projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
skb: allow skb_scrub_packet() to be used by tunnels
[linux-2.6-block.git] / net / ipv6 / ip6_gre.c
CommitLineData
c12b395a 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>
c5441932 41#include <net/ip_tunnels.h>
c12b395a 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
eccc1bb8 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
c12b395a 64#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
65#define IPV6_TCLASS_SHIFT 20
66
67#define HASH_SIZE_SHIFT 5
68#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
69
70static int ip6gre_net_id __read_mostly;
71struct ip6gre_net {
72 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
73
74 struct net_device *fb_tunnel_dev;
75};
76
77static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
78static int ip6gre_tunnel_init(struct net_device *dev);
79static void ip6gre_tunnel_setup(struct net_device *dev);
80static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
81static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
82
83/* Tunnel hash table */
84
85/*
86 4 hash tables:
87
88 3: (remote,local)
89 2: (remote,*)
90 1: (*,local)
91 0: (*,*)
92
93 We require exact key match i.e. if a key is present in packet
94 it will match only tunnel with the same key; if it is not present,
95 it will match only keyless tunnel.
96
97 All keysless packets, if not matched configured keyless tunnels
98 will match fallback tunnel.
99 */
100
101#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
102static u32 HASH_ADDR(const struct in6_addr *addr)
103{
104 u32 hash = ipv6_addr_hash(addr);
105
106 return hash_32(hash, HASH_SIZE_SHIFT);
107}
108
109#define tunnels_r_l tunnels[3]
110#define tunnels_r tunnels[2]
111#define tunnels_l tunnels[1]
112#define tunnels_wc tunnels[0]
c12b395a 113
c12b395a 114/* Given src, dst and key, find appropriate for input tunnel. */
115
116static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
117 const struct in6_addr *remote, const struct in6_addr *local,
118 __be32 key, __be16 gre_proto)
119{
120 struct net *net = dev_net(dev);
121 int link = dev->ifindex;
122 unsigned int h0 = HASH_ADDR(remote);
123 unsigned int h1 = HASH_KEY(key);
124 struct ip6_tnl *t, *cand = NULL;
125 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
126 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
127 ARPHRD_ETHER : ARPHRD_IP6GRE;
128 int score, cand_score = 4;
129
e086cadc 130 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
c12b395a 131 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
132 !ipv6_addr_equal(remote, &t->parms.raddr) ||
133 key != t->parms.i_key ||
134 !(t->dev->flags & IFF_UP))
135 continue;
136
137 if (t->dev->type != ARPHRD_IP6GRE &&
138 t->dev->type != dev_type)
139 continue;
140
141 score = 0;
142 if (t->parms.link != link)
143 score |= 1;
144 if (t->dev->type != dev_type)
145 score |= 2;
146 if (score == 0)
147 return t;
148
149 if (score < cand_score) {
150 cand = t;
151 cand_score = score;
152 }
153 }
154
e086cadc 155 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
c12b395a 156 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
157 key != t->parms.i_key ||
158 !(t->dev->flags & IFF_UP))
159 continue;
160
161 if (t->dev->type != ARPHRD_IP6GRE &&
162 t->dev->type != dev_type)
163 continue;
164
165 score = 0;
166 if (t->parms.link != link)
167 score |= 1;
168 if (t->dev->type != dev_type)
169 score |= 2;
170 if (score == 0)
171 return t;
172
173 if (score < cand_score) {
174 cand = t;
175 cand_score = score;
176 }
177 }
178
e086cadc 179 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
c12b395a 180 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
181 (!ipv6_addr_equal(local, &t->parms.raddr) ||
182 !ipv6_addr_is_multicast(local))) ||
183 key != t->parms.i_key ||
184 !(t->dev->flags & IFF_UP))
185 continue;
186
187 if (t->dev->type != ARPHRD_IP6GRE &&
188 t->dev->type != dev_type)
189 continue;
190
191 score = 0;
192 if (t->parms.link != link)
193 score |= 1;
194 if (t->dev->type != dev_type)
195 score |= 2;
196 if (score == 0)
197 return t;
198
199 if (score < cand_score) {
200 cand = t;
201 cand_score = score;
202 }
203 }
204
e086cadc 205 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
c12b395a 206 if (t->parms.i_key != key ||
207 !(t->dev->flags & IFF_UP))
208 continue;
209
210 if (t->dev->type != ARPHRD_IP6GRE &&
211 t->dev->type != dev_type)
212 continue;
213
214 score = 0;
215 if (t->parms.link != link)
216 score |= 1;
217 if (t->dev->type != dev_type)
218 score |= 2;
219 if (score == 0)
220 return t;
221
222 if (score < cand_score) {
223 cand = t;
224 cand_score = score;
225 }
226 }
227
228 if (cand != NULL)
229 return cand;
230
231 dev = ign->fb_tunnel_dev;
232 if (dev->flags & IFF_UP)
233 return netdev_priv(dev);
234
235 return NULL;
236}
237
238static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
239 const struct __ip6_tnl_parm *p)
240{
241 const struct in6_addr *remote = &p->raddr;
242 const struct in6_addr *local = &p->laddr;
243 unsigned int h = HASH_KEY(p->i_key);
244 int prio = 0;
245
246 if (!ipv6_addr_any(local))
247 prio |= 1;
248 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
249 prio |= 2;
250 h ^= HASH_ADDR(remote);
251 }
252
253 return &ign->tunnels[prio][h];
254}
255
256static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
257 const struct ip6_tnl *t)
258{
259 return __ip6gre_bucket(ign, &t->parms);
260}
261
262static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
263{
264 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
265
266 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
267 rcu_assign_pointer(*tp, t);
268}
269
270static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
271{
272 struct ip6_tnl __rcu **tp;
273 struct ip6_tnl *iter;
274
275 for (tp = ip6gre_bucket(ign, t);
276 (iter = rtnl_dereference(*tp)) != NULL;
277 tp = &iter->next) {
