1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	GRE over IPv6 protocol decoder.
4  *
5  *	Authors: Dmitry Kozlov (xeb@mail.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/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32 
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/rtnetlink.h>
47 
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/gre.h>
53 #include <net/erspan.h>
54 #include <net/dst_metadata.h>
55 
56 
57 static bool log_ecn_error = true;
58 module_param(log_ecn_error, bool, 0644);
59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60 
61 #define IP6_GRE_HASH_SIZE_SHIFT  5
62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63 
64 static unsigned int ip6gre_net_id __read_mostly;
65 struct ip6gre_net {
66 	struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67 
68 	struct ip6_tnl __rcu *collect_md_tun;
69 	struct ip6_tnl __rcu *collect_md_tun_erspan;
70 	struct net_device *fb_tunnel_dev;
71 };
72 
73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81 
82 /* Tunnel hash table */
83 
84 /*
85    4 hash tables:
86 
87    3: (remote,local)
88    2: (remote,*)
89    1: (*,local)
90    0: (*,*)
91 
92    We require exact key match i.e. if a key is present in packet
93    it will match only tunnel with the same key; if it is not present,
94    it will match only keyless tunnel.
95 
96    All keysless packets, if not matched configured keyless tunnels
97    will match fallback tunnel.
98  */
99 
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
HASH_ADDR(const struct in6_addr * addr)101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103 	u32 hash = ipv6_addr_hash(addr);
104 
105 	return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106 }
107 
108 #define tunnels_r_l	tunnels[3]
109 #define tunnels_r	tunnels[2]
110 #define tunnels_l	tunnels[1]
111 #define tunnels_wc	tunnels[0]
112 
113 /* Given src, dst and key, find appropriate for input tunnel. */
114 
ip6gre_tunnel_lookup(struct net_device * dev,const struct in6_addr * remote,const struct in6_addr * local,__be32 key,__be16 gre_proto)115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116 		const struct in6_addr *remote, const struct in6_addr *local,
117 		__be32 key, __be16 gre_proto)
118 {
119 	struct net *net = dev_net(dev);
120 	int link = dev->ifindex;
121 	unsigned int h0 = HASH_ADDR(remote);
122 	unsigned int h1 = HASH_KEY(key);
123 	struct ip6_tnl *t, *cand = NULL;
124 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125 	int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126 			gre_proto == htons(ETH_P_ERSPAN) ||
127 			gre_proto == htons(ETH_P_ERSPAN2)) ?
128 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
129 	int score, cand_score = 4;
130 	struct net_device *ndev;
131 
132 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
133 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
134 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
135 		    key != t->parms.i_key ||
136 		    !(t->dev->flags & IFF_UP))
137 			continue;
138 
139 		if (t->dev->type != ARPHRD_IP6GRE &&
140 		    t->dev->type != dev_type)
141 			continue;
142 
143 		score = 0;
144 		if (t->parms.link != link)
145 			score |= 1;
146 		if (t->dev->type != dev_type)
147 			score |= 2;
148 		if (score == 0)
149 			return t;
150 
151 		if (score < cand_score) {
152 			cand = t;
153 			cand_score = score;
154 		}
155 	}
156 
157 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
158 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
159 		    key != t->parms.i_key ||
160 		    !(t->dev->flags & IFF_UP))
161 			continue;
162 
163 		if (t->dev->type != ARPHRD_IP6GRE &&
164 		    t->dev->type != dev_type)
165 			continue;
166 
167 		score = 0;
168 		if (t->parms.link != link)
169 			score |= 1;
170 		if (t->dev->type != dev_type)
171 			score |= 2;
172 		if (score == 0)
173 			return t;
174 
175 		if (score < cand_score) {
176 			cand = t;
177 			cand_score = score;
178 		}
179 	}
180 
181 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
182 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
183 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
184 				 !ipv6_addr_is_multicast(local))) ||
185 		    key != t->parms.i_key ||
186 		    !(t->dev->flags & IFF_UP))
187 			continue;
188 
189 		if (t->dev->type != ARPHRD_IP6GRE &&
190 		    t->dev->type != dev_type)
191 			continue;
192 
193 		score = 0;
194 		if (t->parms.link != link)
195 			score |= 1;
196 		if (t->dev->type != dev_type)
197 			score |= 2;
198 		if (score == 0)
199 			return t;
200 
201 		if (score < cand_score) {
202 			cand = t;
203 			cand_score = score;
204 		}
205 	}
206 
207 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
208 		if (t->parms.i_key != key ||
209 		    !(t->dev->flags & IFF_UP))
210 			continue;
211 
212 		if (t->dev->type != ARPHRD_IP6GRE &&
213 		    t->dev->type != dev_type)
214 			continue;
215 
216 		score = 0;
217 		if (t->parms.link != link)
218 			score |= 1;
219 		if (t->dev->type != dev_type)
220 			score |= 2;
221 		if (score == 0)
222 			return t;
223 
224 		if (score < cand_score) {
225 			cand = t;
226 			cand_score = score;
227 		}
228 	}
229 
230 	if (cand)
231 		return cand;
232 
233 	if (gre_proto == htons(ETH_P_ERSPAN) ||
234 	    gre_proto == htons(ETH_P_ERSPAN2))
235 		t = rcu_dereference(ign->collect_md_tun_erspan);
236 	else
237 		t = rcu_dereference(ign->collect_md_tun);
238 
239 	if (t && t->dev->flags & IFF_UP)
240 		return t;
241 
242 	ndev = READ_ONCE(ign->fb_tunnel_dev);
243 	if (ndev && ndev->flags & IFF_UP)
244 		return netdev_priv(ndev);
245 
246 	return NULL;
247 }
248 
__ip6gre_bucket(struct ip6gre_net * ign,const struct __ip6_tnl_parm * p)249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
250 		const struct __ip6_tnl_parm *p)
251 {
252 	const struct in6_addr *remote = &p->raddr;
253 	const struct in6_addr *local = &p->laddr;
254 	unsigned int h = HASH_KEY(p->i_key);
255 	int prio = 0;
256 
257 	if (!ipv6_addr_any(local))
258 		prio |= 1;
259 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
260 		prio |= 2;
261 		h ^= HASH_ADDR(remote);
262 	}
263 
264 	return &ign->tunnels[prio][h];
265 }
266 
ip6gre_tunnel_link_md(struct ip6gre_net * ign,struct ip6_tnl * t)267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269 	if (t->parms.collect_md)
270 		rcu_assign_pointer(ign->collect_md_tun, t);
271 }
272 
ip6erspan_tunnel_link_md(struct ip6gre_net * ign,struct ip6_tnl * t)273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
274 {
275 	if (t->parms.collect_md)
276 		rcu_assign_pointer(ign->collect_md_tun_erspan, t);
277 }
278 
ip6gre_tunnel_unlink_md(struct ip6gre_net * ign,struct ip6_tnl * t)279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
280 {
281 	if (t->parms.collect_md)
282 		rcu_assign_pointer(ign->collect_md_tun, NULL);
283 }
284 
ip6erspan_tunnel_unlink_md(struct ip6gre_net * ign,struct ip6_tnl * t)285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
286 				       struct ip6_tnl *t)
287 {
288 	if (t->parms.collect_md)
289 		rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
290 }
291 
ip6gre_bucket(struct ip6gre_net * ign,const struct ip6_tnl * t)292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
293 		const struct ip6_tnl *t)
294 {
295 	return __ip6gre_bucket(ign, &t->parms);
296 }
297 
ip6gre_tunnel_link(struct ip6gre_net * ign,struct ip6_tnl * t)298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
299 {
300 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
301 
302 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
303 	rcu_assign_pointer(*tp, t);
304 }
305 
ip6gre_tunnel_unlink(struct ip6gre_net * ign,struct ip6_tnl * t)306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
307 {
308 	struct ip6_tnl __rcu **tp;
309 	struct ip6_tnl *iter;
310 
311 	for (tp = ip6gre_bucket(ign, t);
312 	     (iter = rtnl_dereference(*tp)) != NULL;
313 	     tp = &iter->next) {
314 		if (t == iter) {
315 			rcu_assign_pointer(*tp, t->next);
316 			break;
317 		}
318 	}
319 }
320 
ip6gre_tunnel_find(struct net * net,const struct __ip6_tnl_parm * parms,int type)321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
322 					   const struct __ip6_tnl_parm *parms,
323 					   int type)
324 {
325 	const struct in6_addr *remote = &parms->raddr;
326 	const struct in6_addr *local = &parms->laddr;
327 	__be32 key = parms->i_key;
328 	int link = parms->link;
329 	struct ip6_tnl *t;
330 	struct ip6_tnl __rcu **tp;
331 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
332 
333 	for (tp = __ip6gre_bucket(ign, parms);
334 	     (t = rtnl_dereference(*tp)) != NULL;
335 	     tp = &t->next)
336 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
337 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
338 		    key == t->parms.i_key &&
339 		    link == t->parms.link &&
340 		    type == t->dev->type)
341 			break;
342 
343 	return t;
344 }
345 
ip6gre_tunnel_locate(struct net * net,const struct __ip6_tnl_parm * parms,int create)346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
347 		const struct __ip6_tnl_parm *parms, int create)
348 {
349 	struct ip6_tnl *t, *nt;
350 	struct net_device *dev;
351 	char name[IFNAMSIZ];
352 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
353 
354 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
355 	if (t && create)
356 		return NULL;
357 	if (t || !create)
358 		return t;
359 
360 	if (parms->name[0]) {
361 		if (!dev_valid_name(parms->name))
362 			return NULL;
