xref: /linux/net/ipv6/ip6_gre.c (revision ccd0ef38be3b9e0cc6efff407c53c43f648d598b)
1 /*
2  *	GRE over IPv6 protocol decoder.
3  *
4  *	Authors: Dmitry Kozlov (xeb@mail.ru)
5  *
6  *	This program is free software; you can redistribute it and/or
7  *	modify it under the terms of the GNU General Public License
8  *	as published by the Free Software Foundation; either version
9  *	2 of the License, or (at your option) any later version.
10  *
11  */
12 
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/mroute.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
35 #include <linux/hash.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38 
39 #include <net/sock.h>
40 #include <net/ip.h>
41 #include <net/ip_tunnels.h>
42 #include <net/icmp.h>
43 #include <net/protocol.h>
44 #include <net/addrconf.h>
45 #include <net/arp.h>
46 #include <net/checksum.h>
47 #include <net/dsfield.h>
48 #include <net/inet_ecn.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/rtnetlink.h>
53 
54 #include <net/ipv6.h>
55 #include <net/ip6_fib.h>
56 #include <net/ip6_route.h>
57 #include <net/ip6_tunnel.h>
58 
59 
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63 
64 #define HASH_SIZE_SHIFT  5
65 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
66 
67 static int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69 	struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
70 
71 	struct net_device *fb_tunnel_dev;
72 };
73 
74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75 static struct rtnl_link_ops ip6gre_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 
81 /* Tunnel hash table */
82 
83 /*
84    4 hash tables:
85 
86    3: (remote,local)
87    2: (remote,*)
88    1: (*,local)
89    0: (*,*)
90 
91    We require exact key match i.e. if a key is present in packet
92    it will match only tunnel with the same key; if it is not present,
93    it will match only keyless tunnel.
94 
95    All keysless packets, if not matched configured keyless tunnels
96    will match fallback tunnel.
97  */
98 
99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
100 static u32 HASH_ADDR(const struct in6_addr *addr)
101 {
102 	u32 hash = ipv6_addr_hash(addr);
103 
104 	return hash_32(hash, HASH_SIZE_SHIFT);
105 }
106 
107 #define tunnels_r_l	tunnels[3]
108 #define tunnels_r	tunnels[2]
109 #define tunnels_l	tunnels[1]
110 #define tunnels_wc	tunnels[0]
111 
112 /* Given src, dst and key, find appropriate for input tunnel. */
113 
114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
115 		const struct in6_addr *remote, const struct in6_addr *local,
116 		__be32 key, __be16 gre_proto)
117 {
118 	struct net *net = dev_net(dev);
119 	int link = dev->ifindex;
120 	unsigned int h0 = HASH_ADDR(remote);
121 	unsigned int h1 = HASH_KEY(key);
122 	struct ip6_tnl *t, *cand = NULL;
123 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
124 	int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
125 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
126 	int score, cand_score = 4;
127 
128 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
129 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
130 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
131 		    key != t->parms.i_key ||
132 		    !(t->dev->flags & IFF_UP))
133 			continue;
134 
135 		if (t->dev->type != ARPHRD_IP6GRE &&
136 		    t->dev->type != dev_type)
137 			continue;
138 
139 		score = 0;
140 		if (t->parms.link != link)
141 			score |= 1;
142 		if (t->dev->type != dev_type)
143 			score |= 2;
144 		if (score == 0)
145 			return t;
146 
147 		if (score < cand_score) {
148 			cand = t;
149 			cand_score = score;
150 		}
151 	}
152 
153 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
154 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
155 		    key != t->parms.i_key ||
156 		    !(t->dev->flags & IFF_UP))
157 			continue;
158 
159 		if (t->dev->type != ARPHRD_IP6GRE &&
160 		    t->dev->type != dev_type)
161 			continue;
162 
163 		score = 0;
164 		if (t->parms.link != link)
165 			score |= 1;
166 		if (t->dev->type != dev_type)
167 			score |= 2;
168 		if (score == 0)
169 			return t;
170 
171 		if (score < cand_score) {
172 			cand = t;
173 			cand_score = score;
174 		}
175 	}
176 
177 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
178 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
179 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
180 				 !ipv6_addr_is_multicast(local))) ||
181 		    key != t->parms.i_key ||
182 		    !(t->dev->flags & IFF_UP))
183 			continue;
184 
185 		if (t->dev->type != ARPHRD_IP6GRE &&
186 		    t->dev->type != dev_type)
187 			continue;
188 
189 		score = 0;
190 		if (t->parms.link != link)
191 			score |= 1;
192 		if (t->dev->type != dev_type)
193 			score |= 2;
194 		if (score == 0)
195 			return t;
196 
197 		if (score < cand_score) {
198 			cand = t;
199 			cand_score = score;
200 		}
201 	}
202 
203 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
204 		if (t->parms.i_key != key ||
205 		    !(t->dev->flags & IFF_UP))
206 			continue;
207 
208 		if (t->dev->type != ARPHRD_IP6GRE &&
209 		    t->dev->type != dev_type)
210 			continue;
211 
212 		score = 0;
213 		if (t->parms.link != link)
214 			score |= 1;
215 		if (t->dev->type != dev_type)
216 			score |= 2;
217 		if (score == 0)
218 			return t;
219 
220 		if (score < cand_score) {
221 			cand = t;
222 			cand_score = score;
223 		}
224 	}
225 
226 	if (cand != NULL)
227 		return cand;
228 
229 	dev = ign->fb_tunnel_dev;
230 	if (dev->flags & IFF_UP)
231 		return netdev_priv(dev);
232 
233 	return NULL;
234 }
235 
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237 		const struct __ip6_tnl_parm *p)
238 {
239 	const struct in6_addr *remote = &p->raddr;
240 	const struct in6_addr *local = &p->laddr;
241 	unsigned int h = HASH_KEY(p->i_key);
242 	int prio = 0;
243 
244 	if (!ipv6_addr_any(local))
245 		prio |= 1;
