xref: /kernel/linux/linux-5.10/net/ipv6/ip6_gre.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 *	GRE over IPv6 protocol decoder.
4 *
5 *	Authors: Dmitry Kozlov (xeb@mail.ru)
6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/capability.h>
11#include <linux/module.h>
12#include <linux/types.h>
13#include <linux/kernel.h>
14#include <linux/slab.h>
15#include <linux/uaccess.h>
16#include <linux/skbuff.h>
17#include <linux/netdevice.h>
18#include <linux/in.h>
19#include <linux/tcp.h>
20#include <linux/udp.h>
21#include <linux/if_arp.h>
22#include <linux/init.h>
23#include <linux/in6.h>
24#include <linux/inetdevice.h>
25#include <linux/igmp.h>
26#include <linux/netfilter_ipv4.h>
27#include <linux/etherdevice.h>
28#include <linux/if_ether.h>
29#include <linux/hash.h>
30#include <linux/if_tunnel.h>
31#include <linux/ip6_tunnel.h>
32
33#include <net/sock.h>
34#include <net/ip.h>
35#include <net/ip_tunnels.h>
36#include <net/icmp.h>
37#include <net/protocol.h>
38#include <net/addrconf.h>
39#include <net/arp.h>
40#include <net/checksum.h>
41#include <net/dsfield.h>
42#include <net/inet_ecn.h>
43#include <net/xfrm.h>
44#include <net/net_namespace.h>
45#include <net/netns/generic.h>
46#include <net/rtnetlink.h>
47
48#include <net/ipv6.h>
49#include <net/ip6_fib.h>
50#include <net/ip6_route.h>
51#include <net/ip6_tunnel.h>
52#include <net/gre.h>
53#include <net/erspan.h>
54#include <net/dst_metadata.h>
55
56
57static bool log_ecn_error = true;
58module_param(log_ecn_error, bool, 0644);
59MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60
61#define IP6_GRE_HASH_SIZE_SHIFT  5
62#define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63
64static unsigned int ip6gre_net_id __read_mostly;
65struct ip6gre_net {
66	struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67
68	struct ip6_tnl __rcu *collect_md_tun;
69	struct ip6_tnl __rcu *collect_md_tun_erspan;
70	struct net_device *fb_tunnel_dev;
71};
72
73static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75static struct rtnl_link_ops ip6erspan_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);
80static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82/* Tunnel hash table */
83
84/*
85   4 hash tables:
86
87   3: (remote,local)
88   2: (remote,*)
89   1: (*,local)
90   0: (*,*)
91
92   We require exact key match i.e. if a key is present in packet
93   it will match only tunnel with the same key; if it is not present,
94   it will match only keyless tunnel.
95
96   All keysless packets, if not matched configured keyless tunnels
97   will match fallback tunnel.
98 */
99
100#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
101static u32 HASH_ADDR(const struct in6_addr *addr)
102{
103	u32 hash = ipv6_addr_hash(addr);
104
105	return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106}
107
108#define tunnels_r_l	tunnels[3]
109#define tunnels_r	tunnels[2]
110#define tunnels_l	tunnels[1]
111#define tunnels_wc	tunnels[0]
112
113/* Given src, dst and key, find appropriate for input tunnel. */
114
115static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116		const struct in6_addr *remote, const struct in6_addr *local,
117		__be32 key, __be16 gre_proto)
118{
119	struct net *net = dev_net(dev);
120	int link = dev->ifindex;
121	unsigned int h0 = HASH_ADDR(remote);
122	unsigned int h1 = HASH_KEY(key);
123	struct ip6_tnl *t, *cand = NULL;
124	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125	int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126			gre_proto == htons(ETH_P_ERSPAN) ||
127			gre_proto == htons(ETH_P_ERSPAN2)) ?
128		       ARPHRD_ETHER : ARPHRD_IP6GRE;
129	int score, cand_score = 4;
130	struct net_device *ndev;
131
132	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
133		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
134		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
135		    key != t->parms.i_key ||
136		    !(t->dev->flags & IFF_UP))
137			continue;
138
139		if (t->dev->type != ARPHRD_IP6GRE &&
140		    t->dev->type != dev_type)
141			continue;
142
143		score = 0;
144		if (t->parms.link != link)
145			score |= 1;
146		if (t->dev->type != dev_type)
147			score |= 2;
148		if (score == 0)
149			return t;
150
151		if (score < cand_score) {
152			cand = t;
153			cand_score = score;
154		}
155	}
156
157	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
158		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
159		    key != t->parms.i_key ||
160		    !(t->dev->flags & IFF_UP))
161			continue;
162
163		if (t->dev->type != ARPHRD_IP6GRE &&
164		    t->dev->type != dev_type)
165			continue;
166
167		score = 0;
168		if (t->parms.link != link)
169			score |= 1;
170		if (t->dev->type != dev_type)
171			score |= 2;
172		if (score == 0)
173			return t;
174
175		if (score < cand_score) {
176			cand = t;
177			cand_score = score;
178		}
179	}
180
181	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
182		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
183			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
184				 !ipv6_addr_is_multicast(local))) ||
185		    key != t->parms.i_key ||
186		    !(t->dev->flags & IFF_UP))
187			continue;
188
189		if (t->dev->type != ARPHRD_IP6GRE &&
190		    t->dev->type != dev_type)
191			continue;
192
193		score = 0;
194		if (t->parms.link != link)
195			score |= 1;
196		if (t->dev->type != dev_type)
197			score |= 2;
198		if (score == 0)
199			return t;
200
201		if (score < cand_score) {
202			cand = t;
203			cand_score = score;
204		}
205	}
206
207	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
208		if (t->parms.i_key != key ||
209		    !(t->dev->flags & IFF_UP))
210			continue;
211
212		if (t->dev->type != ARPHRD_IP6GRE &&
213		    t->dev->type != dev_type)
214			continue;
215
216		score = 0;
217		if (t->parms.link != link)
218			score |= 1;
219		if (t->dev->type != dev_type)
220			score |= 2;
221		if (score == 0)
222			return t;
223
224		if (score < cand_score) {
225			cand = t;
226			cand_score = score;
227		}
228	}
229
230	if (cand)
231		return cand;
232
233	if (gre_proto == htons(ETH_P_ERSPAN) ||
234	    gre_proto == htons(ETH_P_ERSPAN2))
235		t = rcu_dereference(ign->collect_md_tun_erspan);
236	else
237		t = rcu_dereference(ign->collect_md_tun);
238
239	if (t && t->dev->flags & IFF_UP)
240		return t;
241
242	ndev = READ_ONCE(ign->fb_tunnel_dev);
243	if (ndev && ndev->flags & IFF_UP)
244		return netdev_priv(ndev);
245
246	return NULL;
247}
248
249static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
250		const struct __ip6_tnl_parm *p)
251{
252	const struct in6_addr *remote = &p->raddr;
253	const struct in6_addr *local = &p->laddr;
254	unsigned int h = HASH_KEY(p->i_key);
255	int prio = 0;
256
257	if (!ipv6_addr_any(local))
258		prio |= 1;
259	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
260		prio |= 2;
261		h ^= HASH_ADDR(remote);
262	}
263
264	return &ign->tunnels[prio][h];
265}
266
267static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
268{
269	if (t->parms.collect_md)
270		rcu_assign_pointer(ign->collect_md_tun, t);
271}
272
273static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
274{
275	if (t->parms.collect_md)
276		rcu_assign_pointer(ign->collect_md_tun_erspan, t);
277}
278
279static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
280{
281	if (t->parms.collect_md)
282		rcu_assign_pointer(ign->collect_md_tun, NULL);
283}
284
285static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
286				       struct ip6_tnl *t)
287{
288	if (t->parms.collect_md)
289		rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
290}
291
292static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
293		const struct ip6_tnl *t)
294{
295	return __ip6gre_bucket(ign, &t->parms);
296}
297
298static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
299{
300	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
301
302	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
303	rcu_assign_pointer(*tp, t);
304}
305
306static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
307{
308	struct ip6_tnl __rcu **tp;
309	struct ip6_tnl *iter;
310
311	for (tp = ip6gre_bucket(ign, t);
312	     (iter = rtnl_dereference(*tp)) != NULL;
313	     tp = &iter->next) {
314		if (t == iter) {
315			rcu_assign_pointer(*tp, t->next);
316			break;
317		}
318	}
319}
320
321static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
322					   const struct __ip6_tnl_parm *parms,
323					   int type)
324{
325	const struct in6_addr *remote = &parms->raddr;
326	const struct in6_addr *local = &parms->laddr;
327	__be32 key = parms->i_key;
328	int link = parms->link;
329	struct ip6_tnl *t;
330	struct ip6_tnl __rcu **tp;
331	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
332
333	for (tp = __ip6gre_bucket(ign, parms);
