xref: /linux/net/ipv4/ip_gre.c (revision 9e995c573b63453a904f3157813dc8cde4a6aba4)
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/flow.h>
32 #include <net/sock.h>
33 #include <net/ip.h>
34 #include <net/icmp.h>
35 #include <net/protocol.h>
36 #include <net/ip_tunnels.h>
37 #include <net/arp.h>
38 #include <net/checksum.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/xfrm.h>
42 #include <net/net_namespace.h>
43 #include <net/netns/generic.h>
44 #include <net/rtnetlink.h>
45 #include <net/gre.h>
46 #include <net/dst_metadata.h>
47 #include <net/erspan.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 			/* MUST set options_len before referencing options */
334 			info = &tun_dst->u.tun_info;
335 			info->options_len = sizeof(*md);
336 
337 			/* skb can be uncloned in __iptunnel_pull_header, so
338 			 * old pkt_md is no longer valid and we need to reset
339 			 * it
340 			 */
341 			gh = skb_network_header(skb) +
342 			     skb_network_header_len(skb);
343 			pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
344 							    sizeof(*ershdr));
345 			md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
346 			md->version = ver;
347 			md2 = &md->u.md2;
348 			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
349 						       ERSPAN_V2_MDSIZE);
350 
351 			__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
352 				  info->key.tun_flags);
353 		}
354 
355 		skb_reset_mac_header(skb);
356 		ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
357 		return PACKET_RCVD;
358 	}
359 	return PACKET_REJECT;
360 
361 drop:
362 	kfree_skb(skb);
363 	return PACKET_RCVD;
364 }
365 
__ipgre_rcv(struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct ip_tunnel_net * itn,int hdr_len,bool raw_proto)366 static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
367 		       struct ip_tunnel_net *itn, int hdr_len, bool raw_proto)
368 {
369 	struct metadata_dst *tun_dst = NULL;
370 	const struct iphdr *iph;
371 	struct ip_tunnel *tunnel;
372 
373 	iph = ip_hdr(skb);
374 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
375 				  iph->saddr, iph->daddr, tpi->key);
376 
377 	if (tunnel) {
378 		const struct iphdr *tnl_params;
379 
380 		if (__iptunnel_pull_header(skb, hdr_len, tpi->proto,
381 					   raw_proto, false) < 0)
382 			goto drop;
383 
384 		/* Special case for ipgre_header_parse(), which expects the
385 		 * mac_header to point to the outer IP header.
386 		 */
387 		if (tunnel->dev->header_ops == &ipgre_header_ops)
388 			skb_pop_mac_header(skb);
389 		else
390 			skb_reset_mac_header(skb);
391 
392 		tnl_params = &tunnel->parms.iph;
393 		if (tunnel->collect_md || tnl_params->daddr == 0) {
394 			IP_TUNNEL_DECLARE_FLAGS(flags) = { };
395 			__be64 tun_id;
396 
397 			__set_bit(IP_TUNNEL_CSUM_BIT, flags);
398 			__set_bit(IP_TUNNEL_KEY_BIT, flags);
399 			ip_tunnel_flags_and(flags, tpi->flags, flags);
400 
401 			tun_id = key32_to_tunnel_id(tpi->key);
402 			tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0);
403 			if (!tun_dst)
404 				return PACKET_REJECT;
405 		}
406 
407 		ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
408 		return PACKET_RCVD;
409 	}
410 	return PACKET_NEXT;
411 
412 drop:
413 	kfree_skb(skb);
414 	return PACKET_RCVD;
415 }
416 
ipgre_rcv(struct sk_buff * skb,const struct tnl_ptk_info * tpi,int hdr_len)417 static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
418 		     int hdr_len)
419 {
420 	struct net *net = dev_net(skb->dev);
421 	struct ip_tunnel_net *itn;
422 	int res;
423 
424 	if (tpi->proto == htons(ETH_P_TEB))
425 		itn = net_generic(net, gre_tap_net_id);
426 	else
427 		itn = net_generic(net, ipgre_net_id);
428 
429 	res = __ipgre_rcv(skb, tpi, itn, hdr_len, false);
430 	if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) {
431 		/* ipgre tunnels in collect metadata mode should receive
432 		 * also ETH_P_TEB traffic.
433 		 */
434 		itn = net_generic(net, ipgre_net_id);
435 		res = __ipgre_rcv(skb, tpi, itn, hdr_len, true);
436 	}
437 	return res;
438 }
439 
gre_rcv(struct sk_buff * skb)440 static int gre_rcv(struct sk_buff *skb)
441 {
442 	struct tnl_ptk_info tpi;
443 	bool csum_err = false;
444 	int hdr_len;
445 
446 #ifdef CONFIG_NET_IPGRE_BROADCAST
447 	if (ipv4_is_multicast(ip_hdr(skb)->daddr)) {
448 		/* Looped back packet, drop it! */
449 		if (rt_is_output_route(skb_rtable(skb)))
450 			goto drop;
451 	}
452 #endif
453 
454 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0);
455 	if (hdr_len < 0)
456 		goto drop;
457 
458 	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
459 		     tpi.proto == htons(ETH_P_ERSPAN2))) {
460 		if (erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
461 			return 0;
462 		goto out;
463 	}
464 
465 	if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
466 		return 0;
467 
468 out:
469 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
470 drop:
471 	kfree_skb(skb);
472 	return 0;
473 }
474 
__gre_xmit(struct sk_buff * skb,struct net_device * dev,const struct iphdr * tnl_params,__be16 proto)475 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
476 		       const struct iphdr *tnl_params,
477 		       __be16 proto)
478 {
479 	struct ip_tunnel *tunnel = netdev_priv(dev);
480 	IP_TUNNEL_DECLARE_FLAGS(flags);
481 
482 	ip_tunnel_flags_copy(flags, tunnel->parms.o_flags);
483 
484 	/* Push GRE header. */
485 	gre_build_header(skb, tunnel->tun_hlen,
486 			 flags, proto, tunnel->parms.o_key,
487 			 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
488 			 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
489 
490 	ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
491 }
492 
gre_handle_offloads(struct sk_buff * skb,bool csum)493 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
494 {
495 	return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
496 }
497 
gre_fb_xmit(struct sk_buff * skb,struct net_device * dev,__be16 proto)498 static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
499 			__be16 proto)
500 {
501 	struct ip_tunnel *tunnel = netdev_priv(dev);