246 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
247 		prio |= 2;
248 		h ^= HASH_ADDR(remote);
249 	}
250 
251 	return &ign->tunnels[prio][h];
252 }
253 
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255 		const struct ip6_tnl *t)
256 {
257 	return __ip6gre_bucket(ign, &t->parms);
258 }
259 
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
261 {
262 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
263 
264 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265 	rcu_assign_pointer(*tp, t);
266 }
267 
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
269 {
270 	struct ip6_tnl __rcu **tp;
271 	struct ip6_tnl *iter;
272 
273 	for (tp = ip6gre_bucket(ign, t);
274 	     (iter = rtnl_dereference(*tp)) != NULL;
275 	     tp = &iter->next) {
276 		if (t == iter) {
277 			rcu_assign_pointer(*tp, t->next);
278 			break;
279 		}
280 	}
281 }
282 
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 					   const struct __ip6_tnl_parm *parms,
285 					   int type)
286 {
287 	const struct in6_addr *remote = &parms->raddr;
288 	const struct in6_addr *local = &parms->laddr;
289 	__be32 key = parms->i_key;
290 	int link = parms->link;
291 	struct ip6_tnl *t;
292 	struct ip6_tnl __rcu **tp;
293 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
294 
295 	for (tp = __ip6gre_bucket(ign, parms);
296 	     (t = rtnl_dereference(*tp)) != NULL;
297 	     tp = &t->next)
298 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
299 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
300 		    key == t->parms.i_key &&
301 		    link == t->parms.link &&
302 		    type == t->dev->type)
303 			break;
304 
305 	return t;
306 }
307 
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309 		const struct __ip6_tnl_parm *parms, int create)
310 {
311 	struct ip6_tnl *t, *nt;
312 	struct net_device *dev;
313 	char name[IFNAMSIZ];
314 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315 
316 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317 	if (t && create)
318 		return NULL;
319 	if (t || !create)
320 		return t;
321 
322 	if (parms->name[0])
323 		strlcpy(name, parms->name, IFNAMSIZ);
324 	else
325 		strcpy(name, "ip6gre%d");
326 
327 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328 			   ip6gre_tunnel_setup);
329 	if (!dev)
330 		return NULL;
331 
332 	dev_net_set(dev, net);
333 
334 	nt = netdev_priv(dev);
335 	nt->parms = *parms;
336 	dev->rtnl_link_ops = &ip6gre_link_ops;
337 
338 	nt->dev = dev;
339 	nt->net = dev_net(dev);
340 	ip6gre_tnl_link_config(nt, 1);
341 
342 	if (register_netdevice(dev) < 0)
343 		goto failed_free;
344 
345 	/* Can use a lockless transmit, unless we generate output sequences */
346 	if (!(nt->parms.o_flags & GRE_SEQ))
347 		dev->features |= NETIF_F_LLTX;
348 
349 	dev_hold(dev);
350 	ip6gre_tunnel_link(ign, nt);
351 	return nt;
352 
353 failed_free:
354 	free_netdev(dev);
355 	return NULL;
356 }
357 
358 static void ip6gre_tunnel_uninit(struct net_device *dev)
359 {
360 	struct ip6_tnl *t = netdev_priv(dev);
361 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
362 
363 	ip6gre_tunnel_unlink(ign, t);
364 	dev_put(dev);
365 }
366 
367 
368 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
369 		u8 type, u8 code, int offset, __be32 info)
370 {
371 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
372 	__be16 *p = (__be16 *)(skb->data + offset);
373 	int grehlen = offset + 4;
374 	struct ip6_tnl *t;
375 	__be16 flags;
376 
377 	flags = p[0];
378 	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
379 		if (flags&(GRE_VERSION|GRE_ROUTING))
380 			return;
381 		if (flags&GRE_KEY) {
382 			grehlen += 4;
383 			if (flags&GRE_CSUM)
384 				grehlen += 4;
385 		}
386 	}
387 
388 	/* If only 8 bytes returned, keyed message will be dropped here */
389 	if (!pskb_may_pull(skb, grehlen))
390 		return;
391 	ipv6h = (const struct ipv6hdr *)skb->data;
392 	p = (__be16 *)(skb->data + offset);
393 
394 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
395 				flags & GRE_KEY ?
396 				*(((__be32 *)p) + (grehlen / 4) - 1) : 0,
397 				p[1]);
398 	if (t == NULL)
399 		return;
400 
401 	switch (type) {
402 		__u32 teli;
403 		struct ipv6_tlv_tnl_enc_lim *tel;
404 		__u32 mtu;
405 	case ICMPV6_DEST_UNREACH:
406 		net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
407 				     t->parms.name);
408 		break;
409 	case ICMPV6_TIME_EXCEED:
410 		if (code == ICMPV6_EXC_HOPLIMIT) {
411 			net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
412 					     t->parms.name);
413 		}
414 		break;
415 	case ICMPV6_PARAMPROB:
416 		teli = 0;
417 		if (code == ICMPV6_HDR_FIELD)
418 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
419 
420 		if (teli && teli == info - 2) {
421 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
422 			if (tel->encap_limit == 0) {
423 				net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
424 						     t->parms.name);
425 			}
426 		} else {
427 			net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
428 					     t->parms.name);
429 		}
430 		break;
431 	case ICMPV6_PKT_TOOBIG:
432 		mtu = info - offset;
433 		if (mtu < IPV6_MIN_MTU)
434 			mtu = IPV6_MIN_MTU;
435 		t->dev->mtu = mtu;
436 		break;
437 	}
438 
439 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
440 		t->err_count++;
441 	else
442 		t->err_count = 1;
443 	t->err_time = jiffies;
444 }
445 
446 static int ip6gre_rcv(struct sk_buff *skb)
447 {
448 	const struct ipv6hdr *ipv6h;
449 	u8     *h;
450 	__be16    flags;
451 	__sum16   csum = 0;
452 	__be32 key = 0;
453 	u32    seqno = 0;
454 	struct ip6_tnl *tunnel;
455 	int    offset = 4;
456 	__be16 gre_proto;
457 	int err;
458 
459 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
460 		goto drop;
461 
462 	ipv6h = ipv6_hdr(skb);
463 	h = skb->data;
464 	flags = *(__be16 *)h;
465 
466 	if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
467 		/* - Version must be 0.
468 		   - We do not support routing headers.
