2 * Copyright (c) 2013 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/capability.h>
22 #include <linux/module.h>
23 #include <linux/types.h>
24 #include <linux/kernel.h>
25 #include <linux/slab.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/init.h>
34 #include <linux/in6.h>
35 #include <linux/inetdevice.h>
36 #include <linux/igmp.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/etherdevice.h>
39 #include <linux/if_ether.h>
40 #include <linux/if_vlan.h>
41 #include <linux/rculist.h>
42 #include <linux/err.h>
47 #include <net/protocol.h>
48 #include <net/ip_tunnels.h>
50 #include <net/checksum.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 #include <net/rtnetlink.h>
58 #include <net/dst_metadata.h>
60 #if IS_ENABLED(CONFIG_IPV6)
62 #include <net/ip6_fib.h>
63 #include <net/ip6_route.h>
66 static unsigned int ip_tunnel_hash(__be32 key
, __be32 remote
)
68 return hash_32((__force u32
)key
^ (__force u32
)remote
,
72 static bool ip_tunnel_key_match(const struct ip_tunnel_parm
*p
,
73 __be16 flags
, __be32 key
)
75 if (p
->i_flags
& TUNNEL_KEY
) {
76 if (flags
& TUNNEL_KEY
)
77 return key
== p
->i_key
;
79 /* key expected, none present */
82 return !(flags
& TUNNEL_KEY
);
85 /* Fallback tunnel: no source, no destination, no key, no options
88 We require exact key match i.e. if a key is present in packet
89 it will match only tunnel with the same key; if it is not present,
90 it will match only keyless tunnel.
92 All keysless packets, if not matched configured keyless tunnels
93 will match fallback tunnel.
94 Given src, dst and key, find appropriate for input tunnel.
96 struct ip_tunnel
*ip_tunnel_lookup(struct ip_tunnel_net
*itn
,
97 int link
, __be16 flags
,
98 __be32 remote
, __be32 local
,
102 struct ip_tunnel
*t
, *cand
= NULL
;
103 struct hlist_head
*head
;
105 hash
= ip_tunnel_hash(key
, remote
);
106 head
= &itn
->tunnels
[hash
];
108 hlist_for_each_entry_rcu(t
, head
, hash_node
) {
109 if (local
!= t
->parms
.iph
.saddr
||
110 remote
!= t
->parms
.iph
.daddr
||
111 !(t
->dev
->flags
& IFF_UP
))
114 if (!ip_tunnel_key_match(&t
->parms
, flags
, key
))
117 if (t
->parms
.link
== link
)
123 hlist_for_each_entry_rcu(t
, head
, hash_node
) {
124 if (remote
!= t
->parms
.iph
.daddr
||
125 t
->parms
.iph
.saddr
!= 0 ||
126 !(t
->dev
->flags
& IFF_UP
))
129 if (!ip_tunnel_key_match(&t
->parms
, flags
, key
))
132 if (t
->parms
.link
== link
)
138 hash
= ip_tunnel_hash(key
, 0);
139 head
= &itn
->tunnels
[hash
];
141 hlist_for_each_entry_rcu(t
, head
, hash_node
) {
142 if ((local
!= t
->parms
.iph
.saddr
|| t
->parms
.iph
.daddr
!= 0) &&
143 (local
!= t
->parms
.iph
.daddr
|| !ipv4_is_multicast(local
)))
146 if (!(t
->dev
->flags
& IFF_UP
))
149 if (!ip_tunnel_key_match(&t
->parms
, flags
, key
))
152 if (t
->parms
.link
== link
)
158 if (flags
& TUNNEL_NO_KEY
)
159 goto skip_key_lookup
;
161 hlist_for_each_entry_rcu(t
, head
, hash_node
) {
162 if (t
->parms
.i_key
!= key
||
163 t
->parms
.iph
.saddr
!= 0 ||
164 t
->parms
.iph
.daddr
!= 0 ||
165 !(t
->dev
->flags
& IFF_UP
))
168 if (t
->parms
.link
== link
)
178 t
= rcu_dereference(itn
->collect_md_tun
);
182 if (itn
->fb_tunnel_dev
&& itn
->fb_tunnel_dev
