]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blame_incremental - net/ipv4/ip_gre.c
projects / mirror_ubuntu-focal-kernel.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
UBUNTU: Ubuntu-5.4.0-117.132
[mirror_ubuntu-focal-kernel.git] / net / ipv4 / ip_gre.c
... / ...
CommitLineData
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Linux NET3: GRE over IP protocol decoder.
4 *
5 * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.ru)
6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/capability.h>
11#include <linux/module.h>
12#include <linux/types.h>
13#include <linux/kernel.h>
14#include <linux/slab.h>
15#include <linux/uaccess.h>
16#include <linux/skbuff.h>
17#include <linux/netdevice.h>
18#include <linux/in.h>
19#include <linux/tcp.h>
20#include <linux/udp.h>
21#include <linux/if_arp.h>
22#include <linux/if_vlan.h>
23#include <linux/init.h>
24#include <linux/in6.h>
25#include <linux/inetdevice.h>
26#include <linux/igmp.h>
27#include <linux/netfilter_ipv4.h>
28#include <linux/etherdevice.h>
29#include <linux/if_ether.h>
30
31#include <net/sock.h>
32#include <net/ip.h>
33#include <net/icmp.h>
34#include <net/protocol.h>
35#include <net/ip_tunnels.h>
36#include <net/arp.h>
37#include <net/checksum.h>
38#include <net/dsfield.h>
39#include <net/inet_ecn.h>
40#include <net/xfrm.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
43#include <net/rtnetlink.h>
44#include <net/gre.h>
45#include <net/dst_metadata.h>
46#include <net/erspan.h>
47
48/*
49 Problems & solutions
50 --------------------
51
52 1. The most important issue is detecting local dead loops.
53 They would cause complete host lockup in transmit, which
54 would be "resolved" by stack overflow or, if queueing is enabled,
55 with infinite looping in net_bh.
56
57 We cannot track such dead loops during route installation,
58 it is infeasible task. The most general solutions would be
59 to keep skb->encapsulation counter (sort of local ttl),
60 and silently drop packet when it expires. It is a good
61 solution, but it supposes maintaining new variable in ALL
62 skb, even if no tunneling is used.
63
64 Current solution: xmit_recursion breaks dead loops. This is a percpu
65 counter, since when we enter the first ndo_xmit(), cpu migration is
66 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
67
68 2. Networking dead loops would not kill routers, but would really
69 kill network. IP hop limit plays role of "t->recursion" in this case,
70 if we copy it from packet being encapsulated to upper header.
71 It is very good solution, but it introduces two problems:
72
73 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
74 do not work over tunnels.
75 - traceroute does not work. I planned to relay ICMP from tunnel,
76 so that this problem would be solved and traceroute output
77 would even more informative. This idea appeared to be wrong:
78 only Linux complies to rfc1812 now (yes, guys, Linux is the only
79 true router now :-)), all routers (at least, in neighbourhood of mine)
80 return only 8 bytes of payload. It is the end.
81
82 Hence, if we want that OSPF worked or traceroute said something reasonable,
83 we should search for another solution.
84
85 One of them is to parse packet trying to detect inner encapsulation
86 made by our node. It is difficult or even impossible, especially,
87 taking into account fragmentation. TO be short, ttl is not solution at all.
88
89 Current solution: The solution was UNEXPECTEDLY SIMPLE.
90 We force DF flag on tunnels with preconfigured hop limit,
91 that is ALL. :-) Well, it does not remove the problem completely,
92 but exponential growth of network traffic is changed to linear
93 (branches, that exceed pmtu are pruned) and tunnel mtu
94 rapidly degrades to value <68, where looping stops.
95 Yes, it is not good if there exists a router in the loop,
96 which does not force DF, even when encapsulating packets have DF set.
97 But it is not our problem! Nobody could accuse us, we made
98 all that we could make. Even if it is your gated who injected
99 fatal route to network, even if it were you who configured
100 fatal static route: you are innocent. :-)
101
102 Alexey Kuznetsov.
103 */
104
105static bool log_ecn_error = true;
106module_param(log_ecn_error, bool, 0644);
107MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
108
109static struct rtnl_link_ops ipgre_link_ops __read_mostly;
110static int ipgre_tunnel_init(struct net_device *dev);
111static void erspan_build_header(struct sk_buff *skb,
112 u32 id, u32 index,
113 bool truncate, bool is_ipv4);
114
115static unsigned int ipgre_net_id __read_mostly;
116static unsigned int gre_tap_net_id __read_mostly;
117static unsigned int erspan_net_id __read_mostly;
118
119static int ipgre_err(struct sk_buff *skb, u32 info,
120 const struct tnl_ptk_info *tpi)
121{
122
123 /* All the routers (except for Linux) return only
124 8 bytes of packet payload. It means, that precise relaying of
125 ICMP in the real Internet is absolutely infeasible.
126
127 Moreover, Cisco "wise men" put GRE key to the third word
128 in GRE header. It makes impossible maintaining even soft
129 state for keyed GRE tunnels with enabled checksum. Tell
130 them "thank you".
131
132 Well, I wonder, rfc1812 was written by Cisco employee,
133 what the hell these idiots break standards established
134 by themselves???
135 */
136 struct net *net = dev_net(skb->dev);
137 struct ip_tunnel_net *itn;
138 const struct iphdr *iph;
139 const int type = icmp_hdr(skb)->type;
140 const int code = icmp_hdr(skb)->code;
141 unsigned int data_len = 0;
142 struct ip_tunnel *t;
143
144 if (tpi->proto == htons(ETH_P_TEB))
145 itn = net_generic(net, gre_tap_net_id);
146 else if (tpi->proto == htons(ETH_P_ERSPAN) ||
147 tpi->proto == htons(ETH_P_ERSPAN2))
148 itn = net_generic(net, erspan_net_id);
149 else
150 itn = net_generic(net, ipgre_net_id);
151
152 iph = (const struct iphdr *)(icmp_hdr(skb) + 1);
153 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
154 iph->daddr, iph->saddr, tpi->key);
155
156 if (!t)
157 return -ENOENT;
158
159 switch (type) {
160 default:
161 case ICMP_PARAMETERPROB:
162 return 0;
163
164 case ICMP_DEST_UNREACH:
165 switch (code) {
166 case ICMP_SR_FAILED:
167 case ICMP_PORT_UNREACH:
168 /* Impossible event. */
169 return 0;
170 default:
171 /* All others are translated to HOST_UNREACH.