469 		 */
470 		if (flags&(GRE_VERSION|GRE_ROUTING))
471 			goto drop;
472 
473 		if (flags&GRE_CSUM) {
474 			csum = skb_checksum_simple_validate(skb);
475 			offset += 4;
476 		}
477 		if (flags&GRE_KEY) {
478 			key = *(__be32 *)(h + offset);
479 			offset += 4;
480 		}
481 		if (flags&GRE_SEQ) {
482 			seqno = ntohl(*(__be32 *)(h + offset));
483 			offset += 4;
484 		}
485 	}
486 
487 	gre_proto = *(__be16 *)(h + 2);
488 
489 	tunnel = ip6gre_tunnel_lookup(skb->dev,
490 					  &ipv6h->saddr, &ipv6h->daddr, key,
491 					  gre_proto);
492 	if (tunnel) {
493 		struct pcpu_sw_netstats *tstats;
494 
495 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
496 			goto drop;
497 
498 		if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
499 			tunnel->dev->stats.rx_dropped++;
500 			goto drop;
501 		}
502 
503 		skb->protocol = gre_proto;
504 		/* WCCP version 1 and 2 protocol decoding.
505 		 * - Change protocol to IP
506 		 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
507 		 */
508 		if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
509 			skb->protocol = htons(ETH_P_IP);
510 			if ((*(h + offset) & 0xF0) != 0x40)
511 				offset += 4;
512 		}
513 
514 		skb->mac_header = skb->network_header;
515 		__pskb_pull(skb, offset);
516 		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
517 
518 		if (((flags&GRE_CSUM) && csum) ||
519 		    (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
520 			tunnel->dev->stats.rx_crc_errors++;
521 			tunnel->dev->stats.rx_errors++;
522 			goto drop;
523 		}
524 		if (tunnel->parms.i_flags&GRE_SEQ) {
525 			if (!(flags&GRE_SEQ) ||
526 			    (tunnel->i_seqno &&
527 					(s32)(seqno - tunnel->i_seqno) < 0)) {
528 				tunnel->dev->stats.rx_fifo_errors++;
529 				tunnel->dev->stats.rx_errors++;
530 				goto drop;
531 			}
532 			tunnel->i_seqno = seqno + 1;
533 		}
534 
535 		/* Warning: All skb pointers will be invalidated! */
536 		if (tunnel->dev->type == ARPHRD_ETHER) {
537 			if (!pskb_may_pull(skb, ETH_HLEN)) {
538 				tunnel->dev->stats.rx_length_errors++;
539 				tunnel->dev->stats.rx_errors++;
540 				goto drop;
541 			}
542 
543 			ipv6h = ipv6_hdr(skb);
544 			skb->protocol = eth_type_trans(skb, tunnel->dev);
545 			skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
546 		}
547 
548 		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
549 
550 		skb_reset_network_header(skb);
551 
552 		err = IP6_ECN_decapsulate(ipv6h, skb);
553 		if (unlikely(err)) {
554 			if (log_ecn_error)
555 				net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
556 						     &ipv6h->saddr,
557 						     ipv6_get_dsfield(ipv6h));
558 			if (err > 1) {
559 				++tunnel->dev->stats.rx_frame_errors;
560 				++tunnel->dev->stats.rx_errors;
561 				goto drop;
562 			}
563 		}
564 
565 		tstats = this_cpu_ptr(tunnel->dev->tstats);
566 		u64_stats_update_begin(&tstats->syncp);
567 		tstats->rx_packets++;
568 		tstats->rx_bytes += skb->len;
569 		u64_stats_update_end(&tstats->syncp);
570 
571 		netif_rx(skb);
572 
573 		return 0;
574 	}
575 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
576 
577 drop:
578 	kfree_skb(skb);
579 	return 0;
580 }
581 
582 struct ipv6_tel_txoption {
583 	struct ipv6_txoptions ops;
584 	__u8 dst_opt[8];
585 };
586 
587 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
588 {
589 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
590 
591 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
592 	opt->dst_opt[3] = 1;
593 	opt->dst_opt[4] = encap_limit;
594 	opt->dst_opt[5] = IPV6_TLV_PADN;
595 	opt->dst_opt[6] = 1;
596 
597 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
598 	opt->ops.opt_nflen = 8;
599 }
600 
601 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
602 			 struct net_device *dev,
603 			 __u8 dsfield,
604 			 struct flowi6 *fl6,
605 			 int encap_limit,
606 			 __u32 *pmtu)
607 {
608 	struct ip6_tnl *tunnel = netdev_priv(dev);
609 	struct net *net = tunnel->net;
610 	struct net_device *tdev;    /* Device to other host */
611 	struct ipv6hdr  *ipv6h;     /* Our new IP header */
612 	unsigned int max_headroom = 0; /* The extra header space needed */
613 	int    gre_hlen;
614 	struct ipv6_tel_txoption opt;
615 	int    mtu;
616 	struct dst_entry *dst = NULL, *ndst = NULL;
617 	struct net_device_stats *stats = &tunnel->dev->stats;
618 	int err = -1;
619 	u8 proto;
620 	struct sk_buff *new_skb;
621 
622 	if (dev->type == ARPHRD_ETHER)
623 		IPCB(skb)->flags = 0;
624 
625 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
626 		gre_hlen = 0;
627 		ipv6h = (struct ipv6hdr *)skb->data;
628 		fl6->daddr = ipv6h->daddr;
629 	} else {
630 		gre_hlen = tunnel->hlen;
631 		fl6->daddr = tunnel->parms.raddr;
632 	}
633 
634 	if (!fl6->flowi6_mark)
635 		dst = ip6_tnl_dst_check(tunnel);
636 
637 	if (!dst) {
638 		ndst = ip6_route_output(net, NULL, fl6);
639 
640 		if (ndst->error)
641 			goto tx_err_link_failure;
642 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
643 		if (IS_ERR(ndst)) {
644 			err = PTR_ERR(ndst);
645 			ndst = NULL;
646 			goto tx_err_link_failure;
647 		}
648 		dst = ndst;
649 	}
650 
651 	tdev = dst->dev;
652 
653 	if (tdev == dev) {
654 		stats->collisions++;
655 		net_warn_ratelimited("%s: Local routing loop detected!\n",
656 				     tunnel->parms.name);
657 		goto tx_err_dst_release;
658 	}
659 
660 	mtu = dst_mtu(dst) - sizeof(*ipv6h);
661 	if (encap_limit >= 0) {
662 		max_headroom += 8;
663 		mtu -= 8;
664 	}
665 	if (mtu < IPV6_MIN_MTU)
666 		mtu = IPV6_MIN_MTU;
667 	if (skb_dst(skb))
668 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
669 	if (skb->len > mtu) {
670 		*pmtu = mtu;
671 		err = -EMSGSIZE;
672 		goto tx_err_dst_release;
673 	}
674 
675 	if (tunnel->err_count > 0) {
676 		if (time_before(jiffies,
677 				tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
678 			tunnel->err_count--;
679 
680 			dst_link_failure(skb);
681 		} else
682 			tunnel->err_count = 0;
683 	}
684 
685 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
686 
687 	max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
688 
689 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
690 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
691 		new_skb = skb_realloc_headroom(skb, max_headroom);
692 		if (max_headroom > dev->needed_headroom)
693 			dev->needed_headroom = max_headroom;
694 		if (!new_skb)
695 			goto tx_err_dst_release;
696 
697 		if (skb->sk)
698 			skb_set_owner_w(new_skb, skb->sk);
699 		consume_skb(skb);
700 		skb = new_skb;
701 	}
702 
703 	if (fl6->flowi6_mark) {
704 		skb_dst_set(skb, dst);
705 		ndst = NULL;
706 	} else {
707 		skb_dst_set_noref(skb, dst);
708 	}
709 
710 	proto = NEXTHDR_GRE;
711 	if (encap_limit >= 0) {
712 		init_tel_txopt(&opt, encap_limit);
713 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
714 	}
715 
716 	if (likely(!skb->encapsulation)) {
717 		skb_reset_inner_headers(skb);
718 		skb->encapsulation = 1;
719 	}
720 
721 	skb_push(skb, gre_hlen);
722 	skb_reset_network_header(skb);
723 	skb_set_transport_header(skb, sizeof(*ipv6h));
724 
725 	/*
726 	 *	Push down and install the IP header.