502 	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
503 	struct ip_tunnel_info *tun_info;
504 	const struct ip_tunnel_key *key;
505 	int tunnel_hlen;
506 
507 	tun_info = skb_tunnel_info(skb);
508 	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
509 		     ip_tunnel_info_af(tun_info) != AF_INET))
510 		goto err_free_skb;
511 
512 	key = &tun_info->key;
513 	tunnel_hlen = gre_calc_hlen(key->tun_flags);
514 
515 	if (skb_cow_head(skb, dev->needed_headroom))
516 		goto err_free_skb;
517 
518 	/* Push Tunnel header. */
519 	if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
520 					      tunnel->parms.o_flags)))
521 		goto err_free_skb;
522 
523 	__set_bit(IP_TUNNEL_CSUM_BIT, flags);
524 	__set_bit(IP_TUNNEL_KEY_BIT, flags);
525 	__set_bit(IP_TUNNEL_SEQ_BIT, flags);
526 	ip_tunnel_flags_and(flags, tun_info->key.tun_flags, flags);
527 
528 	gre_build_header(skb, tunnel_hlen, flags, proto,
529 			 tunnel_id_to_key32(tun_info->key.tun_id),
530 			 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
531 			 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
532 
533 	ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
534 
535 	return;
536 
537 err_free_skb:
538 	kfree_skb(skb);
539 	DEV_STATS_INC(dev, tx_dropped);
540 }
541 
erspan_fb_xmit(struct sk_buff * skb,struct net_device * dev)542 static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
543 {
544 	struct ip_tunnel *tunnel = netdev_priv(dev);
545 	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
546 	struct ip_tunnel_info *tun_info;
547 	const struct ip_tunnel_key *key;
548 	struct erspan_metadata *md;
549 	bool truncate = false;
550 	__be16 proto;
551 	int tunnel_hlen;
552 	int version;
553 	int nhoff;
554 
555 	tun_info = skb_tunnel_info(skb);
556 	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
557 		     ip_tunnel_info_af(tun_info) != AF_INET))
558 		goto err_free_skb;
559 
560 	key = &tun_info->key;
561 	if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
562 		goto err_free_skb;
563 	if (tun_info->options_len < sizeof(*md))
564 		goto err_free_skb;
565 	md = ip_tunnel_info_opts(tun_info);
566 
567 	/* ERSPAN has fixed 8 byte GRE header */
568 	version = md->version;
569 	tunnel_hlen = 8 + erspan_hdr_len(version);
570 
571 	if (skb_cow_head(skb, dev->needed_headroom))
572 		goto err_free_skb;
573 
574 	if (gre_handle_offloads(skb, false))
575 		goto err_free_skb;
576 
577 	if (skb->len > dev->mtu + dev->hard_header_len) {
578 		if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
579 			goto err_free_skb;
580 		truncate = true;
581 	}
582 
583 	nhoff = skb_network_offset(skb);
584 	if (skb->protocol == htons(ETH_P_IP) &&
585 	    (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
586 		truncate = true;
587 
588 	if (skb->protocol == htons(ETH_P_IPV6)) {
589 		int thoff;
590 
591 		if (skb_transport_header_was_set(skb))
592 			thoff = skb_transport_offset(skb);
593 		else
594 			thoff = nhoff + sizeof(struct ipv6hdr);
595 		if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
596 			truncate = true;
597 	}
598 
599 	if (version == 1) {
600 		erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
601 				    ntohl(md->u.index), truncate, true);
602 		proto = htons(ETH_P_ERSPAN);
603 	} else if (version == 2) {
604 		erspan_build_header_v2(skb,
605 				       ntohl(tunnel_id_to_key32(key->tun_id)),
606 				       md->u.md2.dir,
607 				       get_hwid(&md->u.md2),
608 				       truncate, true);
609 		proto = htons(ETH_P_ERSPAN2);
610 	} else {
611 		goto err_free_skb;
612 	}
613 
614 	__set_bit(IP_TUNNEL_SEQ_BIT, flags);
615 	gre_build_header(skb, 8, flags, proto, 0,
616 			 htonl(atomic_fetch_inc(&tunnel->o_seqno)));
617 
618 	ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
619 
620 	return;
621 
622 err_free_skb:
623 	kfree_skb(skb);
624 	DEV_STATS_INC(dev, tx_dropped);
625 }
626 
gre_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)627 static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
628 {
629 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
630 	const struct ip_tunnel_key *key;
631 	struct rtable *rt;
632 	struct flowi4 fl4;
633 
634 	if (ip_tunnel_info_af(info) != AF_INET)
635 		return -EINVAL;
636 
637 	key = &info->key;
638 	ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
639 			    tunnel_id_to_key32(key->tun_id),
640 			    key->tos & ~INET_ECN_MASK, dev_net(dev), 0,
641 			    skb->mark, skb_get_hash(skb), key->flow_flags);
642 	rt = ip_route_output_key(dev_net(dev), &fl4);
643 	if (IS_ERR(rt))
644 		return PTR_ERR(rt);
645 
646 	ip_rt_put(rt);
647 	info->key.u.ipv4.src = fl4.saddr;
648 	return 0;
649 }
650 
ipgre_xmit(struct sk_buff * skb,struct net_device * dev)651 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
652 			      struct net_device *dev)
653 {
654 	struct ip_tunnel *tunnel = netdev_priv(dev);
655 	const struct iphdr *tnl_params;
656 
657 	if (!pskb_inet_may_pull(skb))
658 		goto free_skb;
659 
660 	if (tunnel->collect_md) {
661 		gre_fb_xmit(skb, dev, skb->protocol);
662 		return NETDEV_TX_OK;
663 	}
664 
665 	if (dev->header_ops) {
666 		int pull_len = tunnel->hlen + sizeof(struct iphdr);
667 
668 		if (skb_cow_head(skb, 0))
669 			goto free_skb;
670 
671 		if (!pskb_may_pull(skb, pull_len))
672 			goto free_skb;
673 
674 		tnl_params = (const struct iphdr *)skb->data;
675 
676 		/* ip_tunnel_xmit() needs skb->data pointing to gre header. */
677 		skb_pull(skb, pull_len);
678 		skb_reset_mac_header(skb);
679 
680 		if (skb->ip_summed == CHECKSUM_PARTIAL &&
681 		    skb_checksum_start(skb) < skb->data)
682 			goto free_skb;
683 	} else {
684 		if (skb_cow_head(skb, dev->needed_headroom))
685 			goto free_skb;
686 
687 		tnl_params = &tunnel->parms.iph;
688 	}
689 
690 	if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
691 					      tunnel->parms.o_flags)))
692 		goto free_skb;
693 
694 	__gre_xmit(skb, dev, tnl_params, skb->protocol);
695 	return NETDEV_TX_OK;
696 
697 free_skb:
698 	kfree_skb(skb);
699 	DEV_STATS_INC(dev, tx_dropped);