278 if (t == iter) {
279 rcu_assign_pointer(*tp, t->next);
280 break;
281 }
282 }
283}
284
285static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
286 const struct __ip6_tnl_parm *parms,
287 int type)
288{
289 const struct in6_addr *remote = &parms->raddr;
290 const struct in6_addr *local = &parms->laddr;
291 __be32 key = parms->i_key;
292 int link = parms->link;
293 struct ip6_tnl *t;
294 struct ip6_tnl __rcu **tp;
295 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
296
297 for (tp = __ip6gre_bucket(ign, parms);
298 (t = rtnl_dereference(*tp)) != NULL;
299 tp = &t->next)
300 if (ipv6_addr_equal(local, &t->parms.laddr) &&
301 ipv6_addr_equal(remote, &t->parms.raddr) &&
302 key == t->parms.i_key &&
303 link == t->parms.link &&
304 type == t->dev->type)
305 break;
306
307 return t;
308}
309
310static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
311 const struct __ip6_tnl_parm *parms, int create)
312{
313 struct ip6_tnl *t, *nt;
314 struct net_device *dev;
315 char name[IFNAMSIZ];
316 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
317
318 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
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, ip6gre_tunnel_setup);
328 if (!dev)
329 return NULL;
330
331 dev_net_set(dev, net);
332
333 nt = netdev_priv(dev);
334 nt->parms = *parms;
335 dev->rtnl_link_ops = &ip6gre_link_ops;
336
337 nt->dev = dev;
0bd87628 338 nt->net = dev_net(dev);
c12b395a 339 ip6gre_tnl_link_config(nt, 1);
340
341 if (register_netdevice(dev) < 0)
342 goto failed_free;
343
344 /* Can use a lockless transmit, unless we generate output sequences */
345 if (!(nt->parms.o_flags & GRE_SEQ))
346 dev->features |= NETIF_F_LLTX;
347
348 dev_hold(dev);
349 ip6gre_tunnel_link(ign, nt);
350 return nt;
351
352failed_free:
353 free_netdev(dev);
354 return NULL;
355}
356
357static void ip6gre_tunnel_uninit(struct net_device *dev)
358{
359 struct net *net = dev_net(dev);
360 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
361
362 ip6gre_tunnel_unlink(ign, netdev_priv(dev));
363 dev_put(dev);
364}
365
366
367static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
368 u8 type, u8 code, int offset, __be32 info)
369{
370 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
b87fb39e
ED		 371 __be16 *p = (__be16 *)(skb->data + offset);
372 int grehlen = offset + 4;
c12b395a 373 struct ip6_tnl *t;
374 __be16 flags;
375
376 flags = p[0];
377 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
378 if (flags&(GRE_VERSION|GRE_ROUTING))
379 return;
380 if (flags&GRE_KEY) {
381 grehlen += 4;
382 if (flags&GRE_CSUM)
383 grehlen += 4;
384 }
385 }
386
387 /* If only 8 bytes returned, keyed message will be dropped here */
b87fb39e 388 if (!pskb_may_pull(skb, grehlen))
c12b395a 389 return;
b87fb39e
ED		 390 ipv6h = (const struct ipv6hdr *)skb->data;
391 p = (__be16 *)(skb->data + offset);
c12b395a 392
c12b395a 393 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
394 flags & GRE_KEY ?
395 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
396 p[1]);
397 if (t == NULL)
0c5794a6 398 return;
c12b395a 399
400 switch (type) {
401 __u32 teli;
402 struct ipv6_tlv_tnl_enc_lim *tel;
403 __u32 mtu;
404 case ICMPV6_DEST_UNREACH:
405 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
406 t->parms.name);
407 break;
408 case ICMPV6_TIME_EXCEED:
409 if (code == ICMPV6_EXC_HOPLIMIT) {
410 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
411 t->parms.name);
412 }
413 break;
414 case ICMPV6_PARAMPROB:
415 teli = 0;
416 if (code == ICMPV6_HDR_FIELD)
417 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
418
419 if (teli && teli == info - 2) {
420 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
421 if (tel->encap_limit == 0) {
422 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
423 t->parms.name);
424 }
425 } else {
426 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
427 t->parms.name);
428 }
429 break;
430 case ICMPV6_PKT_TOOBIG:
431 mtu = info - offset;
432 if (mtu < IPV6_MIN_MTU)
433 mtu = IPV6_MIN_MTU;
434 t->dev->mtu = mtu;
435 break;
436 }
437
438 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
439 t->err_count++;
440 else
441 t->err_count = 1;
442 t->err_time = jiffies;
c12b395a 443}
444
c12b395a 445static int ip6gre_rcv(struct sk_buff *skb)
446{
447 const struct ipv6hdr *ipv6h;
448 u8 *h;
449 __be16 flags;
450 __sum16 csum = 0;
451 __be32 key = 0;
452 u32 seqno = 0;
453 struct ip6_tnl *tunnel;
454 int offset = 4;
455 __be16 gre_proto;
eccc1bb8 456 int err;
c12b395a 457
458 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
0c5794a6 459 goto drop;
c12b395a 460
461 ipv6h = ipv6_hdr(skb);
462 h = skb->data;
463 flags = *(__be16 *)h;
464
465 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
466 /* - Version must be 0.
467 - We do not support routing headers.
468 */
469 if (flags&(GRE_VERSION|GRE_ROUTING))
0c5794a6 470 goto drop;
c12b395a 471
472 if (flags&GRE_CSUM) {
473 switch (skb->ip_summed) {
474 case CHECKSUM_COMPLETE:
475 csum = csum_fold(skb->csum);
476 if (!csum)
477 break;
478 /* fall through */
479 case CHECKSUM_NONE:
480 skb->csum = 0;
481 csum = __skb_checksum_complete(skb);
482 skb->ip_summed = CHECKSUM_COMPLETE;
483 }
484 offset += 4;
485 }
486 if (flags&GRE_KEY) {
487 key = *(__be32 *)(h + offset);
488 offset += 4;
489 }
490 if (flags&GRE_SEQ) {
491 seqno = ntohl(*(__be32 *)(h + offset));
492 offset += 4;
493 }
494 }
495
496 gre_proto = *(__be16 *)(h + 2);
497
c12b395a 498 tunnel = ip6gre_tunnel_lookup(skb->dev,
499 &ipv6h->saddr, &ipv6h->daddr, key,
500 gre_proto);
501 if (tunnel) {
502 struct pcpu_tstats *tstats;
503
504 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
505 goto drop;
506
507 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
508 tunnel->dev->stats.rx_dropped++;
509 goto drop;
510 }
511
512 secpath_reset(skb);
513
514 skb->protocol = gre_proto;
515 /* WCCP version 1 and 2 protocol decoding.