727 	 */
728 	ipv6h = ipv6_hdr(skb);
729 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
730 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
731 	ipv6h->hop_limit = tunnel->parms.hop_limit;
732 	ipv6h->nexthdr = proto;
733 	ipv6h->saddr = fl6->saddr;
734 	ipv6h->daddr = fl6->daddr;
735 
736 	((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
737 	((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
738 				   htons(ETH_P_TEB) : skb->protocol;
739 
740 	if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
741 		__be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
742 
743 		if (tunnel->parms.o_flags&GRE_SEQ) {
744 			++tunnel->o_seqno;
745 			*ptr = htonl(tunnel->o_seqno);
746 			ptr--;
747 		}
748 		if (tunnel->parms.o_flags&GRE_KEY) {
749 			*ptr = tunnel->parms.o_key;
750 			ptr--;
751 		}
752 		if (tunnel->parms.o_flags&GRE_CSUM) {
753 			*ptr = 0;
754 			*(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
755 				skb->len - sizeof(struct ipv6hdr));
756 		}
757 	}
758 
759 	ip6tunnel_xmit(skb, dev);
760 	if (ndst)
761 		ip6_tnl_dst_store(tunnel, ndst);
762 	return 0;
763 tx_err_link_failure:
764 	stats->tx_carrier_errors++;
765 	dst_link_failure(skb);
766 tx_err_dst_release:
767 	dst_release(ndst);
768 	return err;
769 }
770 
771 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
772 {
773 	struct ip6_tnl *t = netdev_priv(dev);
774 	const struct iphdr  *iph = ip_hdr(skb);
775 	int encap_limit = -1;
776 	struct flowi6 fl6;
777 	__u8 dsfield;
778 	__u32 mtu;
779 	int err;
780 
781 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
782 		encap_limit = t->parms.encap_limit;
783 
784 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
785 	fl6.flowi6_proto = IPPROTO_IPIP;
786 
787 	dsfield = ipv4_get_dsfield(iph);
788 
789 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
790 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
791 					  & IPV6_TCLASS_MASK;
792 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
793 		fl6.flowi6_mark = skb->mark;
794 
795 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
796 	if (err != 0) {
797 		/* XXX: send ICMP error even if DF is not set. */
798 		if (err == -EMSGSIZE)
799 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
800 				  htonl(mtu));
801 		return -1;
802 	}
803 
804 	return 0;
805 }
806 
807 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
808 {
809 	struct ip6_tnl *t = netdev_priv(dev);
810 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
811 	int encap_limit = -1;
812 	__u16 offset;
813 	struct flowi6 fl6;
814 	__u8 dsfield;
815 	__u32 mtu;
816 	int err;
817 
818 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
819 		return -1;
820 
821 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
822 	if (offset > 0) {
823 		struct ipv6_tlv_tnl_enc_lim *tel;
824 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
825 		if (tel->encap_limit == 0) {
826 			icmpv6_send(skb, ICMPV6_PARAMPROB,
827 				    ICMPV6_HDR_FIELD, offset + 2);
828 			return -1;
829 		}
830 		encap_limit = tel->encap_limit - 1;
831 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
832 		encap_limit = t->parms.encap_limit;
833 
834 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
835 	fl6.flowi6_proto = IPPROTO_IPV6;
836 
837 	dsfield = ipv6_get_dsfield(ipv6h);
838 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
839 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
840 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
841 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
842 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
843 		fl6.flowi6_mark = skb->mark;
844 
845 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
846 	if (err != 0) {
847 		if (err == -EMSGSIZE)
848 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
849 		return -1;
850 	}
851 
852 	return 0;
853 }
854 
855 /**
856  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
857  *   @t: the outgoing tunnel device
858  *   @hdr: IPv6 header from the incoming packet
859  *
860  * Description:
861  *   Avoid trivial tunneling loop by checking that tunnel exit-point
862  *   doesn't match source of incoming packet.