172 rfc2003 contains "deep thoughts" about NET_UNREACH,
173 I believe they are just ether pollution. --ANK
174 */
175 break;
176 }
177 break;
178
179 case ICMP_TIME_EXCEEDED:
180 if (code != ICMP_EXC_TTL)
181 return 0;
182 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
183 break;
184
185 case ICMP_REDIRECT:
186 break;
187 }
188
189#if IS_ENABLED(CONFIG_IPV6)
190 if (tpi->proto == htons(ETH_P_IPV6) &&
191 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4 + tpi->hdr_len,
192 type, data_len))
193 return 0;
194#endif
195
196 if (t->parms.iph.daddr == 0 ||
197 ipv4_is_multicast(t->parms.iph.daddr))
198 return 0;
199
200 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
201 return 0;
202
203 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
204 t->err_count++;
205 else
206 t->err_count = 1;
207 t->err_time = jiffies;
208
209 return 0;
210}
211
212static void gre_err(struct sk_buff *skb, u32 info)
213{
214 /* All the routers (except for Linux) return only
215 * 8 bytes of packet payload. It means, that precise relaying of
216 * ICMP in the real Internet is absolutely infeasible.
217 *
218 * Moreover, Cisco "wise men" put GRE key to the third word
219 * in GRE header. It makes impossible maintaining even soft
220 * state for keyed
221 * GRE tunnels with enabled checksum. Tell them "thank you".
222 *
223 * Well, I wonder, rfc1812 was written by Cisco employee,
224 * what the hell these idiots break standards established
225 * by themselves???
226 */
227
228 const struct iphdr *iph = (struct iphdr *)skb->data;
229 const int type = icmp_hdr(skb)->type;
230 const int code = icmp_hdr(skb)->code;
231 struct tnl_ptk_info tpi;
232
233 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IP),
234 iph->ihl * 4) < 0)
235 return;
236
237 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
238 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
239 skb->dev->ifindex, IPPROTO_GRE);
240 return;
241 }
242 if (type == ICMP_REDIRECT) {
243 ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex,
244 IPPROTO_GRE);
245 return;
246 }
247
248 ipgre_err(skb, info, &tpi);
249}
250
251static int erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
252 int gre_hdr_len)
253{
254 struct net *net = dev_net(skb->dev);
255 struct metadata_dst *tun_dst = NULL;
256 struct erspan_base_hdr *ershdr;
257 struct ip_tunnel_net *itn;
258 struct ip_tunnel *tunnel;
259 const struct iphdr *iph;
260 struct erspan_md2 *md2;
261 int ver;
262 int len;
263
264 itn = net_generic(net, erspan_net_id);
265
266 iph = ip_hdr(skb);
267 ershdr = (struct erspan_base_hdr *)(skb->data + gre_hdr_len);
268 ver = ershdr->ver;
269
270 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex,
271 tpi->flags | TUNNEL_KEY,
272 iph->saddr, iph->daddr, tpi->key);
273
274 if (tunnel) {
275 len = gre_hdr_len + erspan_hdr_len(ver);
276 if (unlikely(!pskb_may_pull(skb, len)))
277 return PACKET_REJECT;
278
279 if (__iptunnel_pull_header(skb,
280 len,
281 htons(ETH_P_TEB),
282 false, false) < 0)
283 goto drop;
284
285 if (tunnel->collect_md) {
286 struct erspan_metadata *pkt_md, *md;
287 struct ip_tunnel_info *info;
288 unsigned char *gh;
289 __be64 tun_id;
290 __be16 flags;
291
292 tpi->flags |= TUNNEL_KEY;
293 flags = tpi->flags;
294 tun_id = key32_to_tunnel_id(tpi->key);
295
296 tun_dst = ip_tun_rx_dst(skb, flags,
297 tun_id, sizeof(*md));
298 if (!tun_dst)
299 return PACKET_REJECT;
300
301 /* skb can be uncloned in __iptunnel_pull_header, so
302 * old pkt_md is no longer valid and we need to reset
303 * it
304 */
305 gh = skb_network_header(skb) +
306 skb_network_header_len(skb);
307 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
308 sizeof(*ershdr));
309 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
310 md->version = ver;
311 md2 = &md->u.md2;
312 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
313 ERSPAN_V2_MDSIZE);
314
315 info = &tun_dst->u.tun_info;
316 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
317 info->options_len = sizeof(*md);
318 }
319
320 skb_reset_mac_header(skb);
321 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
322 return PACKET_RCVD;
323 }
324 return PACKET_REJECT;
325
326drop:
327 kfree_skb(skb);
328 return PACKET_RCVD;
329}
330
331static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
332 struct ip_tunnel_net *itn, int hdr_len, bool raw_proto)
333{
334 struct metadata_dst *tun_dst = NULL;
335 const struct iphdr *iph;
336 struct ip_tunnel *tunnel;
337
338 iph = ip_hdr(skb);
339 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
340 iph->saddr, iph->daddr, tpi->key);
341
342 if (tunnel) {
343 if (__iptunnel_pull_header(skb, hdr_len, tpi->proto,
344 raw_proto, false) < 0)
345 goto drop;
346
347 if (tunnel->dev->type != ARPHRD_NONE)
348 skb_pop_mac_header(skb);
349 else
350 skb_reset_mac_header(skb);
351 if (tunnel->collect_md) {
352 __be16 flags;
353 __be64 tun_id;
354
355 flags = tpi->flags & (TUNNEL_CSUM | TUNNEL_KEY);
356 tun_id = key32_to_tunnel_id(tpi->key);
357 tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0);
358 if (!tun_dst)
359 return PACKET_REJECT;
360 }
361
362 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
363 return PACKET_RCVD;
364 }
365 return PACKET_NEXT;
366
367drop:
368 kfree_skb(skb);
369 return PACKET_RCVD;
370}
371
372static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
373 int hdr_len)
374{
375 struct net *net = dev_net(skb->dev);
376 struct ip_tunnel_net *itn;
377 int res;
378
379 if (tpi->proto == htons(ETH_P_TEB))
380 itn = net_generic(net, gre_tap_net_id);
381 else
382 itn = net_generic(net, ipgre_net_id);
383
384 res = __ipgre_rcv(skb, tpi, itn, hdr_len, false);
385 if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) {
386 /* ipgre tunnels in collect metadata mode should receive
387 * also ETH_P_TEB traffic.
