xref: /linux/net/ipv6/ip6_gre.c (revision c75c5ab575af7db707689cdbb5a5c458e9a034bb)
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/icmp.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
44 #include <net/arp.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
48 #include <net/xfrm.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
52 
53 #include <net/ipv6.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.h>
57 
58 
59 static bool log_ecn_error = true;
60 module_param(log_ecn_error, bool, 0644);
61 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
62 
63 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
64 #define IPV6_TCLASS_SHIFT 20
65 
66 #define HASH_SIZE_SHIFT  5
67 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
68 
69 static int ip6gre_net_id __read_mostly;
70 struct ip6gre_net {
71 	struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
72 
73 	struct net_device *fb_tunnel_dev;
74 };
75 
76 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
77 static int ip6gre_tunnel_init(struct net_device *dev);
78 static void ip6gre_tunnel_setup(struct net_device *dev);
79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
80 static void ip6gre_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))&(HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103 	u32 hash = ipv6_addr_hash(addr);
104 
105 	return hash_32(hash, 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 static struct rtnl_link_stats64 *ip6gre_get_stats64(struct net_device *dev,
114 		struct rtnl_link_stats64 *tot)
115 {
116 	int i;
117 
118 	for_each_possible_cpu(i) {
119 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
120 		u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
121 		unsigned int start;
122 
123 		do {
124 			start = u64_stats_fetch_begin_bh(&tstats->syncp);
125 			rx_packets = tstats->rx_packets;
126 			tx_packets = tstats->tx_packets;
127 			rx_bytes = tstats->rx_bytes;
128 			tx_bytes = tstats->tx_bytes;
129 		} while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
130 
131 		tot->rx_packets += rx_packets;
132 		tot->tx_packets += tx_packets;
133 		tot->rx_bytes   += rx_bytes;
134 		tot->tx_bytes   += tx_bytes;
135 	}
136 
137 	tot->multicast = dev->stats.multicast;
138 	tot->rx_crc_errors = dev->stats.rx_crc_errors;
139 	tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
140 	tot->rx_length_errors = dev->stats.rx_length_errors;
141 	tot->rx_frame_errors = dev->stats.rx_frame_errors;
142 	tot->rx_errors = dev->stats.rx_errors;
143 
144 	tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
145 	tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
146 	tot->tx_dropped = dev->stats.tx_dropped;
147 	tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
148 	tot->tx_errors = dev->stats.tx_errors;
149 
150 	return tot;
151 }
152 
153 /* Given src, dst and key, find appropriate for input tunnel. */
154 
155 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
156 		const struct in6_addr *remote, const struct in6_addr *local,
157 		__be32 key, __be16 gre_proto)
158 {
159 	struct net *net = dev_net(dev);
160 	int link = dev->ifindex;
161 	unsigned int h0 = HASH_ADDR(remote);
162 	unsigned int h1 = HASH_KEY(key);
163 	struct ip6_tnl *t, *cand = NULL;
164 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
165 	int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
166 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
167 	int score, cand_score = 4;
168 
169 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
170 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
171 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
172 		    key != t->parms.i_key ||
173 		    !(t->dev->flags & IFF_UP))
174 			continue;
175 
176 		if (t->dev->type != ARPHRD_IP6GRE &&
177 		    t->dev->type != dev_type)
178 			continue;
179 
180 		score = 0;
181 		if (t->parms.link != link)
182 			score |= 1;
183 		if (t->dev->type != dev_type)
184 			score |= 2;
185 		if (score == 0)
186 			return t;
187 
188 		if (score < cand_score) {
189 			cand = t;
190 			cand_score = score;
191 		}
192 	}
193 
194 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
195 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
196 		    key != t->parms.i_key ||
197 		    !(t->dev->flags & IFF_UP))
198 			continue;
199 
200 		if (t->dev->type != ARPHRD_IP6GRE &&
201 		    t->dev->type != dev_type)
202 			continue;
203 
204 		score = 0;
205 		if (t->parms.link != link)
206 			score |= 1;
207 		if (t->dev->type != dev_type)
208 			score |= 2;
209 		if (score == 0)
210 			return t;
211 
212 		if (score < cand_score) {
213 			cand = t;
214 			cand_score = score;
215 		}
216 	}
217 
218 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
219 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
220 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
221 				 !ipv6_addr_is_multicast(local))) ||
222 		    key != t->parms.i_key ||
223 		    !(t->dev->flags & IFF_UP))
224 			continue;
225 
226 		if (t->dev->type != ARPHRD_IP6GRE &&
227 		    t->dev->type != dev_type)
228 			continue;
229 
230 		score = 0;
231 		if (t->parms.link != link)
232 			score |= 1;
233 		if (t->dev->type != dev_type)
234 			score |= 2;
235 		if (score == 0)
236 			return t;
237 
238 		if (score < cand_score) {
239 			cand = t;
240 			cand_score = score;
241 		}
242 	}
243 
244 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
245 		if (t->parms.i_key != key ||
246 		    !(t->dev->flags & IFF_UP))
247 			continue;
248 
249 		if (t->dev->type != ARPHRD_IP6GRE &&
250 		    t->dev->type != dev_type)
251 			continue;
252 
253 		score = 0;
254 		if (t->parms.link != link)
255 			score |= 1;
256 		if (t->dev->type != dev_type)
257 			score |= 2;
258 		if (score == 0)
259 			return t;
260 
261 		if (score < cand_score) {
262 			cand = t;
263 			cand_score = score;
264 		}
265 	}
266 
267 	if (cand != NULL)
268 		return cand;
269 
270 	dev = ign->fb_tunnel_dev;
271 	if (dev->flags & IFF_UP)
272 		return netdev_priv(dev);
273 
274 	return NULL;
275 }
276 
277 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