334	     (t = rtnl_dereference(*tp)) != NULL;
335	     tp = &t->next)
336		if (ipv6_addr_equal(local, &t->parms.laddr) &&
337		    ipv6_addr_equal(remote, &t->parms.raddr) &&
338		    key == t->parms.i_key &&
339		    link == t->parms.link &&
340		    type == t->dev->type)
341			break;
342
343	return t;
344}
345
346static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
347		const struct __ip6_tnl_parm *parms, int create)
348{
349	struct ip6_tnl *t, *nt;
350	struct net_device *dev;
351	char name[IFNAMSIZ];
352	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
353
354	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
355	if (t && create)
356		return NULL;
357	if (t || !create)
358		return t;
359
360	if (parms->name[0]) {
361		if (!dev_valid_name(parms->name))
362			return NULL;
363		strlcpy(name, parms->name, IFNAMSIZ);
364	} else {
365		strcpy(name, "ip6gre%d");
366	}
367	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
368			   ip6gre_tunnel_setup);
369	if (!dev)
370		return NULL;
371
372	dev_net_set(dev, net);
373
374	nt = netdev_priv(dev);
375	nt->parms = *parms;
376	dev->rtnl_link_ops = &ip6gre_link_ops;
377
378	nt->dev = dev;
379	nt->net = dev_net(dev);
380
381	if (register_netdevice(dev) < 0)
382		goto failed_free;
383
384	ip6gre_tnl_link_config(nt, 1);
385
386	/* Can use a lockless transmit, unless we generate output sequences */
387	if (!(nt->parms.o_flags & TUNNEL_SEQ))
388		dev->features |= NETIF_F_LLTX;
389
390	ip6gre_tunnel_link(ign, nt);
391	return nt;
392
393failed_free:
394	free_netdev(dev);
395	return NULL;
396}
397
398static void ip6erspan_tunnel_uninit(struct net_device *dev)
399{
400	struct ip6_tnl *t = netdev_priv(dev);
401	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
402
403	ip6erspan_tunnel_unlink_md(ign, t);
404	ip6gre_tunnel_unlink(ign, t);
405	dst_cache_reset(&t->dst_cache);
406	dev_put(dev);
407}
408
409static void ip6gre_tunnel_uninit(struct net_device *dev)
410{
411	struct ip6_tnl *t = netdev_priv(dev);
412	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
413
414	ip6gre_tunnel_unlink_md(ign, t);
415	ip6gre_tunnel_unlink(ign, t);
416	if (ign->fb_tunnel_dev == dev)
417		WRITE_ONCE(ign->fb_tunnel_dev, NULL);
418	dst_cache_reset(&t->dst_cache);
419	dev_put(dev);
420}
421
422
423static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
424		       u8 type, u8 code, int offset, __be32 info)
425{
426	struct net *net = dev_net(skb->dev);
427	const struct ipv6hdr *ipv6h;
428	struct tnl_ptk_info tpi;
429	struct ip6_tnl *t;
430
431	if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
432			     offset) < 0)
433		return -EINVAL;
434
435	ipv6h = (const struct ipv6hdr *)skb->data;
436	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
437				 tpi.key, tpi.proto);
438	if (!t)
439		return -ENOENT;
440
441	switch (type) {
442	case ICMPV6_DEST_UNREACH:
443		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
444				    t->parms.name);
445		if (code != ICMPV6_PORT_UNREACH)
446			break;
447		return 0;
448	case ICMPV6_TIME_EXCEED:
449		if (code == ICMPV6_EXC_HOPLIMIT) {
450			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
451					    t->parms.name);
452			break;
453		}
454		return 0;
455	case ICMPV6_PARAMPROB: {
456		struct ipv6_tlv_tnl_enc_lim *tel;
457		__u32 teli;
458
459		teli = 0;
460		if (code == ICMPV6_HDR_FIELD)
461			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
462
463		if (teli && teli == be32_to_cpu(info) - 2) {
464			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
465			if (tel->encap_limit == 0) {
466				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
467						    t->parms.name);
468			}
469		} else {
470			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
471					    t->parms.name);
472		}
473		return 0;
474	}
475	case ICMPV6_PKT_TOOBIG:
476		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
477		return 0;
478	case NDISC_REDIRECT:
479		ip6_redirect(skb, net, skb->dev->ifindex, 0,
480			     sock_net_uid(net, NULL));
481		return 0;
482	}
483
484	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
485		t->err_count++;
486	else
487		t->err_count = 1;
488	t->err_time = jiffies;
489
490	return 0;
491}
492
493static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
494{
495	const struct ipv6hdr *ipv6h;
496	struct ip6_tnl *tunnel;
497
498	ipv6h = ipv6_hdr(skb);
499	tunnel = ip6gre_tunnel_lookup(skb->dev,
500				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
501				      tpi->proto);
502	if (tunnel) {
503		if (tunnel->parms.collect_md) {
504			struct metadata_dst *tun_dst;
505			__be64 tun_id;
506			__be16 flags;
507
508			flags = tpi->flags;
509			tun_id = key32_to_tunnel_id(tpi->key);
510
511			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
512			if (!tun_dst)
513				return PACKET_REJECT;
514
515			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
516		} else {
517			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
518		}
519
520		return PACKET_RCVD;
521	}
522
523	return PACKET_REJECT;
524}
525
526static int ip6erspan_rcv(struct sk_buff *skb,
527			 struct tnl_ptk_info *tpi,
528			 int gre_hdr_len)
529{
530	struct erspan_base_hdr *ershdr;
531	const struct ipv6hdr *ipv6h;
532	struct erspan_md2 *md2;
533	struct ip6_tnl *tunnel;
534	u8 ver;
535
536	ipv6h = ipv6_hdr(skb);
537	ershdr = (struct erspan_base_hdr *)skb->data;
538	ver = ershdr->ver;
539
540	tunnel = ip6gre_tunnel_lookup(skb->dev,
541				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
542				      tpi->proto);
543	if (tunnel) {
544		int len = erspan_hdr_len(ver);
545
546		if (unlikely(!pskb_may_pull(skb, len)))
547			return PACKET_REJECT;
548
549		if (__iptunnel_pull_header(skb, len,
550					   htons(ETH_P_TEB),
551					   false, false) < 0)
552			return PACKET_REJECT;
553
554		if (tunnel->parms.collect_md) {
555			struct erspan_metadata *pkt_md, *md;
556			struct metadata_dst *tun_dst;
557			struct ip_tunnel_info *info;
558			unsigned char *gh;
559			__be64 tun_id;
560			__be16 flags;
561
562			tpi->flags |= TUNNEL_KEY;
563			flags = tpi->flags;
564			tun_id = key32_to_tunnel_id(tpi->key);
565
566			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
567						  sizeof(*md));
568			if (!tun_dst)
569				return PACKET_REJECT;
570
571			/* skb can be uncloned in __iptunnel_pull_header, so
572			 * old pkt_md is no longer valid and we need to reset
573			 * it
574			 */
575			gh = skb_network_header(skb) +
576			     skb_network_header_len(skb);
577			pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
578							    sizeof(*ershdr));
579			info = &tun_dst->u.tun_info;
580			md = ip_tunnel_info_opts(info);
581			md->version = ver;
582			md2 = &md->u.md2;
583			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
584						       ERSPAN_V2_MDSIZE);
585			info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
586			info->options_len = sizeof(*md);
587
588			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
589
590		} else {
591			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
592		}
593
594		return PACKET_RCVD;
595	}
596
597	return PACKET_REJECT;
598}
599
600static int gre_rcv(struct sk_buff *skb)
601{
602	struct tnl_ptk_info tpi;
603	bool csum_err = false;
604	int hdr_len;
605
606	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
607	if (hdr_len < 0)
608		goto drop;
609
610	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
611		goto drop;
612
613	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
614		     tpi.proto == htons(ETH_P_ERSPAN2))) {
615		if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
616			return 0;
617		goto out;
618	}
619
620	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
621		return 0;
622
623out:
624	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
625drop:
626	kfree_skb(skb);
627	return 0;
628}
629
630static int gre_handle_offloads(struct sk_buff *skb, bool csum)
631{
632	return iptunnel_handle_offloads(skb,
633					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
634}
635
636static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
637				     struct net_device *dev,
638				     struct flowi6 *fl6, __u8 *dsfield,
639				     int *encap_limit)
640{
641	const struct iphdr *iph = ip_hdr(skb);
642	struct ip6_tnl *t = netdev_priv(dev);
643
644	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
645		*encap_limit = t->parms.encap_limit;
646
647	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
648
649	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
650		*dsfield = ipv4_get_dsfield(iph);
651	else
652		*dsfield = ip6_tclass(t->parms.flowinfo);
653
654	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
655		fl6->flowi6_mark = skb->mark;
656	else
657		fl6->flowi6_mark = t->parms.fwmark;
658