->flags
& IFF_UP
)
183 return netdev_priv(itn
->fb_tunnel_dev
);
187 EXPORT_SYMBOL_GPL(ip_tunnel_lookup
);
189 static struct hlist_head
*ip_bucket(struct ip_tunnel_net
*itn
,
190 struct ip_tunnel_parm
*parms
)
194 __be32 i_key
= parms
->i_key
;
196 if (parms
->iph
.daddr
&& !ipv4_is_multicast(parms
->iph
.daddr
))
197 remote
= parms
->iph
.daddr
;
201 if (!(parms
->i_flags
& TUNNEL_KEY
) && (parms
->i_flags
& VTI_ISVTI
))
204 h
= ip_tunnel_hash(i_key
, remote
);
205 return &itn
->tunnels
[h
];
208 static void ip_tunnel_add(struct ip_tunnel_net
*itn
, struct ip_tunnel
*t
)
210 struct hlist_head
*head
= ip_bucket(itn
, &t
->parms
);
213 rcu_assign_pointer(itn
->collect_md_tun
, t
);
214 hlist_add_head_rcu(&t
->hash_node
, head
);
217 static void ip_tunnel_del(struct ip_tunnel_net
*itn
, struct ip_tunnel
*t
)
220 rcu_assign_pointer(itn
->collect_md_tun
, NULL
);
221 hlist_del_init_rcu(&t
->hash_node
);
224 static struct ip_tunnel
*ip_tunnel_find(struct ip_tunnel_net
*itn
,
225 struct ip_tunnel_parm
*parms
,
228 __be32 remote
= parms
->iph
.daddr
;
229 __be32 local
= parms
->iph
.saddr
;
230 __be32 key
= parms
->i_key
;
231 __be16 flags
= parms
->i_flags
;
232 int link
= parms
->link
;
233 struct ip_tunnel
*t
= NULL
;
234 struct hlist_head
*head
= ip_bucket(itn
, parms
);
236 hlist_for_each_entry_rcu(t
, head
, hash_node
) {
237 if (local
== t
->parms
.iph
.saddr
&&
238 remote
== t
->parms
.iph
.daddr
&&
239 link
== t
->parms
.link
&&
240 type
== t
->dev
->type
&&
241 ip_tunnel_key_match(&t
->parms
, flags
, key
))
247 static struct net_device
*__ip_tunnel_create(struct net
*net
,
248 const struct rtnl_link_ops
*ops
,
249 struct ip_tunnel_parm
*parms
)
252 struct ip_tunnel
*tunnel
;
253 struct net_device
*dev
;
257 if (parms
->name
[0]) {
258 if (!dev_valid_name(parms
->name
))
260 strlcpy(name
, parms
->name
, IFNAMSIZ
);
262 if (strlen(ops
->kind
) > (IFNAMSIZ
- 3))
264 strcpy(name
, ops
->kind
);
269 dev
= alloc_netdev(ops
->priv_size
, name
, NET_NAME_UNKNOWN
, ops
->setup
);
274 dev_net_set(dev
, net
);
276 dev
->rtnl_link_ops
= ops
;
278 tunnel
= netdev_priv(dev
);
279 tunnel
->parms
= *parms
;
282 err
= register_netdevice(dev
);
294 static inline void init_tunnel_flow(struct flowi4
*fl4
,
296 __be32 daddr
, __be32 saddr
,
297 __be32 key
, __u8 tos
, int oif
)
299 memset(fl4
, 0, sizeof(*fl4
));
300 fl4
->flowi4_oif
= oif
;
303 fl4
->flowi4_tos
= tos
;
304 fl4
->flowi4_proto
= proto
;
305 fl4
->fl4_gre_key
= key
;
308 static int ip_tunnel_bind_dev(struct net_device
*dev
)
310 struct net_device
*tdev
= NULL
;
311 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
312 const struct iphdr
*iph
;
313 int hlen
= LL_MAX_HEADER
;
314 int mtu
= ETH_DATA_LEN
;
315 int t_hlen
= tunnel
->hlen
+ sizeof(struct iphdr
);
317 iph
= &tunnel
->parms
.iph
;
319 /* Guess output device to choose reasonable mtu and needed_headroom */
324 init_tunnel_flow(&fl4
, iph
->protocol
, iph
->daddr
,
325 iph
->saddr
, tunnel
->parms
.o_key
,
326 RT_TOS(iph
->tos
), tunnel
->parms
.link
);
327 rt
= ip_route_output_key(tunnel
->net
, &fl4
);
333 if (dev
->type
!= ARPHRD_ETHER
)
334 dev
->flags
|= IFF_POINTOPOINT
;
336 dst_cache_reset(&tunnel
->dst_cache
