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
60static bool log_ecn_error = true;
61module_param(log_ecn_error, bool, 0644);
62MODULE_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
67static int ip6gre_net_id __read_mostly;
68struct ip6gre_net {
69	struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
70
71	struct net_device *fb_tunnel_dev;
72};
73
74static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
76static int ip6gre_tunnel_init(struct net_device *dev);
77static void ip6gre_tunnel_setup(struct net_device *dev);
78static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79static 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))
100static 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
114static 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)
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
236static 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
254static 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
260static 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
268static 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
283static 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
308static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309		const struct __ip6_tnl_parm *parms, int create)
310{
311	struct ip6_tnl *t, *nt;
312	struct net_device *dev;
313	char name[IFNAMSIZ];
314	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315
316	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317	if (t && create)
318		return NULL;
319	if (t || !create)
320		return t;
321
322	if (parms->name[0])
323		strlcpy(name, parms->name, IFNAMSIZ);
324	else
325		strcpy(name, "ip6gre%d");
326
327	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328			   ip6gre_tunnel_setup);
329	if (!dev)
330		return NULL;
331
332	dev_net_set(dev, net);
333
334	nt = netdev_priv(dev);
335	nt->parms = *parms;
336	dev->rtnl_link_ops = &ip6gre_link_ops;
337
338	nt->dev = dev;
339	nt->net = dev_net(dev);
340	ip6gre_tnl_link_config(nt, 1);
341
342	if (register_netdevice(dev) < 0)
343		goto failed_free;
344
345	/* Can use a lockless transmit, unless we generate output sequences */
346	if (!(nt->parms.o_flags & GRE_SEQ))
347		dev->features |= NETIF_F_LLTX;
348
349	dev_hold(dev);
350	ip6gre_tunnel_link(ign, nt);
351	return nt;
352
353failed_free:
354	free_netdev(dev);
355	return NULL;
356}
357
358static void ip6gre_tunnel_uninit(struct net_device *dev)
359{
360	struct ip6_tnl *t = netdev_priv(dev);
361	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
362
363	ip6gre_tunnel_unlink(ign, t);
364	ip6_tnl_dst_reset(t);
365	dev_put(dev);
366}
367
368
369static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
370		u8 type, u8 code, int offset, __be32 info)
371{
372	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
373	__be16 *p = (__be16 *)(skb->data + offset);
374	int grehlen = offset + 4;
375	struct ip6_tnl *t;
376	__be16 flags;
377
378	flags = p[0];
379	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
380		if (flags&(GRE_VERSION|GRE_ROUTING))
381			return;
382		if (flags&GRE_KEY) {
383			grehlen += 4;
384			if (flags&GRE_CSUM)
385				grehlen += 4;
386		}
387	}
388
389	/* If only 8 bytes returned, keyed message will be dropped here */
390	if (!pskb_may_pull(skb, grehlen))
391		return;
392	ipv6h = (const struct ipv6hdr *)skb->data;
393	p = (__be16 *)(skb->data + offset);
394
395	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
396				flags & GRE_KEY ?
397				*(((__be32 *)p) + (grehlen / 4) - 1) : 0,
398				p[1]);
399	if (!t)
400		return;
401
402	switch (type) {
403		__u32 teli;
404		struct ipv6_tlv_tnl_enc_lim *tel;
405		__u32 mtu;
406	case ICMPV6_DEST_UNREACH:
407		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
408				    t->parms.name);
409		break;
410	case ICMPV6_TIME_EXCEED:
411		if (code == ICMPV6_EXC_HOPLIMIT) {
412			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
413					    t->parms.name);
414		}
415		break;
416	case ICMPV6_PARAMPROB:
417		teli = 0;
418		if (code == ICMPV6_HDR_FIELD)
419			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
420
421		if (teli && teli == be32_to_cpu(info) - 2) {
422			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
423			if (tel->encap_limit == 0) {
424				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
425						    t->parms.name);
426			}
427		} else {
428			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
429					    t->parms.name);
430		}
431		break;
432	case ICMPV6_PKT_TOOBIG:
433		mtu = be32_to_cpu(info) - offset;
434		if (mtu < IPV6_MIN_MTU)
435			mtu = IPV6_MIN_MTU;
436		t->dev->mtu = mtu;
437		break;
438	}
439
440	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
441		t->err_count++;
442	else
443		t->err_count = 1;
444	t->err_time = jiffies;
445}
446
447static int ip6gre_rcv(struct sk_buff *skb)
448{
449	const struct ipv6hdr *ipv6h;
450	u8     *h;
451	__be16    flags;
452	__sum16   csum = 0;
453	__be32 key = 0;
454	u32    seqno = 0;
455	struct ip6_tnl *tunnel;
456	int    offset = 4;
457	__be16 gre_proto;
458	int err;
459
460	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
461		goto drop;
462
463	ipv6h = ipv6_hdr(skb);
464	h = skb->data;
465	flags = *(__be16 *)h;
466
467	if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
468		/* - Version must be 0.
469		   - We do not support routing headers.
