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 #include <net/inet_dscp.h>
48
49 /*
50 Problems & solutions
51 --------------------
52
53 1. The most important issue is detecting local dead loops.
54 They would cause complete host lockup in transmit, which
55 would be "resolved" by stack overflow or, if queueing is enabled,
56 with infinite looping in net_bh.
57
58 We cannot track such dead loops during route installation,
59 it is infeasible task. The most general solutions would be
60 to keep skb->encapsulation counter (sort of local ttl),
61 and silently drop packet when it expires. It is a good
62 solution, but it supposes maintaining new variable in ALL
63 skb, even if no tunneling is used.
64
65 Current solution: xmit_recursion breaks dead loops. This is a percpu
66 counter, since when we enter the first ndo_xmit(), cpu migration is
67 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
68
69 2. Networking dead loops would not kill routers, but would really
70 kill network. IP hop limit plays role of "t->recursion" in this case,
71 if we copy it from packet being encapsulated to upper header.
72 It is very good solution, but it introduces two problems:
73
74 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
75 do not work over tunnels.
76 - traceroute does not work. I planned to relay ICMP from tunnel,
77 so that this problem would be solved and traceroute output
78 would even more informative. This idea appeared to be wrong:
79 only Linux complies to rfc1812 now (yes, guys, Linux is the only
80 true router now :-)), all routers (at least, in neighbourhood of mine)
81 return only 8 bytes of payload. It is the end.
82
83 Hence, if we want that OSPF worked or traceroute said something reasonable,
84 we should search for another solution.
85
86 One of them is to parse packet trying to detect inner encapsulation
87 made by our node. It is difficult or even impossible, especially,
88 taking into account fragmentation. TO be short, ttl is not solution at all.
89
90 Current solution: The solution was UNEXPECTEDLY SIMPLE.
91 We force DF flag on tunnels with preconfigured hop limit,
92 that is ALL. :-) Well, it does not remove the problem completely,
93 but exponential growth of network traffic is changed to linear
94 (branches, that exceed pmtu are pruned) and tunnel mtu
95 rapidly degrades to value <68, where looping stops.
96 Yes, it is not good if there exists a router in the loop,
97 which does not force DF, even when encapsulating packets have DF set.
98 But it is not our problem! Nobody could accuse us, we made
99 all that we could make. Even if it is your gated who injected
100 fatal route to network, even if it were you who configured
101 fatal static route: you are innocent. :-)
102
103 Alexey Kuznetsov.
104 */
105
106 static bool log_ecn_error = true;
107 module_param(log_ecn_error, bool, 0644);
108 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
109
110 static struct rtnl_link_ops ipgre_link_ops __read_mostly;
111 static const struct header_ops ipgre_header_ops;
112
113 static int ipgre_tunnel_init(struct net_device *dev);
114 static void erspan_build_header(struct sk_buff *skb,
115 u32 id, u32 index,
116 bool truncate, bool is_ipv4);
117
118 static unsigned int ipgre_net_id __read_mostly;
119 static unsigned int gre_tap_net_id __read_mostly;
120 static unsigned int erspan_net_id __read_mostly;
121
ipgre_err(struct sk_buff * skb,u32 info,const struct tnl_ptk_info * tpi)122 static int ipgre_err(struct sk_buff *skb, u32 info,
123 const struct tnl_ptk_info *tpi)
124 {
125
126 /* All the routers (except for Linux) return only
127 8 bytes of packet payload. It means, that precise relaying of
128 ICMP in the real Internet is absolutely infeasible.
129
130 Moreover, Cisco "wise men" put GRE key to the third word
131 in GRE header. It makes impossible maintaining even soft
132 state for keyed GRE tunnels with enabled checksum. Tell
133 them "thank you".
134
135 Well, I wonder, rfc1812 was written by Cisco employee,
136 what the hell these idiots break standards established
137 by themselves???
138 */
139 struct net *net = dev_net(skb->dev);
140 struct ip_tunnel_net *itn;
141 const struct iphdr *iph;
142 const int type = icmp_hdr(skb)->type;
143 const int code = icmp_hdr(skb)->code;
144 struct ip_tunnel *t;
145
146 if (tpi->proto == htons(ETH_P_TEB))
147 itn = net_generic(net, gre_tap_net_id);
148 else if (tpi->proto == htons(ETH_P_ERSPAN) ||
149 tpi->proto == htons(ETH_P_ERSPAN2))
150 itn = net_generic(net, erspan_net_id);
151 else
152 itn = net_generic(net, ipgre_net_id);
153
154 iph = (const struct iphdr *)(icmp_hdr(skb) + 1);
155 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
156 iph->daddr, iph->saddr, tpi->key);
157
158 if (!t)
159 return -ENOENT;
160
161 switch (type) {
162 default:
163 case ICMP_PARAMETERPROB:
164 return 0;
165
166 case ICMP_DEST_UNREACH:
167 switch (code) {
168 case ICMP_SR_FAILED:
169 case ICMP_PORT_UNREACH:
170 /* Impossible event. */
171 return 0;
172 default:
173 /* All others are translated to HOST_UNREACH.
174 rfc2003 contains "deep thoughts" about NET_UNREACH,
175 I believe they are just ether pollution. --ANK
176 */
177 break;
178 }
179 break;
180
181 case ICMP_TIME_EXCEEDED:
182 if (code != ICMP_EXC_TTL)
183 return 0;
184 break;
185
186 case ICMP_REDIRECT:
187 break;
188 }
189
190 #if IS_ENABLED(CONFIG_IPV6)
191 if (tpi->proto == htons(ETH_P_IPV6)) {
192 unsigned int data_len = 0;
193
194 if (type == ICMP_TIME_EXCEEDED)
195 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
196
197 if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4 + tpi->hdr_len,
198 type, data_len))
199 return 0;
200 }
201 #endif
202
203 if (t->parms.iph.daddr == 0 ||
204 ipv4_is_multicast(t->parms.iph.daddr))
205 return 0;
206
207 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
208 return 0;
209
210 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
211 t->err_count++;
212 else
213 t->err_count = 1;
214 t->err_time = jiffies;
215
216 return 0;
217 }
218
gre_err(struct sk_buff * skb,u32 info)219 static void gre_err(struct sk_buff *skb, u32 info)
220 {
221 /* All the routers (except for Linux) return only
222 * 8 bytes of packet payload. It means, that precise relaying of
223 * ICMP in the real Internet is absolutely infeasible.
224 *
225 * Moreover, Cisco "wise men" put GRE key to the third word
226 * in GRE header. It makes impossible maintaining even soft
227 * state for keyed
228 * GRE tunnels with enabled checksum. Tell them "thank you".
229 *
230 * Well, I wonder, rfc1812 was written by Cisco employee,
231 * what the hell these idiots break standards established
232 * by themselves???
233 */
234
235 const struct iphdr *iph = (struct iphdr *)skb->data;
236 const int type = icmp_hdr(skb)->type;
237 const int code = icmp_hdr(skb)->code;
238 struct tnl_ptk_info tpi;
239
240 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IP),
241 iph->ihl * 4) < 0)
242 return;
243
244 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
245 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
246 skb->dev->ifindex, IPPROTO_GRE);
247 return;
248 }
249 if (type == ICMP_REDIRECT) {
250 ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex,
251 IPPROTO_GRE);
252 return;
253 }
254
255 ipgre_err(skb, info, &tpi);
256 }
257
is_erspan_type1(int gre_hdr_len)258 static bool is_erspan_type1(int gre_hdr_len)
259 {
260 /* Both ERSPAN type I (version 0) and type II (version 1) use
261 * protocol 0x88BE, but the type I has only 4-byte GRE header,
262 * while type II has 8-byte.