700 	return NETDEV_TX_OK;
701 }
702 
erspan_xmit(struct sk_buff * skb,struct net_device * dev)703 static netdev_tx_t erspan_xmit(struct sk_buff *skb,
704 			       struct net_device *dev)
705 {
706 	struct ip_tunnel *tunnel = netdev_priv(dev);
707 	bool truncate = false;
708 	__be16 proto;
709 
710 	if (!pskb_inet_may_pull(skb))
711 		goto free_skb;
712 
713 	if (tunnel->collect_md) {
714 		erspan_fb_xmit(skb, dev);
715 		return NETDEV_TX_OK;
716 	}
717 
718 	if (gre_handle_offloads(skb, false))
719 		goto free_skb;
720 
721 	if (skb_cow_head(skb, dev->needed_headroom))
722 		goto free_skb;
723 
724 	if (skb->len > dev->mtu + dev->hard_header_len) {
725 		if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
726 			goto free_skb;
727 		truncate = true;
728 	}
729 
730 	/* Push ERSPAN header */
731 	if (tunnel->erspan_ver == 0) {
732 		proto = htons(ETH_P_ERSPAN);
733 		__clear_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags);
734 	} else if (tunnel->erspan_ver == 1) {
735 		erspan_build_header(skb, ntohl(tunnel->parms.o_key),
736 				    tunnel->index,
737 				    truncate, true);
738 		proto = htons(ETH_P_ERSPAN);
739 	} else if (tunnel->erspan_ver == 2) {
740 		erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
741 				       tunnel->dir, tunnel->hwid,
742 				       truncate, true);
743 		proto = htons(ETH_P_ERSPAN2);
744 	} else {
745 		goto free_skb;
746 	}
747 
748 	__clear_bit(IP_TUNNEL_KEY_BIT, tunnel->parms.o_flags);
749 	__gre_xmit(skb, dev, &tunnel->parms.iph, proto);
750 	return NETDEV_TX_OK;
751 
752 free_skb:
753 	kfree_skb(skb);
754 	DEV_STATS_INC(dev, tx_dropped);
755 	return NETDEV_TX_OK;
756 }
757 
gre_tap_xmit(struct sk_buff * skb,struct net_device * dev)758 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
759 				struct net_device *dev)
760 {
761 	struct ip_tunnel *tunnel = netdev_priv(dev);
762 
763 	if (!pskb_inet_may_pull(skb))
764 		goto free_skb;
765 
766 	if (tunnel->collect_md) {
767 		gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
768 		return NETDEV_TX_OK;
769 	}
770 
771 	if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
772 					      tunnel->parms.o_flags)))
773 		goto free_skb;
774 
775 	if (skb_cow_head(skb, dev->needed_headroom))
776 		goto free_skb;
777 
778 	__gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
779 	return NETDEV_TX_OK;
780 
781 free_skb:
782 	kfree_skb(skb);
783 	DEV_STATS_INC(dev, tx_dropped);
784 	return NETDEV_TX_OK;
785 }
786 
ipgre_link_update(struct net_device * dev,bool set_mtu)787 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
788 {
789 	struct ip_tunnel *tunnel = netdev_priv(dev);
790 	int len;
791 
792 	len = tunnel->tun_hlen;
793 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
794 	len = tunnel->tun_hlen - len;
795 	tunnel->hlen = tunnel->hlen + len;
796 
797 	if (dev->header_ops)
798 		dev->hard_header_len += len;
799 	else
800 		dev->needed_headroom += len;
801 
802 	if (set_mtu)
803 		WRITE_ONCE(dev->mtu, max_t(int, dev->mtu - len, 68));
804 
805 	if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags) ||
806 	    (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
807 	     tunnel->encap.type != TUNNEL_ENCAP_NONE)) {
808 		dev->features &= ~NETIF_F_GSO_SOFTWARE;
809 		dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
810 	} else {
811 		dev->features |= NETIF_F_GSO_SOFTWARE;
812 		dev->hw_features |= NETIF_F_GSO_SOFTWARE;
813 	}
814 }
815 
ipgre_tunnel_ctl(struct net_device * dev,struct ip_tunnel_parm_kern * p,int cmd)816 static int ipgre_tunnel_ctl(struct net_device *dev,
817 			    struct ip_tunnel_parm_kern *p,
818 			    int cmd)
819 {
820 	__be16 i_flags, o_flags;
821 	int err;
822 
823 	if (!ip_tunnel_flags_is_be16_compat(p->i_flags) ||
824 	    !ip_tunnel_flags_is_be16_compat(p->o_flags))
825 		return -EOVERFLOW;
826 
827 	i_flags = ip_tunnel_flags_to_be16(p->i_flags);
828 	o_flags = ip_tunnel_flags_to_be16(p->o_flags);
829 
830 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
831 		if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
832 		    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
833 		    ((i_flags | o_flags) & (GRE_VERSION | GRE_ROUTING)))
834 			return -EINVAL;
835 	}
836 
837 	gre_flags_to_tnl_flags(p->i_flags, i_flags);
838 	gre_flags_to_tnl_flags(p->o_flags, o_flags);
839 
840 	err = ip_tunnel_ctl(dev, p, cmd);
841 	if (err)
842 		return err;
843 
844 	if (cmd == SIOCCHGTUNNEL) {
845 		struct ip_tunnel *t = netdev_priv(dev);
846 
847 		ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags);
848 		ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags);
849 
850 		if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
851 			ipgre_link_update(dev, true);
852 	}
853 
854 	i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
855 	ip_tunnel_flags_from_be16(p->i_flags, i_flags);
856 	o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
857 	ip_tunnel_flags_from_be16(p->o_flags, o_flags);
858 
859 	return 0;
860 }
861 
862 /* Nice toy. Unfortunately, useless in real life :-)
863    It allows to construct virtual multiprotocol broadcast "LAN"
864    over the Internet, provided multicast routing is tuned.
865 
866 
867    I have no idea was this bicycle invented before me,
868    so that I had to set ARPHRD_IPGRE to a random value.
869    I have an impression, that Cisco could make something similar,
870    but this feature is apparently missing in IOS<=11.2(8).
871 
872    I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
873    with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
874 
875    ping -t 255 224.66.66.66
876 
877    If nobody answers, mbone does not work.
878 
879    ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
880    ip addr add 10.66.66.<somewhat>/24 dev Universe
881    ifconfig Universe up
882    ifconfig Universe add fe80::<Your_real_addr>/10
883    ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
884    ftp 10.66.66.66
885    ...
886    ftp fec0:6666:6666::193.233.7.65
887    ...
