2 * Linux NET3: GRE over IP protocol decoder.
4 * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.ru)
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <asm/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/if_vlan.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
39 #include <net/protocol.h>
40 #include <net/ip_tunnels.h>
42 #include <net/checksum.h>
43 #include <net/dsfield.h>
44 #include <net/inet_ecn.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <net/rtnetlink.h>
50 #include <net/dst_metadata.h>
52 #if IS_ENABLED(CONFIG_IPV6)
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
62 1. The most important issue is detecting local dead loops.
63 They would cause complete host lockup in transmit, which
64 would be "resolved" by stack overflow or, if queueing is enabled,
65 with infinite looping in net_bh.
67 We cannot track such dead loops during route installation,
68 it is infeasible task. The most general solutions would be
69 to keep skb->encapsulation counter (sort of local ttl),
70 and silently drop packet when it expires. It is a good
71 solution, but it supposes maintaining new variable in ALL
72 skb, even if no tunneling is used.
74 Current solution: xmit_recursion breaks dead loops. This is a percpu
75 counter, since when we enter the first ndo_xmit(), cpu migration is
76 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
78 2. Networking dead loops would not kill routers, but would really
79 kill network. IP hop limit plays role of "t->recursion" in this case,
80 if we copy it from packet being encapsulated to upper header.
81 It is very good solution, but it introduces two problems:
83 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
84 do not work over tunnels.
85 - traceroute does not work. I planned to relay ICMP from tunnel,
86 so that this problem would be solved and traceroute output
87 would even more informative. This idea appeared to be wrong:
88 only Linux complies to rfc1812 now (yes, guys, Linux is the only
89 true router now :-)), all routers (at least, in neighbourhood of mine)
90 return only 8 bytes of payload. It is the end.
92 Hence, if we want that OSPF worked or traceroute said something reasonable,
93 we should search for another solution.
95 One of them is to parse packet trying to detect inner encapsulation
96 made by our node. It is difficult or even impossible, especially,
97 taking into account fragmentation. TO be short, ttl is not solution at all.
99 Current solution: The solution was UNEXPECTEDLY SIMPLE.
100 We force DF flag on tunnels with preconfigured hop limit,
101 that is ALL. :-) Well, it does not remove the problem completely,
102 but exponential growth of network traffic is changed to linear
103 (branches, that exceed pmtu are pruned) and tunnel mtu
104 rapidly degrades to value <68, where looping stops.
105 Yes, it is not good if there exists a router in the loop,
106 which does not force DF, even when encapsulating packets have DF set.
107 But it is not our problem! Nobody could accuse us, we made
108 all that we could make. Even if it is your gated who injected
109 fatal route to network, even if it were you who configured
110 fatal static route: you are innocent. :-)
115 static bool log_ecn_error
= true;
116 module_param(log_ecn_error
, bool, 0644);
117 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
119 static struct rtnl_link_ops ipgre_link_ops __read_mostly
;
120 static int ipgre_tunnel_init(struct net_device
*dev
);
122 static int ipgre_net_id __read_mostly
;
123 static int gre_tap_net_id __read_mostly
;
125 static int ip_gre_calc_hlen(__be16 o_flags
)
129 if (o_flags
& TUNNEL_CSUM
)
131 if (o_flags
& TUNNEL_KEY
)
133 if (o_flags
& TUNNEL_SEQ
)
138 static __be16
gre_flags_to_tnl_flags(__be16 flags
)
142 if (flags
& GRE_CSUM
)
143 tflags
|= TUNNEL_CSUM
;
144 if (flags
& GRE_ROUTING
)
145 tflags
|= TUNNEL_ROUTING
;
147 tflags
|= TUNNEL_KEY
;
149 tflags
|= TUNNEL_SEQ
;
150 if (flags
& GRE_STRICT
)
151 tflags
|= TUNNEL_STRICT
;
153 tflags
|= TUNNEL_REC
;
154 if (flags
& GRE_VERSION
)
155 tflags
|= TUNNEL_VERSION
;
160 static __be16
tnl_flags_to_gre_flags(__be16 tflags
)
164 if (tflags
& TUNNEL_CSUM
)
166 if (tflags
& TUNNEL_ROUTING
)
167 flags
|= GRE_ROUTING
;
168 if (tflags
& TUNNEL_KEY
)
170 if (tflags
& TUNNEL_SEQ
)
172 if (tflags
& TUNNEL_STRICT
)
174 if (tflags
& TUNNEL_REC
)
176 if (tflags
& TUNNEL_VERSION
)
177 flags
|= GRE_VERSION
;
182 /* Fills in tpi and returns header length to be pulled. */
183 static int parse_gre_header(struct sk_buff
*skb
, struct tnl_ptk_info
*tpi
,
186 const struct gre_base_hdr
*greh
;
190 if (unlikely(!pskb_may_pull(skb
, sizeof(struct gre_base_hdr
))))
193 greh
= (struct gre_base_hdr
*)skb_transport_header(skb
);
194 if (unlikely(greh
->flags
& (GRE_VERSION
| GRE_ROUTING
)))
197 tpi
->flags
= gre_flags_to_tnl_flags(greh
->flags
);
198 hdr_len
= ip_gre_calc_hlen(tpi
->flags
);
200 if (!pskb_may_pull(skb
, hdr_len
))
203 greh
= (struct gre_base_hdr
*)skb_transport_header(skb
);
204 tpi
->proto
= greh
->protocol
;
206 options
= (__be32
*)(greh
+ 1);
207 if (greh
->flags
& GRE_CSUM
) {
208 if (skb_checksum_simple_validate(skb
)) {
213 skb_checksum_try_convert(skb
, IPPROTO_GRE
, 0,
214 null_compute_pseudo
);
218 if (greh
->flags
& GRE_KEY
) {
224 if (unlikely(greh
->flags
& GRE_SEQ
)) {
230 /* WCCP version 1 and 2 protocol decoding.
