xref: /openbmc/linux/net/ipv6/ip6_gre.c (revision 275876e2)
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 t;
319 
320 	if (parms->name[0])
321 		strlcpy(name, parms->name, IFNAMSIZ);
322 	else
323 		strcpy(name, "ip6gre%d");
324 
325 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
326 			   ip6gre_tunnel_setup);
327 	if (!dev)
328 		return NULL;
329 
330 	dev_net_set(dev, net);
331 
332 	nt = netdev_priv(dev);
333 	nt->parms = *parms;
334 	dev->rtnl_link_ops = &ip6gre_link_ops;
335 
336 	nt->dev = dev;
337 	nt->net = dev_net(dev);
338 	ip6gre_tnl_link_config(nt, 1);
339 
340 	if (register_netdevice(dev) < 0)
341 		goto failed_free;
342 
343 	/* Can use a lockless transmit, unless we generate output sequences */
344 	if (!(nt->parms.o_flags & GRE_SEQ))
345 		dev->features |= NETIF_F_LLTX;
346 
347 	dev_hold(dev);
348 	ip6gre_tunnel_link(ign, nt);
349 	return nt;
350 
351 failed_free:
352 	free_netdev(dev);
353 	return NULL;
354 }
355 
356 static void ip6gre_tunnel_uninit(struct net_device *dev)
357 {
358 	struct ip6_tnl *t = netdev_priv(dev);
359 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
360 
361 	ip6gre_tunnel_unlink(ign, t);
362 	dev_put(dev);
363 }
364 
365 
366 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
367 		u8 type, u8 code, int offset, __be32 info)
368 {
369 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
370 	__be16 *p = (__be16 *)(skb->data + offset);
371 	int grehlen = offset + 4;
372 	struct ip6_tnl *t;
373 	__be16 flags;
374 
375 	flags = p[0];
376 	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
377 		if (flags&(GRE_VERSION|GRE_ROUTING))
378 			return;
379 		if (flags&GRE_KEY) {
380 			grehlen += 4;
381 			if (flags&GRE_CSUM)
382 				grehlen += 4;
383 		}
384 	}
385 
386 	/* If only 8 bytes returned, keyed message will be dropped here */
387 	if (!pskb_may_pull(skb, grehlen))
388 		return;
389 	ipv6h = (const struct ipv6hdr *)skb->data;
390 	p = (__be16 *)(skb->data + offset);
391 
392 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
393 				flags & GRE_KEY ?
394 				*(((__be32 *)p) + (grehlen / 4) - 1) : 0,
395 				p[1]);
396 	if (t == NULL)
397 		return;
398 
399 	switch (type) {
400 		__u32 teli;
401 		struct ipv6_tlv_tnl_enc_lim *tel;
402 		__u32 mtu;
403 	case ICMPV6_DEST_UNREACH:
404 		net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
405 				     t->parms.name);
406 		break;
407 	case ICMPV6_TIME_EXCEED:
408 		if (code == ICMPV6_EXC_HOPLIMIT) {
409 			net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
410 					     t->parms.name);
411 		}
412 		break;
413 	case ICMPV6_PARAMPROB:
414 		teli = 0;
415 		if (code == ICMPV6_HDR_FIELD)
416 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
417 
418 		if (teli && teli == info - 2) {
419 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
420 			if (tel->encap_limit == 0) {
421 				net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
422 						     t->parms.name);
423 			}
424 		} else {
425 			net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
426 					     t->parms.name);
427 		}
428 		break;
429 	case ICMPV6_PKT_TOOBIG:
430 		mtu = info - offset;
431 		if (mtu < IPV6_MIN_MTU)
432 			mtu = IPV6_MIN_MTU;
433 		t->dev->mtu = mtu;
434 		break;
435 	}
436 
437 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
438 		t->err_count++;
439 	else
440 		t->err_count = 1;
441 	t->err_time = jiffies;
442 }
443 
444 static int ip6gre_rcv(struct sk_buff *skb)
445 {
446 	const struct ipv6hdr *ipv6h;
447 	u8     *h;
448 	__be16    flags;
449 	__sum16   csum = 0;
450 	__be32 key = 0;
451 	u32    seqno = 0;
452 	struct ip6_tnl *tunnel;
453 	int    offset = 4;
454 	__be16 gre_proto;
455 	int err;
456 
457 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
458 		goto drop;
459 
460 	ipv6h = ipv6_hdr(skb);
461 	h = skb->data;
462 	flags = *(__be16 *)h;
463 
464 	if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
465 		/* - Version must be 0.
466 		   - We do not support routing headers.
467 		 */
468 		if (flags&(GRE_VERSION|GRE_ROUTING))
469 			goto drop;
470 
471 		if (flags&GRE_CSUM) {
472 			csum = skb_checksum_simple_validate(skb);
473 			offset += 4;
474 		}
475 		if (flags&GRE_KEY) {
476 			key = *(__be32 *)(h + offset);
477 			offset += 4;
478 		}
479 		if (flags&GRE_SEQ) {
480 			seqno = ntohl(*(__be32 *)(h + offset));
481 			offset += 4;
482 		}
483 	}
484 
485 	gre_proto = *(__be16 *)(h + 2);
486 
487 	tunnel = ip6gre_tunnel_lookup(skb->dev,
488 					  &ipv6h->saddr, &ipv6h->daddr, key,
489 					  gre_proto);
490 	if (tunnel) {
491 		struct pcpu_sw_netstats *tstats;
492 
493 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
494 			goto drop;
495 
496 		if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
497 			tunnel->dev->stats.rx_dropped++;
498 			goto drop;
499 		}
500 
501 		skb->protocol = gre_proto;
502 		/* WCCP version 1 and 2 protocol decoding.