516 * - Change protocol to IP
517 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
518 */
519 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
520 skb->protocol = htons(ETH_P_IP);
521 if ((*(h + offset) & 0xF0) != 0x40)
522 offset += 4;
523 }
524
525 skb->mac_header = skb->network_header;
526 __pskb_pull(skb, offset);
527 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
528 skb->pkt_type = PACKET_HOST;
529
530 if (((flags&GRE_CSUM) && csum) ||
531 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
532 tunnel->dev->stats.rx_crc_errors++;
533 tunnel->dev->stats.rx_errors++;
534 goto drop;
535 }
536 if (tunnel->parms.i_flags&GRE_SEQ) {
537 if (!(flags&GRE_SEQ) ||
538 (tunnel->i_seqno &&
539 (s32)(seqno - tunnel->i_seqno) < 0)) {
540 tunnel->dev->stats.rx_fifo_errors++;
541 tunnel->dev->stats.rx_errors++;
542 goto drop;
543 }
544 tunnel->i_seqno = seqno + 1;
545 }
546
547 /* Warning: All skb pointers will be invalidated! */
548 if (tunnel->dev->type == ARPHRD_ETHER) {
549 if (!pskb_may_pull(skb, ETH_HLEN)) {
550 tunnel->dev->stats.rx_length_errors++;
551 tunnel->dev->stats.rx_errors++;
552 goto drop;
553 }
554
555 ipv6h = ipv6_hdr(skb);
556 skb->protocol = eth_type_trans(skb, tunnel->dev);
557 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
558 }
559
eccc1bb8 560 __skb_tunnel_rx(skb, tunnel->dev);
561
562 skb_reset_network_header(skb);
563
564 err = IP6_ECN_decapsulate(ipv6h, skb);
565 if (unlikely(err)) {
566 if (log_ecn_error)
567 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
568 &ipv6h->saddr,
569 ipv6_get_dsfield(ipv6h));
570 if (err > 1) {
571 ++tunnel->dev->stats.rx_frame_errors;
572 ++tunnel->dev->stats.rx_errors;
573 goto drop;
574 }
575 }
576
c12b395a 577 tstats = this_cpu_ptr(tunnel->dev->tstats);
578 u64_stats_update_begin(&tstats->syncp);
579 tstats->rx_packets++;
580 tstats->rx_bytes += skb->len;
581 u64_stats_update_end(&tstats->syncp);
582
c12b395a 583 netif_rx(skb);
584
c12b395a 585 return 0;
586 }
587 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
588
589drop:
c12b395a 590 kfree_skb(skb);
591 return 0;
592}
593
594struct ipv6_tel_txoption {
595 struct ipv6_txoptions ops;
596 __u8 dst_opt[8];
597};
598
599static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
600{
601 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
602
603 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
604 opt->dst_opt[3] = 1;
605 opt->dst_opt[4] = encap_limit;
606 opt->dst_opt[5] = IPV6_TLV_PADN;
607 opt->dst_opt[6] = 1;
608
609 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
610 opt->ops.opt_nflen = 8;
611}
612
613static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
614 struct net_device *dev,
615 __u8 dsfield,
616 struct flowi6 *fl6,
617 int encap_limit,
618 __u32 *pmtu)
619{
620 struct net *net = dev_net(dev);
621 struct ip6_tnl *tunnel = netdev_priv(dev);
622 struct net_device *tdev; /* Device to other host */
623 struct ipv6hdr *ipv6h; /* Our new IP header */
624 unsigned int max_headroom; /* The extra header space needed */
625 int gre_hlen;
626 struct ipv6_tel_txoption opt;
627 int mtu;
628 struct dst_entry *dst = NULL, *ndst = NULL;
629 struct net_device_stats *stats = &tunnel->dev->stats;
630 int err = -1;
631 u8 proto;
c12b395a 632 struct sk_buff *new_skb;
633
634 if (dev->type == ARPHRD_ETHER)
635 IPCB(skb)->flags = 0;
636
637 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
638 gre_hlen = 0;
639 ipv6h = (struct ipv6hdr *)skb->data;
640 fl6->daddr = ipv6h->daddr;
641 } else {
642 gre_hlen = tunnel->hlen;
643 fl6->daddr = tunnel->parms.raddr;
644 }
645
646 if (!fl6->flowi6_mark)
647 dst = ip6_tnl_dst_check(tunnel);
648
649 if (!dst) {
650 ndst = ip6_route_output(net, NULL, fl6);
651
652 if (ndst->error)
653 goto tx_err_link_failure;
654 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
655 if (IS_ERR(ndst)) {
656 err = PTR_ERR(ndst);
657 ndst = NULL;
658 goto tx_err_link_failure;
659 }
660 dst = ndst;
661 }
662
663 tdev = dst->dev;
664
665 if (tdev == dev) {
666 stats->collisions++;
667 net_warn_ratelimited("%s: Local routing loop detected!\n",
668 tunnel->parms.name);
669 goto tx_err_dst_release;
670 }
671
672 mtu = dst_mtu(dst) - sizeof(*ipv6h);
673 if (encap_limit >= 0) {
674 max_headroom += 8;
675 mtu -= 8;
676 }
677 if (mtu < IPV6_MIN_MTU)
678 mtu = IPV6_MIN_MTU;
679 if (skb_dst(skb))
680 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
681 if (skb->len > mtu) {
682 *pmtu = mtu;
683 err = -EMSGSIZE;
684 goto tx_err_dst_release;
685 }
686
687 if (tunnel->err_count > 0) {
688 if (time_before(jiffies,
689 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
690 tunnel->err_count--;
691
692 dst_link_failure(skb);
693 } else
694 tunnel->err_count = 0;
695 }
696
697 max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
698
699 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
700 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
701 new_skb = skb_realloc_headroom(skb, max_headroom);
702 if (max_headroom > dev->needed_headroom)
703 dev->needed_headroom = max_headroom;
704 if (!new_skb)
705 goto tx_err_dst_release;
706
707 if (skb->sk)
708 skb_set_owner_w(new_skb, skb->sk);
709 consume_skb(skb);
710 skb = new_skb;
711 }
712
713 skb_dst_drop(skb);
714
715 if (fl6->flowi6_mark) {
716 skb_dst_set(skb, dst);
717 ndst = NULL;
718 } else {
719 skb_dst_set_noref(skb, dst);
720 }
721
c12b395a 722 proto = NEXTHDR_GRE;
723 if (encap_limit >= 0) {
724 init_tel_txopt(&opt, encap_limit);
725 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
726 }
727
728 skb_push(skb, gre_hlen);
729 skb_reset_network_header(skb);
ae782bb1 730 skb_set_transport_header(skb, sizeof(*ipv6h));
c12b395a 731
732 /*
733 * Push down and install the IP header.