231 * - Change protocol to IP
232 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
234 if (greh
->flags
== 0 && tpi
->proto
== htons(ETH_P_WCCP
)) {
235 tpi
->proto
= htons(ETH_P_IP
);
236 if ((*(u8
*)options
& 0xF0) != 0x40) {
238 if (!pskb_may_pull(skb
, hdr_len
))
245 static void ipgre_err(struct sk_buff
*skb
, u32 info
,
246 const struct tnl_ptk_info
*tpi
)
249 /* All the routers (except for Linux) return only
250 8 bytes of packet payload. It means, that precise relaying of
251 ICMP in the real Internet is absolutely infeasible.
253 Moreover, Cisco "wise men" put GRE key to the third word
254 in GRE header. It makes impossible maintaining even soft
255 state for keyed GRE tunnels with enabled checksum. Tell
258 Well, I wonder, rfc1812 was written by Cisco employee,
259 what the hell these idiots break standards established
262 struct net
*net
= dev_net(skb
->dev
);
263 struct ip_tunnel_net
*itn
;
264 const struct iphdr
*iph
;
265 const int type
= icmp_hdr(skb
)->type
;
266 const int code
= icmp_hdr(skb
)->code
;
271 case ICMP_PARAMETERPROB
:
274 case ICMP_DEST_UNREACH
:
277 case ICMP_PORT_UNREACH
:
278 /* Impossible event. */
281 /* All others are translated to HOST_UNREACH.
282 rfc2003 contains "deep thoughts" about NET_UNREACH,
283 I believe they are just ether pollution. --ANK
289 case ICMP_TIME_EXCEEDED
:
290 if (code
!= ICMP_EXC_TTL
)
298 if (tpi
->proto
== htons(ETH_P_TEB
))
299 itn
= net_generic(net
, gre_tap_net_id
);
301 itn
= net_generic(net
, ipgre_net_id
);
303 iph
= (const struct iphdr
*)(icmp_hdr(skb
) + 1);
304 t
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, tpi
->flags
,
305 iph
->daddr
, iph
->saddr
, tpi
->key
);
310 if (t
->parms
.iph
.daddr
== 0 ||
311 ipv4_is_multicast(t
->parms
.iph
.daddr
))
314 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
317 if (time_before(jiffies
, t
->err_time
+ IPTUNNEL_ERR_TIMEO
))
321 t
->err_time
= jiffies
;
324 static void gre_err(struct sk_buff
*skb
, u32 info
)
326 /* All the routers (except for Linux) return only
327 * 8 bytes of packet payload. It means, that precise relaying of
328 * ICMP in the real Internet is absolutely infeasible.
330 * Moreover, Cisco "wise men" put GRE key to the third word
331 * in GRE header. It makes impossible maintaining even soft
333 * GRE tunnels with enabled checksum. Tell them "thank you".
335 * Well, I wonder, rfc1812 was written by Cisco employee,
336 * what the hell these idiots break standards established
340 const int type
= icmp_hdr(skb
)->type
;
341 const int code
= icmp_hdr(skb
)->code
;
342 struct tnl_ptk_info tpi
;
343 bool csum_err
= false;
345 if (parse_gre_header(skb
, &tpi
, &csum_err
) < 0) {
346 if (!csum_err
) /* ignore csum errors. */
350 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
) {
351 ipv4_update_pmtu(skb
, dev_net(skb
->dev
), info
,
352 skb
->dev
->ifindex
, 0, IPPROTO_GRE
, 0);
355 if (type
== ICMP_REDIRECT
) {
356 ipv4_redirect(skb
, dev_net(skb
->dev
), skb
->dev
->ifindex
, 0,
361 ipgre_err(skb
, info
, &tpi
);
364 static __be64
key_to_tunnel_id(__be32 key
)
367 return (__force __be64
)((__force u32
)key
);
369 return (__force __be64
)((__force u64
)key
<< 32);
373 /* Returns the least-significant 32 bits of a __be64. */
374 static __be32
tunnel_id_to_key(__be64 x
)
377 return (__force __be32
)x
;
379 return (__force __be32
)((__force u64
)x
>> 32);
383 static int ipgre_rcv(struct sk_buff
*skb
, const struct tnl_ptk_info
*tpi
)
385 struct net
*net
= dev_net(skb
->dev
);
386 struct metadata_dst
*tun_dst
= NULL
;
387 struct ip_tunnel_net
*itn
;
388 const struct iphdr
*iph
;
389 struct ip_tunnel
*tunnel
;
391 if (tpi
->proto
== htons(ETH_P_TEB