863  *
864  * Return:
865  *   1 if conflict,
866  *   0 else
867  **/
868 
869 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
870 	const struct ipv6hdr *hdr)
871 {
872 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
873 }
874 
875 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
876 {
877 	struct ip6_tnl *t = netdev_priv(dev);
878 	int encap_limit = -1;
879 	struct flowi6 fl6;
880 	__u32 mtu;
881 	int err;
882 
883 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
884 		encap_limit = t->parms.encap_limit;
885 
886 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
887 	fl6.flowi6_proto = skb->protocol;
888 
889 	err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
890 
891 	return err;
892 }
893 
894 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
895 	struct net_device *dev)
896 {
897 	struct ip6_tnl *t = netdev_priv(dev);
898 	struct net_device_stats *stats = &t->dev->stats;
899 	int ret;
900 
901 	if (!ip6_tnl_xmit_ctl(t))
902 		goto tx_err;
903 
904 	switch (skb->protocol) {
905 	case htons(ETH_P_IP):
906 		ret = ip6gre_xmit_ipv4(skb, dev);
907 		break;
908 	case htons(ETH_P_IPV6):
909 		ret = ip6gre_xmit_ipv6(skb, dev);
910 		break;
911 	default:
912 		ret = ip6gre_xmit_other(skb, dev);
913 		break;
914 	}
915 
916 	if (ret < 0)
917 		goto tx_err;
918 
919 	return NETDEV_TX_OK;
920 
921 tx_err:
922 	stats->tx_errors++;
923 	stats->tx_dropped++;
924 	kfree_skb(skb);
925 	return NETDEV_TX_OK;
926 }
927 
928 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
929 {
930 	struct net_device *dev = t->dev;
931 	struct __ip6_tnl_parm *p = &t->parms;
932 	struct flowi6 *fl6 = &t->fl.u.ip6;
933 	int addend = sizeof(struct ipv6hdr) + 4;
934 
935 	if (dev->type != ARPHRD_ETHER) {
936 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
937 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
938 	}
939 
940 	/* Set up flowi template */
941 	fl6->saddr = p->laddr;
942 	fl6->daddr = p->raddr;
943 	fl6->flowi6_oif = p->link;
944 	fl6->flowlabel = 0;
945 
946 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
947 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
948 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
949 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
950 
951 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
952 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
953 
954 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
955 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
956 		dev->flags |= IFF_POINTOPOINT;
957 	else
958 		dev->flags &= ~IFF_POINTOPOINT;
959 
960 	dev->iflink = p->link;
961 
962 	/* Precalculate GRE options length */
963 	if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
964 		if (t->parms.o_flags&GRE_CSUM)
965 			addend += 4;
966 		if (t->parms.o_flags&GRE_KEY)
967 			addend += 4;
968 		if (t->parms.o_flags&GRE_SEQ)
969 			addend += 4;
970 	}
971 	t->hlen = addend;
972 
973 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
974 		int strict = (ipv6_addr_type(&p->raddr) &
975 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
976 
977 		struct rt6_info *rt = rt6_lookup(t->net,
978 						 &p->raddr, &p->laddr,
979 						 p->link, strict);
980 
981 		if (rt == NULL)
982 			return;
983 
984 		if (rt->dst.dev) {
985 			dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
986 
987 			if (set_mtu) {
988 				dev->mtu = rt->dst.dev->mtu - addend;
989 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
990 					dev->mtu -= 8;
991 
992 				if (dev->mtu < IPV6_MIN_MTU)
993 					dev->mtu = IPV6_MIN_MTU;
994 			}
995 		}
996 		ip6_rt_put(rt);
997 	}
998 }
999 
1000 static int ip6gre_tnl_change(struct ip6_tnl *t,
1001 	const struct __ip6_tnl_parm *p, int set_mtu)
1002 {
1003 	t->parms.laddr = p->laddr;
1004 	t->parms.raddr = p->raddr;
1005 	t->parms.flags = p->flags;
1006 	t->parms.hop_limit = p->hop_limit;
1007 	t->parms.encap_limit = p->encap_limit;
1008 	t->parms.flowinfo = p->flowinfo;
1009 	t->parms.link = p->link;
1010 	t->parms.proto = p->proto;
1011 	t->parms.i_key = p->i_key;
1012 	t->parms.o_key = p->o_key;
1013 	t->parms.i_flags = p->i_flags;
1014 	t->parms.o_flags = p->o_flags;
1015 	ip6_tnl_dst_reset(t);
1016 	ip6gre_tnl_link_config(t, set_mtu);
1017 	return 0;
1018 }
1019 
1020 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1021 	const struct ip6_tnl_parm2 *u)
1022 {
1023 	p->laddr = u->laddr;
1024 	p->raddr = u->raddr;
1025 	p->flags = u->flags;
1026 	p->hop_limit = u->hop_limit;
1027 	p->encap_limit = u->encap_limit;
1028 	p->flowinfo = u->flowinfo;
1029 	p->link = u->link;
1030 	p->i_key = u->i_key;
1031 	p->o_key = u->o_key;
1032 	p->i_flags = u->i_flags;
1033 	p->o_flags = u->o_flags;
1034 	memcpy(p->name, u->name, sizeof(u->name));
1035 }
1036 
1037 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1038 	const struct __ip6_tnl_parm *p)
1039 {
1040 	u->proto = IPPROTO_GRE;
1041 	u->laddr = p->laddr;
1042 	u->raddr = p->raddr;
1043 	u->flags = p->flags;
1044 	u->hop_limit = p->hop_limit;
1045 	u->encap_limit = p->encap_limit;
1046 	u->flowinfo = p->flowinfo;
1047 	u->link = p->link;
1048 	u->i_key = p->i_key;
1049 	u->o_key = p->o_key;
1050 	u->i_flags = p->i_flags;
1051 	u->o_flags = p->o_flags;
1052 	memcpy(u->name, p->name, sizeof(u->name));
1053 }
1054 
1055 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1056 	struct ifreq *ifr, int cmd)
1057 {
1058 	int err = 0;
1059 	struct ip6_tnl_parm2 p;
1060 	struct __ip6_tnl_parm p1;
1061 	struct ip6_tnl *t = netdev_priv(dev);
1062 	struct net *net = t->net;
1063 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1064 
1065 	switch (cmd) {
1066 	case SIOCGETTUNNEL:
1067 		if (dev == ign->fb_tunnel_dev) {
1068 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1069 				err = -EFAULT;
1070 				break;
1071 			}
1072 			ip6gre_tnl_parm_from_user(&p1, &p);
1073 			t = ip6gre_tunnel_locate(net, &p1, 0);
1074 			if (t == NULL)
1075 				t = netdev_priv(dev);
1076 		}
1077 		memset(&p, 0, sizeof(p));
1078 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1079 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1080 			err = -EFAULT;
1081 		break;
1082 
1083 	case SIOCADDTUNNEL:
1084 	case SIOCCHGTUNNEL:
1085 		err = -EPERM;
1086 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1087 			goto done;
1088 
1089 		err = -EFAULT;