363 		strscpy(name, parms->name, IFNAMSIZ);
364 	} else {
365 		strcpy(name, "ip6gre%d");
366 	}
367 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
368 			   ip6gre_tunnel_setup);
369 	if (!dev)
370 		return NULL;
371 
372 	dev_net_set(dev, net);
373 
374 	nt = netdev_priv(dev);
375 	nt->parms = *parms;
376 	dev->rtnl_link_ops = &ip6gre_link_ops;
377 
378 	nt->dev = dev;
379 	nt->net = dev_net(dev);
380 
381 	if (register_netdevice(dev) < 0)
382 		goto failed_free;
383 
384 	ip6gre_tnl_link_config(nt, 1);
385 	ip6gre_tunnel_link(ign, nt);
386 	return nt;
387 
388 failed_free:
389 	free_netdev(dev);
390 	return NULL;
391 }
392 
ip6erspan_tunnel_uninit(struct net_device * dev)393 static void ip6erspan_tunnel_uninit(struct net_device *dev)
394 {
395 	struct ip6_tnl *t = netdev_priv(dev);
396 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
397 
398 	ip6erspan_tunnel_unlink_md(ign, t);
399 	ip6gre_tunnel_unlink(ign, t);
400 	dst_cache_reset(&t->dst_cache);
401 	netdev_put(dev, &t->dev_tracker);
402 }
403 
ip6gre_tunnel_uninit(struct net_device * dev)404 static void ip6gre_tunnel_uninit(struct net_device *dev)
405 {
406 	struct ip6_tnl *t = netdev_priv(dev);
407 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
408 
409 	ip6gre_tunnel_unlink_md(ign, t);
410 	ip6gre_tunnel_unlink(ign, t);
411 	if (ign->fb_tunnel_dev == dev)
412 		WRITE_ONCE(ign->fb_tunnel_dev, NULL);
413 	dst_cache_reset(&t->dst_cache);
414 	netdev_put(dev, &t->dev_tracker);
415 }
416 
417 
ip6gre_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)418 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
419 		       u8 type, u8 code, int offset, __be32 info)
420 {
421 	struct net *net = dev_net(skb->dev);
422 	const struct ipv6hdr *ipv6h;
423 	struct tnl_ptk_info tpi;
424 	struct ip6_tnl *t;
425 
426 	if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
427 			     offset) < 0)
428 		return -EINVAL;
429 
430 	ipv6h = (const struct ipv6hdr *)skb->data;
431 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
432 				 tpi.key, tpi.proto);
433 	if (!t)
434 		return -ENOENT;
435 
436 	switch (type) {
437 	case ICMPV6_DEST_UNREACH:
438 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
439 				    t->parms.name);
440 		if (code != ICMPV6_PORT_UNREACH)
441 			break;
442 		return 0;
443 	case ICMPV6_TIME_EXCEED:
444 		if (code == ICMPV6_EXC_HOPLIMIT) {
445 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
446 					    t->parms.name);
447 			break;
448 		}
449 		return 0;
450 	case ICMPV6_PARAMPROB: {
451 		struct ipv6_tlv_tnl_enc_lim *tel;
452 		__u32 teli;
453 
454 		teli = 0;
455 		if (code == ICMPV6_HDR_FIELD)
456 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
457 
458 		if (teli && teli == be32_to_cpu(info) - 2) {
459 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
460 			if (tel->encap_limit == 0) {
461 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
462 						    t->parms.name);
463 			}
464 		} else {
465 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
466 					    t->parms.name);
467 		}
468 		return 0;
469 	}
470 	case ICMPV6_PKT_TOOBIG:
471 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
472 		return 0;
473 	case NDISC_REDIRECT:
474 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
475 			     sock_net_uid(net, NULL));
476 		return 0;
477 	}
478 
479 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
480 		t->err_count++;
481 	else
482 		t->err_count = 1;
483 	t->err_time = jiffies;
484 
485 	return 0;
486 }
487 
ip6gre_rcv(struct sk_buff * skb,const struct tnl_ptk_info * tpi)488 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
489 {
490 	const struct ipv6hdr *ipv6h;
491 	struct ip6_tnl *tunnel;
492 
493 	ipv6h = ipv6_hdr(skb);
494 	tunnel = ip6gre_tunnel_lookup(skb->dev,
495 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
496 				      tpi->proto);
497 	if (tunnel) {
498 		if (tunnel->parms.collect_md) {
499 			struct metadata_dst *tun_dst;
500 			__be64 tun_id;
501 			__be16 flags;
502 
503 			flags = tpi->flags;
504 			tun_id = key32_to_tunnel_id(tpi->key);
505 
506 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
507 			if (!tun_dst)
508 				return PACKET_REJECT;
509 
510 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
511 		} else {
512 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
513 		}
514 
515 		return PACKET_RCVD;
516 	}
517 
518 	return PACKET_REJECT;
519 }
520 
ip6erspan_rcv(struct sk_buff * skb,struct tnl_ptk_info * tpi,int gre_hdr_len)521 static int ip6erspan_rcv(struct sk_buff *skb,
522 			 struct tnl_ptk_info *tpi,
523 			 int gre_hdr_len)
524 {
525 	struct erspan_base_hdr *ershdr;
526 	const struct ipv6hdr *ipv6h;
527 	struct erspan_md2 *md2;
528 	struct ip6_tnl *tunnel;
529 	u8 ver;
530 
531 	ipv6h = ipv6_hdr(skb);
532 	ershdr = (struct erspan_base_hdr *)skb->data;
533 	ver = ershdr->ver;
534 
535 	tunnel = ip6gre_tunnel_lookup(skb->dev,
536 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
537 				      tpi->proto);
538 	if (tunnel) {
539 		int len = erspan_hdr_len(ver);
540 
541 		if (unlikely(!pskb_may_pull(skb, len)))
542 			return PACKET_REJECT;
543 
544 		if (__iptunnel_pull_header(skb, len,
545 					   htons(ETH_P_TEB),
546 					   false, false) < 0)
547 			return PACKET_REJECT;
548 
549 		if (tunnel->parms.collect_md) {
550 			struct erspan_metadata *pkt_md, *md;
551 			struct metadata_dst *tun_dst;
552 			struct ip_tunnel_info *info;
553 			unsigned char *gh;
554 			__be64 tun_id;
555 			__be16 flags;
556 
557 			tpi->flags |= TUNNEL_KEY;
558 			flags = tpi->flags;
559 			tun_id = key32_to_tunnel_id(tpi->key);
560 
561 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
562 						  sizeof(*md));
563 			if (!tun_dst)
564 				return PACKET_REJECT;
565 
566 			/* skb can be uncloned in __iptunnel_pull_header, so
567 			 * old pkt_md is no longer valid and we need to reset
568 			 * it
569 			 */
570 			gh = skb_network_header(skb) +
571 			     skb_network_header_len(skb);
572 			pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
573 							    sizeof(*ershdr));
574 			info = &tun_dst->u.tun_info;
575 			md = ip_tunnel_info_opts(info);
576 			md->version = ver;
577 			md2 = &md->u.md2;
578 			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
579 						       ERSPAN_V2_MDSIZE);
580 			info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
581 			info->options_len = sizeof(*md);
582 
583 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
584 
585 		} else {
586 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
587 		}
588 
589 		return PACKET_RCVD;
590 	}
591 
592 	return PACKET_REJECT;
593 }
594 
gre_rcv(struct sk_buff * skb)595 static int gre_rcv(struct sk_buff *skb)
596 {
597 	struct tnl_ptk_info tpi;
598 	bool csum_err = false;
599 	int hdr_len;
600 
601 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
602 	if (hdr_len < 0)
603 		goto drop;
604 
605 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
606 		goto drop;
607 
608 	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
609 		     tpi.proto == htons(ETH_P_ERSPAN2))) {
610 		if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
611 			return 0;
612 		goto out;
613 	}
614 
615 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
616 		return 0;
617 
618 out:
619 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
620 drop:
621 	kfree_skb(skb);
622 	return 0;
623 }
624 
gre_handle_offloads(struct sk_buff * skb,bool csum)625 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
626 {
627 	return iptunnel_handle_offloads(skb,
628 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
629 }
630 
prepare_ip6gre_xmit_ipv4(struct sk_buff * skb,struct net_device * dev,struct flowi6 * fl6,__u8 * dsfield,int * encap_limit)631 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
632 				     struct net_device *dev,
633 				     struct flowi6 *fl6, __u8 *dsfield,
634 				     int *encap_limit)
635 {
636 	const struct iphdr *iph = ip_hdr(skb);
637 	struct ip6_tnl *t = netdev_priv(dev);
638 
639 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
640 		*encap_limit = t->parms.encap_limit;
641 
642 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
643 
644 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
645 		*dsfield = ipv4_get_dsfield(iph);
646 	else
647 		*dsfield = ip6_tclass(t->parms.flowinfo);
648 
649 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
650 		fl6->flowi6_mark = skb->mark;
651 	else
652 		fl6->flowi6_mark = t->parms.fwmark;
653 
654 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
655 }
656 
prepare_ip6gre_xmit_ipv6(struct sk_buff * skb,struct net_device * dev,struct flowi6 * fl6,__u8 * dsfield,int * encap_limit)657 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
658 				    struct net_device *dev,
659 				    struct flowi6 *fl6, __u8 *dsfield,
660 				    int *encap_limit)
661 {
662 	struct ipv6hdr *ipv6h;
663 	struct ip6_tnl *t = netdev_priv(dev);
664 	__u16 offset;
665 
666 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
667 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
668 	ipv6h = ipv6_hdr(skb);
669 
670 	if (offset > 0) {
671 		struct ipv6_tlv_tnl_enc_lim *tel;
672 
673 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
674 		if (tel->encap_limit == 0) {
675 			icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
676 					ICMPV6_HDR_FIELD, offset + 2);
677 			return -1;
678 		}
679 		*encap_limit = tel->encap_limit - 1;