))
392 itn
= net_generic(net
, gre_tap_net_id
);
394 itn
= net_generic(net
, ipgre_net_id
);
397 tunnel
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, tpi
->flags
,
398 iph
->saddr
, iph
->daddr
, tpi
->key
);
401 skb_pop_mac_header(skb
);
402 if (tunnel
->collect_md
) {
406 flags
= tpi
->flags
& (TUNNEL_CSUM
| TUNNEL_KEY
);
407 tun_id
= key_to_tunnel_id(tpi
->key
);
408 tun_dst
= ip_tun_rx_dst(skb
, flags
, tun_id
, 0);
410 return PACKET_REJECT
;
413 ip_tunnel_rcv(tunnel
, skb
, tpi
, tun_dst
, log_ecn_error
);
416 return PACKET_REJECT
;
419 static int gre_rcv(struct sk_buff
*skb
)
421 struct tnl_ptk_info tpi
;
422 bool csum_err
= false;
425 #ifdef CONFIG_NET_IPGRE_BROADCAST
426 if (ipv4_is_multicast(ip_hdr(skb
)->daddr
)) {
427 /* Looped back packet, drop it! */
428 if (rt_is_output_route(skb_rtable(skb
)))
433 hdr_len
= parse_gre_header(skb
, &tpi
, &csum_err
);
436 if (iptunnel_pull_header(skb
, hdr_len
, tpi
.proto
, false) < 0)
439 if (ipgre_rcv(skb
, &tpi
) == PACKET_RCVD
)
442 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PORT_UNREACH
, 0);
448 static __sum16
gre_checksum(struct sk_buff
*skb
)
452 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
453 csum
= lco_csum(skb
);
455 csum
= skb_checksum(skb
, 0, skb
->len
, 0);
456 return csum_fold(csum
);
459 static void build_header(struct sk_buff
*skb
, int hdr_len
, __be16 flags
,
460 __be16 proto
, __be32 key
, __be32 seq
)
462 struct gre_base_hdr
*greh
;
464 skb_push(skb
, hdr_len
);
466 skb_reset_transport_header(skb
);
467 greh
= (struct gre_base_hdr
*)skb
->data
;
468 greh
->flags
= tnl_flags_to_gre_flags(flags
);
469 greh
->protocol
= proto
;
471 if (flags
& (TUNNEL_KEY
| TUNNEL_CSUM
| TUNNEL_SEQ
)) {
472 __be32
*ptr
= (__be32
*)(((u8
*)greh
) + hdr_len
- 4);
474 if (flags
& TUNNEL_SEQ
) {
478 if (flags
& TUNNEL_KEY
) {
482 if (flags
& TUNNEL_CSUM
&&
483 !(skb_shinfo(skb
)->gso_type
&
484 (SKB_GSO_GRE
| SKB_GSO_GRE_CSUM
))) {
486 *(__sum16
*)ptr
= gre_checksum(skb
);
491 static void __gre_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
492 const struct iphdr
*tnl_params
,
495 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
497 if (tunnel
->parms
.o_flags
& TUNNEL_SEQ
)
500 /* Push GRE header. */
501 build_header(skb
, tunnel
->tun_hlen
, tunnel
->parms
.o_flags
,
502 proto
, tunnel
->parms
.o_key
, htonl(tunnel
->o_seqno
));
504 skb_set_inner_protocol(skb
, proto
);
505 ip_tunnel_xmit(skb
, dev
, tnl_params
, tnl_params
->protocol
);
508 static struct sk_buff
*gre_handle_offloads(struct sk_buff
*skb
,
511 return iptunnel_handle_offloads(skb
, csum
? SKB_GSO_GRE_CSUM
: SKB_GSO_GRE
);
514 static struct rtable
*gre_get_rt(struct sk_buff
*skb
,
515 struct net_device
*dev
,
517 const struct ip_tunnel_key
*key
)
519 struct net
*net
= dev_net(dev
);
521 memset(fl
, 0, sizeof(*fl
));
522 fl
->daddr
= key
->u
.ipv4
.dst
;
523 fl
->saddr
= key
->u
.ipv4
.src
;
524 fl
->flowi4_tos
= RT_TOS(key
->tos
);
525 fl
->flowi4_mark
= skb
->mark
;
526 fl
->flowi4_proto
= IPPROTO_GRE
;
528 return ip_route_output_key(net
, fl
);
531 static void gre_fb_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
534 struct ip_tunnel_info
*tun_info
;
535 const struct ip_tunnel_key
*key
;
536 struct rtable
*rt
= NULL
;
544 tun_info
= skb_tunnel_info(skb
);
545 if (unlikely(!tun_info
|| !(tun_info
->mode
& IP_TUNNEL_INFO_TX
) ||
546 ip_tunnel_info_af(tun_info
) != AF_INET
))
549 key
= &tun_info
->key
;
550 use_cache
= ip_tunnel_dst_cache_usable(skb
, tun_info
);
552 rt
= dst_cache_get_ip4(&tun_info
->dst_cache
, &fl
.saddr
);
554 rt
= gre_get_rt(skb
, dev
, &fl
, key
);
558 dst_cache_set_ip4(&tun_info
->dst_cache
, &rt
->dst
,