659	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
660}
661
662static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
663				    struct net_device *dev,
664				    struct flowi6 *fl6, __u8 *dsfield,
665				    int *encap_limit)
666{
667	struct ipv6hdr *ipv6h;
668	struct ip6_tnl *t = netdev_priv(dev);
669	__u16 offset;
670
671	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
672	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
673	ipv6h = ipv6_hdr(skb);
674
675	if (offset > 0) {
676		struct ipv6_tlv_tnl_enc_lim *tel;
677
678		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
679		if (tel->encap_limit == 0) {
680			icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
681					ICMPV6_HDR_FIELD, offset + 2);
682			return -1;
683		}
684		*encap_limit = tel->encap_limit - 1;
685	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
686		*encap_limit = t->parms.encap_limit;
687	}
688
689	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
690
691	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
692		*dsfield = ipv6_get_dsfield(ipv6h);
693	else
694		*dsfield = ip6_tclass(t->parms.flowinfo);
695
696	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
697		fl6->flowlabel |= ip6_flowlabel(ipv6h);
698
699	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
700		fl6->flowi6_mark = skb->mark;
701	else
702		fl6->flowi6_mark = t->parms.fwmark;
703
704	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
705
706	return 0;
707}
708
709static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
710{
711	struct ip_tunnel_info *tun_info;
712
713	tun_info = skb_tunnel_info(skb);
714	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
715		return ERR_PTR(-EINVAL);
716
717	return tun_info;
718}
719
720static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
721			       struct net_device *dev, __u8 dsfield,
722			       struct flowi6 *fl6, int encap_limit,
723			       __u32 *pmtu, __be16 proto)
724{
725	struct ip6_tnl *tunnel = netdev_priv(dev);
726	__be16 protocol;
727	__be16 flags;
728
729	if (dev->type == ARPHRD_ETHER)
730		IPCB(skb)->flags = 0;
731
732	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
733		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
734	else
735		fl6->daddr = tunnel->parms.raddr;
736
737	/* Push GRE header. */
738	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
739
740	if (tunnel->parms.collect_md) {
741		struct ip_tunnel_info *tun_info;
742		const struct ip_tunnel_key *key;
743		int tun_hlen;
744
745		tun_info = skb_tunnel_info_txcheck(skb);
746		if (IS_ERR(tun_info) ||
747		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
748			return -EINVAL;
749
750		key = &tun_info->key;
751		memset(fl6, 0, sizeof(*fl6));
752		fl6->flowi6_proto = IPPROTO_GRE;
753		fl6->daddr = key->u.ipv6.dst;
754		fl6->flowlabel = key->label;
755		fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
756		fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
757
758		dsfield = key->tos;
759		flags = key->tun_flags &
760			(TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
761		tun_hlen = gre_calc_hlen(flags);
762
763		if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
764			return -ENOMEM;
765
766		gre_build_header(skb, tun_hlen,
767				 flags, protocol,
768				 tunnel_id_to_key32(tun_info->key.tun_id),
769				 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
770						      : 0);
771
772	} else {
773		if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
774			return -ENOMEM;
775
776		flags = tunnel->parms.o_flags;
777
778		gre_build_header(skb, tunnel->tun_hlen, flags,
779				 protocol, tunnel->parms.o_key,
780				 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
781						      : 0);
782	}
783
784	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
785			    NEXTHDR_GRE);
786}
787
788static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
789{
790	struct ip6_tnl *t = netdev_priv(dev);
791	int encap_limit = -1;
792	struct flowi6 fl6;
793	__u8 dsfield = 0;
794	__u32 mtu;
795	int err;
796
797	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
798
799	if (!t->parms.collect_md)
800		prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
801					 &dsfield, &encap_limit);
802
803	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
804	if (err)
805		return -1;
806
807	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
808			  skb->protocol);
809	if (err != 0) {
810		/* XXX: send ICMP error even if DF is not set. */
811		if (err == -EMSGSIZE)
812			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
813				      htonl(mtu));
814		return -1;
815	}
816
817	return 0;
818}
819
820static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
821{
822	struct ip6_tnl *t = netdev_priv(dev);
823	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
824	int encap_limit = -1;
825	struct flowi6 fl6;
826	__u8 dsfield = 0;
827	__u32 mtu;
828	int err;
829
830	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
831		return -1;
832
833	if (!t->parms.collect_md &&
834	    prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
835		return -1;
836
837	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
838		return -1;
839
840	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
841			  &mtu, skb->protocol);
842	if (err != 0) {
843		if (err == -EMSGSIZE)
844			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
845		return -1;
846	}
847
848	return 0;
849}
850
851/**
852 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
853 *   @t: the outgoing tunnel device
854 *   @hdr: IPv6 header from the incoming packet
855 *
856 * Description:
857 *   Avoid trivial tunneling loop by checking that tunnel exit-point
858 *   doesn't match source of incoming packet.
859 *
860 * Return:
861 *   1 if conflict,
862 *   0 else
863 **/
864
865static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
866	const struct ipv6hdr *hdr)
867{
868	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
869}
870
871static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
872{
873	struct ip6_tnl *t = netdev_priv(dev);
874	int encap_limit = -1;
875	struct flowi6 fl6;
876	__u32 mtu;
877	int err;
878
879	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
880		encap_limit = t->parms.encap_limit;
881
882	if (!t->parms.collect_md)
883		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
884
885	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
886	if (err)
887		return err;
888
889	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
890
891	return err;
892}
893
894static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
895	struct net_device *dev)
896{
897	struct ip6_tnl *t = netdev_priv(dev);
898	struct net_device_stats *stats = &t->dev->stats;
899	int ret;
900
901	if (!pskb_inet_may_pull(skb))
902		goto tx_err;
903
904	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
905		goto tx_err;
906
907	switch (skb->protocol) {
908	case htons(ETH_P_IP):
909		ret = ip6gre_xmit_ipv4(skb, dev);
910		break;
911	case htons(ETH_P_IPV6):
912		ret = ip6gre_xmit_ipv6(skb, dev);
913		break;
914	default:
915		ret = ip6gre_xmit_other(skb, dev);
916		break;
917	}
918
919	if (ret < 0)
920		goto tx_err;
921
922	return NETDEV_TX_OK;
923
924tx_err:
925	if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
926		stats->tx_errors++;
927	stats->tx_dropped++;
928	kfree_skb(skb);
929	return NETDEV_TX_OK;
930}
931
932static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
933					 struct net_device *dev)
934{
935	struct ip_tunnel_info *tun_info = NULL;
936	struct ip6_tnl *t = netdev_priv(dev);
937	struct dst_entry *dst = skb_dst(skb);
938	struct net_device_stats *stats;
939	bool truncate = false;
940	int encap_limit = -1;
941	__u8 dsfield = false;
942	struct flowi6 fl6;
943	int err = -EINVAL;
944	__be16 proto;
945	__u32 mtu;
946	int nhoff;
947
948	if (!pskb_inet_may_pull(skb))
949		goto tx_err;
950
951	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
952		goto tx_err;
953
954	if (gre_handle_offloads(skb, false))
955		goto tx_err;
956
957	if (skb->len > dev->mtu + dev->hard_header_len) {
958		if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
959			goto tx_err;
960		truncate = true;
961	}
962
963	nhoff = skb_network_offset(skb);
964	if (skb->protocol == htons(ETH_P_IP) &&
965	    (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
966		truncate = true;
967
968	if (skb->protocol == htons(ETH_P_IPV6)) {
969		int thoff;
970
971		if (skb_transport_header_was_set(skb))
972			thoff = skb_transport_offset(skb);
973		else
974			thoff = nhoff + sizeof(struct ipv6hdr);
975		if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
976			truncate = true;
977	}
978
979	if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
980		goto tx_err;
981
982	t->parms.o_flags &= ~TUNNEL_KEY;