278 		const struct __ip6_tnl_parm *p)
279 {
280 	const struct in6_addr *remote = &p->raddr;
281 	const struct in6_addr *local = &p->laddr;
282 	unsigned int h = HASH_KEY(p->i_key);
283 	int prio = 0;
284 
285 	if (!ipv6_addr_any(local))
286 		prio |= 1;
287 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
288 		prio |= 2;
289 		h ^= HASH_ADDR(remote);
290 	}
291 
292 	return &ign->tunnels[prio][h];
293 }
294 
295 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
296 		const struct ip6_tnl *t)
297 {
298 	return __ip6gre_bucket(ign, &t->parms);
299 }
300 
301 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
302 {
303 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
304 
305 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
306 	rcu_assign_pointer(*tp, t);
307 }
308 
309 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
310 {
311 	struct ip6_tnl __rcu **tp;
312 	struct ip6_tnl *iter;
313 
314 	for (tp = ip6gre_bucket(ign, t);
315 	     (iter = rtnl_dereference(*tp)) != NULL;
316 	     tp = &iter->next) {
317 		if (t == iter) {
318 			rcu_assign_pointer(*tp, t->next);
319 			break;
320 		}
321 	}
322 }
323 
324 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
325 					   const struct __ip6_tnl_parm *parms,
326 					   int type)
327 {
328 	const struct in6_addr *remote = &parms->raddr;
329 	const struct in6_addr *local = &parms->laddr;
330 	__be32 key = parms->i_key;
331 	int link = parms->link;
332 	struct ip6_tnl *t;
333 	struct ip6_tnl __rcu **tp;
334 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
335 
336 	for (tp = __ip6gre_bucket(ign, parms);
337 	     (t = rtnl_dereference(*tp)) != NULL;
338 	     tp = &t->next)
339 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
340 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
341 		    key == t->parms.i_key &&
342 		    link == t->parms.link &&
343 		    type == t->dev->type)
344 			break;
345 
346 	return t;
347 }
348 
349 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
350 		const struct __ip6_tnl_parm *parms, int create)
351 {
352 	struct ip6_tnl *t, *nt;
353 	struct net_device *dev;
354 	char name[IFNAMSIZ];
355 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
356 
357 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
358 	if (t || !create)
359 		return t;
360 
361 	if (parms->name[0])
362 		strlcpy(name, parms->name, IFNAMSIZ);
363 	else
364 		strcpy(name, "ip6gre%d");
365 
366 	dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
367 	if (!dev)
368 		return NULL;
369 
370 	dev_net_set(dev, net);
371 
372 	nt = netdev_priv(dev);
373 	nt->parms = *parms;
374 	dev->rtnl_link_ops = &ip6gre_link_ops;
375 
376 	nt->dev = dev;
377 	ip6gre_tnl_link_config(nt, 1);
378 
379 	if (register_netdevice(dev) < 0)
380 		goto failed_free;
381 
382 	/* Can use a lockless transmit, unless we generate output sequences */
383 	if (!(nt->parms.o_flags & GRE_SEQ))
384 		dev->features |= NETIF_F_LLTX;
385 
386 	dev_hold(dev);
387 	ip6gre_tunnel_link(ign, nt);
388 	return nt;
389 
390 failed_free:
391 	free_netdev(dev);
392 	return NULL;
393 }
394 
395 static void ip6gre_tunnel_uninit(struct net_device *dev)
396 {
397 	struct net *net = dev_net(dev);
398 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
399 
400 	ip6gre_tunnel_unlink(ign, netdev_priv(dev));
401 	dev_put(dev);
402 }
403 
404 
405 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
406 		u8 type, u8 code, int offset, __be32 info)
407 {
408 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
409 	__be16 *p = (__be16 *)(skb->data + offset);
410 	int grehlen = offset + 4;
411 	struct ip6_tnl *t;
412 	__be16 flags;
413 
414 	flags = p[0];
415 	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
416 		if (flags&(GRE_VERSION|GRE_ROUTING))
417 			return;
418 		if (flags&GRE_KEY) {
419 			grehlen += 4;
420 			if (flags&GRE_CSUM)
421 				grehlen += 4;
422 		}
423 	}
424 
425 	/* If only 8 bytes returned, keyed message will be dropped here */
426 	if (!pskb_may_pull(skb, grehlen))
427 		return;
428 	ipv6h = (const struct ipv6hdr *)skb->data;
429 	p = (__be16 *)(skb->data + offset);
430 
431 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
432 				flags & GRE_KEY ?
433 				*(((__be32 *)p) + (grehlen / 4) - 1) : 0,
434 				p[1]);
435 	if (t == NULL)
436 		return;
437 
438 	switch (type) {
439 		__u32 teli;
440 		struct ipv6_tlv_tnl_enc_lim *tel;
441 		__u32 mtu;
442 	case ICMPV6_DEST_UNREACH:
443 		net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
444 				     t->parms.name);
445 		break;
446 	case ICMPV6_TIME_EXCEED:
447 		if (code == ICMPV6_EXC_HOPLIMIT) {
448 			net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
449 					     t->parms.name);
450 		}
451 		break;
452 	case ICMPV6_PARAMPROB:
453 		teli = 0;
454 		if (code == ICMPV6_HDR_FIELD)
455 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
456 
457 		if (teli && teli == info - 2) {
458 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
459 			if (tel->encap_limit == 0) {
460 				net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
461 						     t->parms.name);
462 			}
463 		} else {
464 			net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
465 					     t->parms.name);
466 		}
467 		break;
468 	case ICMPV6_PKT_TOOBIG:
469 		mtu = info - offset;
470 		if (mtu < IPV6_MIN_MTU)
471 			mtu = IPV6_MIN_MTU;
472 		t->dev->mtu = mtu;
473 		break;
474 	}
475 
476 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
477 		t->err_count++;
478 	else
479 		t->err_count = 1;
480 	t->err_time = jiffies;
481 }
482 
483 static int ip6gre_rcv(struct sk_buff *skb)
484 {
485 	const struct ipv6hdr *ipv6h;
486 	u8     *h;
487 	__be16    flags;
488 	__sum16   csum = 0;
489 	__be32 key = 0;
490 	u32    seqno = 0;
491 	struct ip6_tnl *tunnel;
492 	int    offset = 4;
493 	__be16 gre_proto;
494 	int err;
495 
496 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
497 		goto drop;
498 
499 	ipv6h = ipv6_hdr(skb);
500 	h = skb->data;
501 	flags = *(__be16 *)h;
502 
503 	if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
504 		/* - Version must be 0.
505 		   - We do not support routing headers.