734 */
735 ipv6h = ipv6_hdr(skb);
3e4e4c1f 736 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
c12b395a 737 ipv6h->hop_limit = tunnel->parms.hop_limit;
738 ipv6h->nexthdr = proto;
739 ipv6h->saddr = fl6->saddr;
740 ipv6h->daddr = fl6->daddr;
741
742 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
743 ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
744 htons(ETH_P_TEB) : skb->protocol;
745
746 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
747 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
748
749 if (tunnel->parms.o_flags&GRE_SEQ) {
750 ++tunnel->o_seqno;
751 *ptr = htonl(tunnel->o_seqno);
752 ptr--;
753 }
754 if (tunnel->parms.o_flags&GRE_KEY) {
755 *ptr = tunnel->parms.o_key;
756 ptr--;
757 }
758 if (tunnel->parms.o_flags&GRE_CSUM) {
759 *ptr = 0;
760 *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
761 skb->len - sizeof(struct ipv6hdr));
762 }
763 }
764
e8f72ea4 765 ip6tunnel_xmit(skb, dev);
c12b395a 766 if (ndst)
767 ip6_tnl_dst_store(tunnel, ndst);
c12b395a 768 return 0;
769tx_err_link_failure:
770 stats->tx_carrier_errors++;
771 dst_link_failure(skb);
772tx_err_dst_release:
773 dst_release(ndst);
774 return err;
775}
776
777static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
778{
779 struct ip6_tnl *t = netdev_priv(dev);
780 const struct iphdr *iph = ip_hdr(skb);
781 int encap_limit = -1;
782 struct flowi6 fl6;
783 __u8 dsfield;
784 __u32 mtu;
785 int err;
786
787 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
788 encap_limit = t->parms.encap_limit;
789
790 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
791 fl6.flowi6_proto = IPPROTO_IPIP;
792
793 dsfield = ipv4_get_dsfield(iph);
794
795 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
796 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
797 & IPV6_TCLASS_MASK;
798 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
799 fl6.flowi6_mark = skb->mark;
800
801 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
802 if (err != 0) {
803 /* XXX: send ICMP error even if DF is not set. */
804 if (err == -EMSGSIZE)
805 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
806 htonl(mtu));
807 return -1;
808 }
809
810 return 0;
811}
812
813static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
814{
815 struct ip6_tnl *t = netdev_priv(dev);
816 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
817 int encap_limit = -1;
818 __u16 offset;
819 struct flowi6 fl6;
820 __u8 dsfield;
821 __u32 mtu;
822 int err;
823
824 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
825 return -1;
826
827 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
828 if (offset > 0) {
829 struct ipv6_tlv_tnl_enc_lim *tel;
830 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
831 if (tel->encap_limit == 0) {
832 icmpv6_send(skb, ICMPV6_PARAMPROB,
833 ICMPV6_HDR_FIELD, offset + 2);
834 return -1;
835 }
836 encap_limit = tel->encap_limit - 1;
837 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
838 encap_limit = t->parms.encap_limit;
839
840 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
841 fl6.flowi6_proto = IPPROTO_IPV6;
842
843 dsfield = ipv6_get_dsfield(ipv6h);
844 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
845 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
846 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
847 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
848 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
849 fl6.flowi6_mark = skb->mark;
850
851 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
852 if (err != 0) {
853 if (err == -EMSGSIZE)
854 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
855 return -1;
856 }
857
858 return 0;
859}
860
861/**
862 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
863 * @t: the outgoing tunnel device
864 * @hdr: IPv6 header from the incoming packet
865 *
866 * Description:
867 * Avoid trivial tunneling loop by checking that tunnel exit-point
868 * doesn't match source of incoming packet.