680 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
681 		*encap_limit = t->parms.encap_limit;
682 	}
683 
684 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
685 
686 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
687 		*dsfield = ipv6_get_dsfield(ipv6h);
688 	else
689 		*dsfield = ip6_tclass(t->parms.flowinfo);
690 
691 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
692 		fl6->flowlabel |= ip6_flowlabel(ipv6h);
693 
694 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
695 		fl6->flowi6_mark = skb->mark;
696 	else
697 		fl6->flowi6_mark = t->parms.fwmark;
698 
699 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
700 
701 	return 0;
702 }
703 
prepare_ip6gre_xmit_other(struct sk_buff * skb,struct net_device * dev,struct flowi6 * fl6,__u8 * dsfield,int * encap_limit)704 static int prepare_ip6gre_xmit_other(struct sk_buff *skb,
705 				     struct net_device *dev,
706 				     struct flowi6 *fl6, __u8 *dsfield,
707 				     int *encap_limit)
708 {
709 	struct ip6_tnl *t = netdev_priv(dev);
710 
711 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
712 		*encap_limit = t->parms.encap_limit;
713 
714 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
715 
716 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
717 		*dsfield = 0;
718 	else
719 		*dsfield = ip6_tclass(t->parms.flowinfo);
720 
721 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
722 		fl6->flowi6_mark = skb->mark;
723 	else
724 		fl6->flowi6_mark = t->parms.fwmark;
725 
726 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
727 
728 	return 0;
729 }
730 
skb_tunnel_info_txcheck(struct sk_buff * skb)731 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
732 {
733 	struct ip_tunnel_info *tun_info;
734 
735 	tun_info = skb_tunnel_info(skb);
736 	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
737 		return ERR_PTR(-EINVAL);
738 
739 	return tun_info;
740 }
741 
__gre6_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__be16 proto)742 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
743 			       struct net_device *dev, __u8 dsfield,
744 			       struct flowi6 *fl6, int encap_limit,
745 			       __u32 *pmtu, __be16 proto)
746 {
747 	struct ip6_tnl *tunnel = netdev_priv(dev);
748 	__be16 protocol;
749 	__be16 flags;
750 
751 	if (dev->type == ARPHRD_ETHER)
752 		IPCB(skb)->flags = 0;
753 
754 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
755 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
756 	else
757 		fl6->daddr = tunnel->parms.raddr;
758 
759 	/* Push GRE header. */
760 	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
761 
762 	if (tunnel->parms.collect_md) {
763 		struct ip_tunnel_info *tun_info;
764 		const struct ip_tunnel_key *key;
765 		int tun_hlen;
766 
767 		tun_info = skb_tunnel_info_txcheck(skb);
768 		if (IS_ERR(tun_info) ||
769 		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
770 			return -EINVAL;
771 
772 		key = &tun_info->key;
773 		memset(fl6, 0, sizeof(*fl6));
774 		fl6->flowi6_proto = IPPROTO_GRE;
775 		fl6->daddr = key->u.ipv6.dst;
776 		fl6->flowlabel = key->label;
777 		fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
778 		fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
779 
780 		dsfield = key->tos;
781 		flags = key->tun_flags &
782 			(TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
783 		tun_hlen = gre_calc_hlen(flags);
784 
785 		if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
786 			return -ENOMEM;
787 
788 		gre_build_header(skb, tun_hlen,
789 				 flags, protocol,
790 				 tunnel_id_to_key32(tun_info->key.tun_id),
791 				 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
792 						      : 0);
793 
794 	} else {
795 		if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
796 			return -ENOMEM;
797 
798 		flags = tunnel->parms.o_flags;
799 
800 		gre_build_header(skb, tunnel->tun_hlen, flags,
801 				 protocol, tunnel->parms.o_key,
802 				 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
803 						      : 0);
804 	}
805 
806 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
807 			    NEXTHDR_GRE);
808 }
809 
ip6gre_xmit_ipv4(struct sk_buff * skb,struct net_device * dev)810 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
811 {
812 	struct ip6_tnl *t = netdev_priv(dev);
813 	int encap_limit = -1;
814 	struct flowi6 fl6;
815 	__u8 dsfield = 0;
816 	__u32 mtu;
817 	int err;
818 
819 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
820 
821 	if (!t->parms.collect_md)
822 		prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
823 					 &dsfield, &encap_limit);
824 
825 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
826 	if (err)
827 		return -1;
828 
829 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
830 			  skb->protocol);
831 	if (err != 0) {
832 		/* XXX: send ICMP error even if DF is not set. */
833 		if (err == -EMSGSIZE)
834 			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
835 				      htonl(mtu));
836 		return -1;
837 	}
838 
839 	return 0;
840 }
841 
ip6gre_xmit_ipv6(struct sk_buff * skb,struct net_device * dev)842 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
843 {
844 	struct ip6_tnl *t = netdev_priv(dev);
845 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
846 	int encap_limit = -1;
847 	struct flowi6 fl6;
848 	__u8 dsfield = 0;
849 	__u32 mtu;
850 	int err;
851 
852 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
853 		return -1;
854 
855 	if (!t->parms.collect_md &&
856 	    prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
857 		return -1;
858 
859 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
860 		return -1;
861 
862 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
863 			  &mtu, skb->protocol);
864 	if (err != 0) {
865 		if (err == -EMSGSIZE)
866 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
867 		return -1;
868 	}
869 
870 	return 0;
871 }
872 
873 /**
874  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
875  *   @t: the outgoing tunnel device
876  *   @hdr: IPv6 header from the incoming packet
877  *
878  * Description:
879  *   Avoid trivial tunneling loop by checking that tunnel exit-point
880  *   doesn't match source of incoming packet.
881  *
882  * Return:
883  *   1 if conflict,
884  *   0 else
885  **/
886 
ip6gre_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)887 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
888 	const struct ipv6hdr *hdr)
889 {
890 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
891 }
892 
ip6gre_xmit_other(struct sk_buff * skb,struct net_device * dev)893 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
894 {
895 	struct ip6_tnl *t = netdev_priv(dev);
896 	int encap_limit = -1;
897 	struct flowi6 fl6;
898 	__u8 dsfield = 0;
899 	__u32 mtu;
900 	int err;
901 
902 	if (!t->parms.collect_md &&
903 	    prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit))
904 		return -1;
905 
906 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
907 	if (err)
908 		return err;
909 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol);
910 
911 	return err;
912 }
913 
ip6gre_tunnel_xmit(struct sk_buff * skb,struct net_device * dev)914 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
915 	struct net_device *dev)
916 {
917 	struct ip6_tnl *t = netdev_priv(dev);
918 	struct net_device_stats *stats = &t->dev->stats;
919 	__be16 payload_protocol;
920 	int ret;
921 
922 	if (!pskb_inet_may_pull(skb))
923 		goto tx_err;
924 
925 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
926 		goto tx_err;
927 
928 	payload_protocol = skb_protocol(skb, true);
929 	switch (payload_protocol) {
930 	case htons(ETH_P_IP):
931 		ret = ip6gre_xmit_ipv4(skb, dev);
932 		break;
933 	case htons(ETH_P_IPV6):
934 		ret = ip6gre_xmit_ipv6(skb, dev);
935 		break;
936 	default:
937 		ret = ip6gre_xmit_other(skb, dev);
938 		break;
939 	}
940 
941 	if (ret < 0)
942 		goto tx_err;
943 
944 	return NETDEV_TX_OK;
945 
946 tx_err:
947 	if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
948 		stats->tx_errors++;
949 	stats->tx_dropped++;
950 	kfree_skb(skb);
951 	return NETDEV_TX_OK;
952 }
953 
ip6erspan_tunnel_xmit(struct sk_buff * skb,struct net_device * dev)954 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
955 					 struct net_device *dev)
956 {
957 	struct ip_tunnel_info *tun_info = NULL;
958 	struct ip6_tnl *t = netdev_priv(dev);
959 	struct dst_entry *dst = skb_dst(skb);
960 	struct net_device_stats *stats;
961 	bool truncate = false;
962 	int encap_limit = -1;
963 	__u8 dsfield = false;
964 	struct flowi6 fl6;
965 	int err = -EINVAL;
966 	__be16 proto;
967 	__u32 mtu;
968 	int nhoff;
969 
970 	if (!pskb_inet_may_pull(skb))
971 		goto tx_err;
972 
973 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
974 		goto tx_err;
975 
976 	if (gre_handle_offloads(skb, false))
977 		goto tx_err;
978 
979 	if (skb->len > dev->mtu + dev->hard_header_len) {
980 		pskb_trim(skb, dev->mtu + dev->hard_header_len);
981 		truncate = true;
982 	}
983 
984 	nhoff = skb_network_header(skb) - skb_mac_header(skb);
985 	if (skb->protocol == htons(ETH_P_IP) &&
986 	    (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
987 		truncate = true;
988 
989 	if (skb->protocol == htons(ETH_P_IPV6)) {
990 		int thoff;
991 
992 		if (skb_transport_header_was_set(skb))
993 			thoff = skb_transport_header(skb) - skb_mac_header(skb);
994 		else
995 			thoff = nhoff + sizeof(struct ipv6hdr);
996 		if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