503 		 * - Change protocol to IP
504 		 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
505 		 */
506 		if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
507 			skb->protocol = htons(ETH_P_IP);
508 			if ((*(h + offset) & 0xF0) != 0x40)
509 				offset += 4;
510 		}
511 
512 		skb->mac_header = skb->network_header;
513 		__pskb_pull(skb, offset);
514 		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
515 
516 		if (((flags&GRE_CSUM) && csum) ||
517 		    (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
518 			tunnel->dev->stats.rx_crc_errors++;
519 			tunnel->dev->stats.rx_errors++;
520 			goto drop;
521 		}
522 		if (tunnel->parms.i_flags&GRE_SEQ) {
523 			if (!(flags&GRE_SEQ) ||
524 			    (tunnel->i_seqno &&
525 					(s32)(seqno - tunnel->i_seqno) < 0)) {
526 				tunnel->dev->stats.rx_fifo_errors++;
527 				tunnel->dev->stats.rx_errors++;
528 				goto drop;
529 			}
530 			tunnel->i_seqno = seqno + 1;
531 		}
532 
533 		/* Warning: All skb pointers will be invalidated! */
534 		if (tunnel->dev->type == ARPHRD_ETHER) {
535 			if (!pskb_may_pull(skb, ETH_HLEN)) {
536 				tunnel->dev->stats.rx_length_errors++;
537 				tunnel->dev->stats.rx_errors++;
538 				goto drop;
539 			}
540 
541 			ipv6h = ipv6_hdr(skb);
542 			skb->protocol = eth_type_trans(skb, tunnel->dev);
543 			skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
544 		}
545 
546 		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
547 
548 		skb_reset_network_header(skb);
549 
550 		err = IP6_ECN_decapsulate(ipv6h, skb);
551 		if (unlikely(err)) {
552 			if (log_ecn_error)
553 				net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
554 						     &ipv6h->saddr,
555 						     ipv6_get_dsfield(ipv6h));
556 			if (err > 1) {
557 				++tunnel->dev->stats.rx_frame_errors;
558 				++tunnel->dev->stats.rx_errors;
559 				goto drop;
560 			}
561 		}
562 
563 		tstats = this_cpu_ptr(tunnel->dev->tstats);
564 		u64_stats_update_begin(&tstats->syncp);
565 		tstats->rx_packets++;
566 		tstats->rx_bytes += skb->len;
567 		u64_stats_update_end(&tstats->syncp);
568 
569 		netif_rx(skb);
570 
571 		return 0;
572 	}
573 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
574 
575 drop:
576 	kfree_skb(skb);
577 	return 0;
578 }
579 
580 struct ipv6_tel_txoption {
581 	struct ipv6_txoptions ops;
582 	__u8 dst_opt[8];
583 };
584 
585 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
586 {
587 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
588 
589 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
590 	opt->dst_opt[3] = 1;
591 	opt->dst_opt[4] = encap_limit;
592 	opt->dst_opt[5] = IPV6_TLV_PADN;
593 	opt->dst_opt[6] = 1;
594 
595 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
596 	opt->ops.opt_nflen = 8;
597 }
598 
599 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
600 			 struct net_device *dev,
601 			 __u8 dsfield,
602 			 struct flowi6 *fl6,
603 			 int encap_limit,
604 			 __u32 *pmtu)
605 {
606 	struct ip6_tnl *tunnel = netdev_priv(dev);
607 	struct net *net = tunnel->net;
608 	struct net_device *tdev;    /* Device to other host */
609 	struct ipv6hdr  *ipv6h;     /* Our new IP header */
610 	unsigned int max_headroom = 0; /* The extra header space needed */
611 	int    gre_hlen;
612 	struct ipv6_tel_txoption opt;
613 	int    mtu;
614 	struct dst_entry *dst = NULL, *ndst = NULL;
615 	struct net_device_stats *stats = &tunnel->dev->stats;
616 	int err = -1;
617 	u8 proto;
618 	struct sk_buff *new_skb;
619 
620 	if (dev->type == ARPHRD_ETHER)
621 		IPCB(skb)->flags = 0;
622 
623 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
624 		gre_hlen = 0;
625 		ipv6h = (struct ipv6hdr *)skb->data;
626 		fl6->daddr = ipv6h->daddr;
627 	} else {
628 		gre_hlen = tunnel->hlen;
629 		fl6->daddr = tunnel->parms.raddr;
630 	}
631 
632 	if (!fl6->flowi6_mark)
633 		dst = ip6_tnl_dst_check(tunnel);
634 
635 	if (!dst) {
636 		ndst = ip6_route_output(net, NULL, fl6);
637 
638 		if (ndst->error)
639 			goto tx_err_link_failure;
640 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
641 		if (IS_ERR(ndst)) {
642 			err = PTR_ERR(ndst);
643 			ndst = NULL;
644 			goto tx_err_link_failure;
645 		}
646 		dst = ndst;
647 	}
648 
649 	tdev = dst->dev;
650 
651 	if (tdev == dev) {
652 		stats->collisions++;
653 		net_warn_ratelimited("%s: Local routing loop detected!\n",
654 				     tunnel->parms.name);
655 		goto tx_err_dst_release;
656 	}
657 
658 	mtu = dst_mtu(dst) - sizeof(*ipv6h);
659 	if (encap_limit >= 0) {
660 		max_headroom += 8;
661 		mtu -= 8;
662 	}
663 	if (mtu < IPV6_MIN_MTU)
664 		mtu = IPV6_MIN_MTU;
665 	if (skb_dst(skb))
666 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
667 	if (skb->len > mtu) {
668 		*pmtu = mtu;
669 		err = -EMSGSIZE;
670 		goto tx_err_dst_release;
671 	}
672 
673 	if (tunnel->err_count > 0) {
674 		if (time_before(jiffies,
675 				tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
676 			tunnel->err_count--;
677 
678 			dst_link_failure(skb);
679 		} else
680 			tunnel->err_count = 0;
681 	}
682 
683 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
684 
685 	max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
686 
