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 ret
= RX_HANDLER_ANOTHER
;
290 ipvlan_count_rx(ipvlan
, len
, success
, false);
294 static struct ipvl_addr
*ipvlan_addr_lookup(struct ipvl_port
*port
,
295 void *lyr3h
, int addr_type
,
298 struct ipvl_addr
*addr
= NULL
;
300 if (addr_type
== IPVL_IPV6
) {
301 struct ipv6hdr
*ip6h
;
302 struct in6_addr
*i6addr
;
304 ip6h
= (struct ipv6hdr
*)lyr3h
;
305 i6addr
= use_dest
? &ip6h
->daddr
: &ip6h
->saddr
;
306 addr
= ipvlan_ht_addr_lookup(port
, i6addr
, true);
307 } else if (addr_type
== IPVL_ICMPV6
) {
309 struct in6_addr
*i6addr
;
311 /* Make sure that the NeighborSolicitation ICMPv6 packets
312 * are handled to avoid DAD issue.
314 ndmh
= (struct nd_msg
*)lyr3h
;
315 if (ndmh
->icmph
.icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
) {
316 i6addr
= &ndmh
->target
;
317 addr
= ipvlan_ht_addr_lookup(port
, i6addr
, true);
319 } else if (addr_type
== IPVL_IPV4
) {
323 ip4h
= (struct iphdr
*)lyr3h
;
324 i4addr
= use_dest
? &ip4h
->daddr
: &ip4h
->saddr
;
325 addr
= ipvlan_ht_addr_lookup(port
, i4addr
, false);
326 } else if (addr_type
== IPVL_ARP
) {
328 unsigned char *arp_ptr
;
331 arph
= (struct arphdr
*)lyr3h
;
332 arp_ptr
= (unsigned char *)(arph
+ 1);
334 arp_ptr
+= (2 * port
->dev
->addr_len
) + 4;
336 arp_ptr
+= port
->dev
->addr_len
;
338 memcpy(&dip
, arp_ptr
, 4);
339 addr
= ipvlan_ht_addr_lookup(port
, &dip
, false);
345 static int ipvlan_process_v4_outbound(struct sk_buff
*skb
)
347 const struct iphdr
*ip4h
= ip_hdr(skb
);
348 struct net_device
*dev
= skb
->dev
;
349 struct net
*net
= dev_net(dev
);
351 int err
, ret
= NET_XMIT_DROP
;
352 struct flowi4 fl4
= {
353 .flowi4_oif
= dev
->ifindex
,
354 .flowi4_tos
= RT_TOS(ip4h
->tos
),
355 .flowi4_flags
= FLOWI_FLAG_ANYSRC
,
356 .daddr
= ip4h
->daddr
,
357 .saddr
= ip4h
->saddr
,
360 rt
= ip_route_output_flow(net
, &fl4
, NULL
);
364 if (rt
->rt_type
!= RTN_UNICAST
&& rt
->rt_type
!= RTN_LOCAL
) {
369 skb_dst_set(skb
, &rt
->dst
);
370 err
= ip_local_out(net
, skb
->sk
, skb
);
371 if (unlikely(net_xmit_eval(err
)))
372 dev
->stats
.tx_errors
++;
374 ret
= NET_XMIT_SUCCESS
;
377 dev
->stats
.tx_errors
++;
383 static int ipvlan_process_v6_outbound(struct sk_buff
*skb
)
385 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
386 struct net_device
*dev
= skb
->dev
;
387 struct net
*net
= dev_net(dev
);
388 struct dst_entry
*dst
;
389 int err
, ret
= NET_XMIT_DROP
;
390 struct flowi6 fl6
= {
391 .flowi6_iif
= dev
->ifindex
,
392 .daddr
= ip6h
->daddr
,
393 .saddr
= ip6h
->saddr
,
394 .flowi6_flags
= FLOWI_FLAG_ANYSRC
,
395 .flowlabel
= ip6_flowinfo(ip6h
),
396 .flowi6_mark
= skb
->mark
,
397 .flowi6_proto
= ip6h
->nexthdr
,
400 dst
= ip6_route_output(net
, NULL
, &fl6
);
407 skb_dst_set(skb
, dst
);
408 err
= ip6_local_out(net
, skb
->sk
, skb
);
409 if (unlikely(net_xmit_eval(err
)))
410 dev
->stats
.tx_errors
++;
412 ret
= NET_XMIT_SUCCESS
;
415 dev
->stats
.tx_errors
++;
421 static int ipvlan_process_outbound(struct sk_buff
*skb
,
422 const struct ipvl_dev
*ipvlan
)
424 struct ethhdr
*ethh
= eth_hdr(skb
);
425 int ret
= NET_XMIT_DROP
;
427 /* In this mode we dont care about multicast and broadcast traffic */
428 if (is_multicast_ether_addr(ethh
->h_dest
)) {
429 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
430 ntohs(skb
->protocol
));
435 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
436 * will have L2; which need to discarded and processed further
437 * in the net-ns of the main-device.