983	IPCB(skb)->flags = 0;
984
985	/* For collect_md mode, derive fl6 from the tunnel key,
986	 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
987	 */
988	if (t->parms.collect_md) {
989		const struct ip_tunnel_key *key;
990		struct erspan_metadata *md;
991		__be32 tun_id;
992
993		tun_info = skb_tunnel_info_txcheck(skb);
994		if (IS_ERR(tun_info) ||
995		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
996			goto tx_err;
997
998		key = &tun_info->key;
999		memset(&fl6, 0, sizeof(fl6));
1000		fl6.flowi6_proto = IPPROTO_GRE;
1001		fl6.daddr = key->u.ipv6.dst;
1002		fl6.flowlabel = key->label;
1003		fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1004		fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1005
1006		dsfield = key->tos;
1007		if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1008			goto tx_err;
1009		if (tun_info->options_len < sizeof(*md))
1010			goto tx_err;
1011		md = ip_tunnel_info_opts(tun_info);
1012
1013		tun_id = tunnel_id_to_key32(key->tun_id);
1014		if (md->version == 1) {
1015			erspan_build_header(skb,
1016					    ntohl(tun_id),
1017					    ntohl(md->u.index), truncate,
1018					    false);
1019			proto = htons(ETH_P_ERSPAN);
1020		} else if (md->version == 2) {
1021			erspan_build_header_v2(skb,
1022					       ntohl(tun_id),
1023					       md->u.md2.dir,
1024					       get_hwid(&md->u.md2),
1025					       truncate, false);
1026			proto = htons(ETH_P_ERSPAN2);
1027		} else {
1028			goto tx_err;
1029		}
1030	} else {
1031		switch (skb->protocol) {
1032		case htons(ETH_P_IP):
1033			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1034			prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1035						 &dsfield, &encap_limit);
1036			break;
1037		case htons(ETH_P_IPV6):
1038			if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1039				goto tx_err;
1040			if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1041						     &dsfield, &encap_limit))
1042				goto tx_err;
1043			break;
1044		default:
1045			memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1046			break;
1047		}
1048
1049		if (t->parms.erspan_ver == 1) {
1050			erspan_build_header(skb, ntohl(t->parms.o_key),
1051					    t->parms.index,
1052					    truncate, false);
1053			proto = htons(ETH_P_ERSPAN);
1054		} else if (t->parms.erspan_ver == 2) {
1055			erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1056					       t->parms.dir,
1057					       t->parms.hwid,
1058					       truncate, false);
1059			proto = htons(ETH_P_ERSPAN2);
1060		} else {
1061			goto tx_err;
1062		}
1063
1064		fl6.daddr = t->parms.raddr;
1065	}
1066
1067	/* Push GRE header. */
1068	gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1069
1070	/* TooBig packet may have updated dst->dev's mtu */
1071	if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1072		dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1073
1074	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1075			   NEXTHDR_GRE);
1076	if (err != 0) {
1077		/* XXX: send ICMP error even if DF is not set. */
1078		if (err == -EMSGSIZE) {
1079			if (skb->protocol == htons(ETH_P_IP))
1080				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1081					      ICMP_FRAG_NEEDED, htonl(mtu));
1082			else
1083				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1084		}
1085
1086		goto tx_err;
1087	}
1088	return NETDEV_TX_OK;
1089
1090tx_err:
1091	stats = &t->dev->stats;
1092	if (!IS_ERR(tun_info))
1093		stats->tx_errors++;
1094	stats->tx_dropped++;
1095	kfree_skb(skb);
1096	return NETDEV_TX_OK;
1097}
1098
1099static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1100{
1101	struct net_device *dev = t->dev;
1102	struct __ip6_tnl_parm *p = &t->parms;
1103	struct flowi6 *fl6 = &t->fl.u.ip6;
1104
1105	if (dev->type != ARPHRD_ETHER) {
1106		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1107		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1108	}
1109
1110	/* Set up flowi template */
1111	fl6->saddr = p->laddr;
1112	fl6->daddr = p->raddr;
1113	fl6->flowi6_oif = p->link;
1114	fl6->flowlabel = 0;
1115	fl6->flowi6_proto = IPPROTO_GRE;
1116	fl6->fl6_gre_key = t->parms.o_key;
1117
1118	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1119		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1120	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1121		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1122
1123	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1124	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1125
1126	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1127			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1128		dev->flags |= IFF_POINTOPOINT;
1129	else
1130		dev->flags &= ~IFF_POINTOPOINT;
1131}
1132
1133static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1134					 int t_hlen)
1135{
1136	const struct __ip6_tnl_parm *p = &t->parms;
1137	struct net_device *dev = t->dev;
1138
1139	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1140		int strict = (ipv6_addr_type(&p->raddr) &
1141			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1142
1143		struct rt6_info *rt = rt6_lookup(t->net,
1144						 &p->raddr, &p->laddr,
1145						 p->link, NULL, strict);
1146
1147		if (!rt)
1148			return;
1149
1150		if (rt->dst.dev) {
1151			unsigned short dst_len = rt->dst.dev->hard_header_len +
1152						 t_hlen;
1153
1154			if (t->dev->header_ops)
1155				dev->hard_header_len = dst_len;
1156			else
1157				dev->needed_headroom = dst_len;
1158
1159			if (set_mtu) {
1160				int mtu = rt->dst.dev->mtu - t_hlen;
1161
1162				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1163					mtu -= 8;
1164				if (dev->type == ARPHRD_ETHER)
1165					mtu -= ETH_HLEN;
1166
1167				if (mtu < IPV6_MIN_MTU)
1168					mtu = IPV6_MIN_MTU;
1169				WRITE_ONCE(dev->mtu, mtu);
1170			}
1171		}
1172		ip6_rt_put(rt);
1173	}
1174}
1175
1176static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1177{
1178	int t_hlen;
1179
1180	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1181	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1182
1183	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1184
1185	if (tunnel->dev->header_ops)
1186		tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1187	else
1188		tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1189
1190	return t_hlen;
1191}
1192
1193static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1194{
1195	ip6gre_tnl_link_config_common(t);
1196	ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1197}
1198
1199static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1200				     const struct __ip6_tnl_parm *p)
1201{
1202	t->parms.laddr = p->laddr;
1203	t->parms.raddr = p->raddr;
1204	t->parms.flags = p->flags;
1205	t->parms.hop_limit = p->hop_limit;
1206	t->parms.encap_limit = p->encap_limit;
1207	t->parms.flowinfo = p->flowinfo;
1208	t->parms.link = p->link;
1209	t->parms.proto = p->proto;
1210	t->parms.i_key = p->i_key;
1211	t->parms.o_key = p->o_key;
1212	t->parms.i_flags = p->i_flags;
1213	t->parms.o_flags = p->o_flags;
1214	t->parms.fwmark = p->fwmark;
1215	t->parms.erspan_ver = p->erspan_ver;
1216	t->parms.index = p->index;
1217	t->parms.dir = p->dir;
1218	t->parms.hwid = p->hwid;
1219	dst_cache_reset(&t->dst_cache);
1220}
1221
1222static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1223			     int set_mtu)
1224{
1225	ip6gre_tnl_copy_tnl_parm(t, p);
1226	ip6gre_tnl_link_config(t, set_mtu);
1227	return 0;
1228}
1229
1230static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1231	const struct ip6_tnl_parm2 *u)
1232{
1233	p->laddr = u->laddr;
1234	p->raddr = u->raddr;
1235	p->flags = u->flags;
1236	p->hop_limit = u->hop_limit;
1237	p->encap_limit = u->encap_limit;
1238	p->flowinfo = u->flowinfo;
1239	p->link = u->link;
1240	p->i_key = u->i_key;
1241	p->o_key = u->o_key;
1242	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1243	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1244	memcpy(p->name, u->name, sizeof(u->name));
1245}
1246
1247static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1248	const struct __ip6_tnl_parm *p)
1249{
1250	u->proto = IPPROTO_GRE;
1251	u->laddr = p->laddr;
1252	u->raddr = p->raddr;
1253	u->flags = p->flags;
1254	u->hop_limit = p->hop_limit;
1255	u->encap_limit = p->encap_limit;
1256	u->flowinfo = p->flowinfo;
1257	u->link = p->link;
1258	u->i_key = p->i_key;
1259	u->o_key = p->o_key;
1260	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1261	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1262	memcpy(u->name, p->name, sizeof(u->name));
1263}
1264
1265static int ip6gre_tunnel_ioctl(struct net_device *dev,
1266	struct ifreq *ifr, int cmd)
1267{
1268	int err = 0;
1269	struct ip6_tnl_parm2 p;
1270	struct __ip6_tnl_parm p1;