263 */
264 return gre_hdr_len == 4;
265 }
266
erspan_rcv(struct sk_buff * skb,struct tnl_ptk_info * tpi,int gre_hdr_len)267 static int erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
268 int gre_hdr_len)
269 {
270 struct net *net = dev_net(skb->dev);
271 struct metadata_dst *tun_dst = NULL;
272 struct erspan_base_hdr *ershdr;
273 IP_TUNNEL_DECLARE_FLAGS(flags);
274 struct ip_tunnel_net *itn;
275 struct ip_tunnel *tunnel;
276 const struct iphdr *iph;
277 struct erspan_md2 *md2;
278 int ver;
279 int len;
280
281 ip_tunnel_flags_copy(flags, tpi->flags);
282
283 itn = net_generic(net, erspan_net_id);
284 iph = ip_hdr(skb);
285 if (is_erspan_type1(gre_hdr_len)) {
286 ver = 0;
287 __set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
288 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
289 iph->saddr, iph->daddr, 0);
290 } else {
291 if (unlikely(!pskb_may_pull(skb,
292 gre_hdr_len + sizeof(*ershdr))))
293 return PACKET_REJECT;
294
295 ershdr = (struct erspan_base_hdr *)(skb->data + gre_hdr_len);
296 ver = ershdr->ver;
297 iph = ip_hdr(skb);
298 __set_bit(IP_TUNNEL_KEY_BIT, flags);
299 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
300 iph->saddr, iph->daddr, tpi->key);
301 }
302
303 if (tunnel) {
304 if (is_erspan_type1(gre_hdr_len))
305 len = gre_hdr_len;
306 else
307 len = gre_hdr_len + erspan_hdr_len(ver);
308
309 if (unlikely(!pskb_may_pull(skb, len)))
310 return PACKET_REJECT;
311
312 if (__iptunnel_pull_header(skb,
313 len,
314 htons(ETH_P_TEB),
315 false, false) < 0)
316 goto drop;
317
318 if (tunnel->collect_md) {
319 struct erspan_metadata *pkt_md, *md;
320 struct ip_tunnel_info *info;
321 unsigned char *gh;
322 __be64 tun_id;
323
324 __set_bit(IP_TUNNEL_KEY_BIT, tpi->flags);
325 ip_tunnel_flags_copy(flags, tpi->flags);
326 tun_id = key32_to_tunnel_id(tpi->key);
327
328 tun_dst = ip_tun_rx_dst(skb, flags,
329 tun_id, sizeof(*md));
330 if (!tun_dst)
331 return PACKET_REJECT;
332
333 /* skb can be uncloned in __iptunnel_pull_header, so
334 * old pkt_md is no longer valid and we need to reset
335 * it
336 */
337 gh = skb_network_header(skb) +
338 skb_network_header_len(skb);
339 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
340 sizeof(*ershdr));
341 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
342 md->version = ver;
343 md2 = &md->u.md2;
344 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
345 ERSPAN_V2_MDSIZE);
346
347 info = &tun_dst->u.tun_info;
348 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
349 info->key.tun_flags);
350 info->options_len = sizeof(*md);
351 }
352
353 skb_reset_mac_header(skb);
354 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
355 return PACKET_RCVD;
356 }
357 return PACKET_REJECT;
358
359 drop:
360 kfree_skb(skb);
361 return PACKET_RCVD;
362 }
363
__ipgre_rcv(struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct ip_tunnel_net * itn,int hdr_len,bool raw_proto)364 static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
365 struct ip_tunnel_net *itn, int hdr_len, bool raw_proto)
366 {
367 struct metadata_dst *tun_dst = NULL;
368 const struct iphdr *iph;
369 struct ip_tunnel *tunnel;
370
371 iph = ip_hdr(skb);
372 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
373 iph->saddr, iph->daddr, tpi->key);
374
375 if (tunnel) {
376 const struct iphdr *tnl_params;
377
378 if (__iptunnel_pull_header(skb, hdr_len, tpi->proto,
379 raw_proto, false) < 0)
380 goto drop;
381
382 /* Special case for ipgre_header_parse(), which expects the
383 * mac_header to point to the outer IP header.
384 */
385 if (tunnel->dev->header_ops == &ipgre_header_ops)
386 skb_pop_mac_header(skb);
387 else
388 skb_reset_mac_header(skb);
389
390 tnl_params = &tunnel->parms.iph;
391 if (tunnel->collect_md || tnl_params->daddr == 0) {
392 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
393 __be64 tun_id;
394
395 __set_bit(IP_TUNNEL_CSUM_BIT, flags);
396 __set_bit(IP_TUNNEL_KEY_BIT, flags);
397 ip_tunnel_flags_and(flags, tpi->flags, flags);
398
399 tun_id = key32_to_tunnel_id(tpi->key);
400 tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0);
401 if (!tun_dst)
402 return PACKET_REJECT;
403 }
404
405 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
406 return PACKET_RCVD;
407 }
408 return PACKET_NEXT;
409
410 drop:
411 kfree_skb(skb);
412 return PACKET_RCVD;
413 }
414
ipgre_rcv(struct sk_buff * skb,const struct tnl_ptk_info * tpi,int hdr_len)415 static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
416 int hdr_len)
417 {
418 struct net *net = dev_net(skb->dev);
419 struct ip_tunnel_net *itn;
420 int res;
421
422 if (tpi->proto == htons(ETH_P_TEB))
423 itn = net_generic(net, gre_tap_net_id);
424 else
425 itn = net_generic(net, ipgre_net_id);
426
427 res = __ipgre_rcv(skb, tpi, itn, hdr_len, false);
428 if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) {
429 /* ipgre tunnels in collect metadata mode should receive
430 * also ETH_P_TEB traffic.