869 *
870 * Return:
871 * 1 if conflict,
872 * 0 else
873 **/
874
875static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
876 const struct ipv6hdr *hdr)
877{
878 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
879}
880
881static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
882{
883 struct ip6_tnl *t = netdev_priv(dev);
884 int encap_limit = -1;
885 struct flowi6 fl6;
886 __u32 mtu;
887 int err;
888
889 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
890 encap_limit = t->parms.encap_limit;
891
892 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
893 fl6.flowi6_proto = skb->protocol;
894
895 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
896
897 return err;
898}
899
900static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
901 struct net_device *dev)
902{
903 struct ip6_tnl *t = netdev_priv(dev);
904 struct net_device_stats *stats = &t->dev->stats;
905 int ret;
906
907 if (!ip6_tnl_xmit_ctl(t))
41ab3e31 908 goto tx_err;
c12b395a 909
910 switch (skb->protocol) {
911 case htons(ETH_P_IP):
912 ret = ip6gre_xmit_ipv4(skb, dev);
913 break;
914 case htons(ETH_P_IPV6):
915 ret = ip6gre_xmit_ipv6(skb, dev);
916 break;
917 default:
918 ret = ip6gre_xmit_other(skb, dev);
919 break;
920 }
921
922 if (ret < 0)
923 goto tx_err;
924
925 return NETDEV_TX_OK;
926
927tx_err:
928 stats->tx_errors++;
929 stats->tx_dropped++;
930 kfree_skb(skb);
931 return NETDEV_TX_OK;
932}
933
934static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
935{
936 struct net_device *dev = t->dev;
937 struct __ip6_tnl_parm *p = &t->parms;
938 struct flowi6 *fl6 = &t->fl.u.ip6;
939 int addend = sizeof(struct ipv6hdr) + 4;
940
941 if (dev->type != ARPHRD_ETHER) {
942 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
943 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
944 }
945
946 /* Set up flowi template */
947 fl6->saddr = p->laddr;
948 fl6->daddr = p->raddr;
949 fl6->flowi6_oif = p->link;
950 fl6->flowlabel = 0;
951
952 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
953 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
954 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
955 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
956
957 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
958 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
959
960 if (p->flags&IP6_TNL_F_CAP_XMIT &&
961 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
962 dev->flags |= IFF_POINTOPOINT;
963 else
964 dev->flags &= ~IFF_POINTOPOINT;
965
966 dev->iflink = p->link;
967
968 /* Precalculate GRE options length */
969 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
970 if (t->parms.o_flags&GRE_CSUM)
971 addend += 4;
972 if (t->parms.o_flags&GRE_KEY)
973 addend += 4;
974 if (t->parms.o_flags&GRE_SEQ)
975 addend += 4;
976 }
977
978 if (p->flags & IP6_TNL_F_CAP_XMIT) {
979 int strict = (ipv6_addr_type(&p->raddr) &
980 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
981
982 struct rt6_info *rt = rt6_lookup(dev_net(dev),
983 &p->raddr, &p->laddr,
984 p->link, strict);
985
986 if (rt == NULL)
987 return;
988
989 if (rt->dst.dev) {
990 dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
991
992 if (set_mtu) {
993 dev->mtu = rt->dst.dev->mtu - addend;
994 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
995 dev->mtu -= 8;
996
997 if (dev->mtu < IPV6_MIN_MTU)
998 dev->mtu = IPV6_MIN_MTU;
999 }
1000 }
94e187c0 1001 ip6_rt_put(rt);
c12b395a 1002 }
1003
1004 t->hlen = addend;
1005}
1006
1007static int ip6gre_tnl_change(struct ip6_tnl *t,
1008 const struct __ip6_tnl_parm *p, int set_mtu)
1009{
1010 t->parms.laddr = p->laddr;
1011 t->parms.raddr = p->raddr;
1012 t->parms.flags = p->flags;
1013 t->parms.hop_limit = p->hop_limit;
1014 t->parms.encap_limit = p->encap_limit;
1015 t->parms.flowinfo = p->flowinfo;
1016 t->parms.link = p->link;
1017 t->parms.proto = p->proto;
1018 t->parms.i_key = p->i_key;
1019 t->parms.o_key = p->o_key;
1020 t->parms.i_flags = p->i_flags;
1021 t->parms.o_flags = p->o_flags;
1022 ip6_tnl_dst_reset(t);
1023 ip6gre_tnl_link_config(t, set_mtu);
1024 return 0;
1025}
1026
1027static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1028 const struct ip6_tnl_parm2 *u)
1029{
1030 p->laddr = u->laddr;
1031 p->raddr = u->raddr;
1032 p->flags = u->flags;
1033 p->hop_limit = u->hop_limit;
1034 p->encap_limit = u->encap_limit;
1035 p->flowinfo = u->flowinfo;
1036 p->link = u->link;
1037 p->i_key = u->i_key;
1038 p->o_key = u->o_key;
1039 p->i_flags = u->i_flags;
1040 p->o_flags = u->o_flags;
1041 memcpy(p->name, u->name, sizeof(u->name));
1042}
1043
1044static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1045 const struct __ip6_tnl_parm *p)
1046{
1047 u->proto = IPPROTO_GRE;
1048 u->laddr = p->laddr;
1049 u->raddr = p->raddr;
1050 u->flags = p->flags;
1051 u->hop_limit = p->hop_limit;
1052 u->encap_limit = p->encap_limit;
1053 u->flowinfo = p->flowinfo;
1054 u->link = p->link;
1055 u->i_key = p->i_key;
1056 u->o_key = p->o_key;
1057 u->i_flags = p->i_flags;
1058 u->o_flags = p->o_flags;
1059 memcpy(u->name, p->name, sizeof(u->name));
1060}
1061
1062static int ip6gre_tunnel_ioctl(struct net_device *dev,
1063 struct ifreq *ifr, int cmd)
1064{
1065 int err = 0;
1066 struct ip6_tnl_parm2 p;
1067 struct __ip6_tnl_parm p1;
1068 struct ip6_tnl *t;
1069 struct net *net = dev_net(dev);
1070 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1071
1072 switch (cmd) {
1073 case SIOCGETTUNNEL:
1074 t = NULL;
1075 if (dev == ign->fb_tunnel_dev) {
1076 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1077 err = -EFAULT;
1078 break;
1079 }
1080 ip6gre_tnl_parm_from_user(&p1, &p);
1081 t = ip6gre_tunnel_locate(net, &p1, 0);