997 			truncate = true;
998 	}
999 
1000 	if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
1001 		goto tx_err;
1002 
1003 	t->parms.o_flags &= ~TUNNEL_KEY;
1004 	IPCB(skb)->flags = 0;
1005 
1006 	/* For collect_md mode, derive fl6 from the tunnel key,
1007 	 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
1008 	 */
1009 	if (t->parms.collect_md) {
1010 		const struct ip_tunnel_key *key;
1011 		struct erspan_metadata *md;
1012 		__be32 tun_id;
1013 
1014 		tun_info = skb_tunnel_info_txcheck(skb);
1015 		if (IS_ERR(tun_info) ||
1016 		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
1017 			goto tx_err;
1018 
1019 		key = &tun_info->key;
1020 		memset(&fl6, 0, sizeof(fl6));
1021 		fl6.flowi6_proto = IPPROTO_GRE;
1022 		fl6.daddr = key->u.ipv6.dst;
1023 		fl6.flowlabel = key->label;
1024 		fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1025 		fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1026 
1027 		dsfield = key->tos;
1028 		if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1029 			goto tx_err;
1030 		if (tun_info->options_len < sizeof(*md))
1031 			goto tx_err;
1032 		md = ip_tunnel_info_opts(tun_info);
1033 
1034 		tun_id = tunnel_id_to_key32(key->tun_id);
1035 		if (md->version == 1) {
1036 			erspan_build_header(skb,
1037 					    ntohl(tun_id),
1038 					    ntohl(md->u.index), truncate,
1039 					    false);
1040 		} else if (md->version == 2) {
1041 			erspan_build_header_v2(skb,
1042 					       ntohl(tun_id),
1043 					       md->u.md2.dir,
1044 					       get_hwid(&md->u.md2),
1045 					       truncate, false);
1046 		} else {
1047 			goto tx_err;
1048 		}
1049 	} else {
1050 		switch (skb->protocol) {
1051 		case htons(ETH_P_IP):
1052 			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1053 			prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1054 						 &dsfield, &encap_limit);
1055 			break;
1056 		case htons(ETH_P_IPV6):
1057 			if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1058 				goto tx_err;
1059 			if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1060 						     &dsfield, &encap_limit))
1061 				goto tx_err;
1062 			break;
1063 		default:
1064 			memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1065 			break;
1066 		}
1067 
1068 		if (t->parms.erspan_ver == 1)
1069 			erspan_build_header(skb, ntohl(t->parms.o_key),
1070 					    t->parms.index,
1071 					    truncate, false);
1072 		else if (t->parms.erspan_ver == 2)
1073 			erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1074 					       t->parms.dir,
1075 					       t->parms.hwid,
1076 					       truncate, false);
1077 		else
1078 			goto tx_err;
1079 
1080 		fl6.daddr = t->parms.raddr;
1081 	}
1082 
1083 	/* Push GRE header. */
1084 	proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1085 					   : htons(ETH_P_ERSPAN2);
1086 	gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1087 
1088 	/* TooBig packet may have updated dst->dev's mtu */
1089 	if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1090 		dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1091 
1092 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1093 			   NEXTHDR_GRE);
1094 	if (err != 0) {
1095 		/* XXX: send ICMP error even if DF is not set. */
1096 		if (err == -EMSGSIZE) {
1097 			if (skb->protocol == htons(ETH_P_IP))
1098 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1099 					      ICMP_FRAG_NEEDED, htonl(mtu));
1100 			else
1101 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1102 		}
1103 
1104 		goto tx_err;
1105 	}
1106 	return NETDEV_TX_OK;
1107 
1108 tx_err:
1109 	stats = &t->dev->stats;
1110 	if (!IS_ERR(tun_info))
1111 		stats->tx_errors++;
1112 	stats->tx_dropped++;
1113 	kfree_skb(skb);
1114 	return NETDEV_TX_OK;
1115 }
1116 
ip6gre_tnl_link_config_common(struct ip6_tnl * t)1117 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1118 {
1119 	struct net_device *dev = t->dev;
1120 	struct __ip6_tnl_parm *p = &t->parms;
1121 	struct flowi6 *fl6 = &t->fl.u.ip6;
1122 
1123 	if (dev->type != ARPHRD_ETHER) {
1124 		__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1125 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1126 	}
1127 
1128 	/* Set up flowi template */
1129 	fl6->saddr = p->laddr;
1130 	fl6->daddr = p->raddr;
1131 	fl6->flowi6_oif = p->link;
1132 	fl6->flowlabel = 0;
1133 	fl6->flowi6_proto = IPPROTO_GRE;
1134 	fl6->fl6_gre_key = t->parms.o_key;
1135 
1136 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1137 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1138 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1139 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1140 
1141 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1142 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1143 
1144 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1145 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1146 		dev->flags |= IFF_POINTOPOINT;
1147 	else
1148 		dev->flags &= ~IFF_POINTOPOINT;
1149 }
1150 
ip6gre_tnl_link_config_route(struct ip6_tnl * t,int set_mtu,int t_hlen)1151 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1152 					 int t_hlen)
1153 {
1154 	const struct __ip6_tnl_parm *p = &t->parms;
1155 	struct net_device *dev = t->dev;
1156 
1157 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1158 		int strict = (ipv6_addr_type(&p->raddr) &
1159 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1160 
1161 		struct rt6_info *rt = rt6_lookup(t->net,
1162 						 &p->raddr, &p->laddr,
1163 						 p->link, NULL, strict);
1164 
1165 		if (!rt)
1166 			return;
1167 
1168 		if (rt->dst.dev) {
1169 			unsigned short dst_len = rt->dst.dev->hard_header_len +
1170 						 t_hlen;
1171 
1172 			if (t->dev->header_ops)
1173 				dev->hard_header_len = dst_len;
1174 			else
1175 				dev->needed_headroom = dst_len;
1176 
1177 			if (set_mtu) {
1178 				int mtu = rt->dst.dev->mtu - t_hlen;
1179 
1180 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1181 					mtu -= 8;
1182 				if (dev->type == ARPHRD_ETHER)
1183 					mtu -= ETH_HLEN;
1184 
1185 				if (mtu < IPV6_MIN_MTU)
1186 					mtu = IPV6_MIN_MTU;
1187 				WRITE_ONCE(dev->mtu, mtu);
1188 			}
1189 		}
1190 		ip6_rt_put(rt);
1191 	}
1192 }
1193 
ip6gre_calc_hlen(struct ip6_tnl * tunnel)1194 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1195 {
1196 	int t_hlen;
1197 
1198 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1199 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1200 
1201 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1202 
1203 	if (tunnel->dev->header_ops)
1204 		tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1205 	else
1206 		tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1207 
1208 	return t_hlen;
1209 }
1210 
ip6gre_tnl_link_config(struct ip6_tnl * t,int set_mtu)1211 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1212 {
1213 	ip6gre_tnl_link_config_common(t);
1214 	ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1215 }
1216 
ip6gre_tnl_copy_tnl_parm(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1217 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1218 				     const struct __ip6_tnl_parm *p)
1219 {
1220 	t->parms.laddr = p->laddr;
1221 	t->parms.raddr = p->raddr;
1222 	t->parms.flags = p->flags;
1223 	t->parms.hop_limit = p->hop_limit;
1224 	t->parms.encap_limit = p->encap_limit;
1225 	t->parms.flowinfo = p->flowinfo;
1226 	t->parms.link = p->link;
1227 	t->parms.proto = p->proto;
1228 	t->parms.i_key = p->i_key;
1229 	t->parms.o_key = p->o_key;
1230 	t->parms.i_flags = p->i_flags;
1231 	t->parms.o_flags = p->o_flags;
1232 	t->parms.fwmark = p->fwmark;
1233 	t->parms.erspan_ver = p->erspan_ver;
1234 	t->parms.index = p->index;
1235 	t->parms.dir = p->dir;
1236 	t->parms.hwid = p->hwid;
1237 	dst_cache_reset(&t->dst_cache);
1238 }
1239 
ip6gre_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p,int set_mtu)1240 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1241 			     int set_mtu)
1242 {
1243 	ip6gre_tnl_copy_tnl_parm(t, p);
1244 	ip6gre_tnl_link_config(t, set_mtu);
1245 	return 0;
1246 }
1247 
ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm2 * u)1248 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1249 	const struct ip6_tnl_parm2 *u)
1250 {
1251 	p->laddr = u->laddr;
1252 	p->raddr = u->raddr;
1253 	p->flags = u->flags;
1254 	p->hop_limit = u->hop_limit;
1255 	p->encap_limit = u->encap_limit;
1256 	p->flowinfo = u->flowinfo;
1257 	p->link = u->link;
1258 	p->i_key = u->i_key;
1259 	p->o_key = u->o_key;
1260 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1261 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1262 	memcpy(p->name, u->name, sizeof(u->name));
1263 }
1264 
ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 * u,const struct __ip6_tnl_parm * p)1265 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1266 	const struct __ip6_tnl_parm *p)
1267 {
1268 	u->proto = IPPROTO_GRE;
1269 	u->laddr = p->laddr;
1270 	u->raddr = p->raddr;
1271 	u->flags = p->flags;
1272 	u->hop_limit = p->hop_limit;
1273 	u->encap_limit = p->encap_limit;
1274 	u->flowinfo = p->flowinfo;
1275 	u->link = p->link;
1276 	u->i_key = p->i_key;
1277 	u->o_key = p->o_key;
1278 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1279 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1280 	memcpy(u->name, p->name, sizeof(u->name));
1281 }
1282 
ip6gre_tunnel_siocdevprivate(struct net_device * dev,struct ifreq * ifr,void __user * data,int cmd)1283 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1284 					struct ifreq *ifr, void __user *data,