);
339 if (!tdev
&& tunnel
->parms
.link
)
340 tdev
= __dev_get_by_index(tunnel
->net
, tunnel
->parms
.link
);
343 hlen
= tdev
->hard_header_len
+ tdev
->needed_headroom
;
347 dev
->needed_headroom
= t_hlen
+ hlen
;
348 mtu
-= (dev
->hard_header_len
+ t_hlen
);
350 if (mtu
< IPV4_MIN_MTU
)
356 static struct ip_tunnel
*ip_tunnel_create(struct net
*net
,
357 struct ip_tunnel_net
*itn
,
358 struct ip_tunnel_parm
*parms
)
360 struct ip_tunnel
*nt
;
361 struct net_device
*dev
;
363 BUG_ON(!itn
->fb_tunnel_dev
);
364 dev
= __ip_tunnel_create(net
, itn
->fb_tunnel_dev
->rtnl_link_ops
, parms
);
366 return ERR_CAST(dev
);
368 dev
->mtu
= ip_tunnel_bind_dev(dev
);
370 nt
= netdev_priv(dev
);
371 ip_tunnel_add(itn
, nt
);
375 int ip_tunnel_rcv(struct ip_tunnel
*tunnel
, struct sk_buff
*skb
,
376 const struct tnl_ptk_info
*tpi
, struct metadata_dst
*tun_dst
,
379 struct pcpu_sw_netstats
*tstats
;
380 const struct iphdr
*iph
= ip_hdr(skb
);
383 #ifdef CONFIG_NET_IPGRE_BROADCAST
384 if (ipv4_is_multicast(iph
->daddr
)) {
385 tunnel
->dev
->stats
.multicast
++;
386 skb
->pkt_type
= PACKET_BROADCAST
;
390 if ((!(tpi
->flags
&TUNNEL_CSUM
) && (tunnel
->parms
.i_flags
&TUNNEL_CSUM
)) ||
391 ((tpi
->flags
&TUNNEL_CSUM
) && !(tunnel
->parms
.i_flags
&TUNNEL_CSUM
))) {
392 tunnel
->dev
->stats
.rx_crc_errors
++;
393 tunnel
->dev
->stats
.rx_errors
++;
397 if (tunnel
->parms
.i_flags
&TUNNEL_SEQ
) {
398 if (!(tpi
->flags
&TUNNEL_SEQ
) ||
399 (tunnel
->i_seqno
&& (s32
)(ntohl(tpi
->seq
) - tunnel
->i_seqno
) < 0)) {
400 tunnel
->dev
->stats
.rx_fifo_errors
++;
401 tunnel
->dev
->stats
.rx_errors
++;
404 tunnel
->i_seqno
= ntohl(tpi
->seq
) + 1;
407 skb_reset_network_header(skb
);
409 err
= IP_ECN_decapsulate(iph
, skb
);
412 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
413 &iph
->saddr
, iph
->tos
);
415 ++tunnel
->dev
->stats
.rx_frame_errors
;
416 ++tunnel
->dev
->stats
.rx_errors
;
421 tstats
= this_cpu_ptr(tunnel
->dev
->tstats
);
422 u64_stats_update_begin(&tstats
->syncp
);
423 tstats
->rx_packets
++;
424 tstats
->rx_bytes
+= skb
->len
;
425 u64_stats_update_end(&tstats
->syncp
);
427 skb_scrub_packet(skb
, !net_eq(tunnel
->net
, dev_net(tunnel
->dev
)));
429 if (tunnel
->dev
->type
== ARPHRD_ETHER
) {
430 skb
->protocol
= eth_type_trans(skb
, tunnel
->dev
);
431 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_HLEN
);
433 skb
->dev
= tunnel
->dev
;
437 skb_dst_set(skb
, (struct dst_entry
*)tun_dst
);
439 gro_cells_receive(&tunnel
->gro_cells
, skb
);
446 EXPORT_SYMBOL_GPL(ip_tunnel_rcv
);
448 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops
*ops
,
451 if (num
>= MAX_IPTUN_ENCAP_OPS
)
454 return !cmpxchg((const struct ip_tunnel_encap_ops
**)
458 EXPORT_SYMBOL(ip_tunnel_encap_add_ops
);
460 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops
*ops
,
465 if (num
>= MAX_IPTUN_ENCAP_OPS
)
468 ret
= (cmpxchg((const struct ip_tunnel_encap_ops
**)
470 ops
, NULL
) == ops
) ? 0 : -1;
476 EXPORT_SYMBOL(ip_tunnel_encap_del_ops
);
478 int ip_tunnel_encap_setup(struct ip_tunnel
*t
,
479 struct ip_tunnel_encap
*ipencap
)
483 memset(&t
->encap
, 0, sizeof(t
->encap
));
485 hlen
= ip_encap_hlen(ipencap
);
489 t
->encap
.type