1082 }
1083 if (t == NULL)
1084 t = netdev_priv(dev);
5dbd5068 1085 memset(&p, 0, sizeof(p));
c12b395a 1086 ip6gre_tnl_parm_to_user(&p, &t->parms);
1087 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1088 err = -EFAULT;
1089 break;
1090
1091 case SIOCADDTUNNEL:
1092 case SIOCCHGTUNNEL:
1093 err = -EPERM;
af31f412 1094 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
c12b395a 1095 goto done;
1096
1097 err = -EFAULT;
1098 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1099 goto done;
1100
1101 err = -EINVAL;
1102 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1103 goto done;
1104
1105 if (!(p.i_flags&GRE_KEY))
1106 p.i_key = 0;
1107 if (!(p.o_flags&GRE_KEY))
1108 p.o_key = 0;
1109
1110 ip6gre_tnl_parm_from_user(&p1, &p);
1111 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1112
1113 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1114 if (t != NULL) {
1115 if (t->dev != dev) {
1116 err = -EEXIST;
1117 break;
1118 }
1119 } else {
1120 t = netdev_priv(dev);
1121
1122 ip6gre_tunnel_unlink(ign, t);
1123 synchronize_net();
1124 ip6gre_tnl_change(t, &p1, 1);
1125 ip6gre_tunnel_link(ign, t);
1126 netdev_state_change(dev);
1127 }
1128 }
1129
1130 if (t) {
1131 err = 0;
1132
5dbd5068 1133 memset(&p, 0, sizeof(p));
c12b395a 1134 ip6gre_tnl_parm_to_user(&p, &t->parms);
1135 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1136 err = -EFAULT;
1137 } else
1138 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1139 break;
1140
1141 case SIOCDELTUNNEL:
1142 err = -EPERM;
af31f412 1143 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
c12b395a 1144 goto done;
1145
1146 if (dev == ign->fb_tunnel_dev) {
1147 err = -EFAULT;
1148 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1149 goto done;
1150 err = -ENOENT;
1151 ip6gre_tnl_parm_from_user(&p1, &p);
1152 t = ip6gre_tunnel_locate(net, &p1, 0);
1153 if (t == NULL)
1154 goto done;
1155 err = -EPERM;
1156 if (t == netdev_priv(ign->fb_tunnel_dev))
1157 goto done;
1158 dev = t->dev;
1159 }
1160 unregister_netdevice(dev);
1161 err = 0;
1162 break;
1163
1164 default:
1165 err = -EINVAL;
1166 }
1167
1168done:
1169 return err;
1170}
1171
1172static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1173{
1174 struct ip6_tnl *tunnel = netdev_priv(dev);
1175 if (new_mtu < 68 ||
1176 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1177 return -EINVAL;
1178 dev->mtu = new_mtu;
1179 return 0;
1180}
1181
1182static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1183 unsigned short type,
1184 const void *daddr, const void *saddr, unsigned int len)
1185{
1186 struct ip6_tnl *t = netdev_priv(dev);
1187 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1188 __be16 *p = (__be16 *)(ipv6h+1);
1189
3e4e4c1f 1190 ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
c12b395a 1191 ipv6h->hop_limit = t->parms.hop_limit;
1192 ipv6h->nexthdr = NEXTHDR_GRE;
1193 ipv6h->saddr = t->parms.laddr;
1194 ipv6h->daddr = t->parms.raddr;
1195
1196 p[0] = t->parms.o_flags;
1197 p[1] = htons(type);
1198
1199 /*
1200 * Set the source hardware address.
1201 */
1202
1203 if (saddr)
1204 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1205 if (daddr)
1206 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1207 if (!ipv6_addr_any(&ipv6h->daddr))
1208 return t->hlen;
1209
1210 return -t->hlen;
1211}
1212
c12b395a 1213static const struct header_ops ip6gre_header_ops = {
1214 .create = ip6gre_header,
c12b395a 1215};
1216
1217static const struct net_device_ops ip6gre_netdev_ops = {
1218 .ndo_init = ip6gre_tunnel_init,
1219 .ndo_uninit = ip6gre_tunnel_uninit,
1220 .ndo_start_xmit = ip6gre_tunnel_xmit,
1221 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1222 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
f61dd388 1223 .ndo_get_stats64 = ip_tunnel_get_stats64,
c12b395a 1224};
1225
1226static void ip6gre_dev_free(struct net_device *dev)
1227{
1228 free_percpu(dev->tstats);
1229 free_netdev(dev);
1230}
1231
1232static void ip6gre_tunnel_setup(struct net_device *dev)
1233{
1234 struct ip6_tnl *t;
1235
1236 dev->netdev_ops = &ip6gre_netdev_ops;
1237 dev->destructor = ip6gre_dev_free;
1238
1239 dev->type = ARPHRD_IP6GRE;
1240 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1241 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1242 t = netdev_priv(dev);
1243 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1244 dev->mtu -= 8;
1245 dev->flags |= IFF_NOARP;
1246 dev->iflink = 0;
1247 dev->addr_len = sizeof(struct in6_addr);
1248 dev->features |= NETIF_F_NETNS_LOCAL;
1249 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1250}
1251
1252static int ip6gre_tunnel_init(struct net_device *dev)
1253{
1254 struct ip6_tnl *tunnel;
1255
1256 tunnel = netdev_priv(dev);
1257
1258 tunnel->dev = dev;
0bd87628 1259 tunnel->net = dev_net(dev);
c12b395a 1260 strcpy(tunnel->parms.name, dev->name);
1261
1262 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1263 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1264
1265 if (ipv6_addr_any(&tunnel->parms.raddr))
1266 dev->header_ops = &ip6gre_header_ops;
1267
1268 dev->tstats = alloc_percpu(struct pcpu_tstats);
1269 if (!dev->tstats)
1270 return -ENOMEM;
1271
1272 return 0;
1273}
1274
1275static void ip6gre_fb_tunnel_init(struct net_device *dev)
1276{
1277 struct ip6_tnl *tunnel = netdev_priv(dev);
1278
1279 tunnel->dev = dev;
0bd87628 1280 tunnel->net = dev_net(dev);
c12b395a 1281 strcpy(tunnel->parms.name, dev->name);
1282
1283 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1284
1285 dev_hold(dev);
1286}
1287
1288
1289static struct inet6_protocol ip6gre_protocol __read_mostly = {
1290 .handler = ip6gre_rcv,