1285 					int cmd)
1286 {
1287 	int err = 0;
1288 	struct ip6_tnl_parm2 p;
1289 	struct __ip6_tnl_parm p1;
1290 	struct ip6_tnl *t = netdev_priv(dev);
1291 	struct net *net = t->net;
1292 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1293 
1294 	memset(&p1, 0, sizeof(p1));
1295 
1296 	switch (cmd) {
1297 	case SIOCGETTUNNEL:
1298 		if (dev == ign->fb_tunnel_dev) {
1299 			if (copy_from_user(&p, data, sizeof(p))) {
1300 				err = -EFAULT;
1301 				break;
1302 			}
1303 			ip6gre_tnl_parm_from_user(&p1, &p);
1304 			t = ip6gre_tunnel_locate(net, &p1, 0);
1305 			if (!t)
1306 				t = netdev_priv(dev);
1307 		}
1308 		memset(&p, 0, sizeof(p));
1309 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1310 		if (copy_to_user(data, &p, sizeof(p)))
1311 			err = -EFAULT;
1312 		break;
1313 
1314 	case SIOCADDTUNNEL:
1315 	case SIOCCHGTUNNEL:
1316 		err = -EPERM;
1317 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1318 			goto done;
1319 
1320 		err = -EFAULT;
1321 		if (copy_from_user(&p, data, sizeof(p)))
1322 			goto done;
1323 
1324 		err = -EINVAL;
1325 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1326 			goto done;
1327 
1328 		if (!(p.i_flags&GRE_KEY))
1329 			p.i_key = 0;
1330 		if (!(p.o_flags&GRE_KEY))
1331 			p.o_key = 0;
1332 
1333 		ip6gre_tnl_parm_from_user(&p1, &p);
1334 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1335 
1336 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1337 			if (t) {
1338 				if (t->dev != dev) {
1339 					err = -EEXIST;
1340 					break;
1341 				}
1342 			} else {
1343 				t = netdev_priv(dev);
1344 
1345 				ip6gre_tunnel_unlink(ign, t);
1346 				synchronize_net();
1347 				ip6gre_tnl_change(t, &p1, 1);
1348 				ip6gre_tunnel_link(ign, t);
1349 				netdev_state_change(dev);
1350 			}
1351 		}
1352 
1353 		if (t) {
1354 			err = 0;
1355 
1356 			memset(&p, 0, sizeof(p));
1357 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1358 			if (copy_to_user(data, &p, sizeof(p)))
1359 				err = -EFAULT;
1360 		} else
1361 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1362 		break;
1363 
1364 	case SIOCDELTUNNEL:
1365 		err = -EPERM;
1366 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1367 			goto done;
1368 
1369 		if (dev == ign->fb_tunnel_dev) {
1370 			err = -EFAULT;
1371 			if (copy_from_user(&p, data, sizeof(p)))
1372 				goto done;
1373 			err = -ENOENT;
1374 			ip6gre_tnl_parm_from_user(&p1, &p);
1375 			t = ip6gre_tunnel_locate(net, &p1, 0);
1376 			if (!t)
1377 				goto done;
1378 			err = -EPERM;
1379 			if (t == netdev_priv(ign->fb_tunnel_dev))
1380 				goto done;
1381 			dev = t->dev;
1382 		}
1383 		unregister_netdevice(dev);
1384 		err = 0;
1385 		break;
1386 
1387 	default:
1388 		err = -EINVAL;
1389 	}
1390 
1391 done:
1392 	return err;
1393 }
1394 
ip6gre_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)1395 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1396 			 unsigned short type, const void *daddr,
1397 			 const void *saddr, unsigned int len)
1398 {
1399 	struct ip6_tnl *t = netdev_priv(dev);
1400 	struct ipv6hdr *ipv6h;
1401 	__be16 *p;
1402 
1403 	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1404 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1405 						  t->fl.u.ip6.flowlabel,
1406 						  true, &t->fl.u.ip6));
1407 	ipv6h->hop_limit = t->parms.hop_limit;
1408 	ipv6h->nexthdr = NEXTHDR_GRE;
1409 	ipv6h->saddr = t->parms.laddr;
1410 	ipv6h->daddr = t->parms.raddr;
1411 
1412 	p = (__be16 *)(ipv6h + 1);
1413 	p[0] = t->parms.o_flags;
1414 	p[1] = htons(type);
1415 
1416 	/*
1417 	 *	Set the source hardware address.
1418 	 */
1419 
1420 	if (saddr)
1421 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1422 	if (daddr)
1423 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1424 	if (!ipv6_addr_any(&ipv6h->daddr))
1425 		return t->hlen;
1426 
1427 	return -t->hlen;
1428 }
1429 
1430 static const struct header_ops ip6gre_header_ops = {
1431 	.create	= ip6gre_header,
1432 };
1433 
1434 static const struct net_device_ops ip6gre_netdev_ops = {
1435 	.ndo_init		= ip6gre_tunnel_init,
1436 	.ndo_uninit		= ip6gre_tunnel_uninit,
1437 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1438 	.ndo_siocdevprivate	= ip6gre_tunnel_siocdevprivate,
1439 	.ndo_change_mtu		= ip6_tnl_change_mtu,
1440 	.ndo_get_stats64	= dev_get_tstats64,
1441 	.ndo_get_iflink		= ip6_tnl_get_iflink,
1442 };
1443 
ip6gre_dev_free(struct net_device * dev)1444 static void ip6gre_dev_free(struct net_device *dev)
1445 {
1446 	struct ip6_tnl *t = netdev_priv(dev);
1447 
1448 	gro_cells_destroy(&t->gro_cells);
1449 	dst_cache_destroy(&t->dst_cache);
1450 	free_percpu(dev->tstats);
1451 }
1452 
ip6gre_tunnel_setup(struct net_device * dev)1453 static void ip6gre_tunnel_setup(struct net_device *dev)
1454 {
1455 	dev->netdev_ops = &ip6gre_netdev_ops;
1456 	dev->needs_free_netdev = true;
1457 	dev->priv_destructor = ip6gre_dev_free;
1458 
1459 	dev->type = ARPHRD_IP6GRE;
1460 
1461 	dev->flags |= IFF_NOARP;
1462 	dev->addr_len = sizeof(struct in6_addr);
1463 	netif_keep_dst(dev);
1464 	/* This perm addr will be used as interface identifier by IPv6 */
1465 	dev->addr_assign_type = NET_ADDR_RANDOM;
1466 	eth_random_addr(dev->perm_addr);
1467 }
1468 
1469 #define GRE6_FEATURES (NETIF_F_SG |		\
1470 		       NETIF_F_FRAGLIST |	\
1471 		       NETIF_F_HIGHDMA |	\
1472 		       NETIF_F_HW_CSUM)
1473 
ip6gre_tnl_init_features(struct net_device * dev)1474 static void ip6gre_tnl_init_features(struct net_device *dev)
1475 {
1476 	struct ip6_tnl *nt = netdev_priv(dev);
1477 	__be16 flags;
1478 
1479 	dev->features		|= GRE6_FEATURES | NETIF_F_LLTX;
1480 	dev->hw_features	|= GRE6_FEATURES;
1481 
1482 	flags = nt->parms.o_flags;
1483 
1484 	/* TCP offload with GRE SEQ is not supported, nor can we support 2
1485 	 * levels of outer headers requiring an update.
1486 	 */
1487 	if (flags & TUNNEL_SEQ)
1488 		return;
1489 	if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE)
1490 		return;
1491 
1492 	dev->features |= NETIF_F_GSO_SOFTWARE;
1493 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1494 }
1495 
ip6gre_tunnel_init_common(struct net_device * dev)1496 static int ip6gre_tunnel_init_common(struct net_device *dev)
1497 {
1498 	struct ip6_tnl *tunnel;
1499 	int ret;
1500 	int t_hlen;
1501 
1502 	tunnel = netdev_priv(dev);
1503 
1504 	tunnel->dev = dev;
1505 	tunnel->net = dev_net(dev);
1506 	strcpy(tunnel->parms.name, dev->name);
1507 
1508 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1509 	if (!dev->tstats)
1510 		return -ENOMEM;
1511 
1512 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1513 	if (ret)
1514 		goto cleanup_alloc_pcpu_stats;
1515 
1516 	ret = gro_cells_init(&tunnel->gro_cells, dev);
1517 	if (ret)
1518 		goto cleanup_dst_cache_init;
1519 
1520 	t_hlen = ip6gre_calc_hlen(tunnel);
1521 	dev->mtu = ETH_DATA_LEN - t_hlen;
1522 	if (dev->type == ARPHRD_ETHER)
1523 		dev->mtu -= ETH_HLEN;
1524 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1525 		dev->mtu -= 8;
1526 
1527 	if (tunnel->parms.collect_md) {
1528 		netif_keep_dst(dev);
1529 	}
1530 	ip6gre_tnl_init_features(dev);
1531 
1532 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1533 	return 0;
1534 
1535 cleanup_dst_cache_init:
1536 	dst_cache_destroy(&tunnel->dst_cache);
1537 cleanup_alloc_pcpu_stats:
1538 	free_percpu(dev->tstats);
1539 	dev->tstats = NULL;
1540 	return ret;
1541 }
1542 
ip6gre_tunnel_init(struct net_device * dev)1543 static int ip6gre_tunnel_init(struct net_device *dev)
1544 {
1545 	struct ip6_tnl *tunnel;
1546 	int ret;
1547 
1548 	ret = ip6gre_tunnel_init_common(dev);
1549 	if (ret)
1550 		return ret;
1551 
1552 	tunnel = netdev_priv(dev);
1553 
1554 	if (tunnel->parms.collect_md)
1555 		return 0;
1556 
1557 	__dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1558 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1559 
1560 	if (ipv6_addr_any(&tunnel->parms.raddr))
1561 		dev->header_ops = &ip6gre_header_ops;
1562 
1563 	return 0;
1564 }
1565 
ip6gre_fb_tunnel_init(struct net_device * dev)1566 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1567 {
1568 	struct ip6_tnl *tunnel = netdev_priv(dev);
1569 
1570 	tunnel->dev = dev;
1571 	tunnel->net = dev_net(dev);
1572 	strcpy(tunnel->parms.name, dev->name);
1573 
1574 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1575 }
1576 
1577 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1578 	.handler     = gre_rcv,
1579 	.err_handler = ip6gre_err,
1580 	.flags       = INET6_PROTO_FINAL,
1581 };
1582 
ip6gre_destroy_tunnels(struct net * net,struct list_head * head)1583 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1584 {
1585 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1586 	struct net_device *dev, *aux;
1587 	int prio;
1588 
1589 	for_each_netdev_safe(net, dev, aux)
1590 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1591 		    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1592 		    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1593 			unregister_netdevice_queue(dev, head);
1594 
1595 	for (prio = 0; prio < 4; prio++) {
1596 		int h;
1597 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1598 			struct ip6_tnl *t;
1599 
1600 			t = rtnl_dereference(ign->tunnels[prio][h]);
1601 
1602 			while (t) {
1603 				/* If dev is in the same netns, it has already
1604 				 * been added to the list by the previous loop.