388 */
389 itn = net_generic(net, ipgre_net_id);
390 res = __ipgre_rcv(skb, tpi, itn, hdr_len, true);
391 }
392 return res;
393}
394
395static int gre_rcv(struct sk_buff *skb)
396{
397 struct tnl_ptk_info tpi;
398 bool csum_err = false;
399 int hdr_len;
400
401#ifdef CONFIG_NET_IPGRE_BROADCAST
402 if (ipv4_is_multicast(ip_hdr(skb)->daddr)) {
403 /* Looped back packet, drop it! */
404 if (rt_is_output_route(skb_rtable(skb)))
405 goto drop;
406 }
407#endif
408
409 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0);
410 if (hdr_len < 0)
411 goto drop;
412
413 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
414 tpi.proto == htons(ETH_P_ERSPAN2))) {
415 if (erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
416 return 0;
417 goto out;
418 }
419
420 if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
421 return 0;
422
423out:
424 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
425drop:
426 kfree_skb(skb);
427 return 0;
428}
429
430static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
431 const struct iphdr *tnl_params,
432 __be16 proto)
433{
434 struct ip_tunnel *tunnel = netdev_priv(dev);
435
436 if (tunnel->parms.o_flags & TUNNEL_SEQ)
437 tunnel->o_seqno++;
438
439 /* Push GRE header. */
440 gre_build_header(skb, tunnel->tun_hlen,
441 tunnel->parms.o_flags, proto, tunnel->parms.o_key,
442 htonl(tunnel->o_seqno));
443
444 ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
445}
446
447static int gre_handle_offloads(struct sk_buff *skb, bool csum)
448{
449 return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
450}
451
452static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
453 __be16 proto)
454{
455 struct ip_tunnel *tunnel = netdev_priv(dev);
456 struct ip_tunnel_info *tun_info;
457 const struct ip_tunnel_key *key;
458 int tunnel_hlen;
459 __be16 flags;
460
461 tun_info = skb_tunnel_info(skb);
462 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
463 ip_tunnel_info_af(tun_info) != AF_INET))
464 goto err_free_skb;
465
466 key = &tun_info->key;
467 tunnel_hlen = gre_calc_hlen(key->tun_flags);
468
469 if (skb_cow_head(skb, dev->needed_headroom))
470 goto err_free_skb;
471
472 /* Push Tunnel header. */
473 if (gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM)))
474 goto err_free_skb;
475
476 flags = tun_info->key.tun_flags &
477 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
478 gre_build_header(skb, tunnel_hlen, flags, proto,
479 tunnel_id_to_key32(tun_info->key.tun_id),
480 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) : 0);
481
482 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
483
484 return;
485
486err_free_skb:
487 kfree_skb(skb);
488 dev->stats.tx_dropped++;
489}
490
491static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
492{
493 struct ip_tunnel *tunnel = netdev_priv(dev);
494 struct ip_tunnel_info *tun_info;
495 const struct ip_tunnel_key *key;
496 struct erspan_metadata *md;
497 bool truncate = false;
498 __be16 proto;
499 int tunnel_hlen;
500 int version;
501 int nhoff;
502 int thoff;
503
504 tun_info = skb_tunnel_info(skb);
505 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
506 ip_tunnel_info_af(tun_info) != AF_INET))
507 goto err_free_skb;
508
509 key = &tun_info->key;
510 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
511 goto err_free_skb;
512 if (tun_info->options_len < sizeof(*md))
513 goto err_free_skb;
514 md = ip_tunnel_info_opts(tun_info);
515
516 /* ERSPAN has fixed 8 byte GRE header */
517 version = md->version;
518 tunnel_hlen = 8 + erspan_hdr_len(version);
519
520 if (skb_cow_head(skb, dev->needed_headroom))
521 goto err_free_skb;
522
523 if (gre_handle_offloads(skb, false))
524 goto err_free_skb;
525
526 if (skb->len > dev->mtu + dev->hard_header_len) {
527 pskb_trim(skb, dev->mtu + dev->hard_header_len);
528 truncate = true;
529 }
530
531 nhoff = skb_network_header(skb) - skb_mac_header(skb);
532 if (skb->protocol == htons(ETH_P_IP) &&
533 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
534 truncate = true;
535
536 thoff = skb_transport_header(skb) - skb_mac_header(skb);
537 if (skb->protocol == htons(ETH_P_IPV6) &&
538 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
539 truncate = true;
540
541 if (version == 1) {
542 erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
543 ntohl(md->u.index), truncate, true);
544 proto = htons(ETH_P_ERSPAN);
545 } else if (version == 2) {
546 erspan_build_header_v2(skb,
547 ntohl(tunnel_id_to_key32(key->tun_id)),
548 md->u.md2.dir,
549 get_hwid(&md->u.md2),
550 truncate, true);
551 proto = htons(ETH_P_ERSPAN2);
552 } else {
553 goto err_free_skb;
554 }
555
556 gre_build_header(skb, 8, TUNNEL_SEQ,
557 proto, 0, htonl(tunnel->o_seqno++));
558
559 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
560
561 return;
562
563err_free_skb:
564 kfree_skb(skb);
565 dev->stats.tx_dropped++;
566}
567
568static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
569{
570 struct ip_tunnel_info *info = skb_tunnel_info(skb);
571 const struct ip_tunnel_key *key;
572 struct rtable *rt;
573 struct flowi4 fl4;
574
575 if (ip_tunnel_info_af(info) != AF_INET)
576 return -EINVAL;
577
578 key = &info->key;
579 ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
580 tunnel_id_to_key32(key->tun_id),
581 key->tos & ~INET_ECN_MASK, 0, skb->mark,
582 skb_get_hash(skb));
583 rt = ip_route_output_key(dev_net(dev), &fl4);
584 if (IS_ERR(rt))
585 return PTR_ERR(rt);
586
587 ip_rt_put(rt);
588 info->key.u.ipv4.src = fl4.saddr;
589 return 0;
590}
591
592static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
593 struct net_device *dev)
594{
595 struct ip_tunnel *tunnel = netdev_priv(dev);
596 const struct iphdr *tnl_params;
597
598 if (!pskb_inet_may_pull(skb))
599 goto free_skb;
600
601 if (tunnel->collect_md) {
602 gre_fb_xmit(skb, dev, skb->protocol);
603 return NETDEV_TX_OK;
604 }
605
606 if (dev->header_ops) {
607 const int pull_len = tunnel->hlen + sizeof(struct iphdr);
608
609 if (skb_cow_head(skb, 0))
610 goto free_skb;
611
612 tnl_params = (const struct iphdr *)skb->data;
613
614 if (pull_len > skb_transport_offset(skb))
615 goto free_skb;
616
617 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
618 * to gre header.