888  */
ipgre_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)889 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
890 			unsigned short type,
891 			const void *daddr, const void *saddr, unsigned int len)
892 {
893 	struct ip_tunnel *t = netdev_priv(dev);
894 	struct gre_base_hdr *greh;
895 	struct iphdr *iph;
896 	int needed;
897 
898 	needed = t->hlen + sizeof(*iph);
899 	if (skb_headroom(skb) < needed &&
900 	    pskb_expand_head(skb, HH_DATA_ALIGN(needed - skb_headroom(skb)),
901 			     0, GFP_ATOMIC))
902 		return -needed;
903 
904 	iph = skb_push(skb, needed);
905 	greh = (struct gre_base_hdr *)(iph+1);
906 	greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
907 	greh->protocol = htons(type);
908 
909 	memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
910 
911 	/* Set the source hardware address. */
912 	if (saddr)
913 		memcpy(&iph->saddr, saddr, 4);
914 	if (daddr)
915 		memcpy(&iph->daddr, daddr, 4);
916 	if (iph->daddr)
917 		return t->hlen + sizeof(*iph);
918 
919 	return -(t->hlen + sizeof(*iph));
920 }
921 
ipgre_header_parse(const struct sk_buff * skb,unsigned char * haddr)922 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
923 {
924 	const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
925 	memcpy(haddr, &iph->saddr, 4);
926 	return 4;
927 }
928 
929 static const struct header_ops ipgre_header_ops = {
930 	.create	= ipgre_header,
931 	.parse	= ipgre_header_parse,
932 };
933 
934 #ifdef CONFIG_NET_IPGRE_BROADCAST
ipgre_open(struct net_device * dev)935 static int ipgre_open(struct net_device *dev)
936 {
937 	struct ip_tunnel *t = netdev_priv(dev);
938 
939 	if (ipv4_is_multicast(t->parms.iph.daddr)) {
940 		struct flowi4 fl4 = {
941 			.flowi4_oif = t->parms.link,
942 			.flowi4_dscp = ip4h_dscp(&t->parms.iph),
943 			.flowi4_scope = RT_SCOPE_UNIVERSE,
944 			.flowi4_proto = IPPROTO_GRE,
945 			.saddr = t->parms.iph.saddr,
946 			.daddr = t->parms.iph.daddr,
947 			.fl4_gre_key = t->parms.o_key,
948 		};
949 		struct rtable *rt;
950 
951 		rt = ip_route_output_key(t->net, &fl4);
952 		if (IS_ERR(rt))
953 			return -EADDRNOTAVAIL;
954 		dev = rt->dst.dev;
955 		ip_rt_put(rt);
956 		if (!__in_dev_get_rtnl(dev))
957 			return -EADDRNOTAVAIL;
958 		t->mlink = dev->ifindex;
959 		ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
960 	}
961 	return 0;
962 }
963 
ipgre_close(struct net_device * dev)964 static int ipgre_close(struct net_device *dev)
965 {
966 	struct ip_tunnel *t = netdev_priv(dev);
967 
968 	if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
969 		struct in_device *in_dev;
970 		in_dev = inetdev_by_index(t->net, t->mlink);
971 		if (in_dev)
972 			ip_mc_dec_group(in_dev, t->parms.iph.daddr);
973 	}
974 	return 0;
975 }
976 #endif
977 
978 static const struct net_device_ops ipgre_netdev_ops = {
979 	.ndo_init		= ipgre_tunnel_init,
980 	.ndo_uninit		= ip_tunnel_uninit,
981 #ifdef CONFIG_NET_IPGRE_BROADCAST
982 	.ndo_open		= ipgre_open,
983 	.ndo_stop		= ipgre_close,
984 #endif
985 	.ndo_start_xmit		= ipgre_xmit,
986 	.ndo_siocdevprivate	= ip_tunnel_siocdevprivate,
987 	.ndo_change_mtu		= ip_tunnel_change_mtu,
988 	.ndo_get_stats64	= dev_get_tstats64,
989 	.ndo_get_iflink		= ip_tunnel_get_iflink,
990 	.ndo_tunnel_ctl		= ipgre_tunnel_ctl,
991 };
992 
993 #define GRE_FEATURES (NETIF_F_SG |		\
994 		      NETIF_F_FRAGLIST |	\
995 		      NETIF_F_HIGHDMA |		\
996 		      NETIF_F_HW_CSUM)
997 
ipgre_tunnel_setup(struct net_device * dev)998 static void ipgre_tunnel_setup(struct net_device *dev)
999 {
1000 	dev->netdev_ops		= &ipgre_netdev_ops;
1001 	dev->type		= ARPHRD_IPGRE;
1002 	ip_tunnel_setup(dev, ipgre_net_id);
1003 }
1004 
__gre_tunnel_init(struct net_device * dev)1005 static void __gre_tunnel_init(struct net_device *dev)
1006 {
1007 	struct ip_tunnel *tunnel;
1008 
1009 	tunnel = netdev_priv(dev);
1010 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1011 	tunnel->parms.iph.protocol = IPPROTO_GRE;
1012 
1013 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1014 	dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
1015 
1016 	dev->features		|= GRE_FEATURES;
1017 	dev->hw_features	|= GRE_FEATURES;
1018 
1019 	/* TCP offload with GRE SEQ is not supported, nor can we support 2
1020 	 * levels of outer headers requiring an update.
1021 	 */
1022 	if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags))
1023 		return;
1024 	if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
1025 	    tunnel->encap.type != TUNNEL_ENCAP_NONE)
1026 		return;
1027 
1028 	dev->features |= NETIF_F_GSO_SOFTWARE;
1029 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1030 
1031 	dev->lltx = true;
1032 }
1033 
ipgre_tunnel_init(struct net_device * dev)1034 static int ipgre_tunnel_init(struct net_device *dev)
1035 {
1036 	struct ip_tunnel *tunnel = netdev_priv(dev);
1037 	struct iphdr *iph = &tunnel->parms.iph;
1038 
1039 	__gre_tunnel_init(dev);
1040 
1041 	__dev_addr_set(dev, &iph->saddr, 4);
1042 	memcpy(dev->broadcast, &iph->daddr, 4);
1043 
1044 	dev->flags		= IFF_NOARP;
1045 	netif_keep_dst(dev);
1046 	dev->addr_len		= 4;
1047 
1048 	if (iph->daddr && !tunnel->collect_md) {
1049 #ifdef CONFIG_NET_IPGRE_BROADCAST
1050 		if (ipv4_is_multicast(iph->daddr)) {
1051 			if (!iph->saddr)
1052 				return -EINVAL;
1053 			dev->flags = IFF_BROADCAST;
1054 			dev->header_ops = &ipgre_header_ops;
1055 			dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1056 			dev->needed_headroom = 0;
1057 		}
1058 #endif
1059 	} else if (!tunnel->collect_md) {
1060 		dev->header_ops = &ipgre_header_ops;
1061 		dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1062 		dev->needed_headroom = 0;
1063 	}
1064 
1065 	return ip_tunnel_init(dev);
1066 }
1067 
1068 static const struct gre_protocol ipgre_protocol = {
1069 	.handler     = gre_rcv,
1070 	.err_handler = gre_err,
1071 };
1072 
ipgre_init_net(struct net * net)1073 static int __net_init ipgre_init_net(struct net *net)
1074 {
1075 	return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1076 }
1077 
ipgre_exit_rtnl(struct net * net,struct list_head * dev_to_kill)1078 static void __net_exit ipgre_exit_rtnl(struct net *net,
1079 				       struct list_head *dev_to_kill)
1080 {
1081 	ip_tunnel_delete_net(net, ipgre_net_id, &ipgre_link_ops, dev_to_kill);
1082 }
1083 
1084 static struct pernet_operations ipgre_net_ops = {
1085 	.init = ipgre_init_net,
1086 	.exit_rtnl = ipgre_exit_rtnl,
1087 	.id   = &ipgre_net_id,
1088 	.size = sizeof(struct ip_tunnel_net),
1089 };
1090 
ipgre_tunnel_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1091 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1092 				 struct netlink_ext_ack *extack)
1093 {
1094 	__be16 flags;
1095 
1096 	if (!data)
1097 		return 0;
1098 
1099 	flags = 0;
1100 	if (data[IFLA_GRE_IFLAGS])
1101 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1102 	if (data[IFLA_GRE_OFLAGS])
1103 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1104 	if (flags & (GRE_VERSION|GRE_ROUTING))
1105 		return -EINVAL;
1106 
1107 	if (data[IFLA_GRE_COLLECT_METADATA] &&
1108 	    data[IFLA_GRE_ENCAP_TYPE] &&
1109 	    nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1110 		return -EINVAL;
1111 
1112 	return 0;
1113 }
1114 
ipgre_tap_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1115 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1116 			      struct netlink_ext_ack *extack)
1117 {
1118 	__be32 daddr;
1119 
1120 	if (tb[IFLA_ADDRESS]) {
1121 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1122 			return -EINVAL;
1123 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1124 			return -EADDRNOTAVAIL;
1125 	}
1126 
1127 	if (!data)
1128 		goto out;
1129 
1130 	if (data[IFLA_GRE_REMOTE]) {
1131 		memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1132 		if (!daddr)
1133 			return -EINVAL;
1134 	}
1135 
1136 out:
1137 	return ipgre_tunnel_validate(tb, data, extack);
1138 }
1139 
erspan_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1140 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1141 			   struct netlink_ext_ack *extack)
1142 {
1143 	__be16 flags = 0;
1144 	int ret;
1145 
1146 	if (!data)
1147 		return 0;
1148 
1149 	ret = ipgre_tap_validate(tb, data, extack);
1150 	if (ret)
1151 		return ret;
1152 
1153 	if (data[IFLA_GRE_ERSPAN_VER] &&
1154 	    nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1155 		return 0;
1156 
1157 	/* ERSPAN type II/III should only have GRE sequence and key flag */
1158 	if (data[IFLA_GRE_OFLAGS])
1159 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1160 	if (data[IFLA_GRE_IFLAGS])
1161 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1162 	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1163 	    flags != (GRE_SEQ | GRE_KEY))
1164 		return -EINVAL;
1165 
1166 	/* ERSPAN Session ID only has 10-bit. Since we reuse
1167 	 * 32-bit key field as ID, check it's range.