= ipencap
->type
;
490 t
->encap
.sport
= ipencap
->sport
;
491 t
->encap
.dport
= ipencap
->dport
;
492 t
->encap
.flags
= ipencap
->flags
;
494 t
->encap_hlen
= hlen
;
495 t
->hlen
= t
->encap_hlen
+ t
->tun_hlen
;
499 EXPORT_SYMBOL_GPL(ip_tunnel_encap_setup
);
501 static int tnl_update_pmtu(struct net_device
*dev
, struct sk_buff
*skb
,
502 struct rtable
*rt
, __be16 df
,
503 const struct iphdr
*inner_iph
)
505 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
506 int pkt_size
= skb
->len
- tunnel
->hlen
- dev
->hard_header_len
;
510 mtu
= dst_mtu(&rt
->dst
) - dev
->hard_header_len
511 - sizeof(struct iphdr
) - tunnel
->hlen
;
513 mtu
= skb_dst(skb
) ? dst_mtu(skb_dst(skb
)) : dev
->mtu
;
516 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
, skb
, mtu
);
518 if (skb
->protocol
== htons(ETH_P_IP
)) {
519 if (!skb_is_gso(skb
) &&
520 (inner_iph
->frag_off
& htons(IP_DF
)) &&
522 memset(IPCB(skb
), 0, sizeof(*IPCB(skb
)));
523 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
, htonl(mtu
));
527 #if IS_ENABLED(CONFIG_IPV6)
528 else if (skb
->protocol
== htons(ETH_P_IPV6
)) {
529 struct rt6_info
*rt6
= (struct rt6_info
*)skb_dst(skb
);
531 if (rt6
&& mtu
< dst_mtu(skb_dst(skb
)) &&
532 mtu
>= IPV6_MIN_MTU
) {
533 if ((tunnel
->parms
.iph
.daddr
&&
534 !ipv4_is_multicast(tunnel
->parms
.iph
.daddr
)) ||
535 rt6
->rt6i_dst
.plen
== 128) {
536 rt6
->rt6i_flags
|= RTF_MODIFIED
;
537 dst_metric_set(skb_dst(skb
), RTAX_MTU
, mtu
);
541 if (!skb_is_gso(skb
) && mtu
>= IPV6_MIN_MTU
&&
543 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
551 void ip_md_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
, u8 proto
)
553 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
554 u32 headroom
= sizeof(struct iphdr
);
555 struct ip_tunnel_info
*tun_info
;
556 const struct ip_tunnel_key
*key
;
557 const struct iphdr
*inner_iph
;
563 tun_info
= skb_tunnel_info(skb
);
564 if (unlikely(!tun_info
|| !(tun_info
->mode
& IP_TUNNEL_INFO_TX
) ||
565 ip_tunnel_info_af(tun_info
) != AF_INET
))
567 key
= &tun_info
->key
;
568 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
569 inner_iph
= (const struct iphdr
*)skb_inner_network_header(skb
);
572 if (skb
->protocol
== htons(ETH_P_IP
))
573 tos
= inner_iph
->tos
;
574 else if (skb
->protocol
== htons(ETH_P_IPV6
))
575 tos
= ipv6_get_dsfield((const struct ipv6hdr
*)inner_iph
);
577 init_tunnel_flow(&fl4
, proto
, key
->u
.ipv4
.dst
, key
->u
.ipv4
.src
, 0,
578 RT_TOS(tos
), tunnel
->parms
.link
);
579 if (tunnel
->encap
.type
!= TUNNEL_ENCAP_NONE
)
581 rt
= ip_route_output_key(tunnel
->net
, &fl4
);
583 dev
->stats
.tx_carrier_errors
++;
586 if (rt
->dst
.dev
== dev
) {
588 dev
->stats
.collisions
++;
591 tos
= ip_tunnel_ecn_encap(tos
, inner_iph
, skb
);
594 if (skb
->protocol
== htons(ETH_P_IP
))
595 ttl
= inner_iph
->ttl
;
596 else if (skb
->protocol
== htons(ETH_P_IPV6
))
597 ttl
= ((const struct ipv6hdr
*)inner_iph
)->hop_limit
;
599 ttl
= ip4_dst_hoplimit(&rt
->dst
);
601 if (key
->tun_flags
& TUNNEL_DONT_FRAGMENT
)
603 else if (skb
->protocol
== htons(ETH_P_IP
))
604 df
= inner_iph
->frag_off
& htons(IP_DF
);
605 headroom
+= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