431 */
432 itn = net_generic(net, ipgre_net_id);
433 res = __ipgre_rcv(skb, tpi, itn, hdr_len, true);
434 }
435 return res;
436 }
437
gre_rcv(struct sk_buff * skb)438 static int gre_rcv(struct sk_buff *skb)
439 {
440 struct tnl_ptk_info tpi;
441 bool csum_err = false;
442 int hdr_len;
443
444 #ifdef CONFIG_NET_IPGRE_BROADCAST
445 if (ipv4_is_multicast(ip_hdr(skb)->daddr)) {
446 /* Looped back packet, drop it! */
447 if (rt_is_output_route(skb_rtable(skb)))
448 goto drop;
449 }
450 #endif
451
452 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0);
453 if (hdr_len < 0)
454 goto drop;
455
456 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
457 tpi.proto == htons(ETH_P_ERSPAN2))) {
458 if (erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
459 return 0;
460 goto out;
461 }
462
463 if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
464 return 0;
465
466 out:
467 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
468 drop:
469 kfree_skb(skb);
470 return 0;
471 }
472
__gre_xmit(struct sk_buff * skb,struct net_device * dev,const struct iphdr * tnl_params,__be16 proto)473 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
474 const struct iphdr *tnl_params,
475 __be16 proto)
476 {
477 struct ip_tunnel *tunnel = netdev_priv(dev);
478 IP_TUNNEL_DECLARE_FLAGS(flags);
479
480 ip_tunnel_flags_copy(flags, tunnel->parms.o_flags);
481
482 /* Push GRE header. */
483 gre_build_header(skb, tunnel->tun_hlen,
484 flags, proto, tunnel->parms.o_key,
485 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
486 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
487
488 ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
489 }
490
gre_handle_offloads(struct sk_buff * skb,bool csum)491 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
492 {
493 return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
494 }
495
gre_fb_xmit(struct sk_buff * skb,struct net_device * dev,__be16 proto)496 static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
497 __be16 proto)
498 {
499 struct ip_tunnel *tunnel = netdev_priv(dev);
500 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
501 struct ip_tunnel_info *tun_info;
502 const struct ip_tunnel_key *key;
503 int tunnel_hlen;
504
505 tun_info = skb_tunnel_info(skb);
506 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
507 ip_tunnel_info_af(tun_info) != AF_INET))
508 goto err_free_skb;
509
510 key = &tun_info->key;
511 tunnel_hlen = gre_calc_hlen(key->tun_flags);
512
513 if (skb_cow_head(skb, dev->needed_headroom))
514 goto err_free_skb;
515
516 /* Push Tunnel header. */
517 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
518 tunnel->parms.o_flags)))
519 goto err_free_skb;
520
521 __set_bit(IP_TUNNEL_CSUM_BIT, flags);
522 __set_bit(IP_TUNNEL_KEY_BIT, flags);
523 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
524 ip_tunnel_flags_and(flags, tun_info->key.tun_flags, flags);
525
526 gre_build_header(skb, tunnel_hlen, flags, proto,
527 tunnel_id_to_key32(tun_info->key.tun_id),
528 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
529 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
530
531 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
532
533 return;
534
535 err_free_skb:
536 kfree_skb(skb);
537 DEV_STATS_INC(dev, tx_dropped);
538 }
539
erspan_fb_xmit(struct sk_buff * skb,struct net_device * dev)540 static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
541 {
542 struct ip_tunnel *tunnel = netdev_priv(dev);
543 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
544 struct ip_tunnel_info *tun_info;
545 const struct ip_tunnel_key *key;
546 struct erspan_metadata *md;
547 bool truncate = false;
548 __be16 proto;
549 int tunnel_hlen;
550 int version;
551 int nhoff;
552
553 tun_info = skb_tunnel_info(skb);
554 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
555 ip_tunnel_info_af(tun_info) != AF_INET))
556 goto err_free_skb;
557
558 key = &tun_info->key;
559 if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
560 goto err_free_skb;
561 if (tun_info->options_len < sizeof(*md))
562 goto err_free_skb;
563 md = ip_tunnel_info_opts(tun_info);
564
565 /* ERSPAN has fixed 8 byte GRE header */
566 version = md->version;
567 tunnel_hlen = 8 + erspan_hdr_len(version);
568
569 if (skb_cow_head(skb, dev->needed_headroom))
570 goto err_free_skb;
571
572 if (gre_handle_offloads(skb, false))
573 goto err_free_skb;
574
575 if (skb->len > dev->mtu + dev->hard_header_len) {
576 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
577 goto err_free_skb;
578 truncate = true;
579 }
580
581 nhoff = skb_network_offset(skb);
582 if (skb->protocol == htons(ETH_P_IP) &&
583 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
584 truncate = true;
585
586 if (skb->protocol == htons(ETH_P_IPV6)) {
587 int thoff;
588
589 if (skb_transport_header_was_set(skb))
590 thoff = skb_transport_offset(skb);
591 else
592 thoff = nhoff + sizeof(struct ipv6hdr);
593 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
594 truncate = true;
595 }
596
597 if (version == 1) {
598 erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
599 ntohl(md->u.index), truncate, true);
600 proto = htons(ETH_P_ERSPAN);
601 } else if (version == 2) {
602 erspan_build_header_v2(skb,
603 ntohl(tunnel_id_to_key32(key->tun_id)),
604 md->u.md2.dir,
605 get_hwid(&md->u.md2),
606 truncate, true);
607 proto = htons(ETH_P_ERSPAN2);
608 } else {
609 goto err_free_skb;
610 }
611
612 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
613 gre_build_header(skb, 8, flags, proto, 0,
614 htonl(atomic_fetch_inc(&tunnel->o_seqno)));
615
616 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
617
618 return;
619
620 err_free_skb:
621 kfree_skb(skb);
622 DEV_STATS_INC(dev, tx_dropped);
623 }
624
gre_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)625 static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
626 {
627 struct ip_tunnel_info *info = skb_tunnel_info(skb);
628 const struct ip_tunnel_key *key;
629 struct rtable *rt;
630 struct flowi4 fl4;
631
632 if (ip_tunnel_info_af(info) != AF_INET)
633 return -EINVAL;
634
635 key = &info->key;
636 ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
637 tunnel_id_to_key32(key->tun_id),
638 key->tos & ~INET_ECN_MASK, dev_net(dev), 0,
639 skb->mark, skb_get_hash(skb), key->flow_flags);
640 rt = ip_route_output_key(dev_net(dev), &fl4);
641 if (IS_ERR(rt))
642 return PTR_ERR(rt);
643
644 ip_rt_put(rt);
645 info->key.u.ipv4.src = fl4.saddr;
646 return 0;
647 }
648
ipgre_xmit(struct sk_buff * skb,struct net_device * dev)649 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
650 struct net_device *dev)
651 {
652 struct ip_tunnel *tunnel = netdev_priv(dev);
653 const struct iphdr *tnl_params;
654