619 */
620 skb_pull(skb, pull_len);
621 skb_reset_mac_header(skb);
622 } else {
623 if (skb_cow_head(skb, dev->needed_headroom))
624 goto free_skb;
625
626 tnl_params = &tunnel->parms.iph;
627 }
628
629 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
630 goto free_skb;
631
632 __gre_xmit(skb, dev, tnl_params, skb->protocol);
633 return NETDEV_TX_OK;
634
635free_skb:
636 kfree_skb(skb);
637 dev->stats.tx_dropped++;
638 return NETDEV_TX_OK;
639}
640
641static netdev_tx_t erspan_xmit(struct sk_buff *skb,
642 struct net_device *dev)
643{
644 struct ip_tunnel *tunnel = netdev_priv(dev);
645 bool truncate = false;
646 __be16 proto;
647
648 if (!pskb_inet_may_pull(skb))
649 goto free_skb;
650
651 if (tunnel->collect_md) {
652 erspan_fb_xmit(skb, dev);
653 return NETDEV_TX_OK;
654 }
655
656 if (gre_handle_offloads(skb, false))
657 goto free_skb;
658
659 if (skb_cow_head(skb, dev->needed_headroom))
660 goto free_skb;
661
662 if (skb->len > dev->mtu + dev->hard_header_len) {
663 pskb_trim(skb, dev->mtu + dev->hard_header_len);
664 truncate = true;
665 }
666
667 /* Push ERSPAN header */
668 if (tunnel->erspan_ver == 1) {
669 erspan_build_header(skb, ntohl(tunnel->parms.o_key),
670 tunnel->index,
671 truncate, true);
672 proto = htons(ETH_P_ERSPAN);
673 } else if (tunnel->erspan_ver == 2) {
674 erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
675 tunnel->dir, tunnel->hwid,
676 truncate, true);
677 proto = htons(ETH_P_ERSPAN2);
678 } else {
679 goto free_skb;
680 }
681
682 tunnel->parms.o_flags &= ~TUNNEL_KEY;
683 __gre_xmit(skb, dev, &tunnel->parms.iph, proto);
684 return NETDEV_TX_OK;
685
686free_skb:
687 kfree_skb(skb);
688 dev->stats.tx_dropped++;
689 return NETDEV_TX_OK;
690}
691
692static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
693 struct net_device *dev)
694{
695 struct ip_tunnel *tunnel = netdev_priv(dev);
696
697 if (!pskb_inet_may_pull(skb))
698 goto free_skb;
699
700 if (tunnel->collect_md) {
701 gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
702 return NETDEV_TX_OK;
703 }
704
705 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
706 goto free_skb;
707
708 if (skb_cow_head(skb, dev->needed_headroom))
709 goto free_skb;
710
711 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
712 return NETDEV_TX_OK;
713
714free_skb:
715 kfree_skb(skb);
716 dev->stats.tx_dropped++;
717 return NETDEV_TX_OK;
718}
719
720static void ipgre_link_update(struct net_device *dev, bool set_mtu)
721{
722 struct ip_tunnel *tunnel = netdev_priv(dev);
723 int len;
724
725 len = tunnel->tun_hlen;
726 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
727 len = tunnel->tun_hlen - len;
728 tunnel->hlen = tunnel->hlen + len;
729
730 if (dev->header_ops)
731 dev->hard_header_len += len;
732 else
733 dev->needed_headroom += len;
734
735 if (set_mtu)
736 dev->mtu = max_t(int, dev->mtu - len, 68);
737
738 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
739 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
740 tunnel->encap.type == TUNNEL_ENCAP_NONE) {
741 dev->features |= NETIF_F_GSO_SOFTWARE;
742 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
743 } else {
744 dev->features &= ~NETIF_F_GSO_SOFTWARE;
745 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
746 }
747 dev->features |= NETIF_F_LLTX;
748 } else {
749 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
750 dev->features &= ~(NETIF_F_LLTX | NETIF_F_GSO_SOFTWARE);
751 }
752}
753
754static int ipgre_tunnel_ioctl(struct net_device *dev,
755 struct ifreq *ifr, int cmd)
756{
757 struct ip_tunnel_parm p;
758 int err;
759
760 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
761 return -EFAULT;
762
763 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
764 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
765 p.iph.ihl != 5 || (p.iph.frag_off & htons(~IP_DF)) ||
766 ((p.i_flags | p.o_flags) & (GRE_VERSION | GRE_ROUTING)))
767 return -EINVAL;
768 }
769
770 p.i_flags = gre_flags_to_tnl_flags(p.i_flags);
771 p.o_flags = gre_flags_to_tnl_flags(p.o_flags);
772
773 err = ip_tunnel_ioctl(dev, &p, cmd);
774 if (err)
775 return err;
776
777 if (cmd == SIOCCHGTUNNEL) {
778 struct ip_tunnel *t = netdev_priv(dev);
779
780 t->parms.i_flags = p.i_flags;
781 t->parms.o_flags = p.o_flags;
782
783 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
784 ipgre_link_update(dev, true);
785 }
786
787 p.i_flags = gre_tnl_flags_to_gre_flags(p.i_flags);
788 p.o_flags = gre_tnl_flags_to_gre_flags(p.o_flags);
789
790 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
791 return -EFAULT;
792
793 return 0;
794}
795
796/* Nice toy. Unfortunately, useless in real life :-)
797 It allows to construct virtual multiprotocol broadcast "LAN"
798 over the Internet, provided multicast routing is tuned.
799
800
801 I have no idea was this bicycle invented before me,
802 so that I had to set ARPHRD_IPGRE to a random value.
803 I have an impression, that Cisco could make something similar,
804 but this feature is apparently missing in IOS<=11.2(8).
805
806 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
807 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
808
809 ping -t 255 224.66.66.66
810
811 If nobody answers, mbone does not work.
812
813 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
814 ip addr add 10.66.66.<somewhat>/24 dev Universe
815 ifconfig Universe up
816 ifconfig Universe add fe80::<Your_real_addr>/10
817 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
818 ftp 10.66.66.66
819 ...
820 ftp fec0:6666:6666::193.233.7.65
821 ...