562 tunnel_hlen
= ip_gre_calc_hlen(key
->tun_flags
);
564 min_headroom
= LL_RESERVED_SPACE(rt
->dst
.dev
) + rt
->dst
.header_len
565 + tunnel_hlen
+ sizeof(struct iphdr
);
566 if (skb_headroom(skb
) < min_headroom
|| skb_header_cloned(skb
)) {
567 int head_delta
= SKB_DATA_ALIGN(min_headroom
-
570 err
= pskb_expand_head(skb
, max_t(int, head_delta
, 0),
576 /* Push Tunnel header. */
577 skb
= gre_handle_offloads(skb
, !!(tun_info
->key
.tun_flags
& TUNNEL_CSUM
));
583 flags
= tun_info
->key
.tun_flags
& (TUNNEL_CSUM
| TUNNEL_KEY
);
584 build_header(skb
, tunnel_hlen
, flags
, proto
,
585 tunnel_id_to_key(tun_info
->key
.tun_id
), 0);
587 df
= key
->tun_flags
& TUNNEL_DONT_FRAGMENT
? htons(IP_DF
) : 0;
589 iptunnel_xmit(skb
->sk
, rt
, skb
, fl
.saddr
, key
->u
.ipv4
.dst
, IPPROTO_GRE
,
590 key
->tos
, key
->ttl
, df
, false);
597 dev
->stats
.tx_dropped
++;
600 static int gre_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
)
602 struct ip_tunnel_info
*info
= skb_tunnel_info(skb
);
606 if (ip_tunnel_info_af(info
) != AF_INET
)
609 rt
= gre_get_rt(skb
, dev
, &fl4
, &info
->key
);
614 info
->key
.u
.ipv4
.src
= fl4
.saddr
;
618 static netdev_tx_t
ipgre_xmit(struct sk_buff
*skb
,
619 struct net_device
*dev
)
621 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
622 const struct iphdr
*tnl_params
;
624 if (tunnel
->collect_md
) {
625 gre_fb_xmit(skb
, dev
, skb
->protocol
);
629 if (dev
->header_ops
) {
630 /* Need space for new headers */
631 if (skb_cow_head(skb
, dev
->needed_headroom
-
632 (tunnel
->hlen
+ sizeof(struct iphdr
))))
635 tnl_params
= (const struct iphdr
*)skb
->data
;
637 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
640 skb_pull(skb
, tunnel
->hlen
+ sizeof(struct iphdr
));
641 skb_reset_mac_header(skb
);
643 if (skb_cow_head(skb
, dev
->needed_headroom
))
646 tnl_params
= &tunnel
->parms
.iph
;
649 skb
= gre_handle_offloads(skb
, !!(tunnel
->parms
.o_flags
&TUNNEL_CSUM
));
653 __gre_xmit(skb
, dev
, tnl_params
, skb
->protocol
);
659 dev
->stats
.tx_dropped
++;
663 static netdev_tx_t
gre_tap_xmit(struct sk_buff
*skb
,
664 struct net_device
*dev
)
666 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
668 if (tunnel
->collect_md
) {
669 gre_fb_xmit(skb
, dev
, htons(ETH_P_TEB
));
673 skb
= gre_handle_offloads(skb
, !!(tunnel
->parms
.o_flags
&TUNNEL_CSUM
));
677 if (skb_cow_head(skb
, dev
->needed_headroom
))
680 __gre_xmit(skb
, dev
, &tunnel
->parms
.iph
, htons(ETH_P_TEB
));
686 dev
->stats
.tx_dropped
++;
690 static int ipgre_tunnel_ioctl(struct net_device
*dev
,
691 struct ifreq
*ifr
, int cmd
)
694 struct ip_tunnel_parm p
;
696 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
698 if (cmd
== SIOCADDTUNNEL
|| cmd
== SIOCCHGTUNNEL
) {
699 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_GRE
||
700 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)) ||
701 ((p
.i_flags
|p
.o_flags
)&(GRE_VERSION
|GRE_ROUTING
)))
704 p
.i_flags
= gre_flags_to_tnl_flags(p
.i_flags
);
705 p
.o_flags
= gre_flags_to_tnl_flags(p
.o_flags
);
707 err
= ip_tunnel_ioctl(dev
, &p
, cmd
);
711 p
.i_flags
= tnl_flags_to_gre_flags(p
.i_flags
);
712 p
.o_flags
= tnl_flags_to_gre_flags(p
.o_flags
);
714 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
719 /* Nice toy. Unfortunately, useless in real life :-)
720 It allows to construct virtual multiprotocol broadcast "LAN"
721 over the Internet, provided multicast routing is tuned.
724 I have no idea was this bicycle invented before me,
725 so that I had to set ARPHRD_IPGRE to a random value.
726 I have an impression, that Cisco could make something similar,
727 but this feature is apparently missing in IOS<=11.2(8).