506 		 */
507 		if (flags&(GRE_VERSION|GRE_ROUTING))
508 			goto drop;
509 
510 		if (flags&GRE_CSUM) {
511 			switch (skb->ip_summed) {
512 			case CHECKSUM_COMPLETE:
513 				csum = csum_fold(skb->csum);
514 				if (!csum)
515 					break;
516 				/* fall through */
517 			case CHECKSUM_NONE:
518 				skb->csum = 0;
519 				csum = __skb_checksum_complete(skb);
520 				skb->ip_summed = CHECKSUM_COMPLETE;
521 			}
522 			offset += 4;
523 		}
524 		if (flags&GRE_KEY) {
525 			key = *(__be32 *)(h + offset);
526 			offset += 4;
527 		}
528 		if (flags&GRE_SEQ) {
529 			seqno = ntohl(*(__be32 *)(h + offset));
530 			offset += 4;
531 		}
532 	}
533 
534 	gre_proto = *(__be16 *)(h + 2);
535 
536 	tunnel = ip6gre_tunnel_lookup(skb->dev,
537 					  &ipv6h->saddr, &ipv6h->daddr, key,
538 					  gre_proto);
539 	if (tunnel) {
540 		struct pcpu_tstats *tstats;
541 
542 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
543 			goto drop;
544 
545 		if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
546 			tunnel->dev->stats.rx_dropped++;
547 			goto drop;
548 		}
549 
550 		secpath_reset(skb);
551 
552 		skb->protocol = gre_proto;
553 		/* WCCP version 1 and 2 protocol decoding.
554 		 * - Change protocol to IP
555 		 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
556 		 */
557 		if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
558 			skb->protocol = htons(ETH_P_IP);
559 			if ((*(h + offset) & 0xF0) != 0x40)
560 				offset += 4;
561 		}
562 
563 		skb->mac_header = skb->network_header;
564 		__pskb_pull(skb, offset);
565 		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
566 		skb->pkt_type = PACKET_HOST;
567 
568 		if (((flags&GRE_CSUM) && csum) ||
569 		    (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
570 			tunnel->dev->stats.rx_crc_errors++;
571 			tunnel->dev->stats.rx_errors++;
572 			goto drop;
573 		}
574 		if (tunnel->parms.i_flags&GRE_SEQ) {
575 			if (!(flags&GRE_SEQ) ||
576 			    (tunnel->i_seqno &&
577 					(s32)(seqno - tunnel->i_seqno) < 0)) {
578 				tunnel->dev->stats.rx_fifo_errors++;
579 				tunnel->dev->stats.rx_errors++;
580 				goto drop;
581 			}
582 			tunnel->i_seqno = seqno + 1;
583 		}
584 
585 		/* Warning: All skb pointers will be invalidated! */
586 		if (tunnel->dev->type == ARPHRD_ETHER) {
587 			if (!pskb_may_pull(skb, ETH_HLEN)) {
588 				tunnel->dev->stats.rx_length_errors++;
589 				tunnel->dev->stats.rx_errors++;
590 				goto drop;
591 			}
592 
593 			ipv6h = ipv6_hdr(skb);
594 			skb->protocol = eth_type_trans(skb, tunnel->dev);
595 			skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
596 		}
597 
598 		__skb_tunnel_rx(skb, tunnel->dev);
599 
600 		skb_reset_network_header(skb);
601 
602 		err = IP6_ECN_decapsulate(ipv6h, skb);
603 		if (unlikely(err)) {
604 			if (log_ecn_error)
605 				net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
606 						     &ipv6h->saddr,
607 						     ipv6_get_dsfield(ipv6h));
608 			if (err > 1) {
609 				++tunnel->dev->stats.rx_frame_errors;
610 				++tunnel->dev->stats.rx_errors;
611 				goto drop;
612 			}
613 		}
614 
615 		tstats = this_cpu_ptr(tunnel->dev->tstats);
616 		u64_stats_update_begin(&tstats->syncp);
617 		tstats->rx_packets++;
618 		tstats->rx_bytes += skb->len;
619 		u64_stats_update_end(&tstats->syncp);
620 
621 		netif_rx(skb);
622 
623 		return 0;
624 	}
625 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
626 
627 drop:
628 	kfree_skb(skb);
629 	return 0;
630 }
631 
632 struct ipv6_tel_txoption {
633 	struct ipv6_txoptions ops;
634 	__u8 dst_opt[8];
635 };
636 
637 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
638 {
639 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
640 
641 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
642 	opt->dst_opt[3] = 1;
643 	opt->dst_opt[4] = encap_limit;
644 	opt->dst_opt[5] = IPV6_TLV_PADN;
645 	opt->dst_opt[6] = 1;
646 
647 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
648 	opt->ops.opt_nflen = 8;
649 }
650 
651 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
652 			 struct net_device *dev,
653 			 __u8 dsfield,
654 			 struct flowi6 *fl6,
655 			 int encap_limit,
656 			 __u32 *pmtu)
657 {
658 	struct net *net = dev_net(dev);
659 	struct ip6_tnl *tunnel = netdev_priv(dev);
660 	struct net_device *tdev;    /* Device to other host */
661 	struct ipv6hdr  *ipv6h;     /* Our new IP header */
662 	unsigned int max_headroom;  /* The extra header space needed */
663 	int    gre_hlen;
664 	struct ipv6_tel_txoption opt;
665 	int    mtu;
666 	struct dst_entry *dst = NULL, *ndst = NULL;
667 	struct net_device_stats *stats = &tunnel->dev->stats;
668 	int err = -1;
669 	u8 proto;
670 	int pkt_len;
671 	struct sk_buff *new_skb;
672 
673 	if (dev->type == ARPHRD_ETHER)
674 		IPCB(skb)->flags = 0;
675 
676 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
677 		gre_hlen = 0;
678 		ipv6h = (struct ipv6hdr *)skb->data;
679 		fl6->daddr = ipv6h->daddr;
680 	} else {
681 		gre_hlen = tunnel->hlen;
682 		fl6->daddr = tunnel->parms.raddr;
683 	}
684 
685 	if (!fl6->flowi6_mark)
686 		dst = ip6_tnl_dst_check(tunnel);
687 
688 	if (!dst) {
689 		ndst = ip6_route_output(net, NULL, fl6);
690 
691 		if (ndst->error)
692 			goto tx_err_link_failure;
693 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
694 		if (IS_ERR(ndst)) {
695 			err = PTR_ERR(ndst);
696 			ndst = NULL;
697 			goto tx_err_link_failure;
698 		}
699 		dst = ndst;
700 	}
701 
702 	tdev = dst->dev;
703 
704 	if (tdev == dev) {
705 		stats->collisions++;
706 		net_warn_ratelimited("%s: Local routing loop detected!\n",
707 				     tunnel->parms.name);
708 		goto tx_err_dst_release;
709 	}
710 
711 	mtu = dst_mtu(dst) - sizeof(*ipv6h);
712 	if (encap_limit >= 0) {
713 		max_headroom += 8;
714 		mtu -= 8;
715 	}
716 	if (mtu < IPV6_MIN_MTU)
717 		mtu = IPV6_MIN_MTU;
718 	if (skb_dst(skb))