1168 	 */
1169 	if (data[IFLA_GRE_IKEY] &&
1170 	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1171 		return -EINVAL;
1172 
1173 	if (data[IFLA_GRE_OKEY] &&
1174 	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1175 		return -EINVAL;
1176 
1177 	return 0;
1178 }
1179 
ipgre_netlink_parms(struct net_device * dev,struct nlattr * data[],struct nlattr * tb[],struct ip_tunnel_parm_kern * parms,__u32 * fwmark)1180 static int ipgre_netlink_parms(struct net_device *dev,
1181 				struct nlattr *data[],
1182 				struct nlattr *tb[],
1183 				struct ip_tunnel_parm_kern *parms,
1184 				__u32 *fwmark)
1185 {
1186 	struct ip_tunnel *t = netdev_priv(dev);
1187 
1188 	memset(parms, 0, sizeof(*parms));
1189 
1190 	parms->iph.protocol = IPPROTO_GRE;
1191 
1192 	if (!data)
1193 		return 0;
1194 
1195 	if (data[IFLA_GRE_LINK])
1196 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1197 
1198 	if (data[IFLA_GRE_IFLAGS])
1199 		gre_flags_to_tnl_flags(parms->i_flags,
1200 				       nla_get_be16(data[IFLA_GRE_IFLAGS]));
1201 
1202 	if (data[IFLA_GRE_OFLAGS])
1203 		gre_flags_to_tnl_flags(parms->o_flags,
1204 				       nla_get_be16(data[IFLA_GRE_OFLAGS]));
1205 
1206 	if (data[IFLA_GRE_IKEY])
1207 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1208 
1209 	if (data[IFLA_GRE_OKEY])
1210 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1211 
1212 	if (data[IFLA_GRE_LOCAL])
1213 		parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1214 
1215 	if (data[IFLA_GRE_REMOTE])
1216 		parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1217 
1218 	if (data[IFLA_GRE_TTL])
1219 		parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1220 
1221 	if (data[IFLA_GRE_TOS])
1222 		parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1223 
1224 	if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1225 		if (t->ignore_df)
1226 			return -EINVAL;
1227 		parms->iph.frag_off = htons(IP_DF);
1228 	}
1229 
1230 	if (data[IFLA_GRE_COLLECT_METADATA]) {
1231 		t->collect_md = true;
1232 		if (dev->type == ARPHRD_IPGRE)
1233 			dev->type = ARPHRD_NONE;
1234 	}
1235 
1236 	if (data[IFLA_GRE_IGNORE_DF]) {
1237 		if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1238 		  && (parms->iph.frag_off & htons(IP_DF)))
1239 			return -EINVAL;
1240 		t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1241 	}
1242 
1243 	if (data[IFLA_GRE_FWMARK])
1244 		*fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1245 
1246 	return 0;
1247 }
1248 
erspan_netlink_parms(struct net_device * dev,struct nlattr * data[],struct nlattr * tb[],struct ip_tunnel_parm_kern * parms,__u32 * fwmark)1249 static int erspan_netlink_parms(struct net_device *dev,
1250 				struct nlattr *data[],
1251 				struct nlattr *tb[],
1252 				struct ip_tunnel_parm_kern *parms,
1253 				__u32 *fwmark)
1254 {
1255 	struct ip_tunnel *t = netdev_priv(dev);
1256 	int err;
1257 
1258 	err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1259 	if (err)
1260 		return err;
1261 	if (!data)
1262 		return 0;
1263 
1264 	if (data[IFLA_GRE_ERSPAN_VER]) {
1265 		t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1266 
1267 		if (t->erspan_ver > 2)
1268 			return -EINVAL;
1269 	}
1270 
1271 	if (t->erspan_ver == 1) {
1272 		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1273 			t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1274 			if (t->index & ~INDEX_MASK)
1275 				return -EINVAL;
1276 		}
1277 	} else if (t->erspan_ver == 2) {
1278 		if (data[IFLA_GRE_ERSPAN_DIR]) {
1279 			t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1280 			if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1281 				return -EINVAL;
1282 		}
1283 		if (data[IFLA_GRE_ERSPAN_HWID]) {
1284 			t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1285 			if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1286 				return -EINVAL;
1287 		}
1288 	}
1289 
1290 	return 0;
1291 }
1292 
1293 /* 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)1294 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1295 				      struct ip_tunnel_encap *ipencap)
1296 {
1297 	bool ret = false;
1298 
1299 	memset(ipencap, 0, sizeof(*ipencap));
1300 
1301 	if (!data)
1302 		return ret;
1303 
1304 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1305 		ret = true;
1306 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1307 	}
1308 
1309 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1310 		ret = true;
1311 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1312 	}
1313 
1314 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1315 		ret = true;
1316 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1317 	}
1318 
1319 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1320 		ret = true;
1321 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1322 	}
1323 
1324 	return ret;
1325 }
1326 
gre_tap_init(struct net_device * dev)1327 static int gre_tap_init(struct net_device *dev)
1328 {
1329 	__gre_tunnel_init(dev);
1330 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1331 	netif_keep_dst(dev);
1332 
1333 	return ip_tunnel_init(dev);
1334 }
1335 
1336 static const struct net_device_ops gre_tap_netdev_ops = {
1337 	.ndo_init		= gre_tap_init,
1338 	.ndo_uninit		= ip_tunnel_uninit,
1339 	.ndo_start_xmit		= gre_tap_xmit,
1340 	.ndo_set_mac_address 	= eth_mac_addr,
1341 	.ndo_validate_addr	= eth_validate_addr,
1342 	.ndo_change_mtu		= ip_tunnel_change_mtu,
1343 	.ndo_get_stats64	= dev_get_tstats64,
1344 	.ndo_get_iflink		= ip_tunnel_get_iflink,
1345 	.ndo_fill_metadata_dst	= gre_fill_metadata_dst,
1346 };
1347 
erspan_tunnel_init(struct net_device * dev)1348 static int erspan_tunnel_init(struct net_device *dev)
1349 {
1350 	struct ip_tunnel *tunnel = netdev_priv(dev);
1351 
1352 	if (tunnel->erspan_ver == 0)
1353 		tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1354 	else
1355 		tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1356 
1357 	tunnel->parms.iph.protocol = IPPROTO_GRE;
1358 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1359 		       erspan_hdr_len(tunnel->erspan_ver);
1360 
1361 	dev->features		|= GRE_FEATURES;
1362 	dev->hw_features	|= GRE_FEATURES;
1363 	dev->priv_flags		|= IFF_LIVE_ADDR_CHANGE;
1364 	netif_keep_dst(dev);
1365 
1366 	return ip_tunnel_init(dev);
1367 }
1368 