729 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
730 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
732 ping -t 255 224.66.66.66
734 If nobody answers, mbone does not work.
736 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
737 ip addr add 10.66.66.<somewhat>/24 dev Universe
739 ifconfig Universe add fe80::<Your_real_addr>/10
740 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
743 ftp fec0:6666:6666::193.233.7.65
746 static int ipgre_header(struct sk_buff
*skb
, struct net_device
*dev
,
748 const void *daddr
, const void *saddr
, unsigned int len
)
750 struct ip_tunnel
*t
= netdev_priv(dev
);
752 struct gre_base_hdr
*greh
;
754 iph
= (struct iphdr
*)skb_push(skb
, t
->hlen
+ sizeof(*iph
));
755 greh
= (struct gre_base_hdr
*)(iph
+1);
756 greh
->flags
= tnl_flags_to_gre_flags(t
->parms
.o_flags
);
757 greh
->protocol
= htons(type
);
759 memcpy(iph
, &t
->parms
.iph
, sizeof(struct iphdr
));
761 /* Set the source hardware address. */
763 memcpy(&iph
->saddr
, saddr
, 4);
765 memcpy(&iph
->daddr
, daddr
, 4);
767 return t
->hlen
+ sizeof(*iph
);
769 return -(t
->hlen
+ sizeof(*iph
));
772 static int ipgre_header_parse(const struct sk_buff
*skb
, unsigned char *haddr
)
774 const struct iphdr
*iph
= (const struct iphdr
*) skb_mac_header(skb
);
775 memcpy(haddr
, &iph
->saddr
, 4);
779 static const struct header_ops ipgre_header_ops
= {
780 .create
= ipgre_header
,
781 .parse
= ipgre_header_parse
,
784 #ifdef CONFIG_NET_IPGRE_BROADCAST
785 static int ipgre_open(struct net_device
*dev
)
787 struct ip_tunnel
*t
= netdev_priv(dev
);
789 if (ipv4_is_multicast(t
->parms
.iph
.daddr
)) {
793 rt
= ip_route_output_gre(t
->net
, &fl4
,
797 RT_TOS(t
->parms
.iph
.tos
),
800 return -EADDRNOTAVAIL
;
803 if (!__in_dev_get_rtnl(dev
))
804 return -EADDRNOTAVAIL
;
805 t
->mlink
= dev
->ifindex
;
806 ip_mc_inc_group(__in_dev_get_rtnl(dev
), t
->parms
.iph
.daddr
);
811 static int ipgre_close(struct net_device
*dev
)
813 struct ip_tunnel
*t
= netdev_priv(dev
);
815 if (ipv4_is_multicast(t
->parms
.iph
.daddr
) && t
->mlink
) {
816 struct in_device
*in_dev
;
817 in_dev
= inetdev_by_index(t
->net
, t
->mlink
);
819 ip_mc_dec_group(in_dev
, t
->parms
.iph
.daddr
);
825 static const struct net_device_ops ipgre_netdev_ops
= {
826 .ndo_init
= ipgre_tunnel_init
,
827 .ndo_uninit
= ip_tunnel_uninit
,
828 #ifdef CONFIG_NET_IPGRE_BROADCAST
829 .ndo_open
= ipgre_open
,
830 .ndo_stop
= ipgre_close
,
832 .ndo_start_xmit
= ipgre_xmit
,
833 .ndo_do_ioctl
= ipgre_tunnel_ioctl
,
834 .ndo_change_mtu
= ip_tunnel_change_mtu
,
835 .ndo_get_stats64
= ip_tunnel_get_stats64
,
836 .ndo_get_iflink
= ip_tunnel_get_iflink
,
839 #define GRE_FEATURES (NETIF_F_SG | \
844 static void ipgre_tunnel_setup(struct net_device
*dev
)
846 dev
->netdev_ops
= &ipgre_netdev_ops
;
847 dev
->type
= ARPHRD_IPGRE
;
848 ip_tunnel_setup(dev
, ipgre_net_id
);
851 static void __gre_tunnel_init(struct net_device
*dev
)
853 struct ip_tunnel
*tunnel
;
856 tunnel
= netdev_priv(dev
);
857 tunnel
->tun_hlen
= ip_gre_calc_hlen(tunnel
->parms
.o_flags
);
858 tunnel
->parms
.iph
.protocol
= IPPROTO_GRE
;
860 tunnel
->hlen
= tunnel
->tun_hlen
+ tunnel
->encap_hlen
;
862 t_hlen
= tunnel
->hlen
+ sizeof(struct iphdr
);
864 dev
->needed_headroom
= LL_MAX_HEADER
+ t_hlen
+ 4;
865 dev
->mtu
= ETH_DATA_LEN
- t_hlen
- 4;
867 dev
->features
|= GRE_FEATURES
;
868 dev
->hw_features
|= GRE_FEATURES
;
870 if (!(tunnel
->parms
.o_flags
& TUNNEL_SEQ
)) {
871 /* TCP offload with GRE SEQ is not supported, nor
872 * can we support 2 levels of outer headers requiring
875 if (!(tunnel
->parms
.o_flags
& TUNNEL_CSUM
) ||
876 (tunnel
->encap
.type
== TUNNEL_ENCAP_NONE
)) {
877 dev
->features
|= NETIF_F_GSO_SOFTWARE
;
878 dev
->hw_features
|= NETIF_F_GSO_SOFTWARE
;
881 /* Can use a lockless transmit, unless we generate
884 dev
->features
|= NETIF_F_LLTX
;