1271	struct ip6_tnl *t = netdev_priv(dev);
1272	struct net *net = t->net;
1273	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1274
1275	memset(&p1, 0, sizeof(p1));
1276
1277	switch (cmd) {
1278	case SIOCGETTUNNEL:
1279		if (dev == ign->fb_tunnel_dev) {
1280			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1281				err = -EFAULT;
1282				break;
1283			}
1284			ip6gre_tnl_parm_from_user(&p1, &p);
1285			t = ip6gre_tunnel_locate(net, &p1, 0);
1286			if (!t)
1287				t = netdev_priv(dev);
1288		}
1289		memset(&p, 0, sizeof(p));
1290		ip6gre_tnl_parm_to_user(&p, &t->parms);
1291		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1292			err = -EFAULT;
1293		break;
1294
1295	case SIOCADDTUNNEL:
1296	case SIOCCHGTUNNEL:
1297		err = -EPERM;
1298		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1299			goto done;
1300
1301		err = -EFAULT;
1302		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1303			goto done;
1304
1305		err = -EINVAL;
1306		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1307			goto done;
1308
1309		if (!(p.i_flags&GRE_KEY))
1310			p.i_key = 0;
1311		if (!(p.o_flags&GRE_KEY))
1312			p.o_key = 0;
1313
1314		ip6gre_tnl_parm_from_user(&p1, &p);
1315		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1316
1317		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1318			if (t) {
1319				if (t->dev != dev) {
1320					err = -EEXIST;
1321					break;
1322				}
1323			} else {
1324				t = netdev_priv(dev);
1325
1326				ip6gre_tunnel_unlink(ign, t);
1327				synchronize_net();
1328				ip6gre_tnl_change(t, &p1, 1);
1329				ip6gre_tunnel_link(ign, t);
1330				netdev_state_change(dev);
1331			}
1332		}
1333
1334		if (t) {
1335			err = 0;
1336
1337			memset(&p, 0, sizeof(p));
1338			ip6gre_tnl_parm_to_user(&p, &t->parms);
1339			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1340				err = -EFAULT;
1341		} else
1342			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1343		break;
1344
1345	case SIOCDELTUNNEL:
1346		err = -EPERM;
1347		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1348			goto done;
1349
1350		if (dev == ign->fb_tunnel_dev) {
1351			err = -EFAULT;
1352			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1353				goto done;
1354			err = -ENOENT;
1355			ip6gre_tnl_parm_from_user(&p1, &p);
1356			t = ip6gre_tunnel_locate(net, &p1, 0);
1357			if (!t)
1358				goto done;
1359			err = -EPERM;
1360			if (t == netdev_priv(ign->fb_tunnel_dev))
1361				goto done;
1362			dev = t->dev;
1363		}
1364		unregister_netdevice(dev);
1365		err = 0;
1366		break;
1367
1368	default:
1369		err = -EINVAL;
1370	}
1371
1372done:
1373	return err;
1374}
1375
1376static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1377			 unsigned short type, const void *daddr,
1378			 const void *saddr, unsigned int len)
1379{
1380	struct ip6_tnl *t = netdev_priv(dev);
1381	struct ipv6hdr *ipv6h;
1382	__be16 *p;
1383
1384	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1385	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1386						  t->fl.u.ip6.flowlabel,
1387						  true, &t->fl.u.ip6));
1388	ipv6h->hop_limit = t->parms.hop_limit;
1389	ipv6h->nexthdr = NEXTHDR_GRE;
1390	ipv6h->saddr = t->parms.laddr;
1391	ipv6h->daddr = t->parms.raddr;
1392
1393	p = (__be16 *)(ipv6h + 1);
1394	p[0] = t->parms.o_flags;
1395	p[1] = htons(type);
1396
1397	/*
1398	 *	Set the source hardware address.
1399	 */
1400
1401	if (saddr)
1402		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1403	if (daddr)
1404		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1405	if (!ipv6_addr_any(&ipv6h->daddr))
1406		return t->hlen;
1407
1408	return -t->hlen;
1409}
1410
1411static const struct header_ops ip6gre_header_ops = {
1412	.create	= ip6gre_header,
1413};
1414
1415static const struct net_device_ops ip6gre_netdev_ops = {
1416	.ndo_init		= ip6gre_tunnel_init,
1417	.ndo_uninit		= ip6gre_tunnel_uninit,
1418	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1419	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1420	.ndo_change_mtu		= ip6_tnl_change_mtu,
1421	.ndo_get_stats64	= ip_tunnel_get_stats64,
1422	.ndo_get_iflink		= ip6_tnl_get_iflink,
1423};
1424
1425static void ip6gre_dev_free(struct net_device *dev)
1426{
1427	struct ip6_tnl *t = netdev_priv(dev);
1428
1429	gro_cells_destroy(&t->gro_cells);
1430	dst_cache_destroy(&t->dst_cache);
1431	free_percpu(dev->tstats);
1432}
1433
1434static void ip6gre_tunnel_setup(struct net_device *dev)
1435{
1436	dev->netdev_ops = &ip6gre_netdev_ops;
1437	dev->needs_free_netdev = true;
1438	dev->priv_destructor = ip6gre_dev_free;
1439
1440	dev->type = ARPHRD_IP6GRE;
1441
1442	dev->flags |= IFF_NOARP;
1443	dev->addr_len = sizeof(struct in6_addr);
1444	netif_keep_dst(dev);
1445	/* This perm addr will be used as interface identifier by IPv6 */
1446	dev->addr_assign_type = NET_ADDR_RANDOM;
1447	eth_random_addr(dev->perm_addr);
1448}
1449
1450#define GRE6_FEATURES (NETIF_F_SG |		\
1451		       NETIF_F_FRAGLIST |	\
1452		       NETIF_F_HIGHDMA |	\
1453		       NETIF_F_HW_CSUM)
1454
1455static void ip6gre_tnl_init_features(struct net_device *dev)
1456{
1457	struct ip6_tnl *nt = netdev_priv(dev);
1458
1459	dev->features		|= GRE6_FEATURES;
1460	dev->hw_features	|= GRE6_FEATURES;
1461
1462	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1463		/* TCP offload with GRE SEQ is not supported, nor
1464		 * can we support 2 levels of outer headers requiring
1465		 * an update.
1466		 */
1467		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1468		    nt->encap.type == TUNNEL_ENCAP_NONE) {
1469			dev->features    |= NETIF_F_GSO_SOFTWARE;
1470			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1471		}
1472
1473		/* Can use a lockless transmit, unless we generate
1474		 * output sequences
1475		 */
1476		dev->features |= NETIF_F_LLTX;
1477	}
1478}
1479
1480static int ip6gre_tunnel_init_common(struct net_device *dev)
1481{
1482	struct ip6_tnl *tunnel;
1483	int ret;
1484	int t_hlen;
1485
1486	tunnel = netdev_priv(dev);
1487
1488	tunnel->dev = dev;
1489	tunnel->net = dev_net(dev);
1490	strcpy(tunnel->parms.name, dev->name);
1491
1492	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1493	if (!dev->tstats)
1494		return -ENOMEM;
1495
1496	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1497	if (ret)
1498		goto cleanup_alloc_pcpu_stats;
1499
1500	ret = gro_cells_init(&tunnel->gro_cells, dev);
1501	if (ret)
1502		goto cleanup_dst_cache_init;
1503
1504	t_hlen = ip6gre_calc_hlen(tunnel);
1505	dev->mtu = ETH_DATA_LEN - t_hlen;
1506	if (dev->type == ARPHRD_ETHER)
1507		dev->mtu -= ETH_HLEN;
1508	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1509		dev->mtu -= 8;
1510
1511	if (tunnel->parms.collect_md) {
1512		netif_keep_dst(dev);
1513	}
1514	ip6gre_tnl_init_features(dev);
1515
1516	dev_hold(dev);
1517	return 0;
1518
1519cleanup_dst_cache_init:
1520	dst_cache_destroy(&tunnel->dst_cache);
1521cleanup_alloc_pcpu_stats:
1522	free_percpu(dev->tstats);
1523	dev->tstats = NULL;
1524	return ret;
1525}
1526
1527static int ip6gre_tunnel_init(struct net_device *dev)
1528{
1529	struct ip6_tnl *tunnel;
1530	int ret;
1531
1532	ret = ip6gre_tunnel_init_common(dev);
1533	if (ret)
1534		return ret;
1535
1536	tunnel = netdev_priv(dev);
1537
1538	if (tunnel->parms.collect_md)
1539		return 0;
1540
1541	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1542	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1543
1544	if (ipv6_addr_any(&tunnel->parms.raddr))
1545		dev->header_ops = &ip6gre_header_ops;
1546
1547	return 0;
1548}
1549
1550static void ip6gre_fb_tunnel_init(struct net_device *dev)
1551{
1552	struct ip6_tnl *tunnel = netdev_priv(dev);
1553
1554	tunnel->dev = dev;
1555	tunnel->net = dev_net(dev);
1556	strcpy(tunnel->parms.name, dev->name);
1557
1558	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1559}
1560
1561static struct inet6_protocol ip6gre_protocol __read_mostly = {
1562	.handler     = gre_rcv,
1563	.err_handler = ip6gre_err,
1564	.flags       = INET6_PROTO_FINAL,
1565};
1566
1567static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1568{
1569	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1570	struct net_device *dev, *aux;
1571	int prio;
1572
1573	for_each_netdev_safe(net, dev, aux)
1574		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1575		    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1576		    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1577			unregister_netdevice_queue(dev, head);
1578
1579	for (prio = 0; prio < 4; prio++) {
1580		int h;
1581		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1582			struct ip6_tnl *t;
1583
1584			t = rtnl_dereference(ign->tunnels[prio][h]);
1585
1586			while (t) {
1587				/* If dev is in the same netns, it has already
1588				 * been added to the list by the previous loop.