655 if (!pskb_inet_may_pull(skb))
656 goto free_skb;
657
658 if (tunnel->collect_md) {
659 gre_fb_xmit(skb, dev, skb->protocol);
660 return NETDEV_TX_OK;
661 }
662
663 if (dev->header_ops) {
664 int pull_len = tunnel->hlen + sizeof(struct iphdr);
665
666 if (skb_cow_head(skb, 0))
667 goto free_skb;
668
669 if (!pskb_may_pull(skb, pull_len))
670 goto free_skb;
671
672 tnl_params = (const struct iphdr *)skb->data;
673
674 /* ip_tunnel_xmit() needs skb->data pointing to gre header. */
675 skb_pull(skb, pull_len);
676 skb_reset_mac_header(skb);
677
678 if (skb->ip_summed == CHECKSUM_PARTIAL &&
679 skb_checksum_start(skb) < skb->data)
680 goto free_skb;
681 } else {
682 if (skb_cow_head(skb, dev->needed_headroom))
683 goto free_skb;
684
685 tnl_params = &tunnel->parms.iph;
686 }
687
688 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
689 tunnel->parms.o_flags)))
690 goto free_skb;
691
692 __gre_xmit(skb, dev, tnl_params, skb->protocol);
693 return NETDEV_TX_OK;
694
695 free_skb:
696 kfree_skb(skb);
697 DEV_STATS_INC(dev, tx_dropped);
698 return NETDEV_TX_OK;
699 }
700
erspan_xmit(struct sk_buff * skb,struct net_device * dev)701 static netdev_tx_t erspan_xmit(struct sk_buff *skb,
702 struct net_device *dev)
703 {
704 struct ip_tunnel *tunnel = netdev_priv(dev);
705 bool truncate = false;
706 __be16 proto;
707
708 if (!pskb_inet_may_pull(skb))
709 goto free_skb;
710
711 if (tunnel->collect_md) {
712 erspan_fb_xmit(skb, dev);
713 return NETDEV_TX_OK;
714 }
715
716 if (gre_handle_offloads(skb, false))
717 goto free_skb;
718
719 if (skb_cow_head(skb, dev->needed_headroom))
720 goto free_skb;
721
722 if (skb->len > dev->mtu + dev->hard_header_len) {
723 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
724 goto free_skb;
725 truncate = true;
726 }
727
728 /* Push ERSPAN header */
729 if (tunnel->erspan_ver == 0) {
730 proto = htons(ETH_P_ERSPAN);
731 __clear_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags);
732 } else if (tunnel->erspan_ver == 1) {
733 erspan_build_header(skb, ntohl(tunnel->parms.o_key),
734 tunnel->index,
735 truncate, true);
736 proto = htons(ETH_P_ERSPAN);
737 } else if (tunnel->erspan_ver == 2) {
738 erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
739 tunnel->dir, tunnel->hwid,
740 truncate, true);
741 proto = htons(ETH_P_ERSPAN2);
742 } else {
743 goto free_skb;
744 }
745
746 __clear_bit(IP_TUNNEL_KEY_BIT, tunnel->parms.o_flags);
747 __gre_xmit(skb, dev, &tunnel->parms.iph, proto);
748 return NETDEV_TX_OK;
749
750 free_skb:
751 kfree_skb(skb);
752 DEV_STATS_INC(dev, tx_dropped);
753 return NETDEV_TX_OK;
754 }
755
gre_tap_xmit(struct sk_buff * skb,struct net_device * dev)756 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
757 struct net_device *dev)
758 {
759 struct ip_tunnel *tunnel = netdev_priv(dev);
760
761 if (!pskb_inet_may_pull(skb))
762 goto free_skb;
763
764 if (tunnel->collect_md) {
765 gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
766 return NETDEV_TX_OK;
767 }
768
769 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
770 tunnel->parms.o_flags)))
771 goto free_skb;
772
773 if (skb_cow_head(skb, dev->needed_headroom))
774 goto free_skb;
775
776 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
777 return NETDEV_TX_OK;
778
779 free_skb:
780 kfree_skb(skb);
781 DEV_STATS_INC(dev, tx_dropped);
782 return NETDEV_TX_OK;
783 }
784
ipgre_link_update(struct net_device * dev,bool set_mtu)785 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
786 {
787 struct ip_tunnel *tunnel = netdev_priv(dev);
788 int len;
789
790 len = tunnel->tun_hlen;
791 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
792 len = tunnel->tun_hlen - len;
793 tunnel->hlen = tunnel->hlen + len;
794
795 if (dev->header_ops)
796 dev->hard_header_len += len;
797 else
798 dev->needed_headroom += len;
799
800 if (set_mtu)
801 WRITE_ONCE(dev->mtu, max_t(int, dev->mtu - len, 68));
802
803 if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags) ||
804 (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
805 tunnel->encap.type != TUNNEL_ENCAP_NONE)) {
806 dev->features &= ~NETIF_F_GSO_SOFTWARE;
807 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
808 } else {
809 dev->features |= NETIF_F_GSO_SOFTWARE;
810 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
811 }
812 }
813
ipgre_tunnel_ctl(struct net_device * dev,struct ip_tunnel_parm_kern * p,int cmd)814 static int ipgre_tunnel_ctl(struct net_device *dev,
815 struct ip_tunnel_parm_kern *p,
816 int cmd)
817 {
818 __be16 i_flags, o_flags;
819 int err;
820
821 if (!ip_tunnel_flags_is_be16_compat(p->i_flags) ||
822 !ip_tunnel_flags_is_be16_compat(p->o_flags))
823 return -EOVERFLOW;
824
825 i_flags = ip_tunnel_flags_to_be16(p->i_flags);
826 o_flags = ip_tunnel_flags_to_be16(p->o_flags);
827
828 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
829 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
830 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
831 ((i_flags | o_flags) & (GRE_VERSION | GRE_ROUTING)))
832 return -EINVAL;
833 }
834
835 gre_flags_to_tnl_flags(p->i_flags, i_flags);
836 gre_flags_to_tnl_flags(p->o_flags, o_flags);
837
838 err = ip_tunnel_ctl(dev, p, cmd);
839 if (err)
840 return err;
841
842 if (cmd == SIOCCHGTUNNEL) {
843 struct ip_tunnel *t = netdev_priv(dev);
844
845 ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags);
846 ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags);
847
848 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
849 ipgre_link_update(dev, true);
850 }
851
852 i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
853 ip_tunnel_flags_from_be16(p->i_flags, i_flags);
854 o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
855 ip_tunnel_flags_from_be16(p->o_flags, o_flags);
856
857 return 0;
858 }
859
860 /* Nice toy. Unfortunately, useless in real life :-)
861 It allows to construct virtual multiprotocol broadcast "LAN"
862 over the Internet, provided multicast routing is tuned.
863
864
865 I have no idea was this bicycle invented before me,
866 so that I had to set ARPHRD_IPGRE to a random value.
867 I have an impression, that Cisco could make something similar,
868 but this feature is apparently missing in IOS<=11.2(8).
869
870 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
871 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
872
873 ping -t 255 224.66.66.66
874
875 If nobody answers, mbone does not work.
876
877 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
878 ip addr add 10.66.66.<somewhat>/24 dev Universe
879 ifconfig Universe up
880 ifconfig Universe add fe80::<Your_real_addr>/10
881 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
882 ftp 10.66.66.66
883 ...
884 ftp fec0:6666:6666::193.233.7.65
885 ...