1291 .err_handler = ip6gre_err,
1292 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1293};
1294
1295static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1296 struct list_head *head)
1297{
1298 int prio;
1299
1300 for (prio = 0; prio < 4; prio++) {
1301 int h;
1302 for (h = 0; h < HASH_SIZE; h++) {
1303 struct ip6_tnl *t;
1304
1305 t = rtnl_dereference(ign->tunnels[prio][h]);
1306
1307 while (t != NULL) {
1308 unregister_netdevice_queue(t->dev, head);
1309 t = rtnl_dereference(t->next);
1310 }
1311 }
1312 }
1313}
1314
1315static int __net_init ip6gre_init_net(struct net *net)
1316{
1317 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1318 int err;
1319
1320 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1321 ip6gre_tunnel_setup);
1322 if (!ign->fb_tunnel_dev) {
1323 err = -ENOMEM;
1324 goto err_alloc_dev;
1325 }
1326 dev_net_set(ign->fb_tunnel_dev, net);
1327
1328 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1329 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1330
1331 err = register_netdev(ign->fb_tunnel_dev);
1332 if (err)
1333 goto err_reg_dev;
1334
1335 rcu_assign_pointer(ign->tunnels_wc[0],
1336 netdev_priv(ign->fb_tunnel_dev));
1337 return 0;
1338
1339err_reg_dev:
1340 ip6gre_dev_free(ign->fb_tunnel_dev);
1341err_alloc_dev:
1342 return err;
1343}
1344
1345static void __net_exit ip6gre_exit_net(struct net *net)
1346{
1347 struct ip6gre_net *ign;
1348 LIST_HEAD(list);
1349
1350 ign = net_generic(net, ip6gre_net_id);
1351 rtnl_lock();
1352 ip6gre_destroy_tunnels(ign, &list);
1353 unregister_netdevice_many(&list);
1354 rtnl_unlock();
1355}
1356
1357static struct pernet_operations ip6gre_net_ops = {
1358 .init = ip6gre_init_net,
1359 .exit = ip6gre_exit_net,
1360 .id = &ip6gre_net_id,
1361 .size = sizeof(struct ip6gre_net),
1362};
1363
1364static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1365{
1366 __be16 flags;
1367
1368 if (!data)
1369 return 0;
1370
1371 flags = 0;
1372 if (data[IFLA_GRE_IFLAGS])
1373 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1374 if (data[IFLA_GRE_OFLAGS])
1375 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1376 if (flags & (GRE_VERSION|GRE_ROUTING))
1377 return -EINVAL;
1378
1379 return 0;
1380}
1381
1382static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1383{
1384 struct in6_addr daddr;
1385
1386 if (tb[IFLA_ADDRESS]) {
1387 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1388 return -EINVAL;
1389 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1390 return -EADDRNOTAVAIL;
1391 }
1392
1393 if (!data)
1394 goto out;
1395
1396 if (data[IFLA_GRE_REMOTE]) {
1397 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1398 if (ipv6_addr_any(&daddr))
1399 return -EINVAL;
1400 }
1401
1402out:
1403 return ip6gre_tunnel_validate(tb, data);
1404}
1405
1406
1407static void ip6gre_netlink_parms(struct nlattr *data[],
1408 struct __ip6_tnl_parm *parms)
1409{
1410 memset(parms, 0, sizeof(*parms));
1411
1412 if (!data)
1413 return;
1414
1415 if (data[IFLA_GRE_LINK])
1416 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1417
1418 if (data[IFLA_GRE_IFLAGS])
1419 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1420
1421 if (data[IFLA_GRE_OFLAGS])
1422 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1423
1424 if (data[IFLA_GRE_IKEY])
1425 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1426
1427 if (data[IFLA_GRE_OKEY])
1428 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1429
1430 if (data[IFLA_GRE_LOCAL])
1431 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1432
1433 if (data[IFLA_GRE_REMOTE])
1434 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1435
1436 if (data[IFLA_GRE_TTL])
1437 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1438
1439 if (data[IFLA_GRE_ENCAP_LIMIT])
1440 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1441
1442 if (data[IFLA_GRE_FLOWINFO])
1443 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1444
1445 if (data[IFLA_GRE_FLAGS])
1446 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1447}
1448
1449static int ip6gre_tap_init(struct net_device *dev)
1450{
1451 struct ip6_tnl *tunnel;
1452
1453 tunnel = netdev_priv(dev);
1454
1455 tunnel->dev = dev;
0bd87628 1456 tunnel->net = dev_net(dev);
c12b395a 1457 strcpy(tunnel->parms.name, dev->name);
1458
1459 ip6gre_tnl_link_config(tunnel, 1);
1460
1461 dev->tstats = alloc_percpu(struct pcpu_tstats);
1462 if (!dev->tstats)
1463 return -ENOMEM;
1464
1465 return 0;
1466}
1467
1468static const struct net_device_ops ip6gre_tap_netdev_ops = {
1469 .ndo_init = ip6gre_tap_init,
1470 .ndo_uninit = ip6gre_tunnel_uninit,
1471 .ndo_start_xmit = ip6gre_tunnel_xmit,
1472 .ndo_set_mac_address = eth_mac_addr,
1473 .ndo_validate_addr = eth_validate_addr,
1474 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
f61dd388 1475 .ndo_get_stats64 = ip_tunnel_get_stats64,
c12b395a 1476};
1477
1478static void ip6gre_tap_setup(struct net_device *dev)
1479{
1480
1481 ether_setup(dev);
1482
1483 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1484 dev->destructor = ip6gre_dev_free;
1485
1486 dev->iflink = 0;
1487 dev->features |= NETIF_F_NETNS_LOCAL;
1488}
1489
1490static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1491 struct nlattr *tb[], struct nlattr *data[])
1492{
1493 struct ip6_tnl *nt;
1494 struct net *net = dev_net(dev);
1495 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1496 int err;
1497
1498 nt = netdev_priv(dev);
1499 ip6gre_netlink_parms(data, &nt->parms);
1500
1501 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1502 return -EEXIST;
1503
1504 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1505 eth_hw_addr_random(dev);
1506
1507 nt->dev = dev;