1589				 */
1590				if (!net_eq(dev_net(t->dev), net))
1591					unregister_netdevice_queue(t->dev,
1592								   head);
1593				t = rtnl_dereference(t->next);
1594			}
1595		}
1596	}
1597}
1598
1599static int __net_init ip6gre_init_net(struct net *net)
1600{
1601	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1602	struct net_device *ndev;
1603	int err;
1604
1605	if (!net_has_fallback_tunnels(net))
1606		return 0;
1607	ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1608			    NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1609	if (!ndev) {
1610		err = -ENOMEM;
1611		goto err_alloc_dev;
1612	}
1613	ign->fb_tunnel_dev = ndev;
1614	dev_net_set(ign->fb_tunnel_dev, net);
1615	/* FB netdevice is special: we have one, and only one per netns.
1616	 * Allowing to move it to another netns is clearly unsafe.
1617	 */
1618	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1619
1620
1621	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1622	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1623
1624	err = register_netdev(ign->fb_tunnel_dev);
1625	if (err)
1626		goto err_reg_dev;
1627
1628	rcu_assign_pointer(ign->tunnels_wc[0],
1629			   netdev_priv(ign->fb_tunnel_dev));
1630	return 0;
1631
1632err_reg_dev:
1633	free_netdev(ndev);
1634err_alloc_dev:
1635	return err;
1636}
1637
1638static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1639{
1640	struct net *net;
1641	LIST_HEAD(list);
1642
1643	rtnl_lock();
1644	list_for_each_entry(net, net_list, exit_list)
1645		ip6gre_destroy_tunnels(net, &list);
1646	unregister_netdevice_many(&list);
1647	rtnl_unlock();
1648}
1649
1650static struct pernet_operations ip6gre_net_ops = {
1651	.init = ip6gre_init_net,
1652	.exit_batch = ip6gre_exit_batch_net,
1653	.id   = &ip6gre_net_id,
1654	.size = sizeof(struct ip6gre_net),
1655};
1656
1657static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1658				  struct netlink_ext_ack *extack)
1659{
1660	__be16 flags;
1661
1662	if (!data)
1663		return 0;
1664
1665	flags = 0;
1666	if (data[IFLA_GRE_IFLAGS])
1667		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1668	if (data[IFLA_GRE_OFLAGS])
1669		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1670	if (flags & (GRE_VERSION|GRE_ROUTING))
1671		return -EINVAL;
1672
1673	return 0;
1674}
1675
1676static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1677			       struct netlink_ext_ack *extack)
1678{
1679	struct in6_addr daddr;
1680
1681	if (tb[IFLA_ADDRESS]) {
1682		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1683			return -EINVAL;
1684		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1685			return -EADDRNOTAVAIL;
1686	}
1687
1688	if (!data)
1689		goto out;
1690
1691	if (data[IFLA_GRE_REMOTE]) {
1692		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1693		if (ipv6_addr_any(&daddr))
1694			return -EINVAL;
1695	}
1696
1697out:
1698	return ip6gre_tunnel_validate(tb, data, extack);
1699}
1700
1701static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1702				  struct netlink_ext_ack *extack)
1703{
1704	__be16 flags = 0;
1705	int ret, ver = 0;
1706
1707	if (!data)
1708		return 0;
1709
1710	ret = ip6gre_tap_validate(tb, data, extack);
1711	if (ret)
1712		return ret;
1713
1714	/* ERSPAN should only have GRE sequence and key flag */
1715	if (data[IFLA_GRE_OFLAGS])
1716		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1717	if (data[IFLA_GRE_IFLAGS])
1718		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1719	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1720	    flags != (GRE_SEQ | GRE_KEY))
1721		return -EINVAL;
1722
1723	/* ERSPAN Session ID only has 10-bit. Since we reuse
1724	 * 32-bit key field as ID, check it's range.
1725	 */
1726	if (data[IFLA_GRE_IKEY] &&
1727	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1728		return -EINVAL;
1729
1730	if (data[IFLA_GRE_OKEY] &&
1731	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1732		return -EINVAL;
1733
1734	if (data[IFLA_GRE_ERSPAN_VER]) {
1735		ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1736		if (ver != 1 && ver != 2)
1737			return -EINVAL;
1738	}
1739
1740	if (ver == 1) {
1741		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1742			u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1743
1744			if (index & ~INDEX_MASK)
1745				return -EINVAL;
1746		}
1747	} else if (ver == 2) {
1748		if (data[IFLA_GRE_ERSPAN_DIR]) {
1749			u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1750
1751			if (dir & ~(DIR_MASK >> DIR_OFFSET))
1752				return -EINVAL;
1753		}
1754
1755		if (data[IFLA_GRE_ERSPAN_HWID]) {
1756			u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1757
1758			if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1759				return -EINVAL;
1760		}
1761	}
1762
1763	return 0;
1764}
1765
1766static void ip6erspan_set_version(struct nlattr *data[],
1767				  struct __ip6_tnl_parm *parms)
1768{
1769	if (!data)
1770		return;
1771
1772	parms->erspan_ver = 1;
1773	if (data[IFLA_GRE_ERSPAN_VER])
1774		parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1775
1776	if (parms->erspan_ver == 1) {
1777		if (data[IFLA_GRE_ERSPAN_INDEX])
1778			parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1779	} else if (parms->erspan_ver == 2) {
1780		if (data[IFLA_GRE_ERSPAN_DIR])
1781			parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1782		if (data[IFLA_GRE_ERSPAN_HWID])
1783			parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1784	}
1785}
1786
1787static void ip6gre_netlink_parms(struct nlattr *data[],
1788				struct __ip6_tnl_parm *parms)
1789{
1790	memset(parms, 0, sizeof(*parms));
1791
1792	if (!data)
1793		return;
1794
1795	if (data[IFLA_GRE_LINK])
1796		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1797
1798	if (data[IFLA_GRE_IFLAGS])
1799		parms->i_flags = gre_flags_to_tnl_flags(
1800				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1801
1802	if (data[IFLA_GRE_OFLAGS])
1803		parms->o_flags = gre_flags_to_tnl_flags(
1804				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1805
1806	if (data[IFLA_GRE_IKEY])
1807		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1808
1809	if (data[IFLA_GRE_OKEY])
1810		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1811
1812	if (data[IFLA_GRE_LOCAL])
1813		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1814
1815	if (data[IFLA_GRE_REMOTE])
1816		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1817
1818	if (data[IFLA_GRE_TTL])
1819		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1820
1821	if (data[IFLA_GRE_ENCAP_LIMIT])
1822		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1823
1824	if (data[IFLA_GRE_FLOWINFO])
1825		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1826
1827	if (data[IFLA_GRE_FLAGS])
1828		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1829
1830	if (data[IFLA_GRE_FWMARK])
1831		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1832
1833	if (data[IFLA_GRE_COLLECT_METADATA])
1834		parms->collect_md = true;
1835}
1836
1837static int ip6gre_tap_init(struct net_device *dev)
1838{
1839	int ret;
1840
1841	ret = ip6gre_tunnel_init_common(dev);
1842	if (ret)
1843		return ret;
1844
1845	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1846
1847	return 0;
1848}
1849
1850static const struct net_device_ops ip6gre_tap_netdev_ops = {
1851	.ndo_init = ip6gre_tap_init,
1852	.ndo_uninit = ip6gre_tunnel_uninit,
1853	.ndo_start_xmit = ip6gre_tunnel_xmit,
1854	.ndo_set_mac_address = eth_mac_addr,
1855	.ndo_validate_addr = eth_validate_addr,
1856	.ndo_change_mtu = ip6_tnl_change_mtu,
1857	.ndo_get_stats64 = ip_tunnel_get_stats64,
1858	.ndo_get_iflink = ip6_tnl_get_iflink,
1859};
1860
1861static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1862{
1863	int t_hlen;
1864
1865	tunnel->tun_hlen = 8;
1866	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1867		       erspan_hdr_len(tunnel->parms.erspan_ver);
1868
1869	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1870	tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1871	return t_hlen;
1872}
1873
1874static int ip6erspan_tap_init(struct net_device *dev)
1875{
1876	struct ip6_tnl *tunnel;
1877	int t_hlen;
1878	int ret;