888 static int ipgre_tunnel_init(struct net_device
*dev
)
890 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
891 struct iphdr
*iph
= &tunnel
->parms
.iph
;
893 __gre_tunnel_init(dev
);
895 memcpy(dev
->dev_addr
, &iph
->saddr
, 4);
896 memcpy(dev
->broadcast
, &iph
->daddr
, 4);
898 dev
->flags
= IFF_NOARP
;
902 if (iph
->daddr
&& !tunnel
->collect_md
) {
903 #ifdef CONFIG_NET_IPGRE_BROADCAST
904 if (ipv4_is_multicast(iph
->daddr
)) {
907 dev
->flags
= IFF_BROADCAST
;
908 dev
->header_ops
= &ipgre_header_ops
;
911 } else if (!tunnel
->collect_md
) {
912 dev
->header_ops
= &ipgre_header_ops
;
915 return ip_tunnel_init(dev
);
918 static const struct gre_protocol ipgre_protocol
= {
920 .err_handler
= gre_err
,
923 static int __net_init
ipgre_init_net(struct net
*net
)
925 return ip_tunnel_init_net(net
, ipgre_net_id
, &ipgre_link_ops
, NULL
);
928 static void __net_exit
ipgre_exit_net(struct net
*net
)
930 struct ip_tunnel_net
*itn
= net_generic(net
, ipgre_net_id
);
931 ip_tunnel_delete_net(itn
, &ipgre_link_ops
);
934 static struct pernet_operations ipgre_net_ops
= {
935 .init
= ipgre_init_net
,
936 .exit
= ipgre_exit_net
,
938 .size
= sizeof(struct ip_tunnel_net
),
941 static int ipgre_tunnel_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
949 if (data
[IFLA_GRE_IFLAGS
])
950 flags
|= nla_get_be16(data
[IFLA_GRE_IFLAGS
]);
951 if (data
[IFLA_GRE_OFLAGS
])
952 flags
|= nla_get_be16(data
[IFLA_GRE_OFLAGS
]);
953 if (flags
& (GRE_VERSION
|GRE_ROUTING
))
956 if (data
[IFLA_GRE_COLLECT_METADATA
] &&
957 data
[IFLA_GRE_ENCAP_TYPE
] &&
958 nla_get_u16(data
[IFLA_GRE_ENCAP_TYPE
]) != TUNNEL_ENCAP_NONE
)
964 static int ipgre_tap_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
968 if (tb
[IFLA_ADDRESS
]) {
969 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
)
971 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
])))
972 return -EADDRNOTAVAIL
;
978 if (data
[IFLA_GRE_REMOTE
]) {
979 memcpy(&daddr
, nla_data(data
[IFLA_GRE_REMOTE
]), 4);
985 return ipgre_tunnel_validate(tb
, data
);
988 static void ipgre_netlink_parms(struct net_device
*dev
,
989 struct nlattr
*data
[],
991 struct ip_tunnel_parm
*parms
)
993 memset(parms
, 0, sizeof(*parms
));
995 parms
->iph
.protocol
= IPPROTO_GRE
;
1000 if (data
[IFLA_GRE_LINK
])
1001 parms
->link
= nla_get_u32(data
[IFLA_GRE_LINK
]);
1003 if (data
[IFLA_GRE_IFLAGS
])
1004 parms
->i_flags
= gre_flags_to_tnl_flags(nla_get_be16(data
[IFLA_GRE_IFLAGS
]));
1006 if (data
[IFLA_GRE_OFLAGS
])
1007 parms
->o_flags
= gre_flags_to_tnl_flags(nla_get_be16(data
[IFLA_GRE_OFLAGS
]));
1009 if (data
[IFLA_GRE_IKEY
])
1010 parms
->i_key
= nla_get_be32(data
[IFLA_GRE_IKEY
]);
1012 if (data
[IFLA_GRE_OKEY
])
1013 parms
->o_key
= nla_get_be32(data
[IFLA_GRE_OKEY
]);
1015 if (data
[IFLA_GRE_LOCAL
])
1016 parms
->iph
.saddr
= nla_get_in_addr(data
[IFLA_GRE_LOCAL
]);
1018 if (data
[IFLA_GRE_REMOTE
])
1019 parms
->iph
.daddr
= nla_get_in_addr(data
[IFLA_GRE_REMOTE
]);
1021 if (data
[IFLA_GRE_TTL
])
1022 parms
->iph
.ttl
= nla_get_u8(data
[IFLA_GRE_TTL
]);
1024 if (data
[IFLA_GRE_TOS
])
1025 parms
->iph
.tos
= nla_get_u8(data
[IFLA_GRE_TOS
]);
1027 if (!data
[IFLA_GRE_PMTUDISC
] || nla_get_u8(data
[IFLA_GRE_PMTUDISC
]))
1028 parms
->iph
.frag_off
= htons(IP_DF
);
1030 if (data
[IFLA_GRE_COLLECT_METADATA
]) {
1031 struct ip_tunnel
*t
= netdev_priv(dev
);
1033 t
->collect_md
= true;
1037 /* This function returns true when ENCAP attributes are present in the nl msg */
1038 static bool ipgre_netlink_encap_parms(struct nlattr
*data
[],
1039 struct ip_tunnel_encap
*ipencap
)
1043 memset(ipencap
, 0, sizeof(*ipencap
));
1048 if (data
[IFLA_GRE_ENCAP_TYPE
]) {
1050 ipencap
->type
= nla_get_u16(data
[IFLA_GRE_ENCAP_TYPE
]);
1053 if (data
[IFLA_GRE_ENCAP_FLAGS
]) {
1055 ipencap
->flags
= nla_get_u16(data
[IFLA_GRE_ENCAP_FLAGS
]);
1058 if (data
[IFLA_GRE_ENCAP_SPORT
]) {
1060 ipencap
->sport