470		 */
471		if (flags&(GRE_VERSION|GRE_ROUTING))
472			goto drop;
473
474		if (flags&GRE_CSUM) {
475			csum = skb_checksum_simple_validate(skb);
476			offset += 4;
477		}
478		if (flags&GRE_KEY) {
479			key = *(__be32 *)(h + offset);
480			offset += 4;
481		}
482		if (flags&GRE_SEQ) {
483			seqno = ntohl(*(__be32 *)(h + offset));
484			offset += 4;
485		}
486	}
487
488	gre_proto = *(__be16 *)(h + 2);
489
490	tunnel = ip6gre_tunnel_lookup(skb->dev,
491					  &ipv6h->saddr, &ipv6h->daddr, key,
492					  gre_proto);
493	if (tunnel) {
494		struct pcpu_sw_netstats *tstats;
495
496		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
497			goto drop;
498
499		if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
500			tunnel->dev->stats.rx_dropped++;
501			goto drop;
502		}
503
504		skb->protocol = gre_proto;
505		/* WCCP version 1 and 2 protocol decoding.
506		 * - Change protocol to IPv6
507		 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
508		 */
509		if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
510			skb->protocol = htons(ETH_P_IPV6);
511			if ((*(h + offset) & 0xF0) != 0x40)
512				offset += 4;
513		}
514
515		skb->mac_header = skb->network_header;
516		__pskb_pull(skb, offset);
517		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
518
519		if (((flags&GRE_CSUM) && csum) ||
520		    (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
521			tunnel->dev->stats.rx_crc_errors++;
522			tunnel->dev->stats.rx_errors++;
523			goto drop;
524		}
525		if (tunnel->parms.i_flags&GRE_SEQ) {
526			if (!(flags&GRE_SEQ) ||
527			    (tunnel->i_seqno &&
528					(s32)(seqno - tunnel->i_seqno) < 0)) {
529				tunnel->dev->stats.rx_fifo_errors++;
530				tunnel->dev->stats.rx_errors++;
531				goto drop;
532			}
533			tunnel->i_seqno = seqno + 1;
534		}
535
536		/* Warning: All skb pointers will be invalidated! */
537		if (tunnel->dev->type == ARPHRD_ETHER) {
538			if (!pskb_may_pull(skb, ETH_HLEN)) {
539				tunnel->dev->stats.rx_length_errors++;
540				tunnel->dev->stats.rx_errors++;
541				goto drop;
542			}
543
544			ipv6h = ipv6_hdr(skb);
545			skb->protocol = eth_type_trans(skb, tunnel->dev);
546			skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
547		}
548
549		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
550
551		skb_reset_network_header(skb);
552
553		err = IP6_ECN_decapsulate(ipv6h, skb);
554		if (unlikely(err)) {
555			if (log_ecn_error)
556				net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
557						     &ipv6h->saddr,
558						     ipv6_get_dsfield(ipv6h));
559			if (err > 1) {
560				++tunnel->dev->stats.rx_frame_errors;
561				++tunnel->dev->stats.rx_errors;
562				goto drop;
563			}
564		}
565
566		tstats = this_cpu_ptr(tunnel->dev->tstats);
567		u64_stats_update_begin(&tstats->syncp);
568		tstats->rx_packets++;
569		tstats->rx_bytes += skb->len;
570		u64_stats_update_end(&tstats->syncp);
571
572		netif_rx(skb);
573
574		return 0;
575	}
576	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
577
578drop:
579	kfree_skb(skb);
580	return 0;
581}
582
583struct ipv6_tel_txoption {
584	struct ipv6_txoptions ops;
585	__u8 dst_opt[8];
586};
587
588static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
589{
590	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
591
592	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
593	opt->dst_opt[3] = 1;
594	opt->dst_opt[4] = encap_limit;
595	opt->dst_opt[5] = IPV6_TLV_PADN;
596	opt->dst_opt[6] = 1;
597
598	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
599	opt->ops.opt_nflen = 8;
600}
601
602static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
603			 struct net_device *dev,
604			 __u8 dsfield,
605			 struct flowi6 *fl6,
606			 int encap_limit,
607			 __u32 *pmtu)
608{
609	struct ip6_tnl *tunnel = netdev_priv(dev);
610	struct net *net = tunnel->net;
611	struct net_device *tdev;    /* Device to other host */
612	struct ipv6hdr  *ipv6h;     /* Our new IP header */
613	unsigned int max_headroom = 0; /* The extra header space needed */
614	int    gre_hlen;
615	struct ipv6_tel_txoption opt;
616	int    mtu;
617	struct dst_entry *dst = NULL, *ndst = NULL;
618	struct net_device_stats *stats = &tunnel->dev->stats;
619	int err = -1;
620	u8 proto;
621	struct sk_buff *new_skb;
622	__be16 protocol;
623
624	if (dev->type == ARPHRD_ETHER)
625		IPCB(skb)->flags = 0;
626
627	if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
628		gre_hlen = 0;
629		ipv6h = (struct ipv6hdr *)skb->data;
630		fl6->daddr = ipv6h->daddr;
631	} else {
632		gre_hlen = tunnel->hlen;
633		fl6->daddr = tunnel->parms.raddr;
634	}
635
636	if (!fl6->flowi6_mark)
637		dst = ip6_tnl_dst_get(tunnel);
638
639	if (!dst) {
640		dst = ip6_route_output(net, NULL, fl6);
641
642		if (dst->error)
643			goto tx_err_link_failure;
644		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
645		if (IS_ERR(dst)) {
646			err = PTR_ERR(dst);
647			dst = NULL;
648			goto tx_err_link_failure;
649		}
650		ndst = dst;
651	}
652
653	tdev = dst->dev;
654
655	if (tdev == dev) {
656		stats->collisions++;
657		net_warn_ratelimited("%s: Local routing loop detected!\n",
658				     tunnel->parms.name);
659		goto tx_err_dst_release;
660	}
661
662	mtu = dst_mtu(dst) - sizeof(*ipv6h);
663	if (encap_limit >= 0) {
664		max_headroom += 8;
665		mtu -= 8;
666	}
667	if (mtu < IPV6_MIN_MTU)
668		mtu = IPV6_MIN_MTU;
669	if (skb_dst(skb))
670		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
671	if (skb->len > mtu) {
672		*pmtu = mtu;
673		err = -EMSGSIZE;
674		goto tx_err_dst_release;