;
606 if (headroom
> dev
->needed_headroom
)
607 dev
->needed_headroom
= headroom
;
609 if (skb_cow_head(skb
, dev
->needed_headroom
)) {
613 iptunnel_xmit(NULL
, rt
, skb
, fl4
.saddr
, fl4
.daddr
, proto
, tos
, ttl
,
614 df
, !net_eq(tunnel
->net
, dev_net(dev
)));
617 dev
->stats
.tx_errors
++;
620 dev
->stats
.tx_dropped
++;
624 EXPORT_SYMBOL_GPL(ip_md_tunnel_xmit
);
626 void ip_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
627 const struct iphdr
*tnl_params
, u8 protocol
)
629 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
630 unsigned int inner_nhdr_len
= 0;
631 const struct iphdr
*inner_iph
;
635 struct rtable
*rt
; /* Route to the other host */
636 unsigned int max_headroom
; /* The extra header space needed */
640 /* ensure we can access the inner net header, for several users below */
641 if (skb
->protocol
== htons(ETH_P_IP
))
642 inner_nhdr_len
= sizeof(struct iphdr
);
643 else if (skb
->protocol
== htons(ETH_P_IPV6
))
644 inner_nhdr_len
= sizeof(struct ipv6hdr
);
645 if (unlikely(!pskb_may_pull(skb
, inner_nhdr_len
)))
648 inner_iph
= (const struct iphdr
*)skb_inner_network_header(skb
);
649 connected
= (tunnel
->parms
.iph
.daddr
!= 0);
651 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
653 dst
= tnl_params
->daddr
;
658 dev
->stats
.tx_fifo_errors
++;
662 if (skb
->protocol
== htons(ETH_P_IP
)) {
663 rt
= skb_rtable(skb
);
664 dst
= rt_nexthop(rt
, inner_iph
->daddr
);
666 #if IS_ENABLED(CONFIG_IPV6)
667 else if (skb
->protocol
== htons(ETH_P_IPV6
)) {
668 const struct in6_addr
*addr6
;
669 struct neighbour
*neigh
;
670 bool do_tx_error_icmp
;
673 neigh
= dst_neigh_lookup(skb_dst(skb
),
674 &ipv6_hdr(skb
)->daddr
);
678 addr6
= (const struct in6_addr
*)&neigh
->primary_key
;
679 addr_type
= ipv6_addr_type(addr6
);
681 if (addr_type
== IPV6_ADDR_ANY
) {
682 addr6
= &ipv6_hdr(skb
)->daddr
;
683 addr_type
= ipv6_addr_type(addr6
);
686 if ((addr_type
& IPV6_ADDR_COMPATv4
) == 0)
687 do_tx_error_icmp
= true;
689 do_tx_error_icmp
= false;
690 dst
= addr6
->s6_addr32
[3];
692 neigh_release(neigh
);
693 if (do_tx_error_icmp
)
703 tos
= tnl_params
->tos
;
706 if (skb
->protocol
== htons(ETH_P_IP
)) {
707 tos
= inner_iph
->tos
;
709 } else if (skb
->protocol
== htons(ETH_P_IPV6
)) {
710 tos
= ipv6_get_dsfield((const struct ipv6hdr
*)inner_iph
);
715 init_tunnel_flow(&fl4
, protocol
, dst
, tnl_params
->saddr
,
716 tunnel
->parms
.o_key
, RT_TOS(tos
), tunnel
->parms
.link
);
718 if (ip_tunnel_encap(skb
, tunnel
, &protocol
, &fl4
) < 0)
721 rt
= connected
? dst_cache_get_ip4(&tunnel
->dst_cache
, &fl4
.saddr
) :
725 rt
= ip_route_output_key(tunnel
->net
, &fl4
);
728 dev
->stats
.tx_carrier_errors
++;
732 dst_cache_set_ip4(&tunnel
->dst_cache
, &rt
->dst
,
736 if (rt
->dst
.dev
== dev
) {
738 dev
->stats
.collisions
++;
742 if (tnl_update_pmtu(dev
, skb
, rt
, tnl_params
->frag_off
, inner_iph
)) {
747 if (tunnel
->err_count
> 0) {
748 if (time_before(jiffies
,
749 tunnel
->err_time
+ IPTUNNEL_ERR_TIMEO
)) {
752 dst_link_failure(skb
);
754 tunnel
->err_count
= 0;
757 tos
= ip_tunnel_ecn_encap(tos
, inner_iph
, skb
);
758 ttl
= tnl_params
->ttl
;
760 if (skb
->protocol
== htons(ETH_P_IP
))