1879
1880	tunnel = netdev_priv(dev);
1881
1882	tunnel->dev = dev;
1883	tunnel->net = dev_net(dev);
1884	strcpy(tunnel->parms.name, dev->name);
1885
1886	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1887	if (!dev->tstats)
1888		return -ENOMEM;
1889
1890	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1891	if (ret)
1892		goto cleanup_alloc_pcpu_stats;
1893
1894	ret = gro_cells_init(&tunnel->gro_cells, dev);
1895	if (ret)
1896		goto cleanup_dst_cache_init;
1897
1898	t_hlen = ip6erspan_calc_hlen(tunnel);
1899	dev->mtu = ETH_DATA_LEN - t_hlen;
1900	if (dev->type == ARPHRD_ETHER)
1901		dev->mtu -= ETH_HLEN;
1902	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1903		dev->mtu -= 8;
1904
1905	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1906	ip6erspan_tnl_link_config(tunnel, 1);
1907
1908	dev_hold(dev);
1909	return 0;
1910
1911cleanup_dst_cache_init:
1912	dst_cache_destroy(&tunnel->dst_cache);
1913cleanup_alloc_pcpu_stats:
1914	free_percpu(dev->tstats);
1915	dev->tstats = NULL;
1916	return ret;
1917}
1918
1919static const struct net_device_ops ip6erspan_netdev_ops = {
1920	.ndo_init =		ip6erspan_tap_init,
1921	.ndo_uninit =		ip6erspan_tunnel_uninit,
1922	.ndo_start_xmit =	ip6erspan_tunnel_xmit,
1923	.ndo_set_mac_address =	eth_mac_addr,
1924	.ndo_validate_addr =	eth_validate_addr,
1925	.ndo_change_mtu =	ip6_tnl_change_mtu,
1926	.ndo_get_stats64 =	ip_tunnel_get_stats64,
1927	.ndo_get_iflink =	ip6_tnl_get_iflink,
1928};
1929
1930static void ip6gre_tap_setup(struct net_device *dev)
1931{
1932
1933	ether_setup(dev);
1934
1935	dev->max_mtu = 0;
1936	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1937	dev->needs_free_netdev = true;
1938	dev->priv_destructor = ip6gre_dev_free;
1939
1940	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1941	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1942	netif_keep_dst(dev);
1943}
1944
1945static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1946				       struct ip_tunnel_encap *ipencap)
1947{
1948	bool ret = false;
1949
1950	memset(ipencap, 0, sizeof(*ipencap));
1951
1952	if (!data)
1953		return ret;
1954
1955	if (data[IFLA_GRE_ENCAP_TYPE]) {
1956		ret = true;
1957		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1958	}
1959
1960	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1961		ret = true;
1962		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1963	}
1964
1965	if (data[IFLA_GRE_ENCAP_SPORT]) {
1966		ret = true;
1967		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1968	}
1969
1970	if (data[IFLA_GRE_ENCAP_DPORT]) {
1971		ret = true;
1972		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1973	}
1974
1975	return ret;
1976}
1977
1978static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1979				 struct nlattr *tb[], struct nlattr *data[],
1980				 struct netlink_ext_ack *extack)
1981{
1982	struct ip6_tnl *nt;
1983	struct ip_tunnel_encap ipencap;
1984	int err;
1985
1986	nt = netdev_priv(dev);
1987
1988	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1989		int err = ip6_tnl_encap_setup(nt, &ipencap);
1990
1991		if (err < 0)
1992			return err;
1993	}
1994
1995	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1996		eth_hw_addr_random(dev);
1997
1998	nt->dev = dev;
1999	nt->net = dev_net(dev);
2000
2001	err = register_netdevice(dev);
2002	if (err)
2003		goto out;
2004
2005	if (tb[IFLA_MTU])
2006		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2007
2008out:
2009	return err;
2010}
2011
2012static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2013			  struct nlattr *tb[], struct nlattr *data[],
2014			  struct netlink_ext_ack *extack)
2015{
2016	struct ip6_tnl *nt = netdev_priv(dev);
2017	struct net *net = dev_net(dev);
2018	struct ip6gre_net *ign;
2019	int err;
2020
2021	ip6gre_netlink_parms(data, &nt->parms);
2022	ign = net_generic(net, ip6gre_net_id);
2023
2024	if (nt->parms.collect_md) {
2025		if (rtnl_dereference(ign->collect_md_tun))
2026			return -EEXIST;
2027	} else {
2028		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2029			return -EEXIST;
2030	}
2031
2032	err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2033	if (!err) {
2034		ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2035		ip6gre_tunnel_link_md(ign, nt);
2036		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2037	}
2038	return err;
2039}
2040
2041static struct ip6_tnl *
2042ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2043			 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2044			 struct netlink_ext_ack *extack)
2045{
2046	struct ip6_tnl *t, *nt = netdev_priv(dev);
2047	struct net *net = nt->net;
2048	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2049	struct ip_tunnel_encap ipencap;
2050
2051	if (dev == ign->fb_tunnel_dev)
2052		return ERR_PTR(-EINVAL);
2053
2054	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2055		int err = ip6_tnl_encap_setup(nt, &ipencap);
2056
2057		if (err < 0)
2058			return ERR_PTR(err);
2059	}
2060
2061	ip6gre_netlink_parms(data, p_p);
2062
2063	t = ip6gre_tunnel_locate(net, p_p, 0);
2064
2065	if (t) {
2066		if (t->dev != dev)
2067			return ERR_PTR(-EEXIST);
2068	} else {
2069		t = nt;
2070	}
2071
2072	return t;
2073}
2074
2075static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2076			     struct nlattr *data[],
2077			     struct netlink_ext_ack *extack)
2078{
2079	struct ip6_tnl *t = netdev_priv(dev);
2080	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2081	struct __ip6_tnl_parm p;
2082
2083	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2084	if (IS_ERR(t))
2085		return PTR_ERR(t);
2086
2087	ip6gre_tunnel_unlink_md(ign, t);
2088	ip6gre_tunnel_unlink(ign, t);
2089	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2090	ip6gre_tunnel_link_md(ign, t);
2091	ip6gre_tunnel_link(ign, t);
2092	return 0;
2093}
2094
2095static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2096{
2097	struct net *net = dev_net(dev);
2098	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2099
2100	if (dev != ign->fb_tunnel_dev)
2101		unregister_netdevice_queue(dev, head);
2102}
2103
2104static size_t ip6gre_get_size(const struct net_device *dev)
2105{
2106	return
2107		/* IFLA_GRE_LINK */
2108		nla_total_size(4) +
2109		/* IFLA_GRE_IFLAGS */
2110		nla_total_size(2) +
2111		/* IFLA_GRE_OFLAGS */
2112		nla_total_size(2) +
2113		/* IFLA_GRE_IKEY */
2114		nla_total_size(4) +
2115		/* IFLA_GRE_OKEY */
2116		nla_total_size(4) +
2117		/* IFLA_GRE_LOCAL */
2118		nla_total_size(sizeof(struct in6_addr)) +
2119		/* IFLA_GRE_REMOTE */
2120		nla_total_size(sizeof(struct in6_addr)) +
2121		/* IFLA_GRE_TTL */
2122		nla_total_size(1) +
2123		/* IFLA_GRE_ENCAP_LIMIT */
2124		nla_total_size(1) +
2125		/* IFLA_GRE_FLOWINFO */
2126		nla_total_size(4) +
2127		/* IFLA_GRE_FLAGS */
2128		nla_total_size(4) +
2129		/* IFLA_GRE_ENCAP_TYPE */
2130		nla_total_size(2) +
2131		/* IFLA_GRE_ENCAP_FLAGS */
2132		nla_total_size(2) +
2133		/* IFLA_GRE_ENCAP_SPORT */
2134		nla_total_size(2) +
2135		/* IFLA_GRE_ENCAP_DPORT */
2136		nla_total_size(2) +
2137		/* IFLA_GRE_COLLECT_METADATA */
2138		nla_total_size(0) +
2139		/* IFLA_GRE_FWMARK */
2140		nla_total_size(4) +
2141		/* IFLA_GRE_ERSPAN_INDEX */
2142		nla_total_size(4) +
2143		0;
2144}
2145
2146static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2147{
2148	struct ip6_tnl *t = netdev_priv(dev);
2149	struct __ip6_tnl_parm *p = &t->parms;
2150	__be16 o_flags = p->o_flags;
2151
2152	if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2153		if (!p->collect_md)
2154			o_flags |= TUNNEL_KEY;
2155
2156		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2157			goto nla_put_failure;
2158
2159		if (p->erspan_ver == 1) {
2160			if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2161				goto nla_put_failure;
2162		} else {
2163			if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2164				goto nla_put_failure;
2165			if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2166				goto nla_put_failure;
2167		}
2168	}
2169
2170	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2171	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