675	}
676
677	if (tunnel->err_count > 0) {
678		if (time_before(jiffies,
679				tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
680			tunnel->err_count--;
681
682			dst_link_failure(skb);
683		} else
684			tunnel->err_count = 0;
685	}
686
687	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
688
689	max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
690
691	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
692	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
693		new_skb = skb_realloc_headroom(skb, max_headroom);
694		if (max_headroom > dev->needed_headroom)
695			dev->needed_headroom = max_headroom;
696		if (!new_skb)
697			goto tx_err_dst_release;
698
699		if (skb->sk)
700			skb_set_owner_w(new_skb, skb->sk);
701		consume_skb(skb);
702		skb = new_skb;
703	}
704
705	if (!fl6->flowi6_mark && ndst)
706		ip6_tnl_dst_set(tunnel, ndst);
707	skb_dst_set(skb, dst);
708
709	proto = NEXTHDR_GRE;
710	if (encap_limit >= 0) {
711		init_tel_txopt(&opt, encap_limit);
712		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
713	}
714
715	if (likely(!skb->encapsulation)) {
716		skb_reset_inner_headers(skb);
717		skb->encapsulation = 1;
718	}
719
720	skb_push(skb, gre_hlen);
721	skb_reset_network_header(skb);
722	skb_set_transport_header(skb, sizeof(*ipv6h));
723
724	/*
725	 *	Push down and install the IP header.
726	 */
727	ipv6h = ipv6_hdr(skb);
728	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
729		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
730	ipv6h->hop_limit = tunnel->parms.hop_limit;
731	ipv6h->nexthdr = proto;
732	ipv6h->saddr = fl6->saddr;
733	ipv6h->daddr = fl6->daddr;
734
735	((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
736	protocol = (dev->type == ARPHRD_ETHER) ?
737		    htons(ETH_P_TEB) : skb->protocol;
738	((__be16 *)(ipv6h + 1))[1] = protocol;
739
740	if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
741		__be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
742
743		if (tunnel->parms.o_flags&GRE_SEQ) {
744			++tunnel->o_seqno;
745			*ptr = htonl(tunnel->o_seqno);
746			ptr--;
747		}
748		if (tunnel->parms.o_flags&GRE_KEY) {
749			*ptr = tunnel->parms.o_key;
750			ptr--;
751		}
752		if (tunnel->parms.o_flags&GRE_CSUM) {
753			*ptr = 0;
754			*(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
755				skb->len - sizeof(struct ipv6hdr));
756		}
757	}
758
759	skb_set_inner_protocol(skb, protocol);
760
761	ip6tunnel_xmit(NULL, skb, dev);
762	return 0;
763tx_err_link_failure:
764	stats->tx_carrier_errors++;
765	dst_link_failure(skb);
766tx_err_dst_release:
767	dst_release(dst);
768	return err;
769}
770
771static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
772{
773	struct ip6_tnl *t = netdev_priv(dev);
774	const struct iphdr  *iph = ip_hdr(skb);
775	int encap_limit = -1;
776	struct flowi6 fl6;
777	__u8 dsfield;
778	__u32 mtu;
779	int err;
780
781	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
782
783	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
784		encap_limit = t->parms.encap_limit;
785
786	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
787	fl6.flowi6_proto = IPPROTO_GRE;
788
789	dsfield = ipv4_get_dsfield(iph);
790
791	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
792		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
793					  & IPV6_TCLASS_MASK;
794	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
795		fl6.flowi6_mark = skb->mark;
796
797	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
798	if (err != 0) {
799		/* XXX: send ICMP error even if DF is not set. */
800		if (err == -EMSGSIZE)
801			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
802				  htonl(mtu));
803		return -1;
804	}
805
806	return 0;
807}
808
809static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
810{
811	struct ip6_tnl *t = netdev_priv(dev);
812	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
813	int encap_limit = -1;
814	__u16 offset;
815	struct flowi6 fl6;
816	__u8 dsfield;
817	__u32 mtu;
818	int err;
819
820	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
821		return -1;
822
823	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
824	if (offset > 0) {
825		struct ipv6_tlv_tnl_enc_lim *tel;
826		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
827		if (tel->encap_limit == 0) {
828			icmpv6_send(skb, ICMPV6_PARAMPROB,
829				    ICMPV6_HDR_FIELD, offset + 2);
830			return -1;
831		}
832		encap_limit = tel->encap_limit - 1;
833	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
834		encap_limit = t->parms.encap_limit;
835
836	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
837	fl6.flowi6_proto = IPPROTO_GRE;
838
839	dsfield = ipv6_get_dsfield(ipv6h);
840	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
841		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
842	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
843		fl6.flowlabel |= ip6_flowlabel(ipv6h);
844	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
845		fl6.flowi6_mark = skb->mark;
846
847	err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
848	if (err != 0) {
849		if (err == -EMSGSIZE)
850			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
851		return -1;
852	}
853
854	return 0;
855}
856
857/**
858 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
859 *   @t: the outgoing tunnel device
860 *   @hdr: IPv6 header from the incoming packet
861 *
862 * Description:
863 *   Avoid trivial tunneling loop by checking that tunnel exit-point
864 *   doesn't match source of incoming packet.