719 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
720 	if (skb->len > mtu) {
721 		*pmtu = mtu;
722 		err = -EMSGSIZE;
723 		goto tx_err_dst_release;
724 	}
725 
726 	if (tunnel->err_count > 0) {
727 		if (time_before(jiffies,
728 				tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
729 			tunnel->err_count--;
730 
731 			dst_link_failure(skb);
732 		} else
733 			tunnel->err_count = 0;
734 	}
735 
736 	max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
737 
738 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
739 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
740 		new_skb = skb_realloc_headroom(skb, max_headroom);
741 		if (max_headroom > dev->needed_headroom)
742 			dev->needed_headroom = max_headroom;
743 		if (!new_skb)
744 			goto tx_err_dst_release;
745 
746 		if (skb->sk)
747 			skb_set_owner_w(new_skb, skb->sk);
748 		consume_skb(skb);
749 		skb = new_skb;
750 	}
751 
752 	skb_dst_drop(skb);
753 
754 	if (fl6->flowi6_mark) {
755 		skb_dst_set(skb, dst);
756 		ndst = NULL;
757 	} else {
758 		skb_dst_set_noref(skb, dst);
759 	}
760 
761 	proto = NEXTHDR_GRE;
762 	if (encap_limit >= 0) {
763 		init_tel_txopt(&opt, encap_limit);
764 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
765 	}
766 
767 	skb_push(skb, gre_hlen);
768 	skb_reset_network_header(skb);
769 	skb_set_transport_header(skb, sizeof(*ipv6h));
770 
771 	/*
772 	 *	Push down and install the IP header.
773 	 */
774 	ipv6h = ipv6_hdr(skb);
775 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
776 	ipv6h->hop_limit = tunnel->parms.hop_limit;
777 	ipv6h->nexthdr = proto;
778 	ipv6h->saddr = fl6->saddr;
779 	ipv6h->daddr = fl6->daddr;
780 
781 	((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
782 	((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
783 				   htons(ETH_P_TEB) : skb->protocol;
784 
785 	if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
786 		__be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
787 
788 		if (tunnel->parms.o_flags&GRE_SEQ) {
789 			++tunnel->o_seqno;
790 			*ptr = htonl(tunnel->o_seqno);
791 			ptr--;
792 		}
793 		if (tunnel->parms.o_flags&GRE_KEY) {
794 			*ptr = tunnel->parms.o_key;
795 			ptr--;
796 		}
797 		if (tunnel->parms.o_flags&GRE_CSUM) {
798 			*ptr = 0;
799 			*(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
800 				skb->len - sizeof(struct ipv6hdr));
801 		}
802 	}
803 
804 	nf_reset(skb);
805 	pkt_len = skb->len;
806 	err = ip6_local_out(skb);
807 
808 	if (net_xmit_eval(err) == 0) {
809 		struct pcpu_tstats *tstats = this_cpu_ptr(tunnel->dev->tstats);
810 
811 		tstats->tx_bytes += pkt_len;
812 		tstats->tx_packets++;
813 	} else {
814 		stats->tx_errors++;
815 		stats->tx_aborted_errors++;
816 	}
817 
818 	if (ndst)
819 		ip6_tnl_dst_store(tunnel, ndst);
820 
821 	return 0;
822 tx_err_link_failure:
823 	stats->tx_carrier_errors++;
824 	dst_link_failure(skb);
825 tx_err_dst_release:
826 	dst_release(ndst);
827 	return err;
828 }
829 
830 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
831 {
832 	struct ip6_tnl *t = netdev_priv(dev);
833 	const struct iphdr  *iph = ip_hdr(skb);
834 	int encap_limit = -1;
835 	struct flowi6 fl6;
836 	__u8 dsfield;
837 	__u32 mtu;
838 	int err;
839 
840 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
841 		encap_limit = t->parms.encap_limit;
842 
843 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
844 	fl6.flowi6_proto = IPPROTO_IPIP;
845 
846 	dsfield = ipv4_get_dsfield(iph);
847 
848 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
849 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
850 					  & IPV6_TCLASS_MASK;
851 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
852 		fl6.flowi6_mark = skb->mark;
853 
854 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
855 	if (err != 0) {
856 		/* XXX: send ICMP error even if DF is not set. */
857 		if (err == -EMSGSIZE)
858 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
859 				  htonl(mtu));
860 		return -1;
861 	}
862 
863 	return 0;
864 }
865 
866 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
867 {
868 	struct ip6_tnl *t = netdev_priv(dev);
869 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
870 	int encap_limit = -1;
871 	__u16 offset;
872 	struct flowi6 fl6;
873 	__u8 dsfield;
874 	__u32 mtu;
875 	int err;
876 
877 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
878 		return -1;
879 
880 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
881 	if (offset > 0) {
882 		struct ipv6_tlv_tnl_enc_lim *tel;
883 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
884 		if (tel->encap_limit == 0) {
885 			icmpv6_send(skb, ICMPV6_PARAMPROB,
886 				    ICMPV6_HDR_FIELD, offset + 2);
887 			return -1;
888 		}
889 		encap_limit = tel->encap_limit - 1;
890 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
891 		encap_limit = t->parms.encap_limit;
892 
893 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
894 	fl6.flowi6_proto = IPPROTO_IPV6;
895 
896 	dsfield = ipv6_get_dsfield(ipv6h);
897 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
898 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
899 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
900 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
901 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
902 		fl6.flowi6_mark = skb->mark;
903 
904 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
905 	if (err != 0) {
906 		if (err == -EMSGSIZE)
907 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
908 		return -1;
909 	}
910 
911 	return 0;
912 }
913 
914 /**
915  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
916  *   @t: the outgoing tunnel device
917  *   @hdr: IPv6 header from the incoming packet
918  *
919  * Description:
920  *   Avoid trivial tunneling loop by checking that tunnel exit-point
921  *   doesn't match source of incoming packet.