1605 				 */
1606 				if (!net_eq(dev_net(t->dev), net))
1607 					unregister_netdevice_queue(t->dev,
1608 								   head);
1609 				t = rtnl_dereference(t->next);
1610 			}
1611 		}
1612 	}
1613 }
1614 
ip6gre_init_net(struct net * net)1615 static int __net_init ip6gre_init_net(struct net *net)
1616 {
1617 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1618 	struct net_device *ndev;
1619 	int err;
1620 
1621 	if (!net_has_fallback_tunnels(net))
1622 		return 0;
1623 	ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1624 			    NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1625 	if (!ndev) {
1626 		err = -ENOMEM;
1627 		goto err_alloc_dev;
1628 	}
1629 	ign->fb_tunnel_dev = ndev;
1630 	dev_net_set(ign->fb_tunnel_dev, net);
1631 	/* FB netdevice is special: we have one, and only one per netns.
1632 	 * Allowing to move it to another netns is clearly unsafe.
1633 	 */
1634 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1635 
1636 
1637 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1638 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1639 
1640 	err = register_netdev(ign->fb_tunnel_dev);
1641 	if (err)
1642 		goto err_reg_dev;
1643 
1644 	rcu_assign_pointer(ign->tunnels_wc[0],
1645 			   netdev_priv(ign->fb_tunnel_dev));
1646 	return 0;
1647 
1648 err_reg_dev:
1649 	free_netdev(ndev);
1650 err_alloc_dev:
1651 	return err;
1652 }
1653 
ip6gre_exit_batch_net(struct list_head * net_list)1654 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1655 {
1656 	struct net *net;
1657 	LIST_HEAD(list);
1658 
1659 	rtnl_lock();
1660 	list_for_each_entry(net, net_list, exit_list)
1661 		ip6gre_destroy_tunnels(net, &list);
1662 	unregister_netdevice_many(&list);
1663 	rtnl_unlock();
1664 }
1665 
1666 static struct pernet_operations ip6gre_net_ops = {
1667 	.init = ip6gre_init_net,
1668 	.exit_batch = ip6gre_exit_batch_net,
1669 	.id   = &ip6gre_net_id,
1670 	.size = sizeof(struct ip6gre_net),
1671 };
1672 
ip6gre_tunnel_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1673 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1674 				  struct netlink_ext_ack *extack)
1675 {
1676 	__be16 flags;
1677 
1678 	if (!data)
1679 		return 0;
1680 
1681 	flags = 0;
1682 	if (data[IFLA_GRE_IFLAGS])
1683 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1684 	if (data[IFLA_GRE_OFLAGS])
1685 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1686 	if (flags & (GRE_VERSION|GRE_ROUTING))
1687 		return -EINVAL;
1688 
1689 	return 0;
1690 }
1691 
ip6gre_tap_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1692 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1693 			       struct netlink_ext_ack *extack)
1694 {
1695 	struct in6_addr daddr;
1696 
1697 	if (tb[IFLA_ADDRESS]) {
1698 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1699 			return -EINVAL;
1700 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1701 			return -EADDRNOTAVAIL;
1702 	}
1703 
1704 	if (!data)
1705 		goto out;
1706 
1707 	if (data[IFLA_GRE_REMOTE]) {
1708 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1709 		if (ipv6_addr_any(&daddr))
1710 			return -EINVAL;
1711 	}
1712 
1713 out:
1714 	return ip6gre_tunnel_validate(tb, data, extack);
1715 }
1716 
ip6erspan_tap_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1717 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1718 				  struct netlink_ext_ack *extack)
1719 {
1720 	__be16 flags = 0;
1721 	int ret, ver = 0;
1722 
1723 	if (!data)
1724 		return 0;
1725 
1726 	ret = ip6gre_tap_validate(tb, data, extack);
1727 	if (ret)
1728 		return ret;
1729 
1730 	/* ERSPAN should only have GRE sequence and key flag */
1731 	if (data[IFLA_GRE_OFLAGS])
1732 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1733 	if (data[IFLA_GRE_IFLAGS])
1734 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1735 	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1736 	    flags != (GRE_SEQ | GRE_KEY))
1737 		return -EINVAL;
1738 
1739 	/* ERSPAN Session ID only has 10-bit. Since we reuse
1740 	 * 32-bit key field as ID, check it's range.
1741 	 */
1742 	if (data[IFLA_GRE_IKEY] &&
1743 	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1744 		return -EINVAL;
1745 
1746 	if (data[IFLA_GRE_OKEY] &&
1747 	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1748 		return -EINVAL;
1749 
1750 	if (data[IFLA_GRE_ERSPAN_VER]) {
1751 		ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1752 		if (ver != 1 && ver != 2)
1753 			return -EINVAL;
1754 	}
1755 
1756 	if (ver == 1) {
1757 		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1758 			u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1759 
1760 			if (index & ~INDEX_MASK)
1761 				return -EINVAL;
1762 		}
1763 	} else if (ver == 2) {
1764 		if (data[IFLA_GRE_ERSPAN_DIR]) {
1765 			u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1766 
1767 			if (dir & ~(DIR_MASK >> DIR_OFFSET))
1768 				return -EINVAL;
1769 		}
1770 
1771 		if (data[IFLA_GRE_ERSPAN_HWID]) {
1772 			u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1773 
1774 			if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1775 				return -EINVAL;
1776 		}
1777 	}
1778 
1779 	return 0;
1780 }
1781 
ip6erspan_set_version(struct nlattr * data[],struct __ip6_tnl_parm * parms)1782 static void ip6erspan_set_version(struct nlattr *data[],
1783 				  struct __ip6_tnl_parm *parms)
1784 {
1785 	if (!data)
1786 		return;
1787 
1788 	parms->erspan_ver = 1;
1789 	if (data[IFLA_GRE_ERSPAN_VER])
1790 		parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1791 
1792 	if (parms->erspan_ver == 1) {
1793 		if (data[IFLA_GRE_ERSPAN_INDEX])
1794 			parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1795 	} else if (parms->erspan_ver == 2) {
1796 		if (data[IFLA_GRE_ERSPAN_DIR])
1797 			parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1798 		if (data[IFLA_GRE_ERSPAN_HWID])
1799 			parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1800 	}
1801 }
1802 
ip6gre_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1803 static void ip6gre_netlink_parms(struct nlattr *data[],
1804 				struct __ip6_tnl_parm *parms)
1805 {
1806 	memset(parms, 0, sizeof(*parms));
1807 
1808 	if (!data)
1809 		return;
1810 
1811 	if (data[IFLA_GRE_LINK])
1812 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1813 
1814 	if (data[IFLA_GRE_IFLAGS])
1815 		parms->i_flags = gre_flags_to_tnl_flags(
1816 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1817 
1818 	if (data[IFLA_GRE_OFLAGS])
1819 		parms->o_flags = gre_flags_to_tnl_flags(
1820 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1821 
1822 	if (data[IFLA_GRE_IKEY])
1823 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1824 
1825 	if (data[IFLA_GRE_OKEY])
1826 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1827 
1828 	if (data[IFLA_GRE_LOCAL])
1829 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1830 
1831 	if (data[IFLA_GRE_REMOTE])
1832 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1833 
1834 	if (data[IFLA_GRE_TTL])
1835 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1836 
1837 	if (data[IFLA_GRE_ENCAP_LIMIT])
1838 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1839 
1840 	if (data[IFLA_GRE_FLOWINFO])
1841 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1842 
1843 	if (data[IFLA_GRE_FLAGS])
1844 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1845 
1846 	if (data[IFLA_GRE_FWMARK])
1847 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1848 
1849 	if (data[IFLA_GRE_COLLECT_METADATA])
1850 		parms->collect_md = true;
1851 }
1852 
ip6gre_tap_init(struct net_device * dev)1853 static int ip6gre_tap_init(struct net_device *dev)
1854 {
1855 	int ret;
1856 
1857 	ret = ip6gre_tunnel_init_common(dev);
1858 	if (ret)
1859 		return ret;
1860 
1861 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1862 
1863 	return 0;
1864 }
1865 
1866 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1867 	.ndo_init = ip6gre_tap_init,
1868 	.ndo_uninit = ip6gre_tunnel_uninit,
1869 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1870 	.ndo_set_mac_address = eth_mac_addr,
1871 	.ndo_validate_addr = eth_validate_addr,
1872 	.ndo_change_mtu = ip6_tnl_change_mtu,
1873 	.ndo_get_stats64 = dev_get_tstats64,
1874 	.ndo_get_iflink = ip6_tnl_get_iflink,
1875 };
1876 
ip6erspan_calc_hlen(struct ip6_tnl * tunnel)1877 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1878 {
1879 	int t_hlen;
1880 
1881 	tunnel->tun_hlen = 8;
1882 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1883 		       erspan_hdr_len(tunnel->parms.erspan_ver);
1884 
1885 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1886 	tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1887 	return t_hlen;
1888 }
1889 
ip6erspan_tap_init(struct net_device * dev)1890 static int ip6erspan_tap_init(struct net_device *dev)
1891 {
1892 	struct ip6_tnl *tunnel;
1893 	int t_hlen;