865 *
866 * Return:
867 *   1 if conflict,
868 *   0 else
869 **/
870
871static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
872	const struct ipv6hdr *hdr)
873{
874	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
875}
876
877static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
878{
879	struct ip6_tnl *t = netdev_priv(dev);
880	int encap_limit = -1;
881	struct flowi6 fl6;
882	__u32 mtu;
883	int err;
884
885	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
886		encap_limit = t->parms.encap_limit;
887
888	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
889	fl6.flowi6_proto = skb->protocol;
890
891	err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
892
893	return err;
894}
895
896static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
897	struct net_device *dev)
898{
899	struct ip6_tnl *t = netdev_priv(dev);
900	struct net_device_stats *stats = &t->dev->stats;
901	int ret;
902
903	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
904		goto tx_err;
905
906	switch (skb->protocol) {
907	case htons(ETH_P_IP):
908		ret = ip6gre_xmit_ipv4(skb, dev);
909		break;
910	case htons(ETH_P_IPV6):
911		ret = ip6gre_xmit_ipv6(skb, dev);
912		break;
913	default:
914		ret = ip6gre_xmit_other(skb, dev);
915		break;
916	}
917
918	if (ret < 0)
919		goto tx_err;
920
921	return NETDEV_TX_OK;
922
923tx_err:
924	stats->tx_errors++;
925	stats->tx_dropped++;
926	kfree_skb(skb);
927	return NETDEV_TX_OK;
928}
929
930static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
931{
932	struct net_device *dev = t->dev;
933	struct __ip6_tnl_parm *p = &t->parms;
934	struct flowi6 *fl6 = &t->fl.u.ip6;
935	int addend = sizeof(struct ipv6hdr) + 4;
936
937	if (dev->type != ARPHRD_ETHER) {
938		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
939		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
940	}
941
942	/* Set up flowi template */
943	fl6->saddr = p->laddr;
944	fl6->daddr = p->raddr;
945	fl6->flowi6_oif = p->link;
946	fl6->flowlabel = 0;
947
948	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
949		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
950	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
951		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
952
953	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
954	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
955
956	if (p->flags&IP6_TNL_F_CAP_XMIT &&
957			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
958		dev->flags |= IFF_POINTOPOINT;
959	else
960		dev->flags &= ~IFF_POINTOPOINT;
961
962	/* Precalculate GRE options length */
963	if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
964		if (t->parms.o_flags&GRE_CSUM)
965			addend += 4;
966		if (t->parms.o_flags&GRE_KEY)
967			addend += 4;
968		if (t->parms.o_flags&GRE_SEQ)
969			addend += 4;
970	}
971	t->hlen = addend;
972
973	if (p->flags & IP6_TNL_F_CAP_XMIT) {
974		int strict = (ipv6_addr_type(&p->raddr) &
975			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
976
977		struct rt6_info *rt = rt6_lookup(t->net,
978						 &p->raddr, &p->laddr,
979						 p->link, strict);
980
981		if (!rt)
982			return;
983
984		if (rt->dst.dev) {
985			dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
986
987			if (set_mtu) {
988				dev->mtu = rt->dst.dev->mtu - addend;
989				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
990					dev->mtu -= 8;
991
992				if (dev->mtu < IPV6_MIN_MTU)
993					dev->mtu = IPV6_MIN_MTU;
994			}
995		}
996		ip6_rt_put(rt);
997	}
998}
999
1000static int ip6gre_tnl_change(struct ip6_tnl *t,
1001	const struct __ip6_tnl_parm *p, int set_mtu)
1002{
1003	t->parms.laddr = p->laddr;
1004	t->parms.raddr = p->raddr;
1005	t->parms.flags = p->flags;
1006	t->parms.hop_limit = p->hop_limit;
1007	t->parms.encap_limit = p->encap_limit;
1008	t->parms.flowinfo = p->flowinfo;
1009	t->parms.link = p->link;
1010	t->parms.proto = p->proto;
1011	t->parms.i_key = p->i_key;
1012	t->parms.o_key = p->o_key;
1013	t->parms.i_flags = p->i_flags;
1014	t->parms.o_flags = p->o_flags;
1015	ip6_tnl_dst_reset(t);
1016	ip6gre_tnl_link_config(t, set_mtu);
1017	return 0;
1018}
1019
1020static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1021	const struct ip6_tnl_parm2 *u)
1022{
1023	p->laddr = u->laddr;
1024	p->raddr = u->raddr;
1025	p->flags = u->flags;
1026	p->hop_limit = u->hop_limit;
1027	p->encap_limit = u->encap_limit;
1028	p->flowinfo = u->flowinfo;