922  *
923  * Return:
924  *   1 if conflict,
925  *   0 else
926  **/
927 
928 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
929 	const struct ipv6hdr *hdr)
930 {
931 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
932 }
933 
934 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
935 {
936 	struct ip6_tnl *t = netdev_priv(dev);
937 	int encap_limit = -1;
938 	struct flowi6 fl6;
939 	__u32 mtu;
940 	int err;
941 
942 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
943 		encap_limit = t->parms.encap_limit;
944 
945 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
946 	fl6.flowi6_proto = skb->protocol;
947 
948 	err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
949 
950 	return err;
951 }
952 
953 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
954 	struct net_device *dev)
955 {
956 	struct ip6_tnl *t = netdev_priv(dev);
957 	struct net_device_stats *stats = &t->dev->stats;
958 	int ret;
959 
960 	if (!ip6_tnl_xmit_ctl(t))
961 		goto tx_err;
962 
963 	switch (skb->protocol) {
964 	case htons(ETH_P_IP):
965 		ret = ip6gre_xmit_ipv4(skb, dev);
966 		break;
967 	case htons(ETH_P_IPV6):
968 		ret = ip6gre_xmit_ipv6(skb, dev);
969 		break;
970 	default:
971 		ret = ip6gre_xmit_other(skb, dev);
972 		break;
973 	}
974 
975 	if (ret < 0)
976 		goto tx_err;
977 
978 	return NETDEV_TX_OK;
979 
980 tx_err:
981 	stats->tx_errors++;
982 	stats->tx_dropped++;
983 	kfree_skb(skb);
984 	return NETDEV_TX_OK;
985 }
986 
987 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
988 {
989 	struct net_device *dev = t->dev;
990 	struct __ip6_tnl_parm *p = &t->parms;
991 	struct flowi6 *fl6 = &t->fl.u.ip6;
992 	int addend = sizeof(struct ipv6hdr) + 4;
993 
994 	if (dev->type != ARPHRD_ETHER) {
995 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
996 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
997 	}
998 
999 	/* Set up flowi template */
1000 	fl6->saddr = p->laddr;
1001 	fl6->daddr = p->raddr;
1002 	fl6->flowi6_oif = p->link;
1003 	fl6->flowlabel = 0;
1004 
1005 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1006 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1007 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1008 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1009 
1010 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1011 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1012 
1013 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1014 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1015 		dev->flags |= IFF_POINTOPOINT;
1016 	else
1017 		dev->flags &= ~IFF_POINTOPOINT;
1018 
1019 	dev->iflink = p->link;
1020 
1021 	/* Precalculate GRE options length */
1022 	if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1023 		if (t->parms.o_flags&GRE_CSUM)
1024 			addend += 4;
1025 		if (t->parms.o_flags&GRE_KEY)
1026 			addend += 4;
1027 		if (t->parms.o_flags&GRE_SEQ)
1028 			addend += 4;
1029 	}
1030 
1031 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1032 		int strict = (ipv6_addr_type(&p->raddr) &
1033 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1034 
1035 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1036 						 &p->raddr, &p->laddr,
1037 						 p->link, strict);
1038 
1039 		if (rt == NULL)
1040 			return;
1041 
1042 		if (rt->dst.dev) {
1043 			dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1044 
1045 			if (set_mtu) {
1046 				dev->mtu = rt->dst.dev->mtu - addend;
1047 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1048 					dev->mtu -= 8;
1049 
1050 				if (dev->mtu < IPV6_MIN_MTU)
1051 					dev->mtu = IPV6_MIN_MTU;
1052 			}
1053 		}
1054 		ip6_rt_put(rt);
1055 	}
1056 
1057 	t->hlen = addend;
1058 }
1059 
1060 static int ip6gre_tnl_change(struct ip6_tnl *t,
1061 	const struct __ip6_tnl_parm *p, int set_mtu)
1062 {
1063 	t->parms.laddr = p->laddr;
1064 	t->parms.raddr = p->raddr;
1065 	t->parms.flags = p->flags;
1066 	t->parms.hop_limit = p->hop_limit;
1067 	t->parms.encap_limit = p->encap_limit;
1068 	t->parms.flowinfo = p->flowinfo;
1069 	t->parms.link = p->link;
1070 	t->parms.proto = p->proto;
1071 	t->parms.i_key = p->i_key;
1072 	t->parms.o_key = p->o_key;
1073 	t->parms.i_flags = p->i_flags;
1074 	t->parms.o_flags = p->o_flags;
1075 	ip6_tnl_dst_reset(t);
1076 	ip6gre_tnl_link_config(t, set_mtu);
1077 	return 0;
1078 }
1079 
1080 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1081 	const struct ip6_tnl_parm2 *u)
1082 {
1083 	p->laddr = u->laddr;
1084 	p->raddr = u->raddr;
1085 	p->flags = u->flags;
1086 	p->hop_limit = u->hop_limit;
1087 	p->encap_limit = u->encap_limit;
1088 	p->flowinfo = u->flowinfo;
1089 	p->link = u->link;
1090 	p->i_key = u->i_key;
1091 	p->o_key = u->o_key;
1092 	p->i_flags = u->i_flags;
1093 	p->o_flags = u->o_flags;
1094 	memcpy(p->name, u->name, sizeof(u->name));
1095 }
1096 
1097 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1098 	const struct __ip6_tnl_parm *p)
1099 {
1100 	u->proto = IPPROTO_GRE;
1101 	u->laddr = p->laddr;
1102 	u->raddr = p->raddr;
1103 	u->flags = p->flags;
1104 	u->hop_limit = p->hop_limit;
1105 	u->encap_limit = p->encap_limit;
1106 	u->flowinfo = p->flowinfo;
1107 	u->link = p->link;
1108 	u->i_key = p->i_key;
1109 	u->o_key = p->o_key;
1110 	u->i_flags = p->i_flags;
1111 	u->o_flags = p->o_flags;
1112 	memcpy(u->name, p->name, sizeof(u->name));
1113 }
1114 
1115 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1116 	struct ifreq *ifr, int cmd)
1117 {
1118 	int err = 0;
1119 	struct ip6_tnl_parm2 p;
1120 	struct __ip6_tnl_parm p1;
1121 	struct ip6_tnl *t;
1122 	struct net *net = dev_net(dev);
1123 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1124 
1125 	switch (cmd) {
1126 	case SIOCGETTUNNEL:
1127 		t = NULL;
1128 		if (dev == ign->fb_tunnel_dev) {
1129 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1130 				err = -EFAULT;
1131 				break;
1132 			}