1090 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1091 			goto done;
1092 
1093 		err = -EINVAL;
1094 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1095 			goto done;
1096 
1097 		if (!(p.i_flags&GRE_KEY))
1098 			p.i_key = 0;
1099 		if (!(p.o_flags&GRE_KEY))
1100 			p.o_key = 0;
1101 
1102 		ip6gre_tnl_parm_from_user(&p1, &p);
1103 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1104 
1105 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1106 			if (t != NULL) {
1107 				if (t->dev != dev) {
1108 					err = -EEXIST;
1109 					break;
1110 				}
1111 			} else {
1112 				t = netdev_priv(dev);
1113 
1114 				ip6gre_tunnel_unlink(ign, t);
1115 				synchronize_net();
1116 				ip6gre_tnl_change(t, &p1, 1);
1117 				ip6gre_tunnel_link(ign, t);
1118 				netdev_state_change(dev);
1119 			}
1120 		}
1121 
1122 		if (t) {
1123 			err = 0;
1124 
1125 			memset(&p, 0, sizeof(p));
1126 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1127 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1128 				err = -EFAULT;
1129 		} else
1130 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1131 		break;
1132 
1133 	case SIOCDELTUNNEL:
1134 		err = -EPERM;
1135 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1136 			goto done;
1137 
1138 		if (dev == ign->fb_tunnel_dev) {
1139 			err = -EFAULT;
1140 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1141 				goto done;
1142 			err = -ENOENT;
1143 			ip6gre_tnl_parm_from_user(&p1, &p);
1144 			t = ip6gre_tunnel_locate(net, &p1, 0);
1145 			if (t == NULL)
1146 				goto done;
1147 			err = -EPERM;
1148 			if (t == netdev_priv(ign->fb_tunnel_dev))
1149 				goto done;
1150 			dev = t->dev;
1151 		}
1152 		unregister_netdevice(dev);
1153 		err = 0;
1154 		break;
1155 
1156 	default:
1157 		err = -EINVAL;
1158 	}
1159 
1160 done:
1161 	return err;
1162 }
1163 
1164 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1165 {
1166 	if (new_mtu < 68 ||
1167 	    new_mtu > 0xFFF8 - dev->hard_header_len)
1168 		return -EINVAL;
1169 	dev->mtu = new_mtu;
1170 	return 0;
1171 }
1172 
1173 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1174 			unsigned short type,
1175 			const void *daddr, const void *saddr, unsigned int len)
1176 {
1177 	struct ip6_tnl *t = netdev_priv(dev);
1178 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1179 	__be16 *p = (__be16 *)(ipv6h+1);
1180 
1181 	ip6_flow_hdr(ipv6h, 0,
1182 		     ip6_make_flowlabel(dev_net(dev), skb,
1183 					t->fl.u.ip6.flowlabel, false));
1184 	ipv6h->hop_limit = t->parms.hop_limit;
1185 	ipv6h->nexthdr = NEXTHDR_GRE;
1186 	ipv6h->saddr = t->parms.laddr;
1187 	ipv6h->daddr = t->parms.raddr;
1188 
1189 	p[0]		= t->parms.o_flags;
1190 	p[1]		= htons(type);
1191 
1192 	/*
1193 	 *	Set the source hardware address.
1194 	 */
1195 
1196 	if (saddr)
1197 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1198 	if (daddr)
1199 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1200 	if (!ipv6_addr_any(&ipv6h->daddr))
1201 		return t->hlen;
1202 
1203 	return -t->hlen;
1204 }
1205 
1206 static const struct header_ops ip6gre_header_ops = {
1207 	.create	= ip6gre_header,
1208 };
1209 
1210 static const struct net_device_ops ip6gre_netdev_ops = {
1211 	.ndo_init		= ip6gre_tunnel_init,
1212 	.ndo_uninit		= ip6gre_tunnel_uninit,
1213 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1214 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1215 	.ndo_change_mtu		= ip6gre_tunnel_change_mtu,
1216 	.ndo_get_stats64	= ip_tunnel_get_stats64,
1217 };
1218 
1219 static void ip6gre_dev_free(struct net_device *dev)
1220 {
1221 	free_percpu(dev->tstats);
1222 	free_netdev(dev);
1223 }
1224 
1225 static void ip6gre_tunnel_setup(struct net_device *dev)
1226 {
1227 	struct ip6_tnl *t;
1228 
1229 	dev->netdev_ops = &ip6gre_netdev_ops;
1230 	dev->destructor = ip6gre_dev_free;
1231 
1232 	dev->type = ARPHRD_IP6GRE;
1233 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1234 	dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1235 	t = netdev_priv(dev);
1236 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1237 		dev->mtu -= 8;
1238 	dev->flags |= IFF_NOARP;
1239 	dev->iflink = 0;
1240 	dev->addr_len = sizeof(struct in6_addr);
1241 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1242 }
1243 
1244 static int ip6gre_tunnel_init(struct net_device *dev)
1245 {
1246 	struct ip6_tnl *tunnel;
1247 	int i;
1248 
1249 	tunnel = netdev_priv(dev);
1250 
1251 	tunnel->dev = dev;
1252 	tunnel->net = dev_net(dev);
1253 	strcpy(tunnel->parms.name, dev->name);
1254 
1255 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1256 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1257 
1258 	if (ipv6_addr_any(&tunnel->parms.raddr))
1259 		dev->header_ops = &ip6gre_header_ops;
1260 
1261 	dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1262 	if (!dev->tstats)
1263 		return -ENOMEM;
1264 
1265 	for_each_possible_cpu(i) {
1266 		struct pcpu_sw_netstats *ip6gre_tunnel_stats;
1267 		ip6gre_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1268 		u64_stats_init(&ip6gre_tunnel_stats->syncp);
1269 	}
1270 
1271 
1272 	return 0;
1273 }
1274 
1275 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1276 {
1277 	struct ip6_tnl *tunnel = netdev_priv(dev);
1278 
1279 	tunnel->dev = dev;
1280 	tunnel->net = dev_net(dev);
1281 	strcpy(tunnel->parms.name, dev->name);
1282 
1283 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1284 
1285 	dev_hold(dev);
1286 }
1287 
1288 
1289 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1290 	.handler     = ip6gre_rcv,
1291 	.err_handler = ip6gre_err,
1292 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1293 };
1294 
1295 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1296 {
1297 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1298 	struct net_device *dev, *aux;
1299 	int prio;
1300 
1301 	for_each_netdev_safe(net, dev, aux)
1302 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1303 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1304 			unregister_netdevice_queue(dev, head);
1305 
1306 	for (prio = 0; prio < 4; prio++) {
1307 		int h;
1308 		for (h = 0; h < HASH_SIZE; h++) {
1309 			struct ip6_tnl *t;
1310 
1311 			t = rtnl_dereference(ign->tunnels[prio][h]);
1312 
1313 			while (t != NULL) {
1314 				/* If dev is in the same netns, it has already
1315 				 * been added to the list by the previous loop.