761 ttl
= inner_iph
->ttl
;
762 #if IS_ENABLED(CONFIG_IPV6)
763 else if (skb
->protocol
== htons(ETH_P_IPV6
))
764 ttl
= ((const struct ipv6hdr
*)inner_iph
)->hop_limit
;
767 ttl
= ip4_dst_hoplimit(&rt
->dst
);
770 df
= tnl_params
->frag_off
;
771 if (skb
->protocol
== htons(ETH_P_IP
) && !tunnel
->ignore_df
)
772 df
|= (inner_iph
->frag_off
&htons(IP_DF
));
774 max_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + sizeof(struct iphdr
)
775 + rt
->dst
.header_len
+ ip_encap_hlen(&tunnel
->encap
);
776 if (max_headroom
> dev
->needed_headroom
)
777 dev
->needed_headroom
= max_headroom
;
779 if (skb_cow_head(skb
, dev
->needed_headroom
)) {
781 dev
->stats
.tx_dropped
++;
786 iptunnel_xmit(NULL
, rt
, skb
, fl4
.saddr
, fl4
.daddr
, protocol
, tos
, ttl
,
787 df
, !net_eq(tunnel
->net
, dev_net(dev
)));
790 #if IS_ENABLED(CONFIG_IPV6)
792 dst_link_failure(skb
);
795 dev
->stats
.tx_errors
++;
798 EXPORT_SYMBOL_GPL(ip_tunnel_xmit
);
800 static void ip_tunnel_update(struct ip_tunnel_net
*itn
,
802 struct net_device
*dev
,
803 struct ip_tunnel_parm
*p
,
806 ip_tunnel_del(itn
, t
);
807 t
->parms
.iph
.saddr
= p
->iph
.saddr
;
808 t
->parms
.iph
.daddr
= p
->iph
.daddr
;
809 t
->parms
.i_key
= p
->i_key
;
810 t
->parms
.o_key
= p
->o_key
;
811 if (dev
->type
!= ARPHRD_ETHER
) {
812 memcpy(dev
->dev_addr
, &p
->iph
.saddr
, 4);
813 memcpy(dev
->broadcast
, &p
->iph
.daddr
, 4);
815 ip_tunnel_add(itn
, t
);
817 t
->parms
.iph
.ttl
= p
->iph
.ttl
;
818 t
->parms
.iph
.tos
= p
->iph
.tos
;
819 t
->parms
.iph
.frag_off
= p
->iph
.frag_off
;
821 if (t
->parms
.link
!= p
->link
) {
824 t
->parms
.link
= p
->link
;
825 mtu
= ip_tunnel_bind_dev(dev
);
829 dst_cache_reset(&t
->dst_cache
);
830 netdev_state_change(dev
);
833 int ip_tunnel_ioctl(struct net_device
*dev
, struct ip_tunnel_parm
*p
, int cmd
)
836 struct ip_tunnel
*t
= netdev_priv(dev
);
837 struct net
*net
= t
->net
;
838 struct ip_tunnel_net
*itn
= net_generic(net
, t
->ip_tnl_net_id
);
840 BUG_ON(!itn
->fb_tunnel_dev
);
843 if (dev
== itn
->fb_tunnel_dev
) {
844 t
= ip_tunnel_find(itn
, p
, itn
->fb_tunnel_dev
->type
);
846 t
= netdev_priv(dev
);
848 memcpy(p
, &t
->parms
, sizeof(*p
));
854 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
857 p
->iph
.frag_off
|= htons(IP_DF
);
858 if (!(p
->i_flags
& VTI_ISVTI
)) {
859 if (!(p
->i_flags
& TUNNEL_KEY
))
861 if (!(p
->o_flags
& TUNNEL_KEY
))
865 t
= ip_tunnel_find(itn
, p
, itn
->fb_tunnel_dev
->type
);
867 if (cmd
== SIOCADDTUNNEL
) {
869 t
= ip_tunnel_create(net
, itn
, p
);
870 err
= PTR_ERR_OR_ZERO(t
);
877 if (dev
!= itn
->fb_tunnel_dev
&& cmd
== SIOCCHGTUNNEL
) {
884 unsigned int nflags
= 0;
886 if (ipv4_is_multicast(p
->iph
.daddr
))
887 nflags
= IFF_BROADCAST
;
888 else if (p
->iph
.daddr
)
889 nflags
= IFF_POINTOPOINT
;
891 if ((dev
->flags
^nflags
)&(IFF_POINTOPOINT
|IFF_BROADCAST
)) {
896 t
= netdev_priv(dev
);
902 ip_tunnel_update(itn
, t
, dev
, p
, true);
910 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
913 if (dev
== itn
->fb_tunnel_dev
) {
915 t
= ip_tunnel_find(itn
, p
, itn
->fb_tunnel_dev
->type
);
919 if (t
== netdev_priv(itn
->fb_tunnel_dev
))
923 unregister_netdevice(dev
);