1133 			ip6gre_tnl_parm_from_user(&p1, &p);
1134 			t = ip6gre_tunnel_locate(net, &p1, 0);
1135 		}
1136 		if (t == NULL)
1137 			t = netdev_priv(dev);
1138 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1139 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1140 			err = -EFAULT;
1141 		break;
1142 
1143 	case SIOCADDTUNNEL:
1144 	case SIOCCHGTUNNEL:
1145 		err = -EPERM;
1146 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1147 			goto done;
1148 
1149 		err = -EFAULT;
1150 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1151 			goto done;
1152 
1153 		err = -EINVAL;
1154 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1155 			goto done;
1156 
1157 		if (!(p.i_flags&GRE_KEY))
1158 			p.i_key = 0;
1159 		if (!(p.o_flags&GRE_KEY))
1160 			p.o_key = 0;
1161 
1162 		ip6gre_tnl_parm_from_user(&p1, &p);
1163 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1164 
1165 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1166 			if (t != NULL) {
1167 				if (t->dev != dev) {
1168 					err = -EEXIST;
1169 					break;
1170 				}
1171 			} else {
1172 				t = netdev_priv(dev);
1173 
1174 				ip6gre_tunnel_unlink(ign, t);
1175 				synchronize_net();
1176 				ip6gre_tnl_change(t, &p1, 1);
1177 				ip6gre_tunnel_link(ign, t);
1178 				netdev_state_change(dev);
1179 			}
1180 		}
1181 
1182 		if (t) {
1183 			err = 0;
1184 
1185 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1186 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1187 				err = -EFAULT;
1188 		} else
1189 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1190 		break;
1191 
1192 	case SIOCDELTUNNEL:
1193 		err = -EPERM;
1194 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1195 			goto done;
1196 
1197 		if (dev == ign->fb_tunnel_dev) {
1198 			err = -EFAULT;
1199 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1200 				goto done;
1201 			err = -ENOENT;
1202 			ip6gre_tnl_parm_from_user(&p1, &p);
1203 			t = ip6gre_tunnel_locate(net, &p1, 0);
1204 			if (t == NULL)
1205 				goto done;
1206 			err = -EPERM;
1207 			if (t == netdev_priv(ign->fb_tunnel_dev))
1208 				goto done;
1209 			dev = t->dev;
1210 		}
1211 		unregister_netdevice(dev);
1212 		err = 0;
1213 		break;
1214 
1215 	default:
1216 		err = -EINVAL;
1217 	}
1218 
1219 done:
1220 	return err;
1221 }
1222 
1223 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1224 {
1225 	struct ip6_tnl *tunnel = netdev_priv(dev);
1226 	if (new_mtu < 68 ||
1227 	    new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1228 		return -EINVAL;
1229 	dev->mtu = new_mtu;
1230 	return 0;
1231 }
1232 
1233 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1234 			unsigned short type,
1235 			const void *daddr, const void *saddr, unsigned int len)
1236 {
1237 	struct ip6_tnl *t = netdev_priv(dev);
1238 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1239 	__be16 *p = (__be16 *)(ipv6h+1);
1240 
1241 	ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1242 	ipv6h->hop_limit = t->parms.hop_limit;
1243 	ipv6h->nexthdr = NEXTHDR_GRE;
1244 	ipv6h->saddr = t->parms.laddr;
1245 	ipv6h->daddr = t->parms.raddr;
1246 
1247 	p[0]		= t->parms.o_flags;
1248 	p[1]		= htons(type);
1249 
1250 	/*
1251 	 *	Set the source hardware address.
1252 	 */
1253 
1254 	if (saddr)
1255 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1256 	if (daddr)
1257 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1258 	if (!ipv6_addr_any(&ipv6h->daddr))
1259 		return t->hlen;
1260 
1261 	return -t->hlen;
1262 }
1263 
1264 static const struct header_ops ip6gre_header_ops = {
1265 	.create	= ip6gre_header,
1266 };
1267 
1268 static const struct net_device_ops ip6gre_netdev_ops = {
1269 	.ndo_init		= ip6gre_tunnel_init,
1270 	.ndo_uninit		= ip6gre_tunnel_uninit,
1271 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1272 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1273 	.ndo_change_mtu		= ip6gre_tunnel_change_mtu,
1274 	.ndo_get_stats64	= ip6gre_get_stats64,
1275 };
1276 
1277 static void ip6gre_dev_free(struct net_device *dev)
1278 {
1279 	free_percpu(dev->tstats);
1280 	free_netdev(dev);
1281 }
1282 
1283 static void ip6gre_tunnel_setup(struct net_device *dev)
1284 {
1285 	struct ip6_tnl *t;
1286 
1287 	dev->netdev_ops = &ip6gre_netdev_ops;
1288 	dev->destructor = ip6gre_dev_free;
1289 
1290 	dev->type = ARPHRD_IP6GRE;
1291 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1292 	dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1293 	t = netdev_priv(dev);
1294 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1295 		dev->mtu -= 8;
1296 	dev->flags |= IFF_NOARP;
1297 	dev->iflink = 0;
1298 	dev->addr_len = sizeof(struct in6_addr);
1299 	dev->features |= NETIF_F_NETNS_LOCAL;
1300 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1301 }
1302 
1303 static int ip6gre_tunnel_init(struct net_device *dev)
1304 {
1305 	struct ip6_tnl *tunnel;
1306 
1307 	tunnel = netdev_priv(dev);
1308 
1309 	tunnel->dev = dev;
1310 	strcpy(tunnel->parms.name, dev->name);
1311 
1312 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1313 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1314 
1315 	if (ipv6_addr_any(&tunnel->parms.raddr))
1316 		dev->header_ops = &ip6gre_header_ops;
1317 
1318 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1319 	if (!dev->tstats)
1320 		return -ENOMEM;
1321 
1322 	return 0;
1323 }
1324 
1325 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1326 {
1327 	struct ip6_tnl *tunnel = netdev_priv(dev);
1328 
1329 	tunnel->dev = dev;
1330 	strcpy(tunnel->parms.name, dev->name);
1331 
1332 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1333 
1334 	dev_hold(dev);
1335 }
1336 
1337 
1338 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1339 	.handler     = ip6gre_rcv,
1340 	.err_handler = ip6gre_err,
1341 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1342 };
1343 