687 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
688 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
689 		new_skb = skb_realloc_headroom(skb, max_headroom);
690 		if (max_headroom > dev->needed_headroom)
691 			dev->needed_headroom = max_headroom;
692 		if (!new_skb)
693 			goto tx_err_dst_release;
694 
695 		if (skb->sk)
696 			skb_set_owner_w(new_skb, skb->sk);
697 		consume_skb(skb);
698 		skb = new_skb;
699 	}
700 
701 	if (fl6->flowi6_mark) {
702 		skb_dst_set(skb, dst);
703 		ndst = NULL;
704 	} else {
705 		skb_dst_set_noref(skb, dst);
706 	}
707 
708 	proto = NEXTHDR_GRE;
709 	if (encap_limit >= 0) {
710 		init_tel_txopt(&opt, encap_limit);
711 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
712 	}
713 
714 	if (likely(!skb->encapsulation)) {
715 		skb_reset_inner_headers(skb);
716 		skb->encapsulation = 1;
717 	}
718 
719 	skb_push(skb, gre_hlen);
720 	skb_reset_network_header(skb);
721 	skb_set_transport_header(skb, sizeof(*ipv6h));
722 
723 	/*
724 	 *	Push down and install the IP header.
725 	 */
726 	ipv6h = ipv6_hdr(skb);
727 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
728 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
729 	ipv6h->hop_limit = tunnel->parms.hop_limit;
730 	ipv6h->nexthdr = proto;
731 	ipv6h->saddr = fl6->saddr;
732 	ipv6h->daddr = fl6->daddr;
733 
734 	((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
735 	((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
736 				   htons(ETH_P_TEB) : skb->protocol;
737 
738 	if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
739 		__be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
740 
741 		if (tunnel->parms.o_flags&GRE_SEQ) {
742 			++tunnel->o_seqno;
743 			*ptr = htonl(tunnel->o_seqno);
744 			ptr--;
745 		}
746 		if (tunnel->parms.o_flags&GRE_KEY) {
747 			*ptr = tunnel->parms.o_key;
748 			ptr--;
749 		}
750 		if (tunnel->parms.o_flags&GRE_CSUM) {
751 			*ptr = 0;
752 			*(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
753 				skb->len - sizeof(struct ipv6hdr));
754 		}
755 	}
756 
757 	ip6tunnel_xmit(skb, dev);
758 	if (ndst)
759 		ip6_tnl_dst_store(tunnel, ndst);
760 	return 0;
761 tx_err_link_failure:
762 	stats->tx_carrier_errors++;
763 	dst_link_failure(skb);
764 tx_err_dst_release:
765 	dst_release(ndst);
766 	return err;
767 }
768 
769 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
770 {
771 	struct ip6_tnl *t = netdev_priv(dev);
772 	const struct iphdr  *iph = ip_hdr(skb);
773 	int encap_limit = -1;
774 	struct flowi6 fl6;
775 	__u8 dsfield;
776 	__u32 mtu;
777 	int err;
778 
779 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
780 		encap_limit = t->parms.encap_limit;
781 
782 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
783 	fl6.flowi6_proto = IPPROTO_IPIP;
784 
785 	dsfield = ipv4_get_dsfield(iph);
786 
787 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
788 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
789 					  & IPV6_TCLASS_MASK;
790 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
791 		fl6.flowi6_mark = skb->mark;
792 
793 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
794 	if (err != 0) {
795 		/* XXX: send ICMP error even if DF is not set. */
796 		if (err == -EMSGSIZE)
797 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
798 				  htonl(mtu));
799 		return -1;
800 	}
801 
802 	return 0;
803 }
804 
805 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
806 {
807 	struct ip6_tnl *t = netdev_priv(dev);
808 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
809 	int encap_limit = -1;
810 	__u16 offset;
811 	struct flowi6 fl6;
812 	__u8 dsfield;
813 	__u32 mtu;
814 	int err;
815 
816 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
817 		return -1;
818 
819 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
820 	if (offset > 0) {
821 		struct ipv6_tlv_tnl_enc_lim *tel;
822 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
823 		if (tel->encap_limit == 0) {
824 			icmpv6_send(skb, ICMPV6_PARAMPROB,
825 				    ICMPV6_HDR_FIELD, offset + 2);
826 			return -1;
827 		}
828 		encap_limit = tel->encap_limit - 1;
829 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
830 		encap_limit = t->parms.encap_limit;
831 
832 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
833 	fl6.flowi6_proto = IPPROTO_IPV6;
834 
835 	dsfield = ipv6_get_dsfield(ipv6h);
836 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
837 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
838 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
839 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
840 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
841 		fl6.flowi6_mark = skb->mark;
842 
843 	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
844 	if (err != 0) {
845 		if (err == -EMSGSIZE)
846 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
847 		return -1;
848 	}
849 
850 	return 0;
851 }
852 
853 /**
854  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
855  *   @t: the outgoing tunnel device
856  *   @hdr: IPv6 header from the incoming packet
857  *
858  * Description:
859  *   Avoid trivial tunneling loop by checking that tunnel exit-point
860  *   doesn't match source of incoming packet.