934 EXPORT_SYMBOL_GPL(ip_tunnel_ioctl
);
936 int __ip_tunnel_change_mtu(struct net_device
*dev
, int new_mtu
, bool strict
)
938 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
939 int t_hlen
= tunnel
->hlen
+ sizeof(struct iphdr
);
940 int max_mtu
= 0xFFF8 - dev
->hard_header_len
- t_hlen
;
945 if (new_mtu
> max_mtu
) {
955 EXPORT_SYMBOL_GPL(__ip_tunnel_change_mtu
);
957 int ip_tunnel_change_mtu(struct net_device
*dev
, int new_mtu
)
959 return __ip_tunnel_change_mtu(dev
, new_mtu
, true);
961 EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu
);
963 static void ip_tunnel_dev_free(struct net_device
*dev
)
965 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
967 gro_cells_destroy(&tunnel
->gro_cells
);
968 dst_cache_destroy(&tunnel
->dst_cache
);
969 free_percpu(dev
->tstats
);
973 void ip_tunnel_dellink(struct net_device
*dev
, struct list_head
*head
)
975 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
976 struct ip_tunnel_net
*itn
;
978 itn
= net_generic(tunnel
->net
, tunnel
->ip_tnl_net_id
);
980 if (itn
->fb_tunnel_dev
!= dev
) {
981 ip_tunnel_del(itn
, netdev_priv(dev
));
982 unregister_netdevice_queue(dev
, head
);
985 EXPORT_SYMBOL_GPL(ip_tunnel_dellink
);
987 struct net
*ip_tunnel_get_link_net(const struct net_device
*dev
)
989 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
993 EXPORT_SYMBOL(ip_tunnel_get_link_net
);
995 int ip_tunnel_get_iflink(const struct net_device
*dev
)
997 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
999 return tunnel
->parms
.link
;
1001 EXPORT_SYMBOL(ip_tunnel_get_iflink
);
1003 int ip_tunnel_init_net(struct net
*net
, int ip_tnl_net_id
,
1004 struct rtnl_link_ops
*ops
, char *devname
)
1006 struct ip_tunnel_net
*itn
= net_generic(net
, ip_tnl_net_id
);
1007 struct ip_tunnel_parm parms
;
1010 for (i
= 0; i
< IP_TNL_HASH_SIZE
; i
++)
1011 INIT_HLIST_HEAD(&itn
->tunnels
[i
]);
1014 itn
->fb_tunnel_dev
= NULL
;
1018 memset(&parms
, 0, sizeof(parms
));
1020 strlcpy(parms
.name
, devname
, IFNAMSIZ
);
1023 itn
->fb_tunnel_dev
= __ip_tunnel_create(net
, ops
, &parms
);
1024 /* FB netdevice is special: we have one, and only one per netns.
1025 * Allowing to move it to another netns is clearly unsafe.
1027 if (!IS_ERR(itn
->fb_tunnel_dev
)) {
1028 itn
->fb_tunnel_dev
->features
|= NETIF_F_NETNS_LOCAL
;
1029 itn
->fb_tunnel_dev
->mtu
= ip_tunnel_bind_dev(itn
->fb_tunnel_dev
);
1030 ip_tunnel_add(itn
, netdev_priv(itn
->fb_tunnel_dev
));
1034 return PTR_ERR_OR_ZERO(itn
->fb_tunnel_dev
);
1036 EXPORT_SYMBOL_GPL(ip_tunnel_init_net
);
1038 static void ip_tunnel_destroy(struct ip_tunnel_net
*itn
, struct list_head
*head
,
1039 struct rtnl_link_ops
*ops
)
1041 struct net
*net
= dev_net(itn
->fb_tunnel_dev
);
1042 struct net_device
*dev
, *aux
;
1045 for_each_netdev_safe(net
, dev
, aux
)
1046 if (dev
->rtnl_link_ops
== ops
)
1047 unregister_netdevice_queue(dev
, head
);
1049 for (h
= 0; h
< IP_TNL_HASH_SIZE
; h
++) {
1050 struct ip_tunnel
*t
;
1051 struct hlist_node
*n
;
1052 struct hlist_head
*thead
= &itn
->tunnels
[h
];
1054 hlist_for_each_entry_safe(t
, n
, thead
, hash_node
)
1055 /* If dev is in the same netns, it has already
1056 * been added to the list by the previous loop.