886 */
ipgre_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)887 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
888 unsigned short type,
889 const void *daddr, const void *saddr, unsigned int len)
890 {
891 struct ip_tunnel *t = netdev_priv(dev);
892 struct iphdr *iph;
893 struct gre_base_hdr *greh;
894
895 iph = skb_push(skb, t->hlen + sizeof(*iph));
896 greh = (struct gre_base_hdr *)(iph+1);
897 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
898 greh->protocol = htons(type);
899
900 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
901
902 /* Set the source hardware address. */
903 if (saddr)
904 memcpy(&iph->saddr, saddr, 4);
905 if (daddr)
906 memcpy(&iph->daddr, daddr, 4);
907 if (iph->daddr)
908 return t->hlen + sizeof(*iph);
909
910 return -(t->hlen + sizeof(*iph));
911 }
912
ipgre_header_parse(const struct sk_buff * skb,unsigned char * haddr)913 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
914 {
915 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
916 memcpy(haddr, &iph->saddr, 4);
917 return 4;
918 }
919
920 static const struct header_ops ipgre_header_ops = {
921 .create = ipgre_header,
922 .parse = ipgre_header_parse,
923 };
924
925 #ifdef CONFIG_NET_IPGRE_BROADCAST
ipgre_open(struct net_device * dev)926 static int ipgre_open(struct net_device *dev)
927 {
928 struct ip_tunnel *t = netdev_priv(dev);
929
930 if (ipv4_is_multicast(t->parms.iph.daddr)) {
931 struct flowi4 fl4;
932 struct rtable *rt;
933
934 rt = ip_route_output_gre(t->net, &fl4,
935 t->parms.iph.daddr,
936 t->parms.iph.saddr,
937 t->parms.o_key,
938 t->parms.iph.tos & INET_DSCP_MASK,
939 t->parms.link);
940 if (IS_ERR(rt))
941 return -EADDRNOTAVAIL;
942 dev = rt->dst.dev;
943 ip_rt_put(rt);
944 if (!__in_dev_get_rtnl(dev))
945 return -EADDRNOTAVAIL;
946 t->mlink = dev->ifindex;
947 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
948 }
949 return 0;
950 }
951
ipgre_close(struct net_device * dev)952 static int ipgre_close(struct net_device *dev)
953 {
954 struct ip_tunnel *t = netdev_priv(dev);
955
956 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
957 struct in_device *in_dev;
958 in_dev = inetdev_by_index(t->net, t->mlink);
959 if (in_dev)
960 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
961 }
962 return 0;
963 }
964 #endif
965
966 static const struct net_device_ops ipgre_netdev_ops = {
967 .ndo_init = ipgre_tunnel_init,
968 .ndo_uninit = ip_tunnel_uninit,
969 #ifdef CONFIG_NET_IPGRE_BROADCAST
970 .ndo_open = ipgre_open,
971 .ndo_stop = ipgre_close,
972 #endif
973 .ndo_start_xmit = ipgre_xmit,
974 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
975 .ndo_change_mtu = ip_tunnel_change_mtu,
976 .ndo_get_stats64 = dev_get_tstats64,
977 .ndo_get_iflink = ip_tunnel_get_iflink,
978 .ndo_tunnel_ctl = ipgre_tunnel_ctl,
979 };
980
981 #define GRE_FEATURES (NETIF_F_SG | \
982 NETIF_F_FRAGLIST | \
983 NETIF_F_HIGHDMA | \
984 NETIF_F_HW_CSUM)
985
ipgre_tunnel_setup(struct net_device * dev)986 static void ipgre_tunnel_setup(struct net_device *dev)
987 {
988 dev->netdev_ops = &ipgre_netdev_ops;
989 dev->type = ARPHRD_IPGRE;
990 ip_tunnel_setup(dev, ipgre_net_id);
991 }
992
__gre_tunnel_init(struct net_device * dev)993 static void __gre_tunnel_init(struct net_device *dev)
994 {
995 struct ip_tunnel *tunnel;
996
997 tunnel = netdev_priv(dev);
998 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
999 tunnel->parms.iph.protocol = IPPROTO_GRE;
1000
1001 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1002 dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
1003
1004 dev->features |= GRE_FEATURES;
1005 dev->hw_features |= GRE_FEATURES;
1006
1007 /* TCP offload with GRE SEQ is not supported, nor can we support 2
1008 * levels of outer headers requiring an update.
1009 */
1010 if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags))
1011 return;
1012 if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
1013 tunnel->encap.type != TUNNEL_ENCAP_NONE)
1014 return;
1015
1016 dev->features |= NETIF_F_GSO_SOFTWARE;
1017 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1018
1019 dev->lltx = true;
1020 }
1021
ipgre_tunnel_init(struct net_device * dev)1022 static int ipgre_tunnel_init(struct net_device *dev)
1023 {
1024 struct ip_tunnel *tunnel = netdev_priv(dev);
1025 struct iphdr *iph = &tunnel->parms.iph;
1026
1027 __gre_tunnel_init(dev);
1028
1029 __dev_addr_set(dev, &iph->saddr, 4);
1030 memcpy(dev->broadcast, &iph->daddr, 4);
1031
1032 dev->flags = IFF_NOARP;
1033 netif_keep_dst(dev);
1034 dev->addr_len = 4;
1035
1036 if (iph->daddr && !tunnel->collect_md) {
1037 #ifdef CONFIG_NET_IPGRE_BROADCAST
1038 if (ipv4_is_multicast(iph->daddr)) {
1039 if (!iph->saddr)
1040 return -EINVAL;
1041 dev->flags = IFF_BROADCAST;
1042 dev->header_ops = &ipgre_header_ops;
1043 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1044 dev->needed_headroom = 0;
1045 }
1046 #endif
1047 } else if (!tunnel->collect_md) {
1048 dev->header_ops = &ipgre_header_ops;
1049 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1050 dev->needed_headroom = 0;
1051 }
1052
1053 return ip_tunnel_init(dev);
1054 }
1055
1056 static const struct gre_protocol ipgre_protocol = {
1057 .handler = gre_rcv,
1058 .err_handler = gre_err,
1059 };
1060
ipgre_init_net(struct net * net)1061 static int __net_init ipgre_init_net(struct net *net)
1062 {
1063 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1064 }
1065
ipgre_exit_batch_rtnl(struct list_head * list_net,struct list_head * dev_to_kill)1066 static void __net_exit ipgre_exit_batch_rtnl(struct list_head *list_net,
1067 struct list_head *dev_to_kill)
1068 {
1069 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops,
1070 dev_to_kill);
1071 }
1072
1073 static struct pernet_operations ipgre_net_ops = {
1074 .init = ipgre_init_net,
1075 .exit_batch_rtnl = ipgre_exit_batch_rtnl,
1076 .id = &ipgre_net_id,
1077 .size = sizeof(struct ip_tunnel_net),
1078 };
1079
ipgre_tunnel_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1080 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1081 struct netlink_ext_ack *extack)
1082 {
1083 __be16 flags;
1084
1085 if (!data)
1086 return 0;
1087
1088 flags = 0;
1089 if (data[IFLA_GRE_IFLAGS])
1090 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1091 if (data[IFLA_GRE_OFLAGS])
1092 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1093 if (flags & (GRE_VERSION|GRE_ROUTING))
1094 return -EINVAL;
1095
1096 if (data[IFLA_GRE_COLLECT_METADATA] &&
1097 data[IFLA_GRE_ENCAP_TYPE] &&
1098 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1099 return -EINVAL;
1100
1101 return 0;
1102 }
1103
ipgre_tap_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1104 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1105 struct netlink_ext_ack *extack)
1106 {
1107 __be32 daddr;
1108
1109 if (tb[IFLA_ADDRESS]) {
1110 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1111 return -EINVAL;
1112 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1113 return -EADDRNOTAVAIL;
1114 }
1115
1116 if (!data)
1117 goto out;
1118
1119 if (data[IFLA_GRE_REMOTE]) {
1120 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1121 if (!daddr)
1122 return -EINVAL;
1123 }
1124
1125 out:
1126 return ipgre_tunnel_validate(tb, data, extack);
1127 }
1128
erspan_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1129 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1130 struct netlink_ext_ack *extack)
1131 {
1132 __be16 flags = 0;
1133 int ret;
1134
1135 if (!data)
1136 return 0;
1137
1138 ret = ipgre_tap_validate(tb, data, extack);
1139 if (ret)
1140 return ret;
1141
1142 if (data[IFLA_GRE_ERSPAN_VER] &&
1143 nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1144 return 0;
1145
1146 /* ERSPAN type II/III should only have GRE sequence and key flag */
1147 if (data[IFLA_GRE_OFLAGS])
1148 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1149 if (data[IFLA_GRE_IFLAGS])
1150 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1151 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1152 flags != (GRE_SEQ | GRE_KEY))
1153 return -EINVAL;
1154
1155 /* ERSPAN Session ID only has 10-bit. Since we reuse
1156 * 32-bit key field as ID, check it's range.