861  *
862  * Return:
863  *   1 if conflict,
864  *   0 else
865  **/
866 
867 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
868 	const struct ipv6hdr *hdr)
869 {
870 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
871 }
872 
873 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
874 {
875 	struct ip6_tnl *t = netdev_priv(dev);
876 	int encap_limit = -1;
877 	struct flowi6 fl6;
878 	__u32 mtu;
879 	int err;
880 
881 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
882 		encap_limit = t->parms.encap_limit;
883 
884 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
885 	fl6.flowi6_proto = skb->protocol;
886 
887 	err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
888 
889 	return err;
890 }
891 
892 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
893 	struct net_device *dev)
894 {
895 	struct ip6_tnl *t = netdev_priv(dev);
896 	struct net_device_stats *stats = &t->dev->stats;
897 	int ret;
898 
899 	if (!ip6_tnl_xmit_ctl(t))
900 		goto tx_err;
901 
902 	switch (skb->protocol) {
903 	case htons(ETH_P_IP):
904 		ret = ip6gre_xmit_ipv4(skb, dev);
905 		break;
906 	case htons(ETH_P_IPV6):
907 		ret = ip6gre_xmit_ipv6(skb, dev);
908 		break;
909 	default:
910 		ret = ip6gre_xmit_other(skb, dev);
911 		break;
912 	}
913 
914 	if (ret < 0)
915 		goto tx_err;
916 
917 	return NETDEV_TX_OK;
918 
919 tx_err:
920 	stats->tx_errors++;
921 	stats->tx_dropped++;
922 	kfree_skb(skb);
923 	return NETDEV_TX_OK;
924 }
925 
926 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
927 {
928 	struct net_device *dev = t->dev;
929 	struct __ip6_tnl_parm *p = &t->parms;
930 	struct flowi6 *fl6 = &t->fl.u.ip6;
931 	int addend = sizeof(struct ipv6hdr) + 4;
932 
933 	if (dev->type != ARPHRD_ETHER) {
934 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
935 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
936 	}
937 
938 	/* Set up flowi template */
939 	fl6->saddr = p->laddr;
940 	fl6->daddr = p->raddr;
941 	fl6->flowi6_oif = p->link;
942 	fl6->flowlabel = 0;
943 
944 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
945 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
946 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
947 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
948 
949 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
950 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
951 
952 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
953 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
954 		dev->flags |= IFF_POINTOPOINT;
955 	else
956 		dev->flags &= ~IFF_POINTOPOINT;
957 
958 	dev->iflink = p->link;
959 
960 	/* Precalculate GRE options length */
961 	if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
962 		if (t->parms.o_flags&GRE_CSUM)
963 			addend += 4;
964 		if (t->parms.o_flags&GRE_KEY)
965 			addend += 4;
966 		if (t->parms.o_flags&GRE_SEQ)
967 			addend += 4;
968 	}
969 	t->hlen = addend;
970 
971 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
972 		int strict = (ipv6_addr_type(&p->raddr) &
973 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
974 
975 		struct rt6_info *rt = rt6_lookup(t->net,
976 						 &p->raddr, &p->laddr,
977 						 p->link, strict);
978 
979 		if (rt == NULL)
980 			return;
981 
982 		if (rt->dst.dev) {
983 			dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
984 
985 			if (set_mtu) {
986 				dev->mtu = rt->dst.dev->mtu - addend;
987 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
988 					dev->mtu -= 8;
989 
990 				if (dev->mtu < IPV6_MIN_MTU)
991 					dev->mtu = IPV6_MIN_MTU;
992 			}
993 		}
994 		ip6_rt_put(rt);
995 	}
996 }
997 
998 static int ip6gre_tnl_change(struct ip6_tnl *t,
999 	const struct __ip6_tnl_parm *p, int set_mtu)
1000 {
1001 	t->parms.laddr = p->laddr;
1002 	t->parms.raddr = p->raddr;
1003 	t->parms.flags = p->flags;
1004 	t->parms.hop_limit = p->hop_limit;
1005 	t->parms.encap_limit = p->encap_limit;
1006 	t->parms.flowinfo = p->flowinfo;
1007 	t->parms.link = p->link;
1008 	t->parms.proto = p->proto;
1009 	t->parms.i_key = p->i_key;
1010 	t->parms.o_key = p->o_key;
1011 	t->parms.i_flags = p->i_flags;
1012 	t->parms.o_flags = p->o_flags;
1013 	ip6_tnl_dst_reset(t);
1014 	ip6gre_tnl_link_config(t, set_mtu);
1015 	return 0;
1016 }
1017 
1018 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1019 	const struct ip6_tnl_parm2 *u)
1020 {
1021 	p->laddr = u->laddr;
1022 	p->raddr = u->raddr;
1023 	p->flags = u->flags;
1024 	p->hop_limit = u->hop_limit;
1025 	p->encap_limit = u->encap_limit;
1026 	p->flowinfo = u->flowinfo;