1344 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1345 	struct list_head *head)
1346 {
1347 	int prio;
1348 
1349 	for (prio = 0; prio < 4; prio++) {
1350 		int h;
1351 		for (h = 0; h < HASH_SIZE; h++) {
1352 			struct ip6_tnl *t;
1353 
1354 			t = rtnl_dereference(ign->tunnels[prio][h]);
1355 
1356 			while (t != NULL) {
1357 				unregister_netdevice_queue(t->dev, head);
1358 				t = rtnl_dereference(t->next);
1359 			}
1360 		}
1361 	}
1362 }
1363 
1364 static int __net_init ip6gre_init_net(struct net *net)
1365 {
1366 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1367 	int err;
1368 
1369 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1370 					   ip6gre_tunnel_setup);
1371 	if (!ign->fb_tunnel_dev) {
1372 		err = -ENOMEM;
1373 		goto err_alloc_dev;
1374 	}
1375 	dev_net_set(ign->fb_tunnel_dev, net);
1376 
1377 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1378 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1379 
1380 	err = register_netdev(ign->fb_tunnel_dev);
1381 	if (err)
1382 		goto err_reg_dev;
1383 
1384 	rcu_assign_pointer(ign->tunnels_wc[0],
1385 			   netdev_priv(ign->fb_tunnel_dev));
1386 	return 0;
1387 
1388 err_reg_dev:
1389 	ip6gre_dev_free(ign->fb_tunnel_dev);
1390 err_alloc_dev:
1391 	return err;
1392 }
1393 
1394 static void __net_exit ip6gre_exit_net(struct net *net)
1395 {
1396 	struct ip6gre_net *ign;
1397 	LIST_HEAD(list);
1398 
1399 	ign = net_generic(net, ip6gre_net_id);
1400 	rtnl_lock();
1401 	ip6gre_destroy_tunnels(ign, &list);
1402 	unregister_netdevice_many(&list);
1403 	rtnl_unlock();
1404 }
1405 
1406 static struct pernet_operations ip6gre_net_ops = {
1407 	.init = ip6gre_init_net,
1408 	.exit = ip6gre_exit_net,
1409 	.id   = &ip6gre_net_id,
1410 	.size = sizeof(struct ip6gre_net),
1411 };
1412 
1413 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1414 {
1415 	__be16 flags;
1416 
1417 	if (!data)
1418 		return 0;
1419 
1420 	flags = 0;
1421 	if (data[IFLA_GRE_IFLAGS])
1422 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1423 	if (data[IFLA_GRE_OFLAGS])
1424 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1425 	if (flags & (GRE_VERSION|GRE_ROUTING))
1426 		return -EINVAL;
1427 
1428 	return 0;
1429 }
1430 
1431 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1432 {
1433 	struct in6_addr daddr;
1434 
1435 	if (tb[IFLA_ADDRESS]) {
1436 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1437 			return -EINVAL;
1438 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1439 			return -EADDRNOTAVAIL;
1440 	}
1441 
1442 	if (!data)
1443 		goto out;
1444 
1445 	if (data[IFLA_GRE_REMOTE]) {
1446 		nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1447 		if (ipv6_addr_any(&daddr))
1448 			return -EINVAL;
1449 	}
1450 
1451 out:
1452 	return ip6gre_tunnel_validate(tb, data);
1453 }
1454 
1455 
1456 static void ip6gre_netlink_parms(struct nlattr *data[],
1457 				struct __ip6_tnl_parm *parms)
1458 {
1459 	memset(parms, 0, sizeof(*parms));
1460 
1461 	if (!data)
1462 		return;
1463 
1464 	if (data[IFLA_GRE_LINK])
1465 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1466 
1467 	if (data[IFLA_GRE_IFLAGS])
1468 		parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1469 
1470 	if (data[IFLA_GRE_OFLAGS])
1471 		parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1472 
1473 	if (data[IFLA_GRE_IKEY])
1474 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1475 
1476 	if (data[IFLA_GRE_OKEY])
1477 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1478 
1479 	if (data[IFLA_GRE_LOCAL])
1480 		nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1481 
1482 	if (data[IFLA_GRE_REMOTE])
1483 		nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1484 
1485 	if (data[IFLA_GRE_TTL])
1486 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1487 
1488 	if (data[IFLA_GRE_ENCAP_LIMIT])
1489 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1490 
1491 	if (data[IFLA_GRE_FLOWINFO])
1492 		parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1493 
1494 	if (data[IFLA_GRE_FLAGS])
1495 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1496 }
1497 
1498 static int ip6gre_tap_init(struct net_device *dev)
1499 {
1500 	struct ip6_tnl *tunnel;
1501 
1502 	tunnel = netdev_priv(dev);
1503 
1504 	tunnel->dev = dev;
1505 	strcpy(tunnel->parms.name, dev->name);
1506 
1507 	ip6gre_tnl_link_config(tunnel, 1);
1508 
1509 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1510 	if (!dev->tstats)
1511 		return -ENOMEM;
1512 
1513 	return 0;
1514 }
1515 
1516 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1517 	.ndo_init = ip6gre_tap_init,
1518 	.ndo_uninit = ip6gre_tunnel_uninit,
1519 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1520 	.ndo_set_mac_address = eth_mac_addr,
1521 	.ndo_validate_addr = eth_validate_addr,
1522 	.ndo_change_mtu = ip6gre_tunnel_change_mtu,
1523 	.ndo_get_stats64 = ip6gre_get_stats64,
1524 };
1525 
1526 static void ip6gre_tap_setup(struct net_device *dev)
1527 {
1528 
1529 	ether_setup(dev);
1530 
1531 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1532 	dev->destructor = ip6gre_dev_free;
1533 
1534 	dev->iflink = 0;
1535 	dev->features |= NETIF_F_NETNS_LOCAL;
1536 }
1537 
1538 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1539 	struct nlattr *tb[], struct nlattr *data[])
1540 {
1541 	struct ip6_tnl *nt;
1542 	struct net *net = dev_net(dev);
1543 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1544 	int err;
1545 
1546 	nt = netdev_priv(dev);
1547 	ip6gre_netlink_parms(data, &nt->parms);
1548 
1549 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1550 		return -EEXIST;
1551 
1552 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1553 		eth_hw_addr_random(dev);
1554 
1555 	nt->dev = dev;
1556 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1557 
1558 	/* Can use a lockless transmit, unless we generate output sequences */
1559 	if (!(nt->parms.o_flags & GRE_SEQ))