1157 */
1158 if (data[IFLA_GRE_IKEY] &&
1159 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1160 return -EINVAL;
1161
1162 if (data[IFLA_GRE_OKEY] &&
1163 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1164 return -EINVAL;
1165
1166 return 0;
1167 }
1168
ipgre_netlink_parms(struct net_device * dev,struct nlattr * data[],struct nlattr * tb[],struct ip_tunnel_parm_kern * parms,__u32 * fwmark)1169 static int ipgre_netlink_parms(struct net_device *dev,
1170 struct nlattr *data[],
1171 struct nlattr *tb[],
1172 struct ip_tunnel_parm_kern *parms,
1173 __u32 *fwmark)
1174 {
1175 struct ip_tunnel *t = netdev_priv(dev);
1176
1177 memset(parms, 0, sizeof(*parms));
1178
1179 parms->iph.protocol = IPPROTO_GRE;
1180
1181 if (!data)
1182 return 0;
1183
1184 if (data[IFLA_GRE_LINK])
1185 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1186
1187 if (data[IFLA_GRE_IFLAGS])
1188 gre_flags_to_tnl_flags(parms->i_flags,
1189 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1190
1191 if (data[IFLA_GRE_OFLAGS])
1192 gre_flags_to_tnl_flags(parms->o_flags,
1193 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1194
1195 if (data[IFLA_GRE_IKEY])
1196 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1197
1198 if (data[IFLA_GRE_OKEY])
1199 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1200
1201 if (data[IFLA_GRE_LOCAL])
1202 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1203
1204 if (data[IFLA_GRE_REMOTE])
1205 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1206
1207 if (data[IFLA_GRE_TTL])
1208 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1209
1210 if (data[IFLA_GRE_TOS])
1211 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1212
1213 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1214 if (t->ignore_df)
1215 return -EINVAL;
1216 parms->iph.frag_off = htons(IP_DF);
1217 }
1218
1219 if (data[IFLA_GRE_COLLECT_METADATA]) {
1220 t->collect_md = true;
1221 if (dev->type == ARPHRD_IPGRE)
1222 dev->type = ARPHRD_NONE;
1223 }
1224
1225 if (data[IFLA_GRE_IGNORE_DF]) {
1226 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1227 && (parms->iph.frag_off & htons(IP_DF)))
1228 return -EINVAL;
1229 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1230 }
1231
1232 if (data[IFLA_GRE_FWMARK])
1233 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1234
1235 return 0;
1236 }
1237
erspan_netlink_parms(struct net_device * dev,struct nlattr * data[],struct nlattr * tb[],struct ip_tunnel_parm_kern * parms,__u32 * fwmark)1238 static int erspan_netlink_parms(struct net_device *dev,
1239 struct nlattr *data[],
1240 struct nlattr *tb[],
1241 struct ip_tunnel_parm_kern *parms,
1242 __u32 *fwmark)
1243 {
1244 struct ip_tunnel *t = netdev_priv(dev);
1245 int err;
1246
1247 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1248 if (err)
1249 return err;
1250 if (!data)
1251 return 0;
1252
1253 if (data[IFLA_GRE_ERSPAN_VER]) {
1254 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1255
1256 if (t->erspan_ver > 2)
1257 return -EINVAL;
1258 }
1259
1260 if (t->erspan_ver == 1) {
1261 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1262 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1263 if (t->index & ~INDEX_MASK)
1264 return -EINVAL;
1265 }
1266 } else if (t->erspan_ver == 2) {
1267 if (data[IFLA_GRE_ERSPAN_DIR]) {
1268 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1269 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1270 return -EINVAL;
1271 }
1272 if (data[IFLA_GRE_ERSPAN_HWID]) {
1273 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1274 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1275 return -EINVAL;
1276 }
1277 }
1278
1279 return 0;
1280 }
1281
1282 /* This function returns true when ENCAP attributes are present in the nl msg */
ipgre_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)1283 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1284 struct ip_tunnel_encap *ipencap)
1285 {
1286 bool ret = false;
1287
1288 memset(ipencap, 0, sizeof(*ipencap));
1289
1290 if (!data)
1291 return ret;
1292
1293 if (data[IFLA_GRE_ENCAP_TYPE]) {
1294 ret = true;
1295 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1296 }
1297
1298 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1299 ret = true;
1300 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1301 }
1302
1303 if (data[IFLA_GRE_ENCAP_SPORT]) {
1304 ret = true;
1305 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1306 }
1307
1308 if (data[IFLA_GRE_ENCAP_DPORT]) {
1309 ret = true;
1310 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1311 }
1312
1313 return ret;
1314 }
1315
gre_tap_init(struct net_device * dev)1316 static int gre_tap_init(struct net_device *dev)
1317 {
1318 __gre_tunnel_init(dev);
1319 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1320 netif_keep_dst(dev);
1321
1322 return ip_tunnel_init(dev);
1323 }
1324
1325 static const struct net_device_ops gre_tap_netdev_ops = {
1326 .ndo_init = gre_tap_init,
1327 .ndo_uninit = ip_tunnel_uninit,
1328 .ndo_start_xmit = gre_tap_xmit,
1329 .ndo_set_mac_address = eth_mac_addr,
1330 .ndo_validate_addr = eth_validate_addr,
1331 .ndo_change_mtu = ip_tunnel_change_mtu,
1332 .ndo_get_stats64 = dev_get_tstats64,
1333 .ndo_get_iflink = ip_tunnel_get_iflink,
1334 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1335 };
1336
erspan_tunnel_init(struct net_device * dev)1337 static int erspan_tunnel_init(struct net_device *dev)
1338 {
1339 struct ip_tunnel *tunnel = netdev_priv(dev);
1340
1341 if (tunnel->erspan_ver == 0)
1342 tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1343 else
1344 tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1345
1346 tunnel->parms.iph.protocol = IPPROTO_GRE;
1347 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1348 erspan_hdr_len(tunnel->erspan_ver);
1349
1350 dev->features |= GRE_FEATURES;
1351 dev->hw_features |= GRE_FEATURES;
1352 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1353 netif_keep_dst(dev);
1354
1355 return ip_tunnel_init(dev);
1356 }
1357
1358 static const struct net_device_ops erspan_netdev_ops = {
1359 .ndo_init = erspan_tunnel_init,
1360 .ndo_uninit = ip_tunnel_uninit,
1361 .ndo_start_xmit = erspan_xmit,
1362 .ndo_set_mac_address = eth_mac_addr,
1363 .ndo_validate_addr = eth_validate_addr,
1364 .ndo_change_mtu = ip_tunnel_change_mtu,
1365 .ndo_get_stats64 = dev_get_tstats64,
1366 .ndo_get_iflink = ip_tunnel_get_iflink,
1367 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1368 };
1369
ipgre_tap_setup(struct net_device * dev)1370 static void ipgre_tap_setup(struct net_device *dev)
1371 {
1372 ether_setup(dev);
1373 dev->max_mtu = 0;
1374 dev->netdev_ops = &gre_tap_netdev_ops;
1375 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1376 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1377 ip_tunnel_setup(dev, gre_tap_net_id);
1378 }
1379
1380 static int
ipgre_newlink_encap_setup(struct net_device * dev,struct nlattr * data[])1381 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1382 {
1383 struct ip_tunnel_encap ipencap;
1384
1385 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1386 struct ip_tunnel *t = netdev_priv(dev);
1387 int err = ip_tunnel_encap_setup(t, &ipencap);
1388
1389 if (err < 0)
1390 return err;
1391 }
1392
1393 return 0;
1394 }
1395