1027 	p->link = u->link;
1028 	p->i_key = u->i_key;
1029 	p->o_key = u->o_key;
1030 	p->i_flags = u->i_flags;
1031 	p->o_flags = u->o_flags;
1032 	memcpy(p->name, u->name, sizeof(u->name));
1033 }
1034 
1035 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1036 	const struct __ip6_tnl_parm *p)
1037 {
1038 	u->proto = IPPROTO_GRE;
1039 	u->laddr = p->laddr;
1040 	u->raddr = p->raddr;
1041 	u->flags = p->flags;
1042 	u->hop_limit = p->hop_limit;
1043 	u->encap_limit = p->encap_limit;
1044 	u->flowinfo = p->flowinfo;
1045 	u->link = p->link;
1046 	u->i_key = p->i_key;
1047 	u->o_key = p->o_key;
1048 	u->i_flags = p->i_flags;
1049 	u->o_flags = p->o_flags;
1050 	memcpy(u->name, p->name, sizeof(u->name));
1051 }
1052 
1053 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1054 	struct ifreq *ifr, int cmd)
1055 {
1056 	int err = 0;
1057 	struct ip6_tnl_parm2 p;
1058 	struct __ip6_tnl_parm p1;
1059 	struct ip6_tnl *t = netdev_priv(dev);
1060 	struct net *net = t->net;
1061 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1062 
1063 	switch (cmd) {
1064 	case SIOCGETTUNNEL:
1065 		if (dev == ign->fb_tunnel_dev) {
1066 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1067 				err = -EFAULT;
1068 				break;
1069 			}
1070 			ip6gre_tnl_parm_from_user(&p1, &p);
1071 			t = ip6gre_tunnel_locate(net, &p1, 0);
1072 			if (t == NULL)
1073 				t = netdev_priv(dev);
1074 		}
1075 		memset(&p, 0, sizeof(p));
1076 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1077 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1078 			err = -EFAULT;
1079 		break;
1080 
1081 	case SIOCADDTUNNEL:
1082 	case SIOCCHGTUNNEL:
1083 		err = -EPERM;
1084 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1085 			goto done;
1086 
1087 		err = -EFAULT;
1088 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1089 			goto done;
1090 
1091 		err = -EINVAL;
1092 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1093 			goto done;
1094 
1095 		if (!(p.i_flags&GRE_KEY))
1096 			p.i_key = 0;
1097 		if (!(p.o_flags&GRE_KEY))
1098 			p.o_key = 0;
1099 
1100 		ip6gre_tnl_parm_from_user(&p1, &p);
1101 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1102 
1103 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1104 			if (t != NULL) {
1105 				if (t->dev != dev) {
1106 					err = -EEXIST;
1107 					break;
1108 				}
1109 			} else {
1110 				t = netdev_priv(dev);
1111 
1112 				ip6gre_tunnel_unlink(ign, t);
1113 				synchronize_net();
1114 				ip6gre_tnl_change(t, &p1, 1);
1115 				ip6gre_tunnel_link(ign, t);
1116 				netdev_state_change(dev);
1117 			}
1118 		}
1119 
1120 		if (t) {
1121 			err = 0;
1122 
1123 			memset(&p, 0, sizeof(p));
1124 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1125 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1126 				err = -EFAULT;
1127 		} else
1128 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1129 		break;
1130 
1131 	case SIOCDELTUNNEL:
1132 		err = -EPERM;
1133 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1134 			goto done;
1135 
1136 		if (dev == ign->fb_tunnel_dev) {
1137 			err = -EFAULT;
1138 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1139 				goto done;
1140 			err = -ENOENT;
1141 			ip6gre_tnl_parm_from_user(&p1, &p);
1142 			t = ip6gre_tunnel_locate(net, &p1, 0);
1143 			if (t == NULL)
1144 				goto done;
1145 			err = -EPERM;
1146 			if (t == netdev_priv(ign->fb_tunnel_dev))
1147 				goto done;
1148 			dev = t->dev;
1149 		}
1150 		unregister_netdevice(dev);
1151 		err = 0;
1152 		break;
1153 
1154 	default:
1155 		err = -EINVAL;
1156 	}
1157 
1158 done:
1159 	return err;
1160 }
1161 
1162 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1163 {
1164 	if (new_mtu < 68 ||
1165 	    new_mtu > 0xFFF8 - dev->hard_header_len)
1166 		return -EINVAL;
1167 	dev->mtu = new_mtu;
1168 	return 0;
1169 }
1170 
1171 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1172 			unsigned short type,
1173 			const void *daddr, const void *saddr, unsigned int len)
1174 {
1175 	struct ip6_tnl *t = netdev_priv(dev);
1176 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1177 	__be16 *p = (__be16 *)(ipv6h+1);
1178 
1179 	ip6_flow_hdr(ipv6h, 0,
1180 		     ip6_make_flowlabel(dev_net(dev), skb,
1181 					t->fl.u.ip6.flowlabel, false));
1182 	ipv6h->hop_limit = t->parms.hop_limit;
1183 	ipv6h->nexthdr = NEXTHDR_GRE;
1184 	ipv6h->saddr = t->parms.laddr;
1185 	ipv6h->daddr = t->parms.raddr;
1186 
1187 	p[0]		= t->parms.o_flags;
1188 	p[1]		= htons(type);
1189 
1190 	/*
1191 	 *	Set the source hardware address.