1058 if (!net_eq(dev_net(t
->dev
), net
))
1059 unregister_netdevice_queue(t
->dev
, head
);
1063 void ip_tunnel_delete_net(struct ip_tunnel_net
*itn
, struct rtnl_link_ops
*ops
)
1068 ip_tunnel_destroy(itn
, &list
, ops
);
1069 unregister_netdevice_many(&list
);
1072 EXPORT_SYMBOL_GPL(ip_tunnel_delete_net
);
1074 int ip_tunnel_newlink(struct net_device
*dev
, struct nlattr
*tb
[],
1075 struct ip_tunnel_parm
*p
)
1077 struct ip_tunnel
*nt
;
1078 struct net
*net
= dev_net(dev
);
1079 struct ip_tunnel_net
*itn
;
1083 nt
= netdev_priv(dev
);
1084 itn
= net_generic(net
, nt
->ip_tnl_net_id
);
1086 if (nt
->collect_md
) {
1087 if (rtnl_dereference(itn
->collect_md_tun
))
1090 if (ip_tunnel_find(itn
, p
, dev
->type
))
1096 err
= register_netdevice(dev
);
1100 if (dev
->type
== ARPHRD_ETHER
&& !tb
[IFLA_ADDRESS
])
1101 eth_hw_addr_random(dev
);
1103 mtu
= ip_tunnel_bind_dev(dev
);
1107 ip_tunnel_add(itn
, nt
);
1111 EXPORT_SYMBOL_GPL(ip_tunnel_newlink
);
1113 int ip_tunnel_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
1114 struct ip_tunnel_parm
*p
)
1116 struct ip_tunnel
*t
;
1117 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1118 struct net
*net
= tunnel
->net
;
1119 struct ip_tunnel_net
*itn
= net_generic(net
, tunnel
->ip_tnl_net_id
);
1121 if (dev
== itn
->fb_tunnel_dev
)
1124 t
= ip_tunnel_find(itn
, p
, dev
->type
);
1132 if (dev
->type
!= ARPHRD_ETHER
) {
1133 unsigned int nflags
= 0;
1135 if (ipv4_is_multicast(p
->iph
.daddr
))
1136 nflags
= IFF_BROADCAST
;
1137 else if (p
->iph
.daddr
)
1138 nflags
= IFF_POINTOPOINT
;
1140 if ((dev
->flags
^ nflags
) &
1141 (IFF_POINTOPOINT
| IFF_BROADCAST
))
1146 ip_tunnel_update(itn
, t
, dev
, p
, !tb
[IFLA_MTU
]);
1149 EXPORT_SYMBOL_GPL(ip_tunnel_changelink
);
1151 int ip_tunnel_init(struct net_device
*dev
)
1153 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1154 struct iphdr
*iph
= &tunnel
->parms
.iph
;
1157 dev
->destructor
= ip_tunnel_dev_free
;
1158 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
1162 err
= dst_cache_init(&tunnel
->dst_cache
, GFP_KERNEL
);
1164 free_percpu(dev
->tstats
);
1168 err
= gro_cells_init(&tunnel
->gro_cells
, dev
);
1170 dst_cache_destroy(&tunnel
->dst_cache
);
1171 free_percpu(dev
->tstats
);
1176 tunnel
->net
= dev_net(dev
);
1177 strcpy(tunnel
->parms
.name
, dev
->name
);
1181 if (tunnel
->collect_md
) {
1182 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1183 netif_keep_dst(dev
);
1187 EXPORT_SYMBOL_GPL(ip_tunnel_init
);
1189 void ip_tunnel_uninit(struct net_device
*dev
)
1191 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1192 struct net
*net
= tunnel
->net
;
1193 struct ip_tunnel_net
*itn
;
1195 itn
= net_generic(net
, tunnel
->ip_tnl_net_id
);
1196 /* fb_tunnel_dev will be unregisted in net-exit call. */
1197 if (itn
->fb_tunnel_dev
!= dev
)
1198 ip_tunnel_del(itn
, netdev_priv(dev
));
1200 dst_cache_reset(&tunnel
->dst_cache
);
1202 EXPORT_SYMBOL_GPL(ip_tunnel_uninit
);
1204 /* Do least required initialization, rest of init is done in tunnel_init call */
1205 void ip_tunnel_setup(struct net_device
*dev
, int net_id
)
1207 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1208 tunnel
->ip_tnl_net_id
= net_id
;
1210 EXPORT_SYMBOL_GPL(ip_tunnel_setup
);
1212 MODULE_LICENSE("GPL");