ipgre_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1396 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1397 struct nlattr *tb[], struct nlattr *data[],
1398 struct netlink_ext_ack *extack)
1399 {
1400 struct ip_tunnel_parm_kern p;
1401 __u32 fwmark = 0;
1402 int err;
1403
1404 err = ipgre_newlink_encap_setup(dev, data);
1405 if (err)
1406 return err;
1407
1408 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1409 if (err < 0)
1410 return err;
1411 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1412 }
1413
erspan_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1414 static int erspan_newlink(struct net *src_net, struct net_device *dev,
1415 struct nlattr *tb[], struct nlattr *data[],
1416 struct netlink_ext_ack *extack)
1417 {
1418 struct ip_tunnel_parm_kern p;
1419 __u32 fwmark = 0;
1420 int err;
1421
1422 err = ipgre_newlink_encap_setup(dev, data);
1423 if (err)
1424 return err;
1425
1426 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1427 if (err)
1428 return err;
1429 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1430 }
1431
ipgre_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1432 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1433 struct nlattr *data[],
1434 struct netlink_ext_ack *extack)
1435 {
1436 struct ip_tunnel *t = netdev_priv(dev);
1437 struct ip_tunnel_parm_kern p;
1438 __u32 fwmark = t->fwmark;
1439 int err;
1440
1441 err = ipgre_newlink_encap_setup(dev, data);
1442 if (err)
1443 return err;
1444
1445 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1446 if (err < 0)
1447 return err;
1448
1449 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1450 if (err < 0)
1451 return err;
1452
1453 ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1454 ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1455
1456 ipgre_link_update(dev, !tb[IFLA_MTU]);
1457
1458 return 0;
1459 }
1460
erspan_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1461 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1462 struct nlattr *data[],
1463 struct netlink_ext_ack *extack)
1464 {
1465 struct ip_tunnel *t = netdev_priv(dev);
1466 struct ip_tunnel_parm_kern p;
1467 __u32 fwmark = t->fwmark;
1468 int err;
1469
1470 err = ipgre_newlink_encap_setup(dev, data);
1471 if (err)
1472 return err;
1473
1474 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1475 if (err < 0)
1476 return err;
1477
1478 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1479 if (err < 0)
1480 return err;
1481
1482 ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1483 ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1484
1485 return 0;
1486 }
1487
ipgre_get_size(const struct net_device * dev)1488 static size_t ipgre_get_size(const struct net_device *dev)
1489 {
1490 return
1491 /* IFLA_GRE_LINK */
1492 nla_total_size(4) +
1493 /* IFLA_GRE_IFLAGS */
1494 nla_total_size(2) +
1495 /* IFLA_GRE_OFLAGS */
1496 nla_total_size(2) +
1497 /* IFLA_GRE_IKEY */
1498 nla_total_size(4) +
1499 /* IFLA_GRE_OKEY */
1500 nla_total_size(4) +
1501 /* IFLA_GRE_LOCAL */
1502 nla_total_size(4) +
1503 /* IFLA_GRE_REMOTE */
1504 nla_total_size(4) +
1505 /* IFLA_GRE_TTL */
1506 nla_total_size(1) +
1507 /* IFLA_GRE_TOS */
1508 nla_total_size(1) +
1509 /* IFLA_GRE_PMTUDISC */
1510 nla_total_size(1) +
1511 /* IFLA_GRE_ENCAP_TYPE */
1512 nla_total_size(2) +
1513 /* IFLA_GRE_ENCAP_FLAGS */
1514 nla_total_size(2) +
1515 /* IFLA_GRE_ENCAP_SPORT */
1516 nla_total_size(2) +
1517 /* IFLA_GRE_ENCAP_DPORT */
1518 nla_total_size(2) +
1519 /* IFLA_GRE_COLLECT_METADATA */
1520 nla_total_size(0) +
1521 /* IFLA_GRE_IGNORE_DF */
1522 nla_total_size(1) +
1523 /* IFLA_GRE_FWMARK */
1524 nla_total_size(4) +
1525 /* IFLA_GRE_ERSPAN_INDEX */
1526 nla_total_size(4) +
1527 /* IFLA_GRE_ERSPAN_VER */
1528 nla_total_size(1) +
1529 /* IFLA_GRE_ERSPAN_DIR */
1530 nla_total_size(1) +
1531 /* IFLA_GRE_ERSPAN_HWID */
1532 nla_total_size(2) +
1533 0;
1534 }
1535
ipgre_fill_info(struct sk_buff * skb,const struct net_device * dev)1536 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1537 {
1538 struct ip_tunnel *t = netdev_priv(dev);
1539 struct ip_tunnel_parm_kern *p = &t->parms;
1540 IP_TUNNEL_DECLARE_FLAGS(o_flags);
1541
1542 ip_tunnel_flags_copy(o_flags, p->o_flags);
1543
1544 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1545 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1546 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1547 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1548 gre_tnl_flags_to_gre_flags(o_flags)) ||
1549 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1550 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1551 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1552 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1553 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1554 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1555 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1556 !!(p->iph.frag_off & htons(IP_DF))) ||
1557 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1558 goto nla_put_failure;
1559
1560 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1561 t->encap.type) ||
1562 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1563 t->encap.sport) ||
1564 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1565 t->encap.dport) ||
1566 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1567 t->encap.flags))
1568 goto nla_put_failure;
1569
1570 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1571 goto nla_put_failure;
1572
1573 if (t->collect_md) {
1574 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1575 goto nla_put_failure;
1576 }
1577
1578 return 0;
1579
1580 nla_put_failure:
1581 return -EMSGSIZE;
1582 }
1583
erspan_fill_info(struct sk_buff * skb,const struct net_device * dev)1584 static int erspan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1585 {
1586 struct ip_tunnel *t = netdev_priv(dev);
1587
1588 if (t->erspan_ver <= 2) {
1589 if (t->erspan_ver != 0 && !t->collect_md)
1590 __set_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags);
1591
1592 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1593 goto nla_put_failure;
1594
1595 if (t->erspan_ver == 1) {
1596 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1597 goto nla_put_failure;
1598 } else if (t->erspan_ver == 2) {
1599 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1600 goto nla_put_failure;
1601 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1602 goto nla_put_failure;
1603 }
1604 }
1605
1606 return ipgre_fill_info(skb, dev);
1607
1608 nla_put_failure:
1609 return -EMSGSIZE;
1610 }
1611
erspan_setup(struct net_device * dev)1612 static void erspan_setup(struct net_device *dev)
1613 {
1614 struct ip_tunnel *t = netdev_priv(dev);
1615
1616 ether_setup(dev);
1617 dev->max_mtu = 0;
1618 dev->netdev_ops = &erspan_netdev_ops;
1619 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1620 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1621 ip_tunnel_setup(dev, erspan_net_id);
1622 t->erspan_ver = 1;
1623 }
1624
1625 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1626 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1627 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1628 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1629 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1630 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1631 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