1894 	int ret;
1895 
1896 	tunnel = netdev_priv(dev);
1897 
1898 	tunnel->dev = dev;
1899 	tunnel->net = dev_net(dev);
1900 	strcpy(tunnel->parms.name, dev->name);
1901 
1902 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1903 	if (!dev->tstats)
1904 		return -ENOMEM;
1905 
1906 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1907 	if (ret)
1908 		goto cleanup_alloc_pcpu_stats;
1909 
1910 	ret = gro_cells_init(&tunnel->gro_cells, dev);
1911 	if (ret)
1912 		goto cleanup_dst_cache_init;
1913 
1914 	t_hlen = ip6erspan_calc_hlen(tunnel);
1915 	dev->mtu = ETH_DATA_LEN - t_hlen;
1916 	if (dev->type == ARPHRD_ETHER)
1917 		dev->mtu -= ETH_HLEN;
1918 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1919 		dev->mtu -= 8;
1920 
1921 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1922 	ip6erspan_tnl_link_config(tunnel, 1);
1923 
1924 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1925 	return 0;
1926 
1927 cleanup_dst_cache_init:
1928 	dst_cache_destroy(&tunnel->dst_cache);
1929 cleanup_alloc_pcpu_stats:
1930 	free_percpu(dev->tstats);
1931 	dev->tstats = NULL;
1932 	return ret;
1933 }
1934 
1935 static const struct net_device_ops ip6erspan_netdev_ops = {
1936 	.ndo_init =		ip6erspan_tap_init,
1937 	.ndo_uninit =		ip6erspan_tunnel_uninit,
1938 	.ndo_start_xmit =	ip6erspan_tunnel_xmit,
1939 	.ndo_set_mac_address =	eth_mac_addr,
1940 	.ndo_validate_addr =	eth_validate_addr,
1941 	.ndo_change_mtu =	ip6_tnl_change_mtu,
1942 	.ndo_get_stats64 =	dev_get_tstats64,
1943 	.ndo_get_iflink =	ip6_tnl_get_iflink,
1944 };
1945 
ip6gre_tap_setup(struct net_device * dev)1946 static void ip6gre_tap_setup(struct net_device *dev)
1947 {
1948 
1949 	ether_setup(dev);
1950 
1951 	dev->max_mtu = 0;
1952 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1953 	dev->needs_free_netdev = true;
1954 	dev->priv_destructor = ip6gre_dev_free;
1955 
1956 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1957 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1958 	netif_keep_dst(dev);
1959 }
1960 
ip6gre_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)1961 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1962 				       struct ip_tunnel_encap *ipencap)
1963 {
1964 	bool ret = false;
1965 
1966 	memset(ipencap, 0, sizeof(*ipencap));
1967 
1968 	if (!data)
1969 		return ret;
1970 
1971 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1972 		ret = true;
1973 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1974 	}
1975 
1976 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1977 		ret = true;
1978 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1979 	}
1980 
1981 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1982 		ret = true;
1983 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1984 	}
1985 
1986 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1987 		ret = true;
1988 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1989 	}
1990 
1991 	return ret;
1992 }
1993 
ip6gre_newlink_common(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1994 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1995 				 struct nlattr *tb[], struct nlattr *data[],
1996 				 struct netlink_ext_ack *extack)
1997 {
1998 	struct ip6_tnl *nt;
1999 	struct ip_tunnel_encap ipencap;
2000 	int err;
2001 
2002 	nt = netdev_priv(dev);
2003 
2004 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2005 		int err = ip6_tnl_encap_setup(nt, &ipencap);
2006 
2007 		if (err < 0)
2008 			return err;
2009 	}
2010 
2011 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
2012 		eth_hw_addr_random(dev);
2013 
2014 	nt->dev = dev;
2015 	nt->net = dev_net(dev);
2016 
2017 	err = register_netdevice(dev);
2018 	if (err)
2019 		goto out;
2020 
2021 	if (tb[IFLA_MTU])
2022 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2023 
2024 out:
2025 	return err;
2026 }
2027 
ip6gre_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2028 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2029 			  struct nlattr *tb[], struct nlattr *data[],
2030 			  struct netlink_ext_ack *extack)
2031 {
2032 	struct ip6_tnl *nt = netdev_priv(dev);
2033 	struct net *net = dev_net(dev);
2034 	struct ip6gre_net *ign;
2035 	int err;
2036 
2037 	ip6gre_netlink_parms(data, &nt->parms);
2038 	ign = net_generic(net, ip6gre_net_id);
2039 
2040 	if (nt->parms.collect_md) {
2041 		if (rtnl_dereference(ign->collect_md_tun))
2042 			return -EEXIST;
2043 	} else {
2044 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2045 			return -EEXIST;
2046 	}
2047 
2048 	err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2049 	if (!err) {
2050 		ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2051 		ip6gre_tunnel_link_md(ign, nt);
2052 		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2053 	}
2054 	return err;
2055 }
2056 
2057 static struct ip6_tnl *
ip6gre_changelink_common(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct __ip6_tnl_parm * p_p,struct netlink_ext_ack * extack)2058 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2059 			 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2060 			 struct netlink_ext_ack *extack)
2061 {
2062 	struct ip6_tnl *t, *nt = netdev_priv(dev);
2063 	struct net *net = nt->net;
2064 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2065 	struct ip_tunnel_encap ipencap;
2066 
2067 	if (dev == ign->fb_tunnel_dev)
2068 		return ERR_PTR(-EINVAL);
2069 
2070 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2071 		int err = ip6_tnl_encap_setup(nt, &ipencap);
2072 
2073 		if (err < 0)
2074 			return ERR_PTR(err);
2075 	}
2076 
2077 	ip6gre_netlink_parms(data, p_p);
2078 
2079 	t = ip6gre_tunnel_locate(net, p_p, 0);
2080 
2081 	if (t) {
2082 		if (t->dev != dev)
2083 			return ERR_PTR(-EEXIST);
2084 	} else {
2085 		t = nt;
2086 	}
2087 
2088 	return t;
2089 }
2090 
ip6gre_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2091 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2092 			     struct nlattr *data[],
2093 			     struct netlink_ext_ack *extack)
2094 {
2095 	struct ip6_tnl *t = netdev_priv(dev);
2096 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2097 	struct __ip6_tnl_parm p;
2098 
2099 	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2100 	if (IS_ERR(t))
2101 		return PTR_ERR(t);
2102 
2103 	ip6gre_tunnel_unlink_md(ign, t);
2104 	ip6gre_tunnel_unlink(ign, t);
2105 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2106 	ip6gre_tunnel_link_md(ign, t);
2107 	ip6gre_tunnel_link(ign, t);
2108 	return 0;
2109 }
2110 
ip6gre_dellink(struct net_device * dev,struct list_head * head)2111 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2112 {
2113 	struct net *net = dev_net(dev);
2114 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2115 
2116 	if (dev != ign->fb_tunnel_dev)
2117 		unregister_netdevice_queue(dev, head);
2118 }
2119 
ip6gre_get_size(const struct net_device * dev)2120 static size_t ip6gre_get_size(const struct net_device *dev)
2121 {
2122 	return
2123 		/* IFLA_GRE_LINK */
2124 		nla_total_size(4) +
2125 		/* IFLA_GRE_IFLAGS */
2126 		nla_total_size(2) +
2127 		/* IFLA_GRE_OFLAGS */
2128 		nla_total_size(2) +
2129 		/* IFLA_GRE_IKEY */
2130 		nla_total_size(4) +
2131 		/* IFLA_GRE_OKEY */
2132 		nla_total_size(4) +
2133 		/* IFLA_GRE_LOCAL */
2134 		nla_total_size(sizeof(struct in6_addr)) +
2135 		/* IFLA_GRE_REMOTE */
2136 		nla_total_size(sizeof(struct in6_addr)) +
2137 		/* IFLA_GRE_TTL */
2138 		nla_total_size(1) +
2139 		/* IFLA_GRE_ENCAP_LIMIT */
2140 		nla_total_size(1) +
2141 		/* IFLA_GRE_FLOWINFO */
2142 		nla_total_size(4) +
2143 		/* IFLA_GRE_FLAGS */
2144 		nla_total_size(4) +
2145 		/* IFLA_GRE_ENCAP_TYPE */
2146 		nla_total_size(2) +
2147 		/* IFLA_GRE_ENCAP_FLAGS */
2148 		nla_total_size(2) +
2149 		/* IFLA_GRE_ENCAP_SPORT */
2150 		nla_total_size(2) +
2151 		/* IFLA_GRE_ENCAP_DPORT */
2152 		nla_total_size(2) +
2153 		/* IFLA_GRE_COLLECT_METADATA */
2154 		nla_total_size(0) +
2155 		/* IFLA_GRE_FWMARK */
2156 		nla_total_size(4) +
2157 		/* IFLA_GRE_ERSPAN_INDEX */
2158 		nla_total_size(4) +
2159 		0;
2160 }
2161 
ip6gre_fill_info(struct sk_buff * skb,const struct net_device * dev)2162 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2163 {
2164 	struct ip6_tnl *t = netdev_priv(dev);
2165 	struct __ip6_tnl_parm *p = &t->parms;
2166 	__be16 o_flags = p->o_flags;
2167 
2168 	if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2169 		if (!p->collect_md)
2170 			o_flags |= TUNNEL_KEY;
2171 
2172 		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2173 			goto nla_put_failure;
2174 
2175 		if (p->erspan_ver == 1) {
2176 			if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2177 				goto nla_put_failure;
2178 		} else {
2179 			if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2180 				goto nla_put_failure;
2181 			if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2182 				goto nla_put_failure;