1369 static const struct net_device_ops erspan_netdev_ops = {
1370 	.ndo_init		= erspan_tunnel_init,
1371 	.ndo_uninit		= ip_tunnel_uninit,
1372 	.ndo_start_xmit		= erspan_xmit,
1373 	.ndo_set_mac_address	= eth_mac_addr,
1374 	.ndo_validate_addr	= eth_validate_addr,
1375 	.ndo_change_mtu		= ip_tunnel_change_mtu,
1376 	.ndo_get_stats64	= dev_get_tstats64,
1377 	.ndo_get_iflink		= ip_tunnel_get_iflink,
1378 	.ndo_fill_metadata_dst	= gre_fill_metadata_dst,
1379 };
1380 
ipgre_tap_setup(struct net_device * dev)1381 static void ipgre_tap_setup(struct net_device *dev)
1382 {
1383 	ether_setup(dev);
1384 	dev->max_mtu = 0;
1385 	dev->netdev_ops	= &gre_tap_netdev_ops;
1386 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1387 	dev->priv_flags	|= IFF_LIVE_ADDR_CHANGE;
1388 	ip_tunnel_setup(dev, gre_tap_net_id);
1389 }
1390 
1391 static int
ipgre_newlink_encap_setup(struct net_device * dev,struct nlattr * data[])1392 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1393 {
1394 	struct ip_tunnel_encap ipencap;
1395 
1396 	if (ipgre_netlink_encap_parms(data, &ipencap)) {
1397 		struct ip_tunnel *t = netdev_priv(dev);
1398 		int err = ip_tunnel_encap_setup(t, &ipencap);
1399 
1400 		if (err < 0)
1401 			return err;
1402 	}
1403 
1404 	return 0;
1405 }
1406 
ipgre_newlink(struct net_device * dev,struct rtnl_newlink_params * params,struct netlink_ext_ack * extack)1407 static int ipgre_newlink(struct net_device *dev,
1408 			 struct rtnl_newlink_params *params,
1409 			 struct netlink_ext_ack *extack)
1410 {
1411 	struct nlattr **data = params->data;
1412 	struct nlattr **tb = params->tb;
1413 	struct ip_tunnel_parm_kern p;
1414 	__u32 fwmark = 0;
1415 	int err;
1416 
1417 	err = ipgre_newlink_encap_setup(dev, data);
1418 	if (err)
1419 		return err;
1420 
1421 	err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1422 	if (err < 0)
1423 		return err;
1424 	return ip_tunnel_newlink(params->link_net ? : dev_net(dev), dev, tb, &p,
1425 				 fwmark);
1426 }
1427 
erspan_newlink(struct net_device * dev,struct rtnl_newlink_params * params,struct netlink_ext_ack * extack)1428 static int erspan_newlink(struct net_device *dev,
1429 			  struct rtnl_newlink_params *params,
1430 			  struct netlink_ext_ack *extack)
1431 {
1432 	struct nlattr **data = params->data;
1433 	struct nlattr **tb = params->tb;
1434 	struct ip_tunnel_parm_kern p;
1435 	__u32 fwmark = 0;
1436 	int err;
1437 
1438 	err = ipgre_newlink_encap_setup(dev, data);
1439 	if (err)
1440 		return err;
1441 
1442 	err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1443 	if (err)
1444 		return err;
1445 	return ip_tunnel_newlink(params->link_net ? : dev_net(dev), dev, tb, &p,
1446 				 fwmark);
1447 }
1448 
ipgre_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1449 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1450 			    struct nlattr *data[],
1451 			    struct netlink_ext_ack *extack)
1452 {
1453 	struct ip_tunnel *t = netdev_priv(dev);
1454 	struct ip_tunnel_parm_kern p;
1455 	__u32 fwmark = t->fwmark;
1456 	int err;
1457 
1458 	err = ipgre_newlink_encap_setup(dev, data);
1459 	if (err)
1460 		return err;
1461 
1462 	err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1463 	if (err < 0)
1464 		return err;
1465 
1466 	err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1467 	if (err < 0)
1468 		return err;
1469 
1470 	ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1471 	ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1472 
1473 	ipgre_link_update(dev, !tb[IFLA_MTU]);
1474 
1475 	return 0;
1476 }
1477 
erspan_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1478 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1479 			     struct nlattr *data[],
1480 			     struct netlink_ext_ack *extack)
1481 {
1482 	struct ip_tunnel *t = netdev_priv(dev);
1483 	struct ip_tunnel_parm_kern p;
1484 	__u32 fwmark = t->fwmark;
1485 	int err;
1486 
1487 	err = ipgre_newlink_encap_setup(dev, data);
1488 	if (err)
1489 		return err;
1490 
1491 	err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1492 	if (err < 0)
1493 		return err;
1494 
1495 	err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1496 	if (err < 0)
1497 		return err;
1498 
1499 	ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1500 	ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1501 
1502 	return 0;
1503 }
1504 
ipgre_get_size(const struct net_device * dev)1505 static size_t ipgre_get_size(const struct net_device *dev)
1506 {
1507 	return
1508 		/* IFLA_GRE_LINK */
1509 		nla_total_size(4) +
1510 		/* IFLA_GRE_IFLAGS */
1511 		nla_total_size(2) +
1512 		/* IFLA_GRE_OFLAGS */
1513 		nla_total_size(2) +
1514 		/* IFLA_GRE_IKEY */
1515 		nla_total_size(4) +
1516 		/* IFLA_GRE_OKEY */
1517 		nla_total_size(4) +
1518 		/* IFLA_GRE_LOCAL */
1519 		nla_total_size(4) +
1520 		/* IFLA_GRE_REMOTE */
1521 		nla_total_size(4) +
1522 		/* IFLA_GRE_TTL */
1523 		nla_total_size(1) +
1524 		/* IFLA_GRE_TOS */
1525 		nla_total_size(1) +
1526 		/* IFLA_GRE_PMTUDISC */
1527 		nla_total_size(1) +
1528 		/* IFLA_GRE_ENCAP_TYPE */
1529 		nla_total_size(2) +
1530 		/* IFLA_GRE_ENCAP_FLAGS */
1531 		nla_total_size(2) +
1532 		/* IFLA_GRE_ENCAP_SPORT */
1533 		nla_total_size(2) +
1534 		/* IFLA_GRE_ENCAP_DPORT */
1535 		nla_total_size(2) +
1536 		/* IFLA_GRE_COLLECT_METADATA */
1537 		nla_total_size(0) +
1538 		/* IFLA_GRE_IGNORE_DF */
1539 		nla_total_size(1) +
1540 		/* IFLA_GRE_FWMARK */
1541 		nla_total_size(4) +
1542 		/* IFLA_GRE_ERSPAN_INDEX */
1543 		nla_total_size(4) +
1544 		/* IFLA_GRE_ERSPAN_VER */
1545 		nla_total_size(1) +
1546 		/* IFLA_GRE_ERSPAN_DIR */
1547 		nla_total_size(1) +
1548 		/* IFLA_GRE_ERSPAN_HWID */
1549 		nla_total_size(2) +
1550 		0;
1551 }
1552 
ipgre_fill_info(struct sk_buff * skb,const struct net_device * dev)1553 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1554 {
1555 	struct ip_tunnel *t = netdev_priv(dev);
1556 	struct ip_tunnel_parm_kern *p = &t->parms;
1557 	IP_TUNNEL_DECLARE_FLAGS(o_flags);
1558 
1559 	ip_tunnel_flags_copy(o_flags, p->o_flags);
1560 
1561 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1562 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1563 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1564 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1565 			 gre_tnl_flags_to_gre_flags(o_flags)) ||