1192 	 */
1193 
1194 	if (saddr)
1195 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1196 	if (daddr)
1197 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1198 	if (!ipv6_addr_any(&ipv6h->daddr))
1199 		return t->hlen;
1200 
1201 	return -t->hlen;
1202 }
1203 
1204 static const struct header_ops ip6gre_header_ops = {
1205 	.create	= ip6gre_header,
1206 };
1207 
1208 static const struct net_device_ops ip6gre_netdev_ops = {
1209 	.ndo_init		= ip6gre_tunnel_init,
1210 	.ndo_uninit		= ip6gre_tunnel_uninit,
1211 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1212 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1213 	.ndo_change_mtu		= ip6gre_tunnel_change_mtu,
1214 	.ndo_get_stats64	= ip_tunnel_get_stats64,
1215 };
1216 
1217 static void ip6gre_dev_free(struct net_device *dev)
1218 {
1219 	free_percpu(dev->tstats);
1220 	free_netdev(dev);
1221 }
1222 
1223 static void ip6gre_tunnel_setup(struct net_device *dev)
1224 {
1225 	struct ip6_tnl *t;
1226 
1227 	dev->netdev_ops = &ip6gre_netdev_ops;
1228 	dev->destructor = ip6gre_dev_free;
1229 
1230 	dev->type = ARPHRD_IP6GRE;
1231 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1232 	dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1233 	t = netdev_priv(dev);
1234 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1235 		dev->mtu -= 8;
1236 	dev->flags |= IFF_NOARP;
1237 	dev->iflink = 0;
1238 	dev->addr_len = sizeof(struct in6_addr);
1239 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1240 }
1241 
1242 static int ip6gre_tunnel_init(struct net_device *dev)
1243 {
1244 	struct ip6_tnl *tunnel;
1245 	int i;
1246 
1247 	tunnel = netdev_priv(dev);
1248 
1249 	tunnel->dev = dev;
1250 	tunnel->net = dev_net(dev);
1251 	strcpy(tunnel->parms.name, dev->name);
1252 
1253 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1254 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1255 
1256 	if (ipv6_addr_any(&tunnel->parms.raddr))
1257 		dev->header_ops = &ip6gre_header_ops;
1258 
1259 	dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1260 	if (!dev->tstats)
1261 		return -ENOMEM;
1262 
1263 	for_each_possible_cpu(i) {
1264 		struct pcpu_sw_netstats *ip6gre_tunnel_stats;
1265 		ip6gre_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1266 		u64_stats_init(&ip6gre_tunnel_stats->syncp);
1267 	}
1268 
1269 
1270 	return 0;
1271 }
1272 
1273 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1274 {
1275 	struct ip6_tnl *tunnel = netdev_priv(dev);
1276 
1277 	tunnel->dev = dev;
1278 	tunnel->net = dev_net(dev);
1279 	strcpy(tunnel->parms.name, dev->name);
1280 
1281 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1282 
1283 	dev_hold(dev);
1284 }
1285 
1286 
1287 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1288 	.handler     = ip6gre_rcv,
1289 	.err_handler = ip6gre_err,
1290 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1291 };
1292 
1293 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1294 {
1295 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1296 	struct net_device *dev, *aux;
1297 	int prio;
1298 
1299 	for_each_netdev_safe(net, dev, aux)
1300 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1301 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1302 			unregister_netdevice_queue(dev, head);
1303 
1304 	for (prio = 0; prio < 4; prio++) {
1305 		int h;
1306 		for (h = 0; h < HASH_SIZE; h++) {
1307 			struct ip6_tnl *t;
1308 
1309 			t = rtnl_dereference(ign->tunnels[prio][h]);
1310 
1311 			while (t != NULL) {
1312 				/* If dev is in the same netns, it has already
1313 				 * been added to the list by the previous loop.
1314 				 */
1315 				if (!net_eq(dev_net(t->dev), net))
1316 					unregister_netdevice_queue(t->dev,
1317 								   head);
1318 				t = rtnl_dereference(t->next);
1319 			}
1320 		}
1321 	}
1322 }
1323 
1324 static int __net_init ip6gre_init_net(struct net *net)
1325 {
1326 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1327 	int err;
1328 
1329 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1330 					  NET_NAME_UNKNOWN,
1331 					  ip6gre_tunnel_setup);
1332 	if (!ign->fb_tunnel_dev) {
1333 		err = -ENOMEM;
1334 		goto err_alloc_dev;
1335 	}
1336 	dev_net_set(ign->fb_tunnel_dev, net);
1337 	/* FB netdevice is special: we have one, and only one per netns.
1338 	 * Allowing to move it to another netns is clearly unsafe.