1029	p->link = u->link;
1030	p->i_key = u->i_key;
1031	p->o_key = u->o_key;
1032	p->i_flags = u->i_flags;
1033	p->o_flags = u->o_flags;
1034	memcpy(p->name, u->name, sizeof(u->name));
1035}
1036
1037static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1038	const struct __ip6_tnl_parm *p)
1039{
1040	u->proto = IPPROTO_GRE;
1041	u->laddr = p->laddr;
1042	u->raddr = p->raddr;
1043	u->flags = p->flags;
1044	u->hop_limit = p->hop_limit;
1045	u->encap_limit = p->encap_limit;
1046	u->flowinfo = p->flowinfo;
1047	u->link = p->link;
1048	u->i_key = p->i_key;
1049	u->o_key = p->o_key;
1050	u->i_flags = p->i_flags;
1051	u->o_flags = p->o_flags;
1052	memcpy(u->name, p->name, sizeof(u->name));
1053}
1054
1055static int ip6gre_tunnel_ioctl(struct net_device *dev,
1056	struct ifreq *ifr, int cmd)
1057{
1058	int err = 0;
1059	struct ip6_tnl_parm2 p;
1060	struct __ip6_tnl_parm p1;
1061	struct ip6_tnl *t = netdev_priv(dev);
1062	struct net *net = t->net;
1063	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1064
1065	switch (cmd) {
1066	case SIOCGETTUNNEL:
1067		if (dev == ign->fb_tunnel_dev) {
1068			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1069				err = -EFAULT;
1070				break;
1071			}
1072			ip6gre_tnl_parm_from_user(&p1, &p);
1073			t = ip6gre_tunnel_locate(net, &p1, 0);
1074			if (!t)
1075				t = netdev_priv(dev);
1076		}
1077		memset(&p, 0, sizeof(p));
1078		ip6gre_tnl_parm_to_user(&p, &t->parms);
1079		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1080			err = -EFAULT;
1081		break;
1082
1083	case SIOCADDTUNNEL:
1084	case SIOCCHGTUNNEL:
1085		err = -EPERM;
1086		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1087			goto done;
1088
1089		err = -EFAULT;
1090		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1091			goto done;
1092
1093		err = -EINVAL;
1094		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1095			goto done;
1096
1097		if (!(p.i_flags&GRE_KEY))
1098			p.i_key = 0;
1099		if (!(p.o_flags&GRE_KEY))
1100			p.o_key = 0;
1101
1102		ip6gre_tnl_parm_from_user(&p1, &p);
1103		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1104
1105		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1106			if (t) {
1107				if (t->dev != dev) {
1108					err = -EEXIST;
1109					break;
1110				}
1111			} else {
1112				t = netdev_priv(dev);
1113
1114				ip6gre_tunnel_unlink(ign, t);
1115				synchronize_net();
1116				ip6gre_tnl_change(t, &p1, 1);
1117				ip6gre_tunnel_link(ign, t);
1118				netdev_state_change(dev);
1119			}
1120		}
1121
1122		if (t) {
1123			err = 0;
1124
1125			memset(&p, 0, sizeof(p));
1126			ip6gre_tnl_parm_to_user(&p, &t->parms);
1127			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1128				err = -EFAULT;
1129		} else
1130			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1131		break;
1132
1133	case SIOCDELTUNNEL:
1134		err = -EPERM;
1135		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1136			goto done;
1137
1138		if (dev == ign->fb_tunnel_dev) {
1139			err = -EFAULT;
1140			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1141				goto done;
1142			err = -ENOENT;
1143			ip6gre_tnl_parm_from_user(&p1, &p);
1144			t = ip6gre_tunnel_locate(net, &p1, 0);
1145			if (!t)
1146				goto done;
1147			err = -EPERM;
1148			if (t == netdev_priv(ign->fb_tunnel_dev))
1149				goto done;
1150			dev = t->dev;
1151		}
1152		unregister_netdevice(dev);
1153		err = 0;
1154		break;
1155
1156	default:
1157		err = -EINVAL;
1158	}
1159
1160done:
1161	return err;
1162}
1163
1164static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1165{
1166	if (new_mtu < 68 ||
1167	    new_mtu > 0xFFF8 - dev->hard_header_len)
1168		return -EINVAL;
1169	dev->mtu = new_mtu;
1170	return 0;
1171}
1172
1173static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1174			unsigned short type,
1175			const void *daddr, const void *saddr, unsigned int len)
1176{
1177	struct ip6_tnl *t = netdev_priv(dev);
1178	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1179	__be16 *p = (__be16 *)(ipv6h+1);
1180
1181	ip6_flow_hdr(ipv6h, 0,
1182		     ip6_make_flowlabel(dev_net(dev), skb,
1183					t->fl.u.ip6.flowlabel, true,
1184					&t->fl.u.ip6));
1185	ipv6h->hop_limit = t->parms.hop_limit;
1186	ipv6h->nexthdr = NEXTHDR_GRE;
1187	ipv6h->saddr = t->parms.laddr;
1188	ipv6h->daddr = t->parms.raddr;
1189
1190	p[0]		= t->parms.o_flags;
1191	p[1]		= htons(type);
1192
1193	/*
1194	 *	Set the source hardware address.