439 if (skb_mac_header_was_set(skb
)) {
440 skb_pull(skb
, sizeof(*ethh
));
441 skb
->mac_header
= (typeof(skb
->mac_header
))~0U;
442 skb_reset_network_header(skb
);
445 if (skb
->protocol
== htons(ETH_P_IPV6
))
446 ret
= ipvlan_process_v6_outbound(skb
);
447 else if (skb
->protocol
== htons(ETH_P_IP
))
448 ret
= ipvlan_process_v4_outbound(skb
);
450 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
451 ntohs(skb
->protocol
));
458 static void ipvlan_multicast_enqueue(struct ipvl_port
*port
,
461 if (skb
->protocol
== htons(ETH_P_PAUSE
)) {
466 spin_lock(&port
->backlog
.lock
);
467 if (skb_queue_len(&port
->backlog
) < IPVLAN_QBACKLOG_LIMIT
) {
468 __skb_queue_tail(&port
->backlog
, skb
);
469 spin_unlock(&port
->backlog
.lock
);
470 schedule_work(&port
->wq
);
472 spin_unlock(&port
->backlog
.lock
);
473 atomic_long_inc(&skb
->dev
->rx_dropped
);
478 static int ipvlan_xmit_mode_l3(struct sk_buff
*skb
, struct net_device
*dev
)
480 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
482 struct ipvl_addr
*addr
;
485 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
489 addr
= ipvlan_addr_lookup(ipvlan
->port
, lyr3h
, addr_type
, true);
491 return ipvlan_rcv_frame(addr
, &skb
, true);
494 skb
->dev
= ipvlan
->phy_dev
;
495 return ipvlan_process_outbound(skb
, ipvlan
);
498 static int ipvlan_xmit_mode_l2(struct sk_buff
*skb
, struct net_device
*dev
)
500 const struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
501 struct ethhdr
*eth
= eth_hdr(skb
);
502 struct ipvl_addr
*addr
;
506 if (ether_addr_equal(eth
->h_dest
, eth
->h_source
)) {
507 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
509 addr
= ipvlan_addr_lookup(ipvlan
->port
, lyr3h
, addr_type
, true);
511 return ipvlan_rcv_frame(addr
, &skb
, true);
513 skb
= skb_share_check(skb
, GFP_ATOMIC
);
515 return NET_XMIT_DROP
;
517 /* Packet definitely does not belong to any of the
518 * virtual devices, but the dest is local. So forward
519 * the skb for the main-dev. At the RX side we just return
520 * RX_PASS for it to be processed further on the stack.
522 return dev_forward_skb(ipvlan
->phy_dev
, skb
);
524 } else if (is_multicast_ether_addr(eth
->h_dest
)) {
525 ipvlan_multicast_enqueue(ipvlan
->port
, skb
);
526 return NET_XMIT_SUCCESS
;
529 skb
->dev
= ipvlan
->phy_dev
;
530 return dev_queue_xmit(skb
);
533 int ipvlan_queue_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
535 struct ipvl_dev
*ipvlan
= netdev_priv(dev
);
536 struct ipvl_port
*port
= ipvlan_port_get_rcu_bh(ipvlan
->phy_dev
);
541 if (unlikely(!pskb_may_pull(skb
, sizeof(struct ethhdr
))))
546 return ipvlan_xmit_mode_l2(skb
, dev
);
548 return ipvlan_xmit_mode_l3(skb
, dev
);
551 /* Should not reach here */
552 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
556 return NET_XMIT_DROP
;
559 static bool ipvlan_external_frame(struct sk_buff
*skb
, struct ipvl_port
*port
)
561 struct ethhdr
*eth
= eth_hdr(skb
);
562 struct ipvl_addr
*addr
;
566 if (ether_addr_equal(eth
->h_source
, skb
->dev
->dev_addr
)) {
567 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
571 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, false);
579 static rx_handler_result_t
ipvlan_handle_mode_l3(struct sk_buff
**pskb
,
580 struct ipvl_port
*port
)
584 struct ipvl_addr
*addr
;
585 struct sk_buff
*skb
= *pskb
;
586 rx_handler_result_t ret
= RX_HANDLER_PASS
;
588 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
592 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, true);
594 ret
= ipvlan_rcv_frame(addr
, pskb
, false);
600 static rx_handler_result_t
ipvlan_handle_mode_l2(struct sk_buff
**pskb
,
601 struct ipvl_port
*port
)
603 struct sk_buff
*skb
= *pskb
;
604 struct ethhdr
*eth
= eth_hdr(skb
);
605 rx_handler_result_t ret
= RX_HANDLER_PASS
;
609 if (is_multicast_ether_addr(eth
->h_dest
)) {
610 if (ipvlan_external_frame(skb
, port
)) {
611 struct sk_buff
*nskb
= skb_clone(skb
, GFP_ATOMIC
);
613 /* External frames are queued for device local
614 * distribution, but a copy is given to master
615 * straight away to avoid sending duplicates later
616 * when work-queue processes this frame. This is
617 * achieved by returning RX_HANDLER_PASS.
620 ipvlan_multicast_enqueue(port
, nskb
);
623 struct ipvl_addr
*addr
;
625 lyr3h
= ipvlan_get_L3_hdr(skb
, &addr_type
);
629 addr
= ipvlan_addr_lookup(port
, lyr3h
, addr_type
, true);
631 ret
= ipvlan_rcv_frame(addr
, pskb
, false);
637 rx_handler_result_t
ipvlan_handle_frame(struct sk_buff
**pskb
)
639 struct sk_buff
*skb
= *pskb
;
640 struct ipvl_port
*port
= ipvlan_port_get_rcu(skb
->dev
);
643 return RX_HANDLER_PASS
;
645 switch (port
->mode
) {
647 return ipvlan_handle_mode_l2(pskb
, port
);
649 return ipvlan_handle_mode_l3(pskb
, port
);
652 /* Should not reach here */
653 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
656 return RX_HANDLER_CONSUMED
;