1566 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1567 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1568 	    nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1569 	    nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1570 	    nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1571 	    nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1572 	    nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1573 		       !!(p->iph.frag_off & htons(IP_DF))) ||
1574 	    nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1575 		goto nla_put_failure;
1576 
1577 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1578 			t->encap.type) ||
1579 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1580 			 t->encap.sport) ||
1581 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1582 			 t->encap.dport) ||
1583 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1584 			t->encap.flags))
1585 		goto nla_put_failure;
1586 
1587 	if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1588 		goto nla_put_failure;
1589 
1590 	if (t->collect_md) {
1591 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1592 			goto nla_put_failure;
1593 	}
1594 
1595 	return 0;
1596 
1597 nla_put_failure:
1598 	return -EMSGSIZE;
1599 }
1600 
erspan_fill_info(struct sk_buff * skb,const struct net_device * dev)1601 static int erspan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1602 {
1603 	struct ip_tunnel *t = netdev_priv(dev);
1604 
1605 	if (t->erspan_ver <= 2) {
1606 		if (t->erspan_ver != 0 && !t->collect_md)
1607 			__set_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags);
1608 
1609 		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1610 			goto nla_put_failure;
1611 
1612 		if (t->erspan_ver == 1) {
1613 			if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1614 				goto nla_put_failure;
1615 		} else if (t->erspan_ver == 2) {
1616 			if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1617 				goto nla_put_failure;
1618 			if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1619 				goto nla_put_failure;
1620 		}
1621 	}
1622 
1623 	return ipgre_fill_info(skb, dev);
1624 
1625 nla_put_failure:
1626 	return -EMSGSIZE;
1627 }
1628 
erspan_setup(struct net_device * dev)1629 static void erspan_setup(struct net_device *dev)
1630 {
1631 	struct ip_tunnel *t = netdev_priv(dev);
1632 
1633 	ether_setup(dev);
1634 	dev->max_mtu = 0;
1635 	dev->netdev_ops = &erspan_netdev_ops;
1636 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1637 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1638 	ip_tunnel_setup(dev, erspan_net_id);
1639 	t->erspan_ver = 1;
1640 }
1641 
1642 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1643 	[IFLA_GRE_LINK]		= { .type = NLA_U32 },
1644 	[IFLA_GRE_IFLAGS]	= { .type = NLA_U16 },
1645 	[IFLA_GRE_OFLAGS]	= { .type = NLA_U16 },
1646 	[IFLA_GRE_IKEY]		= { .type = NLA_U32 },
1647 	[IFLA_GRE_OKEY]		= { .type = NLA_U32 },
1648 	[IFLA_GRE_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
1649 	[IFLA_GRE_REMOTE]	= { .len = sizeof_field(struct iphdr, daddr) },
1650 	[IFLA_GRE_TTL]		= { .type = NLA_U8 },
1651 	[IFLA_GRE_TOS]		= { .type = NLA_U8 },
1652 	[IFLA_GRE_PMTUDISC]	= { .type = NLA_U8 },
1653 	[IFLA_GRE_ENCAP_TYPE]	= { .type = NLA_U16 },
1654 	[IFLA_GRE_ENCAP_FLAGS]	= { .type = NLA_U16 },
1655 	[IFLA_GRE_ENCAP_SPORT]	= { .type = NLA_U16 },
1656 	[IFLA_GRE_ENCAP_DPORT]	= { .type = NLA_U16 },
1657 	[IFLA_GRE_COLLECT_METADATA]	= { .type = NLA_FLAG },
1658 	[IFLA_GRE_IGNORE_DF]	= { .type = NLA_U8 },
1659 	[IFLA_GRE_FWMARK]	= { .type = NLA_U32 },
1660 	[IFLA_GRE_ERSPAN_INDEX]	= { .type = NLA_U32 },
1661 	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
1662 	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
1663 	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
1664 };
1665 
1666 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1667 	.kind		= "gre",
1668 	.maxtype	= IFLA_GRE_MAX,
1669 	.policy		= ipgre_policy,
1670 	.priv_size	= sizeof(struct ip_tunnel),
1671 	.setup		= ipgre_tunnel_setup,
1672 	.validate	= ipgre_tunnel_validate,
1673 	.newlink	= ipgre_newlink,
1674 	.changelink	= ipgre_changelink,
1675 	.dellink	= ip_tunnel_dellink,
1676 	.get_size	= ipgre_get_size,
1677 	.fill_info	= ipgre_fill_info,
1678 	.get_link_net	= ip_tunnel_get_link_net,
1679 };
1680 
1681 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1682 	.kind		= "gretap",
1683 	.maxtype	= IFLA_GRE_MAX,
1684 	.policy		= ipgre_policy,
1685 	.priv_size	= sizeof(struct ip_tunnel),
1686 	.setup		= ipgre_tap_setup,
1687 	.validate	= ipgre_tap_validate,
1688 	.newlink	= ipgre_newlink,
1689 	.changelink	= ipgre_changelink,
1690 	.dellink	= ip_tunnel_dellink,
1691 	.get_size	= ipgre_get_size,
1692 	.fill_info	= ipgre_fill_info,
1693 	.get_link_net	= ip_tunnel_get_link_net,
1694 };
1695 
1696 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1697 	.kind		= "erspan",
1698 	.maxtype	= IFLA_GRE_MAX,
1699 	.policy		= ipgre_policy,
1700 	.priv_size	= sizeof(struct ip_tunnel),
1701 	.setup		= erspan_setup,
1702 	.validate	= erspan_validate,
1703 	.newlink	= erspan_newlink,
1704 	.changelink	= erspan_changelink,
1705 	.dellink	= ip_tunnel_dellink,
1706 	.get_size	= ipgre_get_size,
1707 	.fill_info	= erspan_fill_info,
1708 	.get_link_net	= ip_tunnel_get_link_net,
1709 };
1710 
gretap_fb_dev_create(struct net * net,const char * name,u8 name_assign_type)1711 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1712 					u8 name_assign_type)
1713 {
1714 	struct rtnl_newlink_params params = { .src_net = net };
1715 	struct nlattr *tb[IFLA_MAX + 1];
1716 	struct net_device *dev;
1717 	LIST_HEAD(list_kill);
1718 	struct ip_tunnel *t;
1719 	int err;
1720 
1721 	memset(&tb, 0, sizeof(tb));
1722 	params.tb = tb;
1723 
1724 	dev = rtnl_create_link(net, name, name_assign_type,
1725 			       &ipgre_tap_ops, tb, NULL);
1726 	if (IS_ERR(dev))
1727 		return dev;
1728 
1729 	/* Configure flow based GRE device. */
1730 	t = netdev_priv(dev);
1731 	t->collect_md = true;
1732 
1733 	err = ipgre_newlink(dev, &params, NULL);
1734 	if (err < 0) {
1735 		free_netdev(dev);
1736 		return ERR_PTR(err);
1737 	}
1738 
1739 	/* openvswitch users expect packet sizes to be unrestricted,
1740 	 * so set the largest MTU we can.