822 */
823static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
824 unsigned short type,
825 const void *daddr, const void *saddr, unsigned int len)
826{
827 struct ip_tunnel *t = netdev_priv(dev);
828 struct iphdr *iph;
829 struct gre_base_hdr *greh;
830
831 iph = skb_push(skb, t->hlen + sizeof(*iph));
832 greh = (struct gre_base_hdr *)(iph+1);
833 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
834 greh->protocol = htons(type);
835
836 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
837
838 /* Set the source hardware address. */
839 if (saddr)
840 memcpy(&iph->saddr, saddr, 4);
841 if (daddr)
842 memcpy(&iph->daddr, daddr, 4);
843 if (iph->daddr)
844 return t->hlen + sizeof(*iph);
845
846 return -(t->hlen + sizeof(*iph));
847}
848
849static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
850{
851 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
852 memcpy(haddr, &iph->saddr, 4);
853 return 4;
854}
855
856static const struct header_ops ipgre_header_ops = {
857 .create = ipgre_header,
858 .parse = ipgre_header_parse,
859};
860
861#ifdef CONFIG_NET_IPGRE_BROADCAST
862static int ipgre_open(struct net_device *dev)
863{
864 struct ip_tunnel *t = netdev_priv(dev);
865
866 if (ipv4_is_multicast(t->parms.iph.daddr)) {
867 struct flowi4 fl4;
868 struct rtable *rt;
869
870 rt = ip_route_output_gre(t->net, &fl4,
871 t->parms.iph.daddr,
872 t->parms.iph.saddr,
873 t->parms.o_key,
874 RT_TOS(t->parms.iph.tos),
875 t->parms.link);
876 if (IS_ERR(rt))
877 return -EADDRNOTAVAIL;
878 dev = rt->dst.dev;
879 ip_rt_put(rt);
880 if (!__in_dev_get_rtnl(dev))
881 return -EADDRNOTAVAIL;
882 t->mlink = dev->ifindex;
883 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
884 }
885 return 0;
886}
887
888static int ipgre_close(struct net_device *dev)
889{
890 struct ip_tunnel *t = netdev_priv(dev);
891
892 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
893 struct in_device *in_dev;
894 in_dev = inetdev_by_index(t->net, t->mlink);
895 if (in_dev)
896 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
897 }
898 return 0;
899}
900#endif
901
902static const struct net_device_ops ipgre_netdev_ops = {
903 .ndo_init = ipgre_tunnel_init,
904 .ndo_uninit = ip_tunnel_uninit,
905#ifdef CONFIG_NET_IPGRE_BROADCAST
906 .ndo_open = ipgre_open,
907 .ndo_stop = ipgre_close,
908#endif
909 .ndo_start_xmit = ipgre_xmit,
910 .ndo_do_ioctl = ipgre_tunnel_ioctl,
911 .ndo_change_mtu = ip_tunnel_change_mtu,
912 .ndo_get_stats64 = ip_tunnel_get_stats64,
913 .ndo_get_iflink = ip_tunnel_get_iflink,
914};
915
916#define GRE_FEATURES (NETIF_F_SG | \
917 NETIF_F_FRAGLIST | \
918 NETIF_F_HIGHDMA | \
919 NETIF_F_HW_CSUM)
920
921static void ipgre_tunnel_setup(struct net_device *dev)
922{
923 dev->netdev_ops = &ipgre_netdev_ops;
924 dev->type = ARPHRD_IPGRE;
925 ip_tunnel_setup(dev, ipgre_net_id);
926}
927
928static void __gre_tunnel_init(struct net_device *dev)
929{
930 struct ip_tunnel *tunnel;
931
932 tunnel = netdev_priv(dev);
933 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
934 tunnel->parms.iph.protocol = IPPROTO_GRE;
935
936 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
937 dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
938
939 dev->features |= GRE_FEATURES;
940 dev->hw_features |= GRE_FEATURES;
941
942 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
943 /* TCP offload with GRE SEQ is not supported, nor
944 * can we support 2 levels of outer headers requiring
945 * an update.
946 */
947 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
948 (tunnel->encap.type == TUNNEL_ENCAP_NONE)) {
949 dev->features |= NETIF_F_GSO_SOFTWARE;
950 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
951 }
952
953 /* Can use a lockless transmit, unless we generate
954 * output sequences
955 */
956 dev->features |= NETIF_F_LLTX;
957 }
958}
959
960static int ipgre_tunnel_init(struct net_device *dev)
961{
962 struct ip_tunnel *tunnel = netdev_priv(dev);
963 struct iphdr *iph = &tunnel->parms.iph;
964
965 __gre_tunnel_init(dev);
966
967 memcpy(dev->dev_addr, &iph->saddr, 4);
968 memcpy(dev->broadcast, &iph->daddr, 4);
969
970 dev->flags = IFF_NOARP;
971 netif_keep_dst(dev);
972 dev->addr_len = 4;
973
974 if (iph->daddr && !tunnel->collect_md) {
975#ifdef CONFIG_NET_IPGRE_BROADCAST
976 if (ipv4_is_multicast(iph->daddr)) {
977 if (!iph->saddr)
978 return -EINVAL;
979 dev->flags = IFF_BROADCAST;
980 dev->header_ops = &ipgre_header_ops;
981 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
982 dev->needed_headroom = 0;
983 }
984#endif
985 } else if (!tunnel->collect_md) {
986 dev->header_ops = &ipgre_header_ops;
987 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
988 dev->needed_headroom = 0;
989 }
990
991 return ip_tunnel_init(dev);
992}
993
994static const struct gre_protocol ipgre_protocol = {
995 .handler = gre_rcv,
996 .err_handler = gre_err,
997};
998
999static int __net_init ipgre_init_net(struct net *net)
1000{
1001 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1002}
1003
1004static void __net_exit ipgre_exit_batch_net(struct list_head *list_net)
1005{
1006 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops);
1007}
1008
1009static struct pernet_operations ipgre_net_ops = {
1010 .init = ipgre_init_net,
1011 .exit_batch = ipgre_exit_batch_net,
1012 .id = &ipgre_net_id,
1013 .size = sizeof(struct ip_tunnel_net),
1014};
1015
1016static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1017 struct netlink_ext_ack *extack)
1018{
1019 __be16 flags;
1020
1021 if (!data)
1022 return 0;
1023
1024 flags = 0;
1025 if (data[IFLA_GRE_IFLAGS])
1026 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1027 if (data[IFLA_GRE_OFLAGS])
1028 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1029 if (flags & (GRE_VERSION|GRE_ROUTING))
1030 return -EINVAL;
1031
1032 if (data[IFLA_GRE_COLLECT_METADATA] &&
1033 data[IFLA_GRE_ENCAP_TYPE] &&
1034 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1035 return -EINVAL;
1036
1037 return 0;
1038}
1039
1040static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1041 struct netlink_ext_ack *extack)
1042{
1043 __be32 daddr;
1044
1045 if (tb[IFLA_ADDRESS]) {
1046 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1047 return -EINVAL;
1048 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1049 return -EADDRNOTAVAIL;
1050 }
1051
1052 if (!data)
1053 goto out;
1054
1055 if (data[IFLA_GRE_REMOTE]) {
1056 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1057 if (!daddr)
1058 return -EINVAL;
1059 }
1060
1061out:
1062 return ipgre_tunnel_validate(tb, data, extack);
1063}
1064
1065static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1066 struct netlink_ext_ack *extack)
1067{
1068 __be16 flags = 0;
1069 int ret;
1070
1071 if (!data)
1072 return 0;
1073
1074 ret = ipgre_tap_validate(tb, data, extack);
1075 if (ret)
1076 return ret;
1077
1078 /* ERSPAN should only have GRE sequence and key flag */
1079 if (data[IFLA_GRE_OFLAGS])
1080 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1081 if (data[IFLA_GRE_IFLAGS])
1082 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1083 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1084 flags != (GRE_SEQ | GRE_KEY))
1085 return -EINVAL;
1086
1087 /* ERSPAN Session ID only has 10-bit. Since we reuse
1088 * 32-bit key field as ID, check it's range.