1195	 */
1196
1197	if (saddr)
1198		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1199	if (daddr)
1200		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1201	if (!ipv6_addr_any(&ipv6h->daddr))
1202		return t->hlen;
1203
1204	return -t->hlen;
1205}
1206
1207static const struct header_ops ip6gre_header_ops = {
1208	.create	= ip6gre_header,
1209};
1210
1211static const struct net_device_ops ip6gre_netdev_ops = {
1212	.ndo_init		= ip6gre_tunnel_init,
1213	.ndo_uninit		= ip6gre_tunnel_uninit,
1214	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1215	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1216	.ndo_change_mtu		= ip6gre_tunnel_change_mtu,
1217	.ndo_get_stats64	= ip_tunnel_get_stats64,
1218	.ndo_get_iflink		= ip6_tnl_get_iflink,
1219};
1220
1221static void ip6gre_dev_free(struct net_device *dev)
1222{
1223	struct ip6_tnl *t = netdev_priv(dev);
1224
1225	ip6_tnl_dst_destroy(t);
1226	free_percpu(dev->tstats);
1227	free_netdev(dev);
1228}
1229
1230static void ip6gre_tunnel_setup(struct net_device *dev)
1231{
1232	struct ip6_tnl *t;
1233
1234	dev->netdev_ops = &ip6gre_netdev_ops;
1235	dev->destructor = ip6gre_dev_free;
1236
1237	dev->type = ARPHRD_IP6GRE;
1238	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1239	dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1240	t = netdev_priv(dev);
1241	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1242		dev->mtu -= 8;
1243	dev->flags |= IFF_NOARP;
1244	dev->addr_len = sizeof(struct in6_addr);
1245	netif_keep_dst(dev);
1246}
1247
1248static int ip6gre_tunnel_init_common(struct net_device *dev)
1249{
1250	struct ip6_tnl *tunnel;
1251	int ret;
1252
1253	tunnel = netdev_priv(dev);
1254
1255	tunnel->dev = dev;
1256	tunnel->net = dev_net(dev);
1257	strcpy(tunnel->parms.name, dev->name);
1258
1259	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1260	if (!dev->tstats)
1261		return -ENOMEM;
1262
1263	ret = ip6_tnl_dst_init(tunnel);
1264	if (ret) {
1265		free_percpu(dev->tstats);
1266		dev->tstats = NULL;
1267		return ret;
1268	}
1269
1270	return 0;
1271}
1272
1273static int ip6gre_tunnel_init(struct net_device *dev)
1274{
1275	struct ip6_tnl *tunnel;
1276	int ret;
1277
1278	ret = ip6gre_tunnel_init_common(dev);
1279	if (ret)
1280		return ret;
1281
1282	tunnel = netdev_priv(dev);
1283
1284	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1285	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1286
1287	if (ipv6_addr_any(&tunnel->parms.raddr))
1288		dev->header_ops = &ip6gre_header_ops;
1289
1290	return 0;
1291}
1292
1293static void ip6gre_fb_tunnel_init(struct net_device *dev)
1294{
1295	struct ip6_tnl *tunnel = netdev_priv(dev);
1296
1297	tunnel->dev = dev;
1298	tunnel->net = dev_net(dev);
1299	strcpy(tunnel->parms.name, dev->name);
1300
1301	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1302
1303	dev_hold(dev);
1304}
1305
1306
1307static struct inet6_protocol ip6gre_protocol __read_mostly = {
1308	.handler     = ip6gre_rcv,
1309	.err_handler = ip6gre_err,
1310	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1311};
1312
1313static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1314{
1315	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1316	struct net_device *dev, *aux;
1317	int prio;
1318
1319	for_each_netdev_safe(net, dev, aux)
1320		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1321		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1322			unregister_netdevice_queue(dev, head);
1323
1324	for (prio = 0; prio < 4; prio++) {
1325		int h;
1326		for (h = 0; h < HASH_SIZE; h++) {
1327			struct ip6_tnl *t;
1328
1329			t = rtnl_dereference(ign->tunnels[prio][h]);
1330
1331			while (t) {
1332				/* If dev is in the same netns, it has already
1333				 * been added to the list by the previous loop.
1334				 */
1335				if (!net_eq(dev_net(t->dev), net))
1336					unregister_netdevice_queue(t->dev,
1337								   head);
1338				t = rtnl_dereference(t->next);
1339			}
1340		}
1341	}
1342}
1343
1344static int __net_init ip6gre_init_net(struct net *net)
1345{
1346	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1347	int err;
1348
1349	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1350					  NET_NAME_UNKNOWN,
1351					  ip6gre_tunnel_setup);
1352	if (!ign->fb_tunnel_dev) {
1353		err = -ENOMEM;
1354		goto err_alloc_dev;
1355	}
1356	dev_net_set(ign->fb_tunnel_dev, net);
1357	/* FB netdevice is special: we have one, and only one per netns.
1358	 * Allowing to move it to another netns is clearly unsafe.