1339 	 */
1340 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1341 
1342 
1343 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1344 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1345 
1346 	err = register_netdev(ign->fb_tunnel_dev);
1347 	if (err)
1348 		goto err_reg_dev;
1349 
1350 	rcu_assign_pointer(ign->tunnels_wc[0],
1351 			   netdev_priv(ign->fb_tunnel_dev));
1352 	return 0;
1353 
1354 err_reg_dev:
1355 	ip6gre_dev_free(ign->fb_tunnel_dev);
1356 err_alloc_dev:
1357 	return err;
1358 }
1359 
1360 static void __net_exit ip6gre_exit_net(struct net *net)
1361 {
1362 	LIST_HEAD(list);
1363 
1364 	rtnl_lock();
1365 	ip6gre_destroy_tunnels(net, &list);
1366 	unregister_netdevice_many(&list);
1367 	rtnl_unlock();
1368 }
1369 
1370 static struct pernet_operations ip6gre_net_ops = {
1371 	.init = ip6gre_init_net,
1372 	.exit = ip6gre_exit_net,
1373 	.id   = &ip6gre_net_id,
1374 	.size = sizeof(struct ip6gre_net),
1375 };
1376 
1377 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1378 {
1379 	__be16 flags;
1380 
1381 	if (!data)
1382 		return 0;
1383 
1384 	flags = 0;
1385 	if (data[IFLA_GRE_IFLAGS])
1386 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1387 	if (data[IFLA_GRE_OFLAGS])
1388 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1389 	if (flags & (GRE_VERSION|GRE_ROUTING))
1390 		return -EINVAL;
1391 
1392 	return 0;
1393 }
1394 
1395 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1396 {
1397 	struct in6_addr daddr;
1398 
1399 	if (tb[IFLA_ADDRESS]) {
1400 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1401 			return -EINVAL;
1402 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1403 			return -EADDRNOTAVAIL;
1404 	}
1405 
1406 	if (!data)
1407 		goto out;
1408 
1409 	if (data[IFLA_GRE_REMOTE]) {
1410 		nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1411 		if (ipv6_addr_any(&daddr))
1412 			return -EINVAL;
1413 	}
1414 
1415 out:
1416 	return ip6gre_tunnel_validate(tb, data);
1417 }
1418 
1419 
1420 static void ip6gre_netlink_parms(struct nlattr *data[],
1421 				struct __ip6_tnl_parm *parms)
1422 {
1423 	memset(parms, 0, sizeof(*parms));
1424 
1425 	if (!data)
1426 		return;
1427 
1428 	if (data[IFLA_GRE_LINK])
1429 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1430 
1431 	if (data[IFLA_GRE_IFLAGS])
1432 		parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1433 
1434 	if (data[IFLA_GRE_OFLAGS])
1435 		parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1436 
1437 	if (data[IFLA_GRE_IKEY])
1438 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1439 
1440 	if (data[IFLA_GRE_OKEY])
1441 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1442 
1443 	if (data[IFLA_GRE_LOCAL])
1444 		nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1445 
1446 	if (data[IFLA_GRE_REMOTE])
1447 		nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1448 
1449 	if (data[IFLA_GRE_TTL])
1450 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1451 
1452 	if (data[IFLA_GRE_ENCAP_LIMIT])
1453 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1454 
1455 	if (data[IFLA_GRE_FLOWINFO])
1456 		parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1457 
1458 	if (data[IFLA_GRE_FLAGS])
1459 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1460 }
1461 
1462 static int ip6gre_tap_init(struct net_device *dev)
1463 {
1464 	struct ip6_tnl *tunnel;
1465 
1466 	tunnel = netdev_priv(dev);
1467 
1468 	tunnel->dev = dev;
1469 	tunnel->net = dev_net(dev);
1470 	strcpy(tunnel->parms.name, dev->name);
1471 
1472 	ip6gre_tnl_link_config(tunnel, 1);
1473 
1474 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1475 	if (!dev->tstats)
1476 		return -ENOMEM;
1477 
1478 	return 0;
1479 }
1480 
1481 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1482 	.ndo_init = ip6gre_tap_init,
1483 	.ndo_uninit = ip6gre_tunnel_uninit,
1484 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1485 	.ndo_set_mac_address = eth_mac_addr,
1486 	.ndo_validate_addr = eth_validate_addr,
1487 	.ndo_change_mtu = ip6gre_tunnel_change_mtu,
1488 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1489 };
1490 
1491 static void ip6gre_tap_setup(struct net_device *dev)
1492 {
1493 
1494 	ether_setup(dev);
1495 
1496 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1497 	dev->destructor = ip6gre_dev_free;
1498 
1499 	dev->iflink = 0;
1500 	dev->features |= NETIF_F_NETNS_LOCAL;
1501 }
1502 
1503 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1504 	struct nlattr *tb[], struct nlattr *data[])
1505 {
1506 	struct ip6_tnl *nt;
1507 	struct net *net = dev_net(dev);
1508 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1509 	int err;
1510 
1511 	nt = netdev_priv(dev);
1512 	ip6gre_netlink_parms(data, &nt->parms);
1513 
1514 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1515 		return -EEXIST;
1516 
1517 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1518 		eth_hw_addr_random(dev);
1519 
1520 	nt->dev = dev;
1521 	nt->net = dev_net(dev);
1522 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1523 
1524 	/* Can use a lockless transmit, unless we generate output sequences */
1525 	if (!(nt->parms.o_flags & GRE_SEQ))
1526 		dev->features |= NETIF_F_LLTX;
1527 
1528 	err = register_netdevice(dev);
1529 	if (err)
1530 		goto out;
1531 
1532 	dev_hold(dev);
1533 	ip6gre_tunnel_link(ign, nt);
1534 
1535 out:
1536 	return err;
1537 }
1538 
1539 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1540 			    struct nlattr *data[])
1541 {
1542 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1543 	struct net *net = nt->net;
1544 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1545 	struct __ip6_tnl_parm p;