1089 */
1090 if (data[IFLA_GRE_IKEY] &&
1091 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1092 return -EINVAL;
1093
1094 if (data[IFLA_GRE_OKEY] &&
1095 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1096 return -EINVAL;
1097
1098 return 0;
1099}
1100
1101static int ipgre_netlink_parms(struct net_device *dev,
1102 struct nlattr *data[],
1103 struct nlattr *tb[],
1104 struct ip_tunnel_parm *parms,
1105 __u32 *fwmark)
1106{
1107 struct ip_tunnel *t = netdev_priv(dev);
1108
1109 memset(parms, 0, sizeof(*parms));
1110
1111 parms->iph.protocol = IPPROTO_GRE;
1112
1113 if (!data)
1114 return 0;
1115
1116 if (data[IFLA_GRE_LINK])
1117 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1118
1119 if (data[IFLA_GRE_IFLAGS])
1120 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
1121
1122 if (data[IFLA_GRE_OFLAGS])
1123 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
1124
1125 if (data[IFLA_GRE_IKEY])
1126 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1127
1128 if (data[IFLA_GRE_OKEY])
1129 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1130
1131 if (data[IFLA_GRE_LOCAL])
1132 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1133
1134 if (data[IFLA_GRE_REMOTE])
1135 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1136
1137 if (data[IFLA_GRE_TTL])
1138 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1139
1140 if (data[IFLA_GRE_TOS])
1141 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1142
1143 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1144 if (t->ignore_df)
1145 return -EINVAL;
1146 parms->iph.frag_off = htons(IP_DF);
1147 }
1148
1149 if (data[IFLA_GRE_COLLECT_METADATA]) {
1150 t->collect_md = true;
1151 if (dev->type == ARPHRD_IPGRE)
1152 dev->type = ARPHRD_NONE;
1153 }
1154
1155 if (data[IFLA_GRE_IGNORE_DF]) {
1156 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1157 && (parms->iph.frag_off & htons(IP_DF)))
1158 return -EINVAL;
1159 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1160 }
1161
1162 if (data[IFLA_GRE_FWMARK])
1163 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1164
1165 return 0;
1166}
1167
1168static int erspan_netlink_parms(struct net_device *dev,
1169 struct nlattr *data[],
1170 struct nlattr *tb[],
1171 struct ip_tunnel_parm *parms,
1172 __u32 *fwmark)
1173{
1174 struct ip_tunnel *t = netdev_priv(dev);
1175 int err;
1176
1177 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1178 if (err)
1179 return err;
1180 if (!data)
1181 return 0;
1182
1183 if (data[IFLA_GRE_ERSPAN_VER]) {
1184 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1185
1186 if (t->erspan_ver != 1 && t->erspan_ver != 2)
1187 return -EINVAL;
1188 }
1189
1190 if (t->erspan_ver == 1) {
1191 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1192 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1193 if (t->index & ~INDEX_MASK)
1194 return -EINVAL;
1195 }
1196 } else if (t->erspan_ver == 2) {
1197 if (data[IFLA_GRE_ERSPAN_DIR]) {
1198 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1199 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1200 return -EINVAL;
1201 }
1202 if (data[IFLA_GRE_ERSPAN_HWID]) {
1203 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1204 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1205 return -EINVAL;
1206 }
1207 }
1208
1209 return 0;
1210}
1211
1212/* This function returns true when ENCAP attributes are present in the nl msg */
1213static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1214 struct ip_tunnel_encap *ipencap)
1215{
1216 bool ret = false;
1217
1218 memset(ipencap, 0, sizeof(*ipencap));
1219
1220 if (!data)
1221 return ret;
1222
1223 if (data[IFLA_GRE_ENCAP_TYPE]) {
1224 ret = true;
1225 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1226 }
1227
1228 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1229 ret = true;
1230 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1231 }
1232
1233 if (data[IFLA_GRE_ENCAP_SPORT]) {
1234 ret = true;
1235 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1236 }
1237
1238 if (data[IFLA_GRE_ENCAP_DPORT]) {
1239 ret = true;
1240 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1241 }
1242
1243 return ret;
1244}
1245
1246static int gre_tap_init(struct net_device *dev)
1247{
1248 __gre_tunnel_init(dev);
1249 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1250 netif_keep_dst(dev);
1251
1252 return ip_tunnel_init(dev);
1253}
1254
1255static const struct net_device_ops gre_tap_netdev_ops = {
1256 .ndo_init = gre_tap_init,
1257 .ndo_uninit = ip_tunnel_uninit,
1258 .ndo_start_xmit = gre_tap_xmit,
1259 .ndo_set_mac_address = eth_mac_addr,
1260 .ndo_validate_addr = eth_validate_addr,
1261 .ndo_change_mtu = ip_tunnel_change_mtu,
1262 .ndo_get_stats64 = ip_tunnel_get_stats64,
1263 .ndo_get_iflink = ip_tunnel_get_iflink,
1264 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1265};
1266
1267static int erspan_tunnel_init(struct net_device *dev)
1268{
1269 struct ip_tunnel *tunnel = netdev_priv(dev);
1270
1271 tunnel->tun_hlen = 8;
1272 tunnel->parms.iph.protocol = IPPROTO_GRE;
1273 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1274 erspan_hdr_len(tunnel->erspan_ver);
1275
1276 dev->features |= GRE_FEATURES;
1277 dev->hw_features |= GRE_FEATURES;
1278 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1279 netif_keep_dst(dev);
1280
1281 return ip_tunnel_init(dev);
1282}
1283
1284static const struct net_device_ops erspan_netdev_ops = {
1285 .ndo_init = erspan_tunnel_init,
1286 .ndo_uninit = ip_tunnel_uninit,
1287 .ndo_start_xmit = erspan_xmit,
1288 .ndo_set_mac_address = eth_mac_addr,
1289 .ndo_validate_addr = eth_validate_addr,
1290 .ndo_change_mtu = ip_tunnel_change_mtu,
1291 .ndo_get_stats64 = ip_tunnel_get_stats64,
1292 .ndo_get_iflink = ip_tunnel_get_iflink,
1293 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1294};
1295
1296static void ipgre_tap_setup(struct net_device *dev)
1297{
1298 ether_setup(dev);
1299 dev->max_mtu = 0;
1300 dev->netdev_ops = &gre_tap_netdev_ops;
1301 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1302 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1303 ip_tunnel_setup(dev, gre_tap_net_id);
1304}
1305
1306static int
1307ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1308{
1309 struct ip_tunnel_encap ipencap;
1310
1311 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1312 struct ip_tunnel *t = netdev_priv(dev);
1313 int err = ip_tunnel_encap_setup(t, &ipencap);
1314
1315 if (err < 0)
1316 return err;
1317 }
1318
1319 return 0;
1320}
1321