1316 				 */
1317 				if (!net_eq(dev_net(t->dev), net))
1318 					unregister_netdevice_queue(t->dev,
1319 								   head);
1320 				t = rtnl_dereference(t->next);
1321 			}
1322 		}
1323 	}
1324 }
1325 
1326 static int __net_init ip6gre_init_net(struct net *net)
1327 {
1328 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1329 	int err;
1330 
1331 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1332 					  NET_NAME_UNKNOWN,
1333 					  ip6gre_tunnel_setup);
1334 	if (!ign->fb_tunnel_dev) {
1335 		err = -ENOMEM;
1336 		goto err_alloc_dev;
1337 	}
1338 	dev_net_set(ign->fb_tunnel_dev, net);
1339 	/* FB netdevice is special: we have one, and only one per netns.
1340 	 * Allowing to move it to another netns is clearly unsafe.
1341 	 */
1342 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1343 
1344 
1345 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1346 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1347 
1348 	err = register_netdev(ign->fb_tunnel_dev);
1349 	if (err)
1350 		goto err_reg_dev;
1351 
1352 	rcu_assign_pointer(ign->tunnels_wc[0],
1353 			   netdev_priv(ign->fb_tunnel_dev));
1354 	return 0;
1355 
1356 err_reg_dev:
1357 	ip6gre_dev_free(ign->fb_tunnel_dev);
1358 err_alloc_dev:
1359 	return err;
1360 }
1361 
1362 static void __net_exit ip6gre_exit_net(struct net *net)
1363 {
1364 	LIST_HEAD(list);
1365 
1366 	rtnl_lock();
1367 	ip6gre_destroy_tunnels(net, &list);
1368 	unregister_netdevice_many(&list);
1369 	rtnl_unlock();
1370 }
1371 
1372 static struct pernet_operations ip6gre_net_ops = {
1373 	.init = ip6gre_init_net,
1374 	.exit = ip6gre_exit_net,
1375 	.id   = &ip6gre_net_id,
1376 	.size = sizeof(struct ip6gre_net),
1377 };
1378 
1379 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1380 {
1381 	__be16 flags;
1382 
1383 	if (!data)
1384 		return 0;
1385 
1386 	flags = 0;
1387 	if (data[IFLA_GRE_IFLAGS])
1388 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1389 	if (data[IFLA_GRE_OFLAGS])
1390 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1391 	if (flags & (GRE_VERSION|GRE_ROUTING))
1392 		return -EINVAL;
1393 
1394 	return 0;
1395 }
1396 
1397 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1398 {
1399 	struct in6_addr daddr;
1400 
1401 	if (tb[IFLA_ADDRESS]) {
1402 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1403 			return -EINVAL;
1404 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1405 			return -EADDRNOTAVAIL;
1406 	}
1407 
1408 	if (!data)
1409 		goto out;
1410 
1411 	if (data[IFLA_GRE_REMOTE]) {
1412 		nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1413 		if (ipv6_addr_any(&daddr))
1414 			return -EINVAL;
1415 	}
1416 
1417 out:
1418 	return ip6gre_tunnel_validate(tb, data);
1419 }
1420 
1421 
1422 static void ip6gre_netlink_parms(struct nlattr *data[],
1423 				struct __ip6_tnl_parm *parms)
1424 {
1425 	memset(parms, 0, sizeof(*parms));
1426 
1427 	if (!data)
1428 		return;
1429 
1430 	if (data[IFLA_GRE_LINK])
1431 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1432 
1433 	if (data[IFLA_GRE_IFLAGS])
1434 		parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1435 
1436 	if (data[IFLA_GRE_OFLAGS])
1437 		parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1438 
1439 	if (data[IFLA_GRE_IKEY])
1440 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1441 
1442 	if (data[IFLA_GRE_OKEY])
1443 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1444 
1445 	if (data[IFLA_GRE_LOCAL])
1446 		nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1447 
1448 	if (data[IFLA_GRE_REMOTE])
1449 		nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1450 
1451 	if (data[IFLA_GRE_TTL])
1452 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1453 
1454 	if (data[IFLA_GRE_ENCAP_LIMIT])
1455 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1456 
1457 	if (data[IFLA_GRE_FLOWINFO])
1458 		parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1459 
1460 	if (data[IFLA_GRE_FLAGS])
1461 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1462 }
1463 
1464 static int ip6gre_tap_init(struct net_device *dev)
1465 {
1466 	struct ip6_tnl *tunnel;
1467 
1468 	tunnel = netdev_priv(dev);
1469 
1470 	tunnel->dev = dev;
1471 	tunnel->net = dev_net(dev);
1472 	strcpy(tunnel->parms.name, dev->name);
1473 
1474 	ip6gre_tnl_link_config(tunnel, 1);
1475 
1476 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1477 	if (!dev->tstats)
1478 		return -ENOMEM;
1479 
1480 	return 0;
1481 }
1482 
1483 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1484 	.ndo_init = ip6gre_tap_init,
1485 	.ndo_uninit = ip6gre_tunnel_uninit,
1486 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1487 	.ndo_set_mac_address = eth_mac_addr,
1488 	.ndo_validate_addr = eth_validate_addr,
1489 	.ndo_change_mtu = ip6gre_tunnel_change_mtu,
1490 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1491 };
1492 
1493 static void ip6gre_tap_setup(struct net_device *dev)
1494 {
1495 
1496 	ether_setup(dev);
1497 
1498 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1499 	dev->destructor = ip6gre_dev_free;
1500 
1501 	dev->iflink = 0;
1502 	dev->features |= NETIF_F_NETNS_LOCAL;
1503 }
1504 
1505 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1506 	struct nlattr *tb[], struct nlattr *data[])
1507 {
1508 	struct ip6_tnl *nt;
1509 	struct net *net = dev_net(dev);
1510 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1511 	int err;
1512 
1513 	nt = netdev_priv(dev);
1514 	ip6gre_netlink_parms(data, &nt->parms);
1515 
1516 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1517 		return -EEXIST;
1518 
1519 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1520 		eth_hw_addr_random(dev);
1521 
1522 	nt->dev = dev;
1523 	nt->net = dev_net(dev);
1524 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1525 
1526 	/* Can use a lockless transmit, unless we generate output sequences */
1527 	if (!(nt->parms.o_flags & GRE_SEQ))
1528 		dev->features |= NETIF_F_LLTX;
1529 
1530 	err = register_netdevice(dev);
1531 	if (err)
1532 		goto out;
1533 
1534 	dev_hold(dev);
1535 	ip6gre_tunnel_link(ign, nt);
