkt_type
= PACKET_HOST
;
282 if (dev_forward_skb(ipvlan
->dev
, skb
) == NET_RX_SUCCESS
)
285 if (!ether_addr_equal_64bits(eth_hdr(skb
)->h_dest
,
286 ipvlan
->phy_dev
->dev_addr
))
287 skb
->pkt_type
= PACKET_OTHERHOST
;
289 ret
= RX_HANDLER_ANOTHER
;
294 ipvlan_count_rx(ipvlan
, len
, success
, false);
298 static struct ipvl_addr
*ipvlan_addr_lookup(struct ipvl_port
*port
,
299 void *lyr3h
, int addr_type
,
302 struct ipvl_addr
*addr
= NULL
;
304 if (addr_type
== IPVL_IPV6
) {
305 struct ipv6hdr
*ip6h
;
306 struct in6_addr
*i6addr
;
308 ip6h
= (struct ipv6hdr
*)lyr3h
;
309 i6addr
= use_dest
? &ip6h
->daddr
: &ip6h
->saddr
;
310 addr
= ipvlan_ht_addr_lookup(port
, i6addr
, true);
311 } else if (addr_type
== IPVL_ICMPV6
) {
313 struct in6_addr
*i6addr
;
315 /* Make sure that the NeighborSolicitation ICMPv6 packets
316 * are handled to avoid DAD issue.
318 ndmh
= (struct nd_msg
*)lyr3h
;
319 if (ndmh
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
) {
320 i6addr
= &ndmh
->target
;
321 addr
= ipvlan_ht_addr_lookup(port
, i6addr
, true);
323 } else if (addr_type
== IPVL_IPV4
) {
327 ip4h
= (struct iphdr
*)lyr3h
;
328 i4addr
= use_dest
? &ip4h
->daddr
: &ip4h
->saddr
;
329 addr
= ipvlan_ht_addr_lookup(port
, i4addr
, false);
330 } else if (addr_type
== IPVL_ARP
) {
332 unsigned char *arp_ptr
;
335 arph
= (struct arphdr
*)lyr3h
;
336 arp_ptr
= (unsigned char *)(arph
+ 1);
338 arp_ptr
+= (2 * port
->dev
->addr_len
) + 4;
340 arp_ptr
+= port
->dev
->addr_len
;
342 memcpy(&dip
, arp_ptr
, 4);
343 addr
= ipvlan_ht_addr_lookup(port
, &dip
, false);
349 static int ipvlan_process_v4_outbound(struct sk_buff
*skb
)
351 const struct iphdr
*ip4h
= ip_hdr(skb
);
352 struct net_device
*dev
= skb
->dev
;
353 struct net
*net
= dev_net(dev
);
355 int err
, ret
= NET_XMIT_DROP
;
356 struct flowi4 fl4
= {
357 .flowi4_oif
= dev
->ifindex
,
358 .flowi4_tos
= RT_TOS(ip4h
->tos
),
359 .flowi4_flags
= FLOWI_FLAG_ANYSRC
,
360 .flowi4_mark
= skb
->mark
,
361 .daddr
= ip4h
->daddr
,
362 .saddr
= ip4h
->saddr
,
365 rt
= ip_route_output_flow(net
, &fl4
, NULL
);
369 if (rt
->rt_type
!= RTN_UNICAST
&& rt
->rt_type
!= RTN_LOCAL
) {
374 skb_dst_set(skb
, &rt
->dst
);
375 err
= ip_local_out(net
, skb
->sk
, skb
);
376 if (unlikely(net_xmit_eval(err
)))
377 dev
->stats
.tx_errors
++;
379 ret
= NET_XMIT_SUCCESS
;
382 dev
->stats
.tx_errors
++;
388 static int ipvlan_process_v6_outbound(struct sk_buff
*skb
)
390 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
391 struct net_device
*dev
= skb
->dev
;
392 struct net
*net
= dev_net(dev
);
393 struct dst_entry
*dst
;
394 int err
, ret
= NET_XMIT_DROP
;
395 struct flowi6 fl6
= {
396 .flowi6_oif
= dev
->ifindex
,
397 .daddr
= ip6h
->daddr
,
398 .saddr
= ip6h
->saddr
,
399 .flowi6_flags
= FLOWI_FLAG_ANYSRC
,
400 .flowlabel
= ip6_flowinfo(ip6h
),
401 .flowi6_mark
= skb
->mark
,
402 .flowi6_proto
= ip6h
->nexthdr
,
405 dst
= ip6_route_output(net
, NULL
, &fl6
);
412 skb_dst_set(skb
, dst
);
413 err
= ip6_local_out(net
, skb
->sk
, skb
);
414 if (unlikely(net_xmit_eval(err
)))
415 dev
->stats
.tx_errors
++;
417 ret
= NET_XMIT_SUCCESS
;
420 dev
->stats
.tx_errors
++;
426 static int ipvlan_process_outbound(struct sk_buff
*skb
,
427 const struct ipvl_dev
*ipvlan
)
429 struct ethhdr
*ethh
= eth_hdr(skb
);
430 int ret
= NET_XMIT_DROP
;
432 /* In this mode we dont care about multicast and broadcast traffic */
433 if (is_multicast_ether_addr(ethh
->h_dest
)) {
434 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
435 ntohs(skb
->protocol
));
440 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
441 * will have L2; which need to discarded and processed further
442 * in the net-ns of the main-device.