1560 		dev->features |= NETIF_F_LLTX;
1561 
1562 	err = register_netdevice(dev);
1563 	if (err)
1564 		goto out;
1565 
1566 	dev_hold(dev);
1567 	ip6gre_tunnel_link(ign, nt);
1568 
1569 out:
1570 	return err;
1571 }
1572 
1573 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1574 			    struct nlattr *data[])
1575 {
1576 	struct ip6_tnl *t, *nt;
1577 	struct net *net = dev_net(dev);
1578 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1579 	struct __ip6_tnl_parm p;
1580 
1581 	if (dev == ign->fb_tunnel_dev)
1582 		return -EINVAL;
1583 
1584 	nt = netdev_priv(dev);
1585 	ip6gre_netlink_parms(data, &p);
1586 
1587 	t = ip6gre_tunnel_locate(net, &p, 0);
1588 
1589 	if (t) {
1590 		if (t->dev != dev)
1591 			return -EEXIST;
1592 	} else {
1593 		t = nt;
1594 
1595 		ip6gre_tunnel_unlink(ign, t);
1596 		ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1597 		ip6gre_tunnel_link(ign, t);
1598 		netdev_state_change(dev);
1599 	}
1600 
1601 	return 0;
1602 }
1603 
1604 static size_t ip6gre_get_size(const struct net_device *dev)
1605 {
1606 	return
1607 		/* IFLA_GRE_LINK */
1608 		nla_total_size(4) +
1609 		/* IFLA_GRE_IFLAGS */
1610 		nla_total_size(2) +
1611 		/* IFLA_GRE_OFLAGS */
1612 		nla_total_size(2) +
1613 		/* IFLA_GRE_IKEY */
1614 		nla_total_size(4) +
1615 		/* IFLA_GRE_OKEY */
1616 		nla_total_size(4) +
1617 		/* IFLA_GRE_LOCAL */
1618 		nla_total_size(sizeof(struct in6_addr)) +
1619 		/* IFLA_GRE_REMOTE */
1620 		nla_total_size(sizeof(struct in6_addr)) +
1621 		/* IFLA_GRE_TTL */
1622 		nla_total_size(1) +
1623 		/* IFLA_GRE_TOS */
1624 		nla_total_size(1) +
1625 		/* IFLA_GRE_ENCAP_LIMIT */
1626 		nla_total_size(1) +
1627 		/* IFLA_GRE_FLOWINFO */
1628 		nla_total_size(4) +
1629 		/* IFLA_GRE_FLAGS */
1630 		nla_total_size(4) +
1631 		0;
1632 }
1633 
1634 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1635 {
1636 	struct ip6_tnl *t = netdev_priv(dev);
1637 	struct __ip6_tnl_parm *p = &t->parms;
1638 
1639 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1640 	    nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1641 	    nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1642 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1643 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1644 	    nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1645 	    nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1646 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1647 	    /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1648 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1649 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1650 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1651 		goto nla_put_failure;
1652 	return 0;
1653 
1654 nla_put_failure:
1655 	return -EMSGSIZE;
1656 }
1657 
1658 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1659 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1660 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1661 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1662 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1663 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1664 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1665 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1666 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1667 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1668 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1669 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1670 };
1671 
1672 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1673 	.kind		= "ip6gre",
1674 	.maxtype	= IFLA_GRE_MAX,
1675 	.policy		= ip6gre_policy,
1676 	.priv_size	= sizeof(struct ip6_tnl),
1677 	.setup		= ip6gre_tunnel_setup,
1678 	.validate	= ip6gre_tunnel_validate,
1679 	.newlink	= ip6gre_newlink,
1680 	.changelink	= ip6gre_changelink,
1681 	.get_size	= ip6gre_get_size,
1682 	.fill_info	= ip6gre_fill_info,
1683 };
1684 
1685 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1686 	.kind		= "ip6gretap",
1687 	.maxtype	= IFLA_GRE_MAX,
1688 	.policy		= ip6gre_policy,
1689 	.priv_size	= sizeof(struct ip6_tnl),
1690 	.setup		= ip6gre_tap_setup,
1691 	.validate	= ip6gre_tap_validate,
1692 	.newlink	= ip6gre_newlink,
1693 	.changelink	= ip6gre_changelink,
1694 	.get_size	= ip6gre_get_size,
1695 	.fill_info	= ip6gre_fill_info,
1696 };
1697 
1698 /*
1699  *	And now the modules code and kernel interface.
1700  */
1701 
1702 static int __init ip6gre_init(void)
1703 {
1704 	int err;
1705 
1706 	pr_info("GRE over IPv6 tunneling driver\n");
1707 
1708 	err = register_pernet_device(&ip6gre_net_ops);
1709 	if (err < 0)
1710 		return err;
1711 
1712 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1713 	if (err < 0) {
1714 		pr_info("%s: can't add protocol\n", __func__);
1715 		goto add_proto_failed;
1716 	}
1717 
1718 	err = rtnl_link_register(&ip6gre_link_ops);
1719 	if (err < 0)
1720 		goto rtnl_link_failed;
1721 
1722 	err = rtnl_link_register(&ip6gre_tap_ops);
1723 	if (err < 0)
1724 		goto tap_ops_failed;
1725 
1726 out:
1727 	return err;
1728 
1729 tap_ops_failed:
1730 	rtnl_link_unregister(&ip6gre_link_ops);
1731 rtnl_link_failed:
1732 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1733 add_proto_failed:
1734 	unregister_pernet_device(&ip6gre_net_ops);
1735 	goto out;
1736 }
1737 
1738 static void __exit ip6gre_fini(void)
1739 {
1740 	rtnl_link_unregister(&ip6gre_tap_ops);
1741 	rtnl_link_unregister(&ip6gre_link_ops);
1742 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1743 	unregister_pernet_device(&ip6gre_net_ops);
1744 }
1745 
1746 module_init(ip6gre_init);
1747 module_exit(ip6gre_fini);
1748 MODULE_LICENSE("GPL");
1749 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1750 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1751 MODULE_ALIAS_RTNL_LINK("ip6gre");
1752 MODULE_ALIAS_NETDEV("ip6gre0");
1753