2183 		}
2184 	}
2185 
2186 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2187 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
2188 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2189 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
2190 			 gre_tnl_flags_to_gre_flags(o_flags)) ||
2191 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2192 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2193 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2194 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2195 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2196 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2197 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2198 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2199 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2200 		goto nla_put_failure;
2201 
2202 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2203 			t->encap.type) ||
2204 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2205 			 t->encap.sport) ||
2206 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2207 			 t->encap.dport) ||
2208 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2209 			t->encap.flags))
2210 		goto nla_put_failure;
2211 
2212 	if (p->collect_md) {
2213 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2214 			goto nla_put_failure;
2215 	}
2216 
2217 	return 0;
2218 
2219 nla_put_failure:
2220 	return -EMSGSIZE;
2221 }
2222 
2223 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2224 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
2225 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2226 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2227 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2228 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2229 	[IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2230 	[IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2231 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
2232 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2233 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2234 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2235 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2236 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2237 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2238 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2239 	[IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2240 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2241 	[IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2242 	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
2243 	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
2244 	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
2245 };
2246 
ip6erspan_tap_setup(struct net_device * dev)2247 static void ip6erspan_tap_setup(struct net_device *dev)
2248 {
2249 	ether_setup(dev);
2250 
2251 	dev->max_mtu = 0;
2252 	dev->netdev_ops = &ip6erspan_netdev_ops;
2253 	dev->needs_free_netdev = true;
2254 	dev->priv_destructor = ip6gre_dev_free;
2255 
2256 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2257 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2258 	netif_keep_dst(dev);
2259 }
2260 
ip6erspan_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2261 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2262 			     struct nlattr *tb[], struct nlattr *data[],
2263 			     struct netlink_ext_ack *extack)
2264 {
2265 	struct ip6_tnl *nt = netdev_priv(dev);
2266 	struct net *net = dev_net(dev);
2267 	struct ip6gre_net *ign;
2268 	int err;
2269 
2270 	ip6gre_netlink_parms(data, &nt->parms);
2271 	ip6erspan_set_version(data, &nt->parms);
2272 	ign = net_generic(net, ip6gre_net_id);
2273 
2274 	if (nt->parms.collect_md) {
2275 		if (rtnl_dereference(ign->collect_md_tun_erspan))
2276 			return -EEXIST;
2277 	} else {
2278 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2279 			return -EEXIST;
2280 	}
2281 
2282 	err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2283 	if (!err) {
2284 		ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2285 		ip6erspan_tunnel_link_md(ign, nt);
2286 		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2287 	}
2288 	return err;
2289 }
2290 
ip6erspan_tnl_link_config(struct ip6_tnl * t,int set_mtu)2291 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2292 {
2293 	ip6gre_tnl_link_config_common(t);
2294 	ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2295 }
2296 
ip6erspan_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p,int set_mtu)2297 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2298 				const struct __ip6_tnl_parm *p, int set_mtu)
2299 {
2300 	ip6gre_tnl_copy_tnl_parm(t, p);
2301 	ip6erspan_tnl_link_config(t, set_mtu);
2302 	return 0;
2303 }
2304 
ip6erspan_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2305 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2306 				struct nlattr *data[],
2307 				struct netlink_ext_ack *extack)
2308 {
2309 	struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2310 	struct __ip6_tnl_parm p;
2311 	struct ip6_tnl *t;
2312 
2313 	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2314 	if (IS_ERR(t))
2315 		return PTR_ERR(t);
2316 
2317 	ip6erspan_set_version(data, &p);
2318 	ip6gre_tunnel_unlink_md(ign, t);
2319 	ip6gre_tunnel_unlink(ign, t);
2320 	ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2321 	ip6erspan_tunnel_link_md(ign, t);
2322 	ip6gre_tunnel_link(ign, t);
2323 	return 0;
2324 }
2325 
2326 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2327 	.kind		= "ip6gre",
2328 	.maxtype	= IFLA_GRE_MAX,
2329 	.policy		= ip6gre_policy,
2330 	.priv_size	= sizeof(struct ip6_tnl),
2331 	.setup		= ip6gre_tunnel_setup,
2332 	.validate	= ip6gre_tunnel_validate,
2333 	.newlink	= ip6gre_newlink,
2334 	.changelink	= ip6gre_changelink,
2335 	.dellink	= ip6gre_dellink,
2336 	.get_size	= ip6gre_get_size,
2337 	.fill_info	= ip6gre_fill_info,
2338 	.get_link_net	= ip6_tnl_get_link_net,
2339 };
2340 
2341 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2342 	.kind		= "ip6gretap",
2343 	.maxtype	= IFLA_GRE_MAX,
2344 	.policy		= ip6gre_policy,
2345 	.priv_size	= sizeof(struct ip6_tnl),
2346 	.setup		= ip6gre_tap_setup,
2347 	.validate	= ip6gre_tap_validate,
2348 	.newlink	= ip6gre_newlink,
2349 	.changelink	= ip6gre_changelink,
2350 	.get_size	= ip6gre_get_size,
2351 	.fill_info	= ip6gre_fill_info,
2352 	.get_link_net	= ip6_tnl_get_link_net,
2353 };
2354 
2355 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2356 	.kind		= "ip6erspan",
2357 	.maxtype	= IFLA_GRE_MAX,
2358 	.policy		= ip6gre_policy,
2359 	.priv_size	= sizeof(struct ip6_tnl),
2360 	.setup		= ip6erspan_tap_setup,
2361 	.validate	= ip6erspan_tap_validate,
2362 	.newlink	= ip6erspan_newlink,
2363 	.changelink	= ip6erspan_changelink,
2364 	.get_size	= ip6gre_get_size,
2365 	.fill_info	= ip6gre_fill_info,
2366 	.get_link_net	= ip6_tnl_get_link_net,
2367 };
2368 
2369 /*
2370  *	And now the modules code and kernel interface.
2371  */
2372 
ip6gre_init(void)2373 static int __init ip6gre_init(void)
2374 {
2375 	int err;
2376 
2377 	pr_info("GRE over IPv6 tunneling driver\n");
2378 
2379 	err = register_pernet_device(&ip6gre_net_ops);
2380 	if (err < 0)
2381 		return err;
2382 
2383 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2384 	if (err < 0) {
2385 		pr_info("%s: can't add protocol\n", __func__);
2386 		goto add_proto_failed;
2387 	}
2388 
2389 	err = rtnl_link_register(&ip6gre_link_ops);
2390 	if (err < 0)
2391 		goto rtnl_link_failed;
2392 
2393 	err = rtnl_link_register(&ip6gre_tap_ops);
2394 	if (err < 0)
2395 		goto tap_ops_failed;
2396 
2397 	err = rtnl_link_register(&ip6erspan_tap_ops);
2398 	if (err < 0)
2399 		goto erspan_link_failed;
2400 
2401 out:
2402 	return err;
2403 
2404 erspan_link_failed:
2405 	rtnl_link_unregister(&ip6gre_tap_ops);
2406 tap_ops_failed:
2407 	rtnl_link_unregister(&ip6gre_link_ops);
2408 rtnl_link_failed:
2409 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2410 add_proto_failed:
2411 	unregister_pernet_device(&ip6gre_net_ops);
2412 	goto out;
2413 }
2414 
ip6gre_fini(void)2415 static void __exit ip6gre_fini(void)
2416 {
2417 	rtnl_link_unregister(&ip6gre_tap_ops);
2418 	rtnl_link_unregister(&ip6gre_link_ops);
2419 	rtnl_link_unregister(&ip6erspan_tap_ops);
2420 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2421 	unregister_pernet_device(&ip6gre_net_ops);
2422 }
2423 
2424 module_init(ip6gre_init);
2425 module_exit(ip6gre_fini);
2426 MODULE_LICENSE("GPL");
2427 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2428 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2429 MODULE_ALIAS_RTNL_LINK("ip6gre");
2430 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2431 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2432 MODULE_ALIAS_NETDEV("ip6gre0");
2433