1741 	 */
1742 	err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1743 	if (err)
1744 		goto out;
1745 
1746 	err = rtnl_configure_link(dev, NULL, 0, NULL);
1747 	if (err < 0)
1748 		goto out;
1749 
1750 	return dev;
1751 out:
1752 	ip_tunnel_dellink(dev, &list_kill);
1753 	unregister_netdevice_many(&list_kill);
1754 	return ERR_PTR(err);
1755 }
1756 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1757 
ipgre_tap_init_net(struct net * net)1758 static int __net_init ipgre_tap_init_net(struct net *net)
1759 {
1760 	return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1761 }
1762 
ipgre_tap_exit_rtnl(struct net * net,struct list_head * dev_to_kill)1763 static void __net_exit ipgre_tap_exit_rtnl(struct net *net,
1764 					   struct list_head *dev_to_kill)
1765 {
1766 	ip_tunnel_delete_net(net, gre_tap_net_id, &ipgre_tap_ops, dev_to_kill);
1767 }
1768 
1769 static struct pernet_operations ipgre_tap_net_ops = {
1770 	.init = ipgre_tap_init_net,
1771 	.exit_rtnl = ipgre_tap_exit_rtnl,
1772 	.id   = &gre_tap_net_id,
1773 	.size = sizeof(struct ip_tunnel_net),
1774 };
1775 
erspan_init_net(struct net * net)1776 static int __net_init erspan_init_net(struct net *net)
1777 {
1778 	return ip_tunnel_init_net(net, erspan_net_id,
1779 				  &erspan_link_ops, "erspan0");
1780 }
1781 
erspan_exit_rtnl(struct net * net,struct list_head * dev_to_kill)1782 static void __net_exit erspan_exit_rtnl(struct net *net,
1783 					struct list_head *dev_to_kill)
1784 {
1785 	ip_tunnel_delete_net(net, erspan_net_id, &erspan_link_ops, dev_to_kill);
1786 }
1787 
1788 static struct pernet_operations erspan_net_ops = {
1789 	.init = erspan_init_net,
1790 	.exit_rtnl = erspan_exit_rtnl,
1791 	.id   = &erspan_net_id,
1792 	.size = sizeof(struct ip_tunnel_net),
1793 };
1794 
ipgre_init(void)1795 static int __init ipgre_init(void)
1796 {
1797 	int err;
1798 
1799 	pr_info("GRE over IPv4 tunneling driver\n");
1800 
1801 	err = register_pernet_device(&ipgre_net_ops);
1802 	if (err < 0)
1803 		return err;
1804 
1805 	err = register_pernet_device(&ipgre_tap_net_ops);
1806 	if (err < 0)
1807 		goto pnet_tap_failed;
1808 
1809 	err = register_pernet_device(&erspan_net_ops);
1810 	if (err < 0)
1811 		goto pnet_erspan_failed;
1812 
1813 	err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1814 	if (err < 0) {
1815 		pr_info("%s: can't add protocol\n", __func__);
1816 		goto add_proto_failed;
1817 	}
1818 
1819 	err = rtnl_link_register(&ipgre_link_ops);
1820 	if (err < 0)
1821 		goto rtnl_link_failed;
1822 
1823 	err = rtnl_link_register(&ipgre_tap_ops);
1824 	if (err < 0)
1825 		goto tap_ops_failed;
1826 
1827 	err = rtnl_link_register(&erspan_link_ops);
1828 	if (err < 0)
1829 		goto erspan_link_failed;
1830 
1831 	return 0;
1832 
1833 erspan_link_failed:
1834 	rtnl_link_unregister(&ipgre_tap_ops);
1835 tap_ops_failed:
1836 	rtnl_link_unregister(&ipgre_link_ops);
1837 rtnl_link_failed:
1838 	gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1839 add_proto_failed:
1840 	unregister_pernet_device(&erspan_net_ops);
1841 pnet_erspan_failed:
1842 	unregister_pernet_device(&ipgre_tap_net_ops);
1843 pnet_tap_failed:
1844 	unregister_pernet_device(&ipgre_net_ops);
1845 	return err;
1846 }
1847 
ipgre_fini(void)1848 static void __exit ipgre_fini(void)
1849 {
1850 	rtnl_link_unregister(&ipgre_tap_ops);
1851 	rtnl_link_unregister(&ipgre_link_ops);
1852 	rtnl_link_unregister(&erspan_link_ops);
1853 	gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1854 	unregister_pernet_device(&ipgre_tap_net_ops);
1855 	unregister_pernet_device(&ipgre_net_ops);
1856 	unregister_pernet_device(&erspan_net_ops);
1857 }
1858 
1859 module_init(ipgre_init);
1860 module_exit(ipgre_fini);
1861 MODULE_DESCRIPTION("IPv4 GRE tunnels over IP library");
1862 MODULE_LICENSE("GPL");
1863 MODULE_ALIAS_RTNL_LINK("gre");
1864 MODULE_ALIAS_RTNL_LINK("gretap");
1865 MODULE_ALIAS_RTNL_LINK("erspan");
1866 MODULE_ALIAS_NETDEV("gre0");
1867 MODULE_ALIAS_NETDEV("gretap0");
1868 MODULE_ALIAS_NETDEV("erspan0");
1869