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