1359	 */
1360	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1361
1362
1363	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1364	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1365
1366	err = register_netdev(ign->fb_tunnel_dev);
1367	if (err)
1368		goto err_reg_dev;
1369
1370	rcu_assign_pointer(ign->tunnels_wc[0],
1371			   netdev_priv(ign->fb_tunnel_dev));
1372	return 0;
1373
1374err_reg_dev:
1375	ip6gre_dev_free(ign->fb_tunnel_dev);
1376err_alloc_dev:
1377	return err;
1378}
1379
1380static void __net_exit ip6gre_exit_net(struct net *net)
1381{
1382	LIST_HEAD(list);
1383
1384	rtnl_lock();
1385	ip6gre_destroy_tunnels(net, &list);
1386	unregister_netdevice_many(&list);
1387	rtnl_unlock();
1388}
1389
1390static struct pernet_operations ip6gre_net_ops = {
1391	.init = ip6gre_init_net,
1392	.exit = ip6gre_exit_net,
1393	.id   = &ip6gre_net_id,
1394	.size = sizeof(struct ip6gre_net),
1395};
1396
1397static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1398{
1399	__be16 flags;
1400
1401	if (!data)
1402		return 0;
1403
1404	flags = 0;
1405	if (data[IFLA_GRE_IFLAGS])
1406		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1407	if (data[IFLA_GRE_OFLAGS])
1408		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1409	if (flags & (GRE_VERSION|GRE_ROUTING))
1410		return -EINVAL;
1411
1412	return 0;
1413}
1414
1415static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1416{
1417	struct in6_addr daddr;
1418
1419	if (tb[IFLA_ADDRESS]) {
1420		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1421			return -EINVAL;
1422		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1423			return -EADDRNOTAVAIL;
1424	}
1425
1426	if (!data)
1427		goto out;
1428
1429	if (data[IFLA_GRE_REMOTE]) {
1430		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1431		if (ipv6_addr_any(&daddr))
1432			return -EINVAL;
1433	}
1434
1435out:
1436	return ip6gre_tunnel_validate(tb, data);
1437}
1438
1439
1440static void ip6gre_netlink_parms(struct nlattr *data[],
1441				struct __ip6_tnl_parm *parms)
1442{
1443	memset(parms, 0, sizeof(*parms));
1444
1445	if (!data)
1446		return;
1447
1448	if (data[IFLA_GRE_LINK])
1449		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1450
1451	if (data[IFLA_GRE_IFLAGS])
1452		parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1453
1454	if (data[IFLA_GRE_OFLAGS])
1455		parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1456
1457	if (data[IFLA_GRE_IKEY])
1458		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1459
1460	if (data[IFLA_GRE_OKEY])
1461		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1462
1463	if (data[IFLA_GRE_LOCAL])
1464		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1465
1466	if (data[IFLA_GRE_REMOTE])
1467		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1468
1469	if (data[IFLA_GRE_TTL])
1470		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1471
1472	if (data[IFLA_GRE_ENCAP_LIMIT])
1473		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1474
1475	if (data[IFLA_GRE_FLOWINFO])
1476		parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1477
1478	if (data[IFLA_GRE_FLAGS])
1479		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1480}
1481
1482static int ip6gre_tap_init(struct net_device *dev)
1483{
1484	struct ip6_tnl *tunnel;
1485	int ret;
1486
1487	ret = ip6gre_tunnel_init_common(dev);
1488	if (ret)
1489		return ret;
1490
1491	tunnel = netdev_priv(dev);
1492
1493	ip6gre_tnl_link_config(tunnel, 1);
1494
1495	return 0;
1496}
1497
1498static const struct net_device_ops ip6gre_tap_netdev_ops = {
1499	.ndo_init = ip6gre_tap_init,
1500	.ndo_uninit = ip6gre_tunnel_uninit,
1501	.ndo_start_xmit = ip6gre_tunnel_xmit,
1502	.ndo_set_mac_address = eth_mac_addr,
1503	.ndo_validate_addr = eth_validate_addr,
1504	.ndo_change_mtu = ip6gre_tunnel_change_mtu,
1505	.ndo_get_stats64 = ip_tunnel_get_stats64,
1506	.ndo_get_iflink = ip6_tnl_get_iflink,
1507};
1508
1509static void ip6gre_tap_setup(struct net_device *dev)
1510{
1511
1512	ether_setup(dev);
1513
1514	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1515	dev->destructor = ip6gre_dev_free;
1516
1517	dev->features |= NETIF_F_NETNS_LOCAL;
1518}
1519
1520static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1521	struct nlattr *tb[], struct nlattr *data[])
1522{
1523	struct ip6_tnl *nt;
1524	struct net *net = dev_net(dev);
1525	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1526	int err;
1527
1528	nt = netdev_priv(dev);
1529	ip6gre_netlink_parms(data, &nt->parms);
1530
1531	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1532		return -EEXIST;
1533
1534	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1535		eth_hw_addr_random(dev);
1536
1537	nt->dev = dev;
1538	nt->net = dev_net(dev);
1539	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1540
1541	/* Can use a lockless transmit, unless we generate output sequences */
1542	if (!(nt->parms.o_flags & GRE_SEQ))
1543		dev->features |= NETIF_F_LLTX;
1544
1545	err = register_netdevice(dev);
1546	if (err)
1547		goto out;
1548
1549	dev_hold(dev);
1550	ip6gre_tunnel_link(ign, nt);
1551
1552out:
1553	return err;
1554}
1555
1556static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1557			    struct nlattr *data[])
1558{
1559	struct ip6_tnl *t, *nt = netdev_priv(dev);
1560	struct net *net = nt->net;
1561	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1562	struct __ip6_tnl_parm p;
1563
1564	if (dev == ign->fb_tunnel_dev)
1565		return -EINVAL;
1566
1567	ip6gre_netlink_parms(data, &p);
1568