444 if (skb_mac_header_was_set(skb
)) {
445 skb_pull(skb
, sizeof(*ethh
));
446 skb
->mac_header
= (typeof(skb
->mac_header
))~0U;
447 skb_reset_network_header(skb
);
450 if (skb
->protocol
== htons(ETH_P_IPV6
))
451 ret
= ipvlan_process_v6_outbound(skb
);
452 else if (skb
->protocol
== htons(ETH_P_IP
))
453 ret
= ipvlan_process_v4_outbound(skb
);
455 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
456 ntohs(skb
->protocol
));
463 static void ipvlan_multicast_enqueue(struct ipvl_port
*port
,
466 if (skb
->protocol
== htons(ETH_P_PAUSE
)) {
471 spin_lock(&port
->backlog
.lock
);
472 if (skb_queue_len(&port
->backlog
) < IPVLAN_QBACKLOG_LIMIT
) {
473 __skb_queue_tail(&port
->backlog
, skb
);
474 spin_unlock(&port
->backlog
.lock
);
475 schedule_work(&port
->wq
);
477 spin_unlock(&port
->backlog
.lock
);
478 atomic_long_inc(&skb
->dev
->rx_dropped
);
483 static int ipvlan_xmit_mode_l3(struct sk_buff
*skb
, struct net_device
*dev
)
485 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
487 struct ipvl_addr
*addr
;
490 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
494 addr
= ipvlan_addr_lookup(ipvlan
->port
, lyr3h
, addr_type
, true);
496 return ipvlan_rcv_frame(addr
, &skb
, true);
499 skb
->dev
= ipvlan
->phy_dev
;
500 return ipvlan_process_outbound(skb
, ipvlan
);
503 static int ipvlan_xmit_mode_l2(struct sk_buff
*skb
, struct net_device
*dev
)
505 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
506 struct ethhdr
*eth
= eth_hdr(skb
);
507 struct ipvl_addr
*addr
;
511 if (ether_addr_equal(eth
->h_dest
, eth
->h_source
)) {
512 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
514 addr
= ipvlan_addr_lookup(ipvlan
->port
, lyr3h
, addr_type
, true);
516 return ipvlan_rcv_frame(addr
, &skb
, true);
518 skb
= skb_share_check(skb
, GFP_ATOMIC
);
520 return NET_XMIT_DROP
;
522 /* Packet definitely does not belong to any of the
523 * virtual devices, but the dest is local. So forward
524 * the skb for the main-dev. At the RX side we just return
525 * RX_PASS for it to be processed further on the stack.
527 return dev_forward_skb(ipvlan
->phy_dev
, skb
);
529 } else if (is_multicast_ether_addr(eth
->h_dest
)) {
530 ipvlan_multicast_enqueue(ipvlan
->port
, skb
);
531 return NET_XMIT_SUCCESS
;
534 skb
->dev
= ipvlan
->phy_dev
;
535 return dev_queue_xmit(skb
);
538 int ipvlan_queue_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
540 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
541 struct ipvl_port
*port
= ipvlan_port_get_rcu_bh(ipvlan
->phy_dev
);
546 if (unlikely(!pskb_may_pull(skb
, sizeof(struct ethhdr
))))
551 return ipvlan_xmit_mode_l2(skb
, dev
);
553 return ipvlan_xmit_mode_l3(skb
, dev
);
556 /* Should not reach here */
557 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
561 return NET_XMIT_DROP
;
564 static bool ipvlan_external_frame(struct sk_buff
*skb
, struct ipvl_port
*port
)
566 struct ethhdr
*eth
= eth_hdr(skb
);
567 struct ipvl_addr
*addr
;
571 if (ether_addr_equal(eth
->h_source
, skb
->dev
->dev_addr
)) {
572 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
576 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, false);
584 static rx_handler_result_t
ipvlan_handle_mode_l3(struct sk_buff
**pskb
,
585 struct ipvl_port
*port
)
589 struct ipvl_addr
*addr
;
590 struct sk_buff
*skb
= *pskb
;
591 rx_handler_result_t ret
= RX_HANDLER_PASS
;
593 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
597 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, true);
599 ret
= ipvlan_rcv_frame(addr
, pskb
, false);
605 static rx_handler_result_t
ipvlan_handle_mode_l2(struct sk_buff
**pskb
,
606 struct ipvl_port
*port
)
608 struct sk_buff
*skb
= *pskb
;
609 struct ethhdr
*eth
= eth_hdr(skb
);
610 rx_handler_result_t ret
= RX_HANDLER_PASS
;
614 if (is_multicast_ether_addr(eth
->h_dest
)) {
615 if (ipvlan_external_frame(skb
, port
)) {
616 struct sk_buff
*nskb
= skb_clone(skb
, GFP_ATOMIC
);
618 /* External frames are queued for device local
619 * distribution, but a copy is given to master
620 * straight away to avoid sending duplicates later
621 * when work-queue processes this frame. This is
622 * achieved by returning RX_HANDLER_PASS.
625 ipvlan_multicast_enqueue(port
, nskb
);
628 struct ipvl_addr
*addr
;
630 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
634 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, true);
636 ret
= ipvlan_rcv_frame(addr
, pskb
, false);
642 rx_handler_result_t
ipvlan_handle_frame(struct sk_buff
**pskb
)
644 struct sk_buff
*skb
= *pskb
;
645 struct ipvl_port
*port
= ipvlan_port_get_rcu(skb
->dev
);
648 return RX_HANDLER_PASS
;
650 switch (port
->mode
) {
652 return ipvlan_handle_mode_l2(pskb
, port
);
654 return ipvlan_handle_mode_l3(pskb
, port
);
657 /* Should not reach here */
658 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
661 return RX_HANDLER_CONSUMED
;