1322static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1323 struct nlattr *tb[], struct nlattr *data[],
1324 struct netlink_ext_ack *extack)
1325{
1326 struct ip_tunnel_parm p;
1327 __u32 fwmark = 0;
1328 int err;
1329
1330 err = ipgre_newlink_encap_setup(dev, data);
1331 if (err)
1332 return err;
1333
1334 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1335 if (err < 0)
1336 return err;
1337 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1338}
1339
1340static int erspan_newlink(struct net *src_net, struct net_device *dev,
1341 struct nlattr *tb[], struct nlattr *data[],
1342 struct netlink_ext_ack *extack)
1343{
1344 struct ip_tunnel_parm p;
1345 __u32 fwmark = 0;
1346 int err;
1347
1348 err = ipgre_newlink_encap_setup(dev, data);
1349 if (err)
1350 return err;
1351
1352 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1353 if (err)
1354 return err;
1355 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1356}
1357
1358static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1359 struct nlattr *data[],
1360 struct netlink_ext_ack *extack)
1361{
1362 struct ip_tunnel *t = netdev_priv(dev);
1363 __u32 fwmark = t->fwmark;
1364 struct ip_tunnel_parm p;
1365 int err;
1366
1367 err = ipgre_newlink_encap_setup(dev, data);
1368 if (err)
1369 return err;
1370
1371 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1372 if (err < 0)
1373 return err;
1374
1375 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1376 if (err < 0)
1377 return err;
1378
1379 t->parms.i_flags = p.i_flags;
1380 t->parms.o_flags = p.o_flags;
1381
1382 ipgre_link_update(dev, !tb[IFLA_MTU]);
1383
1384 return 0;
1385}
1386
1387static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1388 struct nlattr *data[],
1389 struct netlink_ext_ack *extack)
1390{
1391 struct ip_tunnel *t = netdev_priv(dev);
1392 __u32 fwmark = t->fwmark;
1393 struct ip_tunnel_parm p;
1394 int err;
1395
1396 err = ipgre_newlink_encap_setup(dev, data);
1397 if (err)
1398 return err;
1399
1400 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1401 if (err < 0)
1402 return err;
1403
1404 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1405 if (err < 0)
1406 return err;
1407
1408 t->parms.i_flags = p.i_flags;
1409 t->parms.o_flags = p.o_flags;
1410
1411 return 0;
1412}
1413
1414static size_t ipgre_get_size(const struct net_device *dev)
1415{
1416 return
1417 /* IFLA_GRE_LINK */
1418 nla_total_size(4) +
1419 /* IFLA_GRE_IFLAGS */
1420 nla_total_size(2) +
1421 /* IFLA_GRE_OFLAGS */
1422 nla_total_size(2) +
1423 /* IFLA_GRE_IKEY */
1424 nla_total_size(4) +
1425 /* IFLA_GRE_OKEY */
1426 nla_total_size(4) +
1427 /* IFLA_GRE_LOCAL */
1428 nla_total_size(4) +
1429 /* IFLA_GRE_REMOTE */
1430 nla_total_size(4) +
1431 /* IFLA_GRE_TTL */
1432 nla_total_size(1) +
1433 /* IFLA_GRE_TOS */
1434 nla_total_size(1) +
1435 /* IFLA_GRE_PMTUDISC */
1436 nla_total_size(1) +
1437 /* IFLA_GRE_ENCAP_TYPE */
1438 nla_total_size(2) +
1439 /* IFLA_GRE_ENCAP_FLAGS */
1440 nla_total_size(2) +
1441 /* IFLA_GRE_ENCAP_SPORT */
1442 nla_total_size(2) +
1443 /* IFLA_GRE_ENCAP_DPORT */
1444 nla_total_size(2) +
1445 /* IFLA_GRE_COLLECT_METADATA */
1446 nla_total_size(0) +
1447 /* IFLA_GRE_IGNORE_DF */
1448 nla_total_size(1) +
1449 /* IFLA_GRE_FWMARK */
1450 nla_total_size(4) +
1451 /* IFLA_GRE_ERSPAN_INDEX */
1452 nla_total_size(4) +
1453 /* IFLA_GRE_ERSPAN_VER */
1454 nla_total_size(1) +
1455 /* IFLA_GRE_ERSPAN_DIR */
1456 nla_total_size(1) +
1457 /* IFLA_GRE_ERSPAN_HWID */
1458 nla_total_size(2) +
1459 0;
1460}
1461
1462static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1463{
1464 struct ip_tunnel *t = netdev_priv(dev);
1465 struct ip_tunnel_parm *p = &t->parms;
1466 __be16 o_flags = p->o_flags;
1467
1468 if (t->erspan_ver == 1 || t->erspan_ver == 2) {
1469 if (!t->collect_md)
1470 o_flags |= TUNNEL_KEY;
1471
1472 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1473 goto nla_put_failure;
1474
1475 if (t->erspan_ver == 1) {
1476 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1477 goto nla_put_failure;
1478 } else {
1479 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1480 goto nla_put_failure;
1481 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1482 goto nla_put_failure;
1483 }
1484 }
1485
1486 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1487 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1488 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1489 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1490 gre_tnl_flags_to_gre_flags(o_flags)) ||
1491 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1492 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1493 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1494 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1495 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1496 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1497 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1498 !!(p->iph.frag_off & htons(IP_DF))) ||
1499 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1500 goto nla_put_failure;
1501
1502 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1503 t->encap.type) ||
1504 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1505 t->encap.sport) ||
1506 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1507 t->encap.dport) ||
1508 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1509 t->encap.flags))
1510 goto nla_put_failure;
1511
1512 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1513 goto nla_put_failure;
1514
1515 if (t->collect_md) {
1516 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1517 goto nla_put_failure;
1518 }
1519
1520 return 0;
1521
1522nla_put_failure:
1523 return -EMSGSIZE;
1524}
1525
1526static void erspan_setup(struct net_device *dev)
1527{
1528 struct ip_tunnel *t = netdev_priv(dev);
1529
1530 ether_setup(dev);
1531 dev->max_mtu = 0;
1532 dev->netdev_ops = &erspan_netdev_ops;
1533 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1534 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1535 ip_tunnel_setup(dev, erspan_net_id);
1536 t->erspan_ver = 1;
1537}
1538