1546 
1547 	if (dev == ign->fb_tunnel_dev)
1548 		return -EINVAL;
1549 
1550 	ip6gre_netlink_parms(data, &p);
1551 
1552 	t = ip6gre_tunnel_locate(net, &p, 0);
1553 
1554 	if (t) {
1555 		if (t->dev != dev)
1556 			return -EEXIST;
1557 	} else {
1558 		t = nt;
1559 
1560 		ip6gre_tunnel_unlink(ign, t);
1561 		ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1562 		ip6gre_tunnel_link(ign, t);
1563 		netdev_state_change(dev);
1564 	}
1565 
1566 	return 0;
1567 }
1568 
1569 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1570 {
1571 	struct net *net = dev_net(dev);
1572 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1573 
1574 	if (dev != ign->fb_tunnel_dev)
1575 		unregister_netdevice_queue(dev, head);
1576 }
1577 
1578 static size_t ip6gre_get_size(const struct net_device *dev)
1579 {
1580 	return
1581 		/* IFLA_GRE_LINK */
1582 		nla_total_size(4) +
1583 		/* IFLA_GRE_IFLAGS */
1584 		nla_total_size(2) +
1585 		/* IFLA_GRE_OFLAGS */
1586 		nla_total_size(2) +
1587 		/* IFLA_GRE_IKEY */
1588 		nla_total_size(4) +
1589 		/* IFLA_GRE_OKEY */
1590 		nla_total_size(4) +
1591 		/* IFLA_GRE_LOCAL */
1592 		nla_total_size(sizeof(struct in6_addr)) +
1593 		/* IFLA_GRE_REMOTE */
1594 		nla_total_size(sizeof(struct in6_addr)) +
1595 		/* IFLA_GRE_TTL */
1596 		nla_total_size(1) +
1597 		/* IFLA_GRE_TOS */
1598 		nla_total_size(1) +
1599 		/* IFLA_GRE_ENCAP_LIMIT */
1600 		nla_total_size(1) +
1601 		/* IFLA_GRE_FLOWINFO */
1602 		nla_total_size(4) +
1603 		/* IFLA_GRE_FLAGS */
1604 		nla_total_size(4) +
1605 		0;
1606 }
1607 
1608 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1609 {
1610 	struct ip6_tnl *t = netdev_priv(dev);
1611 	struct __ip6_tnl_parm *p = &t->parms;
1612 
1613 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1614 	    nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1615 	    nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1616 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1617 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1618 	    nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1619 	    nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1620 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1621 	    /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1622 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1623 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1624 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1625 		goto nla_put_failure;
1626 	return 0;
1627 
1628 nla_put_failure:
1629 	return -EMSGSIZE;
1630 }
1631 
1632 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1633 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1634 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1635 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1636 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1637 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1638 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1639 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1640 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1641 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1642 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1643 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1644 };
1645 
1646 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1647 	.kind		= "ip6gre",
1648 	.maxtype	= IFLA_GRE_MAX,
1649 	.policy		= ip6gre_policy,
1650 	.priv_size	= sizeof(struct ip6_tnl),
1651 	.setup		= ip6gre_tunnel_setup,
1652 	.validate	= ip6gre_tunnel_validate,
1653 	.newlink	= ip6gre_newlink,
1654 	.changelink	= ip6gre_changelink,
1655 	.dellink	= ip6gre_dellink,
1656 	.get_size	= ip6gre_get_size,
1657 	.fill_info	= ip6gre_fill_info,
1658 };
1659 
1660 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1661 	.kind		= "ip6gretap",
1662 	.maxtype	= IFLA_GRE_MAX,
1663 	.policy		= ip6gre_policy,
1664 	.priv_size	= sizeof(struct ip6_tnl),
1665 	.setup		= ip6gre_tap_setup,
1666 	.validate	= ip6gre_tap_validate,
1667 	.newlink	= ip6gre_newlink,
1668 	.changelink	= ip6gre_changelink,
1669 	.get_size	= ip6gre_get_size,
1670 	.fill_info	= ip6gre_fill_info,
1671 };
1672 
1673 /*
1674  *	And now the modules code and kernel interface.
1675  */
1676 
1677 static int __init ip6gre_init(void)
1678 {
1679 	int err;
1680 
1681 	pr_info("GRE over IPv6 tunneling driver\n");
1682 
1683 	err = register_pernet_device(&ip6gre_net_ops);
1684 	if (err < 0)
1685 		return err;
1686 
1687 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1688 	if (err < 0) {
1689 		pr_info("%s: can't add protocol\n", __func__);
1690 		goto add_proto_failed;
1691 	}
1692 
1693 	err = rtnl_link_register(&ip6gre_link_ops);
1694 	if (err < 0)
1695 		goto rtnl_link_failed;
1696 
1697 	err = rtnl_link_register(&ip6gre_tap_ops);
1698 	if (err < 0)
1699 		goto tap_ops_failed;
1700 
1701 out:
1702 	return err;
1703 
1704 tap_ops_failed:
1705 	rtnl_link_unregister(&ip6gre_link_ops);
1706 rtnl_link_failed:
1707 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1708 add_proto_failed:
1709 	unregister_pernet_device(&ip6gre_net_ops);
1710 	goto out;
1711 }
1712 
1713 static void __exit ip6gre_fini(void)
1714 {
1715 	rtnl_link_unregister(&ip6gre_tap_ops);
1716 	rtnl_link_unregister(&ip6gre_link_ops);
1717 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1718 	unregister_pernet_device(&ip6gre_net_ops);
1719 }
1720 
1721 module_init(ip6gre_init);
1722 module_exit(ip6gre_fini);
1723 MODULE_LICENSE("GPL");
1724 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1725 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1726 MODULE_ALIAS_RTNL_LINK("ip6gre");
1727 MODULE_ALIAS_NETDEV("ip6gre0");
1728