2172			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2173	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
2174			 gre_tnl_flags_to_gre_flags(o_flags)) ||
2175	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2176	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2177	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2178	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2179	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2180	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2181	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2182	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2183	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2184		goto nla_put_failure;
2185
2186	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2187			t->encap.type) ||
2188	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2189			 t->encap.sport) ||
2190	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2191			 t->encap.dport) ||
2192	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2193			t->encap.flags))
2194		goto nla_put_failure;
2195
2196	if (p->collect_md) {
2197		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2198			goto nla_put_failure;
2199	}
2200
2201	return 0;
2202
2203nla_put_failure:
2204	return -EMSGSIZE;
2205}
2206
2207static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2208	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
2209	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2210	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2211	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2212	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2213	[IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2214	[IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2215	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
2216	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2217	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2218	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2219	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2220	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2221	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2222	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2223	[IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2224	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2225	[IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2226	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
2227	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
2228	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
2229};
2230
2231static void ip6erspan_tap_setup(struct net_device *dev)
2232{
2233	ether_setup(dev);
2234
2235	dev->max_mtu = 0;
2236	dev->netdev_ops = &ip6erspan_netdev_ops;
2237	dev->needs_free_netdev = true;
2238	dev->priv_destructor = ip6gre_dev_free;
2239
2240	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2241	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2242	netif_keep_dst(dev);
2243}
2244
2245static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2246			     struct nlattr *tb[], struct nlattr *data[],
2247			     struct netlink_ext_ack *extack)
2248{
2249	struct ip6_tnl *nt = netdev_priv(dev);
2250	struct net *net = dev_net(dev);
2251	struct ip6gre_net *ign;
2252	int err;
2253
2254	ip6gre_netlink_parms(data, &nt->parms);
2255	ip6erspan_set_version(data, &nt->parms);
2256	ign = net_generic(net, ip6gre_net_id);
2257
2258	if (nt->parms.collect_md) {
2259		if (rtnl_dereference(ign->collect_md_tun_erspan))
2260			return -EEXIST;
2261	} else {
2262		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2263			return -EEXIST;
2264	}
2265
2266	err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2267	if (!err) {
2268		ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2269		ip6erspan_tunnel_link_md(ign, nt);
2270		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2271	}
2272	return err;
2273}
2274
2275static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2276{
2277	ip6gre_tnl_link_config_common(t);
2278	ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2279}
2280
2281static int ip6erspan_tnl_change(struct ip6_tnl *t,
2282				const struct __ip6_tnl_parm *p, int set_mtu)
2283{
2284	ip6gre_tnl_copy_tnl_parm(t, p);
2285	ip6erspan_tnl_link_config(t, set_mtu);
2286	return 0;
2287}
2288
2289static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2290				struct nlattr *data[],
2291				struct netlink_ext_ack *extack)
2292{
2293	struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2294	struct __ip6_tnl_parm p;
2295	struct ip6_tnl *t;
2296
2297	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2298	if (IS_ERR(t))
2299		return PTR_ERR(t);
2300
2301	ip6erspan_set_version(data, &p);
2302	ip6gre_tunnel_unlink_md(ign, t);
2303	ip6gre_tunnel_unlink(ign, t);
2304	ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2305	ip6erspan_tunnel_link_md(ign, t);
2306	ip6gre_tunnel_link(ign, t);
2307	return 0;
2308}
2309
2310static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2311	.kind		= "ip6gre",
2312	.maxtype	= IFLA_GRE_MAX,
2313	.policy		= ip6gre_policy,
2314	.priv_size	= sizeof(struct ip6_tnl),
2315	.setup		= ip6gre_tunnel_setup,
2316	.validate	= ip6gre_tunnel_validate,
2317	.newlink	= ip6gre_newlink,
2318	.changelink	= ip6gre_changelink,
2319	.dellink	= ip6gre_dellink,
2320	.get_size	= ip6gre_get_size,
2321	.fill_info	= ip6gre_fill_info,
2322	.get_link_net	= ip6_tnl_get_link_net,
2323};
2324
2325static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2326	.kind		= "ip6gretap",
2327	.maxtype	= IFLA_GRE_MAX,
2328	.policy		= ip6gre_policy,
2329	.priv_size	= sizeof(struct ip6_tnl),
2330	.setup		= ip6gre_tap_setup,
2331	.validate	= ip6gre_tap_validate,
2332	.newlink	= ip6gre_newlink,
2333	.changelink	= ip6gre_changelink,
2334	.get_size	= ip6gre_get_size,
2335	.fill_info	= ip6gre_fill_info,
2336	.get_link_net	= ip6_tnl_get_link_net,
2337};
2338
2339static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2340	.kind		= "ip6erspan",
2341	.maxtype	= IFLA_GRE_MAX,
2342	.policy		= ip6gre_policy,
2343	.priv_size	= sizeof(struct ip6_tnl),
2344	.setup		= ip6erspan_tap_setup,
2345	.validate	= ip6erspan_tap_validate,
2346	.newlink	= ip6erspan_newlink,
2347	.changelink	= ip6erspan_changelink,
2348	.get_size	= ip6gre_get_size,
2349	.fill_info	= ip6gre_fill_info,
2350	.get_link_net	= ip6_tnl_get_link_net,
2351};
2352
2353/*
2354 *	And now the modules code and kernel interface.
2355 */
2356
2357static int __init ip6gre_init(void)
2358{
2359	int err;
2360
2361	pr_info("GRE over IPv6 tunneling driver\n");
2362
2363	err = register_pernet_device(&ip6gre_net_ops);
2364	if (err < 0)
2365		return err;
2366
2367	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2368	if (err < 0) {
2369		pr_info("%s: can't add protocol\n", __func__);
2370		goto add_proto_failed;
2371	}
2372
2373	err = rtnl_link_register(&ip6gre_link_ops);
2374	if (err < 0)
2375		goto rtnl_link_failed;
2376
2377	err = rtnl_link_register(&ip6gre_tap_ops);
2378	if (err < 0)
2379		goto tap_ops_failed;
2380
2381	err = rtnl_link_register(&ip6erspan_tap_ops);
2382	if (err < 0)
2383		goto erspan_link_failed;
2384
2385out:
2386	return err;
2387
2388erspan_link_failed:
2389	rtnl_link_unregister(&ip6gre_tap_ops);
2390tap_ops_failed:
2391	rtnl_link_unregister(&ip6gre_link_ops);
2392rtnl_link_failed:
2393	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2394add_proto_failed:
2395	unregister_pernet_device(&ip6gre_net_ops);
2396	goto out;
2397}
2398
2399static void __exit ip6gre_fini(void)
2400{
2401	rtnl_link_unregister(&ip6gre_tap_ops);
2402	rtnl_link_unregister(&ip6gre_link_ops);
2403	rtnl_link_unregister(&ip6erspan_tap_ops);
2404	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2405	unregister_pernet_device(&ip6gre_net_ops);
2406}
2407
2408module_init(ip6gre_init);
2409module_exit(ip6gre_fini);
2410MODULE_LICENSE("GPL");
2411MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2412MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2413MODULE_ALIAS_RTNL_LINK("ip6gre");
2414MODULE_ALIAS_RTNL_LINK("ip6gretap");
2415MODULE_ALIAS_RTNL_LINK("ip6erspan");
2416MODULE_ALIAS_NETDEV("ip6gre0");
2417