1569	t = ip6gre_tunnel_locate(net, &p, 0);
1570
1571	if (t) {
1572		if (t->dev != dev)
1573			return -EEXIST;
1574	} else {
1575		t = nt;
1576	}
1577
1578	ip6gre_tunnel_unlink(ign, t);
1579	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1580	ip6gre_tunnel_link(ign, t);
1581	return 0;
1582}
1583
1584static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1585{
1586	struct net *net = dev_net(dev);
1587	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1588
1589	if (dev != ign->fb_tunnel_dev)
1590		unregister_netdevice_queue(dev, head);
1591}
1592
1593static size_t ip6gre_get_size(const struct net_device *dev)
1594{
1595	return
1596		/* IFLA_GRE_LINK */
1597		nla_total_size(4) +
1598		/* IFLA_GRE_IFLAGS */
1599		nla_total_size(2) +
1600		/* IFLA_GRE_OFLAGS */
1601		nla_total_size(2) +
1602		/* IFLA_GRE_IKEY */
1603		nla_total_size(4) +
1604		/* IFLA_GRE_OKEY */
1605		nla_total_size(4) +
1606		/* IFLA_GRE_LOCAL */
1607		nla_total_size(sizeof(struct in6_addr)) +
1608		/* IFLA_GRE_REMOTE */
1609		nla_total_size(sizeof(struct in6_addr)) +
1610		/* IFLA_GRE_TTL */
1611		nla_total_size(1) +
1612		/* IFLA_GRE_TOS */
1613		nla_total_size(1) +
1614		/* IFLA_GRE_ENCAP_LIMIT */
1615		nla_total_size(1) +
1616		/* IFLA_GRE_FLOWINFO */
1617		nla_total_size(4) +
1618		/* IFLA_GRE_FLAGS */
1619		nla_total_size(4) +
1620		0;
1621}
1622
1623static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1624{
1625	struct ip6_tnl *t = netdev_priv(dev);
1626	struct __ip6_tnl_parm *p = &t->parms;
1627
1628	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1629	    nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1630	    nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1631	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1632	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1633	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1634	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1635	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1636	    /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1637	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1638	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1639	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1640		goto nla_put_failure;
1641	return 0;
1642
1643nla_put_failure:
1644	return -EMSGSIZE;
1645}
1646
1647static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1648	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1649	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1650	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1651	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1652	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1653	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1654	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1655	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1656	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1657	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1658	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1659};
1660
1661static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1662	.kind		= "ip6gre",
1663	.maxtype	= IFLA_GRE_MAX,
1664	.policy		= ip6gre_policy,
1665	.priv_size	= sizeof(struct ip6_tnl),
1666	.setup		= ip6gre_tunnel_setup,
1667	.validate	= ip6gre_tunnel_validate,
1668	.newlink	= ip6gre_newlink,
1669	.changelink	= ip6gre_changelink,
1670	.dellink	= ip6gre_dellink,
1671	.get_size	= ip6gre_get_size,
1672	.fill_info	= ip6gre_fill_info,
1673	.get_link_net	= ip6_tnl_get_link_net,
1674};
1675
1676static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1677	.kind		= "ip6gretap",
1678	.maxtype	= IFLA_GRE_MAX,
1679	.policy		= ip6gre_policy,
1680	.priv_size	= sizeof(struct ip6_tnl),
1681	.setup		= ip6gre_tap_setup,
1682	.validate	= ip6gre_tap_validate,
1683	.newlink	= ip6gre_newlink,
1684	.changelink	= ip6gre_changelink,
1685	.get_size	= ip6gre_get_size,
1686	.fill_info	= ip6gre_fill_info,
1687	.get_link_net	= ip6_tnl_get_link_net,
1688};
1689
1690/*
1691 *	And now the modules code and kernel interface.
1692 */
1693
1694static int __init ip6gre_init(void)
1695{
1696	int err;
1697
1698	pr_info("GRE over IPv6 tunneling driver\n");
1699
1700	err = register_pernet_device(&ip6gre_net_ops);
1701	if (err < 0)
1702		return err;
1703
1704	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1705	if (err < 0) {
1706		pr_info("%s: can't add protocol\n", __func__);
1707		goto add_proto_failed;
1708	}
1709
1710	err = rtnl_link_register(&ip6gre_link_ops);
1711	if (err < 0)
1712		goto rtnl_link_failed;
1713
1714	err = rtnl_link_register(&ip6gre_tap_ops);
1715	if (err < 0)
1716		goto tap_ops_failed;
1717
1718out:
1719	return err;
1720
1721tap_ops_failed:
1722	rtnl_link_unregister(&ip6gre_link_ops);
1723rtnl_link_failed:
1724	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1725add_proto_failed:
1726	unregister_pernet_device(&ip6gre_net_ops);
1727	goto out;
1728}
1729
1730static void __exit ip6gre_fini(void)
1731{
1732	rtnl_link_unregister(&ip6gre_tap_ops);
1733	rtnl_link_unregister(&ip6gre_link_ops);
1734	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1735	unregister_pernet_device(&ip6gre_net_ops);
1736}
1737
1738module_init(ip6gre_init);
1739module_exit(ip6gre_fini);
1740MODULE_LICENSE("GPL");
1741MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1742MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1743MODULE_ALIAS_RTNL_LINK("ip6gre");
1744MODULE_ALIAS_RTNL_LINK("ip6gretap");
1745MODULE_ALIAS_NETDEV("ip6gre0");
1746