= nla_get_be16(data
[IFLA_GRE_ENCAP_SPORT
]);
1063 if (data
[IFLA_GRE_ENCAP_DPORT
]) {
1065 ipencap
->dport
= nla_get_be16(data
[IFLA_GRE_ENCAP_DPORT
]);
1071 static int gre_tap_init(struct net_device
*dev
)
1073 __gre_tunnel_init(dev
);
1074 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1076 return ip_tunnel_init(dev
);
1079 static const struct net_device_ops gre_tap_netdev_ops
= {
1080 .ndo_init
= gre_tap_init
,
1081 .ndo_uninit
= ip_tunnel_uninit
,
1082 .ndo_start_xmit
= gre_tap_xmit
,
1083 .ndo_set_mac_address
= eth_mac_addr
,
1084 .ndo_validate_addr
= eth_validate_addr
,
1085 .ndo_change_mtu
= ip_tunnel_change_mtu
,
1086 .ndo_get_stats64
= ip_tunnel_get_stats64
,
1087 .ndo_get_iflink
= ip_tunnel_get_iflink
,
1088 .ndo_fill_metadata_dst
= gre_fill_metadata_dst
,
1091 static void ipgre_tap_setup(struct net_device
*dev
)
1094 dev
->netdev_ops
= &gre_tap_netdev_ops
;
1095 dev
->priv_flags
&= ~IFF_TX_SKB_SHARING
;
1096 dev
->priv_flags
|= IFF_LIVE_ADDR_CHANGE
;
1097 ip_tunnel_setup(dev
, gre_tap_net_id
);
1100 static int ipgre_newlink(struct net
*src_net
, struct net_device
*dev
,
1101 struct nlattr
*tb
[], struct nlattr
*data
[])
1103 struct ip_tunnel_parm p
;
1104 struct ip_tunnel_encap ipencap
;
1106 if (ipgre_netlink_encap_parms(data
, &ipencap
)) {
1107 struct ip_tunnel
*t
= netdev_priv(dev
);
1108 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
1114 ipgre_netlink_parms(dev
, data
, tb
, &p
);
1115 return ip_tunnel_newlink(dev
, tb
, &p
);
1118 static int ipgre_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
1119 struct nlattr
*data
[])
1121 struct ip_tunnel_parm p
;
1122 struct ip_tunnel_encap ipencap
;
1124 if (ipgre_netlink_encap_parms(data
, &ipencap
)) {
1125 struct ip_tunnel
*t
= netdev_priv(dev
);
1126 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
1132 ipgre_netlink_parms(dev
, data
, tb
, &p
);
1133 return ip_tunnel_changelink(dev
, tb
, &p
);
1136 static size_t ipgre_get_size(const struct net_device
*dev
)
1141 /* IFLA_GRE_IFLAGS */
1143 /* IFLA_GRE_OFLAGS */
1149 /* IFLA_GRE_LOCAL */
1151 /* IFLA_GRE_REMOTE */
1157 /* IFLA_GRE_PMTUDISC */
1159 /* IFLA_GRE_ENCAP_TYPE */
1161 /* IFLA_GRE_ENCAP_FLAGS */
1163 /* IFLA_GRE_ENCAP_SPORT */
1165 /* IFLA_GRE_ENCAP_DPORT */
1167 /* IFLA_GRE_COLLECT_METADATA */
1172 static int ipgre_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1174 struct ip_tunnel
*t
= netdev_priv(dev
);
1175 struct ip_tunnel_parm
*p
= &t
->parms
;
1177 if (nla_put_u32(skb
, IFLA_GRE_LINK
, p
->link
) ||
1178 nla_put_be16(skb
, IFLA_GRE_IFLAGS
, tnl_flags_to_gre_flags(p
->i_flags
)) ||
1179 nla_put_be16(skb
, IFLA_GRE_OFLAGS
, tnl_flags_to_gre_flags(p
->o_flags
)) ||
1180 nla_put_be32(skb
, IFLA_GRE_IKEY
, p
->i_key
) ||
1181 nla_put_be32(skb
, IFLA_GRE_OKEY
, p
->o_key
) ||
1182 nla_put_in_addr(skb
, IFLA_GRE_LOCAL
, p
->iph
.saddr
) ||
1183 nla_put_in_addr(skb
, IFLA_GRE_REMOTE
, p
->iph
.daddr
) ||
1184 nla_put_u8(skb
, IFLA_GRE_TTL
, p
->iph
.ttl
) ||
1185 nla_put_u8(skb
, IFLA_GRE_TOS
, p
->iph
.tos
) ||
1186 nla_put_u8(skb
, IFLA_GRE_PMTUDISC
,
1187 !!(p
->iph
.frag_off
& htons(IP_DF
))))
1188 goto nla_put_failure
;
1190 if (nla_put_u16(skb
, IFLA_GRE_ENCAP_TYPE
,
1192 nla_put_be16(skb
, IFLA_GRE_ENCAP_SPORT
,
1194 nla_put_be16(skb
, IFLA_GRE_ENCAP_DPORT
,
1196 nla_put_u16(skb
, IFLA_GRE_ENCAP_FLAGS
,
1198 goto nla_put_failure
;
1200 if (t
->collect_md
) {
1201 if (nla_put_flag(skb
, IFLA_GRE_COLLECT_METADATA
))
1202 goto nla_put_failure
;
1211 static const struct nla_policy ipgre_policy
[IFLA_GRE_MAX
+ 1] = {
1212 [IFLA_GRE_LINK
] = { .type
= NLA_U32
},
1213 [IFLA_GRE_IFLAGS
] = { .type
= NLA_U16
},
1214 [IFLA_GRE_OFLAGS
] = { .type
= NLA_U16
},
1215 [IFLA_GRE_IKEY
] = { .type
= NLA_U32
},
1216 [IFLA_GRE_OKEY
] = { .type
= NLA_U32
},
1217 [IFLA_GRE_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
1218 [IFLA_GRE_REMOTE