1536 
1537 out:
1538 	return err;
1539 }
1540 
1541 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1542 			    struct nlattr *data[])
1543 {
1544 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1545 	struct net *net = nt->net;
1546 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1547 	struct __ip6_tnl_parm p;
1548 
1549 	if (dev == ign->fb_tunnel_dev)
1550 		return -EINVAL;
1551 
1552 	ip6gre_netlink_parms(data, &p);
1553 
1554 	t = ip6gre_tunnel_locate(net, &p, 0);
1555 
1556 	if (t) {
1557 		if (t->dev != dev)
1558 			return -EEXIST;
1559 	} else {
1560 		t = nt;
1561 
1562 		ip6gre_tunnel_unlink(ign, t);
1563 		ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1564 		ip6gre_tunnel_link(ign, t);
1565 		netdev_state_change(dev);
1566 	}
1567 
1568 	return 0;
1569 }
1570 
1571 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1572 {
1573 	struct net *net = dev_net(dev);
1574 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1575 
1576 	if (dev != ign->fb_tunnel_dev)
1577 		unregister_netdevice_queue(dev, head);
1578 }
1579 
1580 static size_t ip6gre_get_size(const struct net_device *dev)
1581 {
1582 	return
1583 		/* IFLA_GRE_LINK */
1584 		nla_total_size(4) +
1585 		/* IFLA_GRE_IFLAGS */
1586 		nla_total_size(2) +
1587 		/* IFLA_GRE_OFLAGS */
1588 		nla_total_size(2) +
1589 		/* IFLA_GRE_IKEY */
1590 		nla_total_size(4) +
1591 		/* IFLA_GRE_OKEY */
1592 		nla_total_size(4) +
1593 		/* IFLA_GRE_LOCAL */
1594 		nla_total_size(sizeof(struct in6_addr)) +
1595 		/* IFLA_GRE_REMOTE */
1596 		nla_total_size(sizeof(struct in6_addr)) +
1597 		/* IFLA_GRE_TTL */
1598 		nla_total_size(1) +
1599 		/* IFLA_GRE_TOS */
1600 		nla_total_size(1) +
1601 		/* IFLA_GRE_ENCAP_LIMIT */
1602 		nla_total_size(1) +
1603 		/* IFLA_GRE_FLOWINFO */
1604 		nla_total_size(4) +
1605 		/* IFLA_GRE_FLAGS */
1606 		nla_total_size(4) +
1607 		0;
1608 }
1609 
1610 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1611 {
1612 	struct ip6_tnl *t = netdev_priv(dev);
1613 	struct __ip6_tnl_parm *p = &t->parms;
1614 
1615 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1616 	    nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1617 	    nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1618 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1619 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1620 	    nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1621 	    nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1622 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1623 	    /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1624 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1625 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1626 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1627 		goto nla_put_failure;
1628 	return 0;
1629 
1630 nla_put_failure:
1631 	return -EMSGSIZE;
1632 }
1633 
1634 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1635 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1636 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1637 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1638 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1639 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1640 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1641 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1642 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1643 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1644 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1645 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1646 };
1647 
1648 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1649 	.kind		= "ip6gre",
1650 	.maxtype	= IFLA_GRE_MAX,
1651 	.policy		= ip6gre_policy,
1652 	.priv_size	= sizeof(struct ip6_tnl),
1653 	.setup		= ip6gre_tunnel_setup,
1654 	.validate	= ip6gre_tunnel_validate,
1655 	.newlink	= ip6gre_newlink,
1656 	.changelink	= ip6gre_changelink,
1657 	.dellink	= ip6gre_dellink,
1658 	.get_size	= ip6gre_get_size,
1659 	.fill_info	= ip6gre_fill_info,
1660 };
1661 
1662 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1663 	.kind		= "ip6gretap",
1664 	.maxtype	= IFLA_GRE_MAX,
1665 	.policy		= ip6gre_policy,
1666 	.priv_size	= sizeof(struct ip6_tnl),
1667 	.setup		= ip6gre_tap_setup,
1668 	.validate	= ip6gre_tap_validate,
1669 	.newlink	= ip6gre_newlink,
1670 	.changelink	= ip6gre_changelink,
1671 	.get_size	= ip6gre_get_size,
1672 	.fill_info	= ip6gre_fill_info,
1673 };
1674 
1675 /*
1676  *	And now the modules code and kernel interface.
1677  */
1678 
1679 static int __init ip6gre_init(void)
1680 {
1681 	int err;
1682 
1683 	pr_info("GRE over IPv6 tunneling driver\n");
1684 
1685 	err = register_pernet_device(&ip6gre_net_ops);
1686 	if (err < 0)
1687 		return err;
1688 
1689 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1690 	if (err < 0) {
1691 		pr_info("%s: can't add protocol\n", __func__);
1692 		goto add_proto_failed;
1693 	}
1694 
1695 	err = rtnl_link_register(&ip6gre_link_ops);
1696 	if (err < 0)
1697 		goto rtnl_link_failed;
1698 
1699 	err = rtnl_link_register(&ip6gre_tap_ops);
1700 	if (err < 0)
1701 		goto tap_ops_failed;
1702 
1703 out:
1704 	return err;
1705 
1706 tap_ops_failed:
1707 	rtnl_link_unregister(&ip6gre_link_ops);
1708 rtnl_link_failed:
1709 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1710 add_proto_failed:
1711 	unregister_pernet_device(&ip6gre_net_ops);
1712 	goto out;
1713 }
1714 
1715 static void __exit ip6gre_fini(void)
1716 {
1717 	rtnl_link_unregister(&ip6gre_tap_ops);
1718 	rtnl_link_unregister(&ip6gre_link_ops);
1719 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1720 	unregister_pernet_device(&ip6gre_net_ops);
1721 }
1722 
1723 module_init(ip6gre_init);
1724 module_exit(ip6gre_fini);
1725 MODULE_LICENSE("GPL");
1726 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1727 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1728 MODULE_ALIAS_RTNL_LINK("ip6gre");
1729 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1730 MODULE_ALIAS_NETDEV("ip6gre0");
1731