1632 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
1633 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1634 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1635 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1636 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1637 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1638 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1639 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1640 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1641 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1642 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1643 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1644 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1645 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1646 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1647 };
1648
1649 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1650 .kind = "gre",
1651 .maxtype = IFLA_GRE_MAX,
1652 .policy = ipgre_policy,
1653 .priv_size = sizeof(struct ip_tunnel),
1654 .setup = ipgre_tunnel_setup,
1655 .validate = ipgre_tunnel_validate,
1656 .newlink = ipgre_newlink,
1657 .changelink = ipgre_changelink,
1658 .dellink = ip_tunnel_dellink,
1659 .get_size = ipgre_get_size,
1660 .fill_info = ipgre_fill_info,
1661 .get_link_net = ip_tunnel_get_link_net,
1662 };
1663
1664 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1665 .kind = "gretap",
1666 .maxtype = IFLA_GRE_MAX,
1667 .policy = ipgre_policy,
1668 .priv_size = sizeof(struct ip_tunnel),
1669 .setup = ipgre_tap_setup,
1670 .validate = ipgre_tap_validate,
1671 .newlink = ipgre_newlink,
1672 .changelink = ipgre_changelink,
1673 .dellink = ip_tunnel_dellink,
1674 .get_size = ipgre_get_size,
1675 .fill_info = ipgre_fill_info,
1676 .get_link_net = ip_tunnel_get_link_net,
1677 };
1678
1679 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1680 .kind = "erspan",
1681 .maxtype = IFLA_GRE_MAX,
1682 .policy = ipgre_policy,
1683 .priv_size = sizeof(struct ip_tunnel),
1684 .setup = erspan_setup,
1685 .validate = erspan_validate,
1686 .newlink = erspan_newlink,
1687 .changelink = erspan_changelink,
1688 .dellink = ip_tunnel_dellink,
1689 .get_size = ipgre_get_size,
1690 .fill_info = erspan_fill_info,
1691 .get_link_net = ip_tunnel_get_link_net,
1692 };
1693
gretap_fb_dev_create(struct net * net,const char * name,u8 name_assign_type)1694 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1695 u8 name_assign_type)
1696 {
1697 struct nlattr *tb[IFLA_MAX + 1];
1698 struct net_device *dev;
1699 LIST_HEAD(list_kill);
1700 struct ip_tunnel *t;
1701 int err;
1702
1703 memset(&tb, 0, sizeof(tb));
1704
1705 dev = rtnl_create_link(net, name, name_assign_type,
1706 &ipgre_tap_ops, tb, NULL);
1707 if (IS_ERR(dev))
1708 return dev;
1709
1710 /* Configure flow based GRE device. */
1711 t = netdev_priv(dev);
1712 t->collect_md = true;
1713
1714 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1715 if (err < 0) {
1716 free_netdev(dev);
1717 return ERR_PTR(err);
1718 }
1719
1720 /* openvswitch users expect packet sizes to be unrestricted,
1721 * so set the largest MTU we can.
1722 */
1723 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1724 if (err)
1725 goto out;
1726
1727 err = rtnl_configure_link(dev, NULL, 0, NULL);
1728 if (err < 0)
1729 goto out;
1730
1731 return dev;
1732 out:
1733 ip_tunnel_dellink(dev, &list_kill);
1734 unregister_netdevice_many(&list_kill);
1735 return ERR_PTR(err);
1736 }
1737 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1738
ipgre_tap_init_net(struct net * net)1739 static int __net_init ipgre_tap_init_net(struct net *net)
1740 {
1741 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1742 }
1743
ipgre_tap_exit_batch_rtnl(struct list_head * list_net,struct list_head * dev_to_kill)1744 static void __net_exit ipgre_tap_exit_batch_rtnl(struct list_head *list_net,
1745 struct list_head *dev_to_kill)
1746 {
1747 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops,
1748 dev_to_kill);
1749 }
1750
1751 static struct pernet_operations ipgre_tap_net_ops = {
1752 .init = ipgre_tap_init_net,
1753 .exit_batch_rtnl = ipgre_tap_exit_batch_rtnl,
1754 .id = &gre_tap_net_id,
1755 .size = sizeof(struct ip_tunnel_net),
1756 };
1757
erspan_init_net(struct net * net)1758 static int __net_init erspan_init_net(struct net *net)
1759 {
1760 return ip_tunnel_init_net(net, erspan_net_id,
1761 &erspan_link_ops, "erspan0");
1762 }
1763
erspan_exit_batch_rtnl(struct list_head * net_list,struct list_head * dev_to_kill)1764 static void __net_exit erspan_exit_batch_rtnl(struct list_head *net_list,
1765 struct list_head *dev_to_kill)
1766 {
1767 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops,
1768 dev_to_kill);
1769 }
1770
1771 static struct pernet_operations erspan_net_ops = {
1772 .init = erspan_init_net,
1773 .exit_batch_rtnl = erspan_exit_batch_rtnl,
1774 .id = &erspan_net_id,
1775 .size = sizeof(struct ip_tunnel_net),
1776 };
1777
ipgre_init(void)1778 static int __init ipgre_init(void)
1779 {
1780 int err;
1781
1782 pr_info("GRE over IPv4 tunneling driver\n");
1783
1784 err = register_pernet_device(&ipgre_net_ops);
1785 if (err < 0)
1786 return err;
1787
1788 err = register_pernet_device(&ipgre_tap_net_ops);
1789 if (err < 0)
1790 goto pnet_tap_failed;
1791
1792 err = register_pernet_device(&erspan_net_ops);
1793 if (err < 0)
1794 goto pnet_erspan_failed;
1795
1796 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1797 if (err < 0) {
1798 pr_info("%s: can't add protocol\n", __func__);
1799 goto add_proto_failed;
1800 }
1801
1802 err = rtnl_link_register(&ipgre_link_ops);
1803 if (err < 0)
1804 goto rtnl_link_failed;
1805
1806 err = rtnl_link_register(&ipgre_tap_ops);
1807 if (err < 0)
1808 goto tap_ops_failed;
1809
1810 err = rtnl_link_register(&erspan_link_ops);
1811 if (err < 0)
1812 goto erspan_link_failed;
1813
1814 return 0;
1815
1816 erspan_link_failed:
1817 rtnl_link_unregister(&ipgre_tap_ops);
1818 tap_ops_failed:
1819 rtnl_link_unregister(&ipgre_link_ops);
1820 rtnl_link_failed:
1821 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1822 add_proto_failed:
1823 unregister_pernet_device(&erspan_net_ops);
1824 pnet_erspan_failed:
1825 unregister_pernet_device(&ipgre_tap_net_ops);
1826 pnet_tap_failed:
1827 unregister_pernet_device(&ipgre_net_ops);
1828 return err;
1829 }
1830
ipgre_fini(void)1831 static void __exit ipgre_fini(void)
1832 {
1833 rtnl_link_unregister(&ipgre_tap_ops);
1834 rtnl_link_unregister(&ipgre_link_ops);
1835 rtnl_link_unregister(&erspan_link_ops);
1836 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1837 unregister_pernet_device(&ipgre_tap_net_ops);
1838 unregister_pernet_device(&ipgre_net_ops);
1839 unregister_pernet_device(&erspan_net_ops);
1840 }
1841
1842 module_init(ipgre_init);
1843 module_exit(ipgre_fini);
1844 MODULE_DESCRIPTION("IPv4 GRE tunnels over IP library");
1845 MODULE_LICENSE("GPL");
1846 MODULE_ALIAS_RTNL_LINK("gre");
1847 MODULE_ALIAS_RTNL_LINK("gretap");
1848 MODULE_ALIAS_RTNL_LINK("erspan");
1849 MODULE_ALIAS_NETDEV("gre0");
1850 MODULE_ALIAS_NETDEV("gretap0");
1851 MODULE_ALIAS_NETDEV("erspan0");
1852