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
1219 [IFLA_GRE_TTL
] = { .type
= NLA_U8
},
1220 [IFLA_GRE_TOS
] = { .type
= NLA_U8
},
1221 [IFLA_GRE_PMTUDISC
] = { .type
= NLA_U8
},
1222 [IFLA_GRE_ENCAP_TYPE
] = { .type
= NLA_U16
},
1223 [IFLA_GRE_ENCAP_FLAGS
] = { .type
= NLA_U16
},
1224 [IFLA_GRE_ENCAP_SPORT
] = { .type
= NLA_U16
},
1225 [IFLA_GRE_ENCAP_DPORT
] = { .type
= NLA_U16
},
1226 [IFLA_GRE_COLLECT_METADATA
] = { .type
= NLA_FLAG
},
1229 static struct rtnl_link_ops ipgre_link_ops __read_mostly
= {
1231 .maxtype
= IFLA_GRE_MAX
,
1232 .policy
= ipgre_policy
,
1233 .priv_size
= sizeof(struct ip_tunnel
),
1234 .setup
= ipgre_tunnel_setup
,
1235 .validate
= ipgre_tunnel_validate
,
1236 .newlink
= ipgre_newlink
,
1237 .changelink
= ipgre_changelink
,
1238 .dellink
= ip_tunnel_dellink
,
1239 .get_size
= ipgre_get_size
,
1240 .fill_info
= ipgre_fill_info
,
1241 .get_link_net
= ip_tunnel_get_link_net
,
1244 static struct rtnl_link_ops ipgre_tap_ops __read_mostly
= {
1246 .maxtype
= IFLA_GRE_MAX
,
1247 .policy
= ipgre_policy
,
1248 .priv_size
= sizeof(struct ip_tunnel
),
1249 .setup
= ipgre_tap_setup
,
1250 .validate
= ipgre_tap_validate
,
1251 .newlink
= ipgre_newlink
,
1252 .changelink
= ipgre_changelink
,
1253 .dellink
= ip_tunnel_dellink
,
1254 .get_size
= ipgre_get_size
,
1255 .fill_info
= ipgre_fill_info
,
1256 .get_link_net
= ip_tunnel_get_link_net
,
1259 struct net_device
*gretap_fb_dev_create(struct net
*net
, const char *name
,
1260 u8 name_assign_type
)
1262 struct nlattr
*tb
[IFLA_MAX
+ 1];
1263 struct net_device
*dev
;
1264 struct ip_tunnel
*t
;
1267 memset(&tb
, 0, sizeof(tb
));
1269 dev
= rtnl_create_link(net
, name
, name_assign_type
,
1270 &ipgre_tap_ops
, tb
);
1274 /* Configure flow based GRE device. */
1275 t
= netdev_priv(dev
);
1276 t
->collect_md
= true;
1278 err
= ipgre_newlink(net
, dev
, tb
, NULL
);
1282 /* openvswitch users expect packet sizes to be unrestricted,
1283 * so set the largest MTU we can.
1285 err
= __ip_tunnel_change_mtu(dev
, IP_MAX_MTU
, false);
1292 return ERR_PTR(err
);
1294 EXPORT_SYMBOL_GPL(gretap_fb_dev_create
);
1296 static int __net_init
ipgre_tap_init_net(struct net
*net
)
1298 return ip_tunnel_init_net(net
, gre_tap_net_id
, &ipgre_tap_ops
, "gretap0");
1301 static void __net_exit
ipgre_tap_exit_net(struct net
*net
)
1303 struct ip_tunnel_net
*itn
= net_generic(net
, gre_tap_net_id
);
1304 ip_tunnel_delete_net(itn
, &ipgre_tap_ops
);
1307 static struct pernet_operations ipgre_tap_net_ops
= {
1308 .init
= ipgre_tap_init_net
,
1309 .exit
= ipgre_tap_exit_net
,
1310 .id
= &gre_tap_net_id
,
1311 .size
= sizeof(struct ip_tunnel_net
),
1314 static int __init
ipgre_init(void)
1318 pr_info("GRE over IPv4 tunneling driver\n");
1320 err
= register_pernet_device(&ipgre_net_ops
);
1324 err
= register_pernet_device(&ipgre_tap_net_ops
);
1326 goto pnet_tap_faied
;
1328 err
= gre_add_protocol(&ipgre_protocol
, GREPROTO_CISCO
);
1330 pr_info("%s: can't add protocol\n", __func__
);
1331 goto add_proto_failed
;
1334 err
= rtnl_link_register(&ipgre_link_ops
);
1336 goto rtnl_link_failed
;
1338 err
= rtnl_link_register(&ipgre_tap_ops
);
1340 goto tap_ops_failed
;
1345 rtnl_link_unregister(&ipgre_link_ops
);
1347 gre_del_protocol(&ipgre_protocol
, GREPROTO_CISCO
);
1349 unregister_pernet_device(&ipgre_tap_net_ops
);
1351 unregister_pernet_device(&ipgre_net_ops
);
1355 static void __exit
ipgre_fini(void)
1357 rtnl_link_unregister(&ipgre_tap_ops
);
1358 rtnl_link_unregister(&ipgre_link_ops
);
1359 gre_del_protocol(&ipgre_protocol
, GREPROTO_CISCO
);
1360 unregister_pernet_device(&ipgre_tap_net_ops
);
1361 unregister_pernet_device(&ipgre_net_ops
);
1364 module_init(ipgre_init
);
1365 module_exit(ipgre_fini
);
1366 MODULE_LICENSE("GPL");
1367 MODULE_ALIAS_RTNL_LINK("gre");
1368 MODULE_ALIAS_RTNL_LINK("gretap");
1369 MODULE_ALIAS_NETDEV("gre0");
1370 MODULE_ALIAS_NETDEV("gretap0");