1539static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1540 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1541 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1542 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1543 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1544 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1545 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1546 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1547 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1548 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1549 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1550 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1551 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1552 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1553 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1554 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1555 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1556 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1557 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1558 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1559 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1560 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1561};
1562
1563static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1564 .kind = "gre",
1565 .maxtype = IFLA_GRE_MAX,
1566 .policy = ipgre_policy,
1567 .priv_size = sizeof(struct ip_tunnel),
1568 .setup = ipgre_tunnel_setup,
1569 .validate = ipgre_tunnel_validate,
1570 .newlink = ipgre_newlink,
1571 .changelink = ipgre_changelink,
1572 .dellink = ip_tunnel_dellink,
1573 .get_size = ipgre_get_size,
1574 .fill_info = ipgre_fill_info,
1575 .get_link_net = ip_tunnel_get_link_net,
1576};
1577
1578static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1579 .kind = "gretap",
1580 .maxtype = IFLA_GRE_MAX,
1581 .policy = ipgre_policy,
1582 .priv_size = sizeof(struct ip_tunnel),
1583 .setup = ipgre_tap_setup,
1584 .validate = ipgre_tap_validate,
1585 .newlink = ipgre_newlink,
1586 .changelink = ipgre_changelink,
1587 .dellink = ip_tunnel_dellink,
1588 .get_size = ipgre_get_size,
1589 .fill_info = ipgre_fill_info,
1590 .get_link_net = ip_tunnel_get_link_net,
1591};
1592
1593static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1594 .kind = "erspan",
1595 .maxtype = IFLA_GRE_MAX,
1596 .policy = ipgre_policy,
1597 .priv_size = sizeof(struct ip_tunnel),
1598 .setup = erspan_setup,
1599 .validate = erspan_validate,
1600 .newlink = erspan_newlink,
1601 .changelink = erspan_changelink,
1602 .dellink = ip_tunnel_dellink,
1603 .get_size = ipgre_get_size,
1604 .fill_info = ipgre_fill_info,
1605 .get_link_net = ip_tunnel_get_link_net,
1606};
1607
1608struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1609 u8 name_assign_type)
1610{
1611 struct nlattr *tb[IFLA_MAX + 1];
1612 struct net_device *dev;
1613 LIST_HEAD(list_kill);
1614 struct ip_tunnel *t;
1615 int err;
1616
1617 memset(&tb, 0, sizeof(tb));
1618
1619 dev = rtnl_create_link(net, name, name_assign_type,
1620 &ipgre_tap_ops, tb, NULL);
1621 if (IS_ERR(dev))
1622 return dev;
1623
1624 /* Configure flow based GRE device. */
1625 t = netdev_priv(dev);
1626 t->collect_md = true;
1627
1628 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1629 if (err < 0) {
1630 free_netdev(dev);
1631 return ERR_PTR(err);
1632 }
1633
1634 /* openvswitch users expect packet sizes to be unrestricted,
1635 * so set the largest MTU we can.
1636 */
1637 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1638 if (err)
1639 goto out;
1640
1641 err = rtnl_configure_link(dev, NULL);
1642 if (err < 0)
1643 goto out;
1644
1645 return dev;
1646out:
1647 ip_tunnel_dellink(dev, &list_kill);
1648 unregister_netdevice_many(&list_kill);
1649 return ERR_PTR(err);
1650}
1651EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1652
1653static int __net_init ipgre_tap_init_net(struct net *net)
1654{
1655 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1656}
1657
1658static void __net_exit ipgre_tap_exit_batch_net(struct list_head *list_net)
1659{
1660 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops);
1661}
1662
1663static struct pernet_operations ipgre_tap_net_ops = {
1664 .init = ipgre_tap_init_net,
1665 .exit_batch = ipgre_tap_exit_batch_net,
1666 .id = &gre_tap_net_id,
1667 .size = sizeof(struct ip_tunnel_net),
1668};
1669
1670static int __net_init erspan_init_net(struct net *net)
1671{
1672 return ip_tunnel_init_net(net, erspan_net_id,
1673 &erspan_link_ops, "erspan0");
1674}
1675
1676static void __net_exit erspan_exit_batch_net(struct list_head *net_list)
1677{
1678 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops);
1679}
1680
1681static struct pernet_operations erspan_net_ops = {
1682 .init = erspan_init_net,
1683 .exit_batch = erspan_exit_batch_net,
1684 .id = &erspan_net_id,
1685 .size = sizeof(struct ip_tunnel_net),
1686};
1687
1688static int __init ipgre_init(void)
1689{
1690 int err;
1691
1692 pr_info("GRE over IPv4 tunneling driver\n");
1693
1694 err = register_pernet_device(&ipgre_net_ops);
1695 if (err < 0)
1696 return err;
1697
1698 err = register_pernet_device(&ipgre_tap_net_ops);
1699 if (err < 0)
1700 goto pnet_tap_failed;
1701
1702 err = register_pernet_device(&erspan_net_ops);
1703 if (err < 0)
1704 goto pnet_erspan_failed;
1705
1706 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1707 if (err < 0) {
1708 pr_info("%s: can't add protocol\n", __func__);
1709 goto add_proto_failed;
1710 }
1711
1712 err = rtnl_link_register(&ipgre_link_ops);
1713 if (err < 0)
1714 goto rtnl_link_failed;
1715
1716 err = rtnl_link_register(&ipgre_tap_ops);
1717 if (err < 0)
1718 goto tap_ops_failed;
1719
1720 err = rtnl_link_register(&erspan_link_ops);
1721 if (err < 0)
1722 goto erspan_link_failed;
1723
1724 return 0;
1725
1726erspan_link_failed:
1727 rtnl_link_unregister(&ipgre_tap_ops);
1728tap_ops_failed:
1729 rtnl_link_unregister(&ipgre_link_ops);
1730rtnl_link_failed:
1731 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1732add_proto_failed:
1733 unregister_pernet_device(&erspan_net_ops);
1734pnet_erspan_failed:
1735 unregister_pernet_device(&ipgre_tap_net_ops);
1736pnet_tap_failed:
1737 unregister_pernet_device(&ipgre_net_ops);
1738 return err;
1739}
1740
1741static void __exit ipgre_fini(void)
1742{
1743 rtnl_link_unregister(&ipgre_tap_ops);
1744 rtnl_link_unregister(&ipgre_link_ops);
1745 rtnl_link_unregister(&erspan_link_ops);
1746 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1747 unregister_pernet_device(&ipgre_tap_net_ops);
1748 unregister_pernet_device(&ipgre_net_ops);
1749 unregister_pernet_device(&erspan_net_ops);
1750}
1751
1752module_init(ipgre_init);
1753module_exit(ipgre_fini);
1754MODULE_LICENSE("GPL");
1755MODULE_ALIAS_RTNL_LINK("gre");
1756MODULE_ALIAS_RTNL_LINK("gretap");
1757MODULE_ALIAS_RTNL_LINK("erspan");
1758MODULE_ALIAS_NETDEV("gre0");
1759MODULE_ALIAS_NETDEV("gretap0");
1760MODULE_ALIAS_NETDEV("erspan0");
Ubuntu Focal Linux kernel mirror