0bd87628 1508 nt->net = dev_net(dev);
c12b395a 1509 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1510
1511 /* Can use a lockless transmit, unless we generate output sequences */
1512 if (!(nt->parms.o_flags & GRE_SEQ))
1513 dev->features |= NETIF_F_LLTX;
1514
1515 err = register_netdevice(dev);
1516 if (err)
1517 goto out;
1518
1519 dev_hold(dev);
1520 ip6gre_tunnel_link(ign, nt);
1521
1522out:
1523 return err;
1524}
1525
1526static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1527 struct nlattr *data[])
1528{
1529 struct ip6_tnl *t, *nt;
1530 struct net *net = dev_net(dev);
1531 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1532 struct __ip6_tnl_parm p;
1533
1534 if (dev == ign->fb_tunnel_dev)
1535 return -EINVAL;
1536
1537 nt = netdev_priv(dev);
1538 ip6gre_netlink_parms(data, &p);
1539
1540 t = ip6gre_tunnel_locate(net, &p, 0);
1541
1542 if (t) {
1543 if (t->dev != dev)
1544 return -EEXIST;
1545 } else {
1546 t = nt;
1547
1548 ip6gre_tunnel_unlink(ign, t);
1549 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1550 ip6gre_tunnel_link(ign, t);
1551 netdev_state_change(dev);
1552 }
1553
1554 return 0;
1555}
1556
1557static size_t ip6gre_get_size(const struct net_device *dev)
1558{
1559 return
1560 /* IFLA_GRE_LINK */
1561 nla_total_size(4) +
1562 /* IFLA_GRE_IFLAGS */
1563 nla_total_size(2) +
1564 /* IFLA_GRE_OFLAGS */
1565 nla_total_size(2) +
1566 /* IFLA_GRE_IKEY */
1567 nla_total_size(4) +
1568 /* IFLA_GRE_OKEY */
1569 nla_total_size(4) +
1570 /* IFLA_GRE_LOCAL */
a3754133 1571 nla_total_size(sizeof(struct in6_addr)) +
c12b395a 1572 /* IFLA_GRE_REMOTE */
a3754133 1573 nla_total_size(sizeof(struct in6_addr)) +
c12b395a 1574 /* IFLA_GRE_TTL */
1575 nla_total_size(1) +
1576 /* IFLA_GRE_TOS */
1577 nla_total_size(1) +
1578 /* IFLA_GRE_ENCAP_LIMIT */
1579 nla_total_size(1) +
1580 /* IFLA_GRE_FLOWINFO */
1581 nla_total_size(4) +
1582 /* IFLA_GRE_FLAGS */
1583 nla_total_size(4) +
1584 0;
1585}
1586
1587static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1588{
1589 struct ip6_tnl *t = netdev_priv(dev);
1590 struct __ip6_tnl_parm *p = &t->parms;
1591
1592 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1593 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1594 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1595 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1596 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
a3754133
ND		 1597 nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1598 nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
c12b395a 1599 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1600 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1601 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1602 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1603 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1604 goto nla_put_failure;
1605 return 0;
1606
1607nla_put_failure:
1608 return -EMSGSIZE;
1609}
1610
1611static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1612 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1613 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1614 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1615 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1616 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1617 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1618 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1619 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1620 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1621 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1622 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1623};
1624
1625static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1626 .kind = "ip6gre",
1627 .maxtype = IFLA_GRE_MAX,
1628 .policy = ip6gre_policy,
1629 .priv_size = sizeof(struct ip6_tnl),
1630 .setup = ip6gre_tunnel_setup,
1631 .validate = ip6gre_tunnel_validate,
1632 .newlink = ip6gre_newlink,
1633 .changelink = ip6gre_changelink,
1634 .get_size = ip6gre_get_size,
1635 .fill_info = ip6gre_fill_info,
1636};
1637
1638static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1639 .kind = "ip6gretap",
1640 .maxtype = IFLA_GRE_MAX,
1641 .policy = ip6gre_policy,
1642 .priv_size = sizeof(struct ip6_tnl),
1643 .setup = ip6gre_tap_setup,
1644 .validate = ip6gre_tap_validate,
1645 .newlink = ip6gre_newlink,
1646 .changelink = ip6gre_changelink,
1647 .get_size = ip6gre_get_size,
1648 .fill_info = ip6gre_fill_info,
1649};
1650
1651/*
1652 * And now the modules code and kernel interface.
1653 */
1654
1655static int __init ip6gre_init(void)
1656{
1657 int err;
1658
1659 pr_info("GRE over IPv6 tunneling driver\n");
1660
1661 err = register_pernet_device(&ip6gre_net_ops);
1662 if (err < 0)
1663 return err;
1664
1665 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1666 if (err < 0) {
1667 pr_info("%s: can't add protocol\n", __func__);
1668 goto add_proto_failed;
1669 }
1670
1671 err = rtnl_link_register(&ip6gre_link_ops);
1672 if (err < 0)
1673 goto rtnl_link_failed;
1674
1675 err = rtnl_link_register(&ip6gre_tap_ops);
1676 if (err < 0)
1677 goto tap_ops_failed;
1678
1679out:
1680 return err;
1681
1682tap_ops_failed:
1683 rtnl_link_unregister(&ip6gre_link_ops);
1684rtnl_link_failed:
1685 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1686add_proto_failed:
1687 unregister_pernet_device(&ip6gre_net_ops);
1688 goto out;
1689}
1690
1691static void __exit ip6gre_fini(void)
1692{
1693 rtnl_link_unregister(&ip6gre_tap_ops);
1694 rtnl_link_unregister(&ip6gre_link_ops);
1695 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1696 unregister_pernet_device(&ip6gre_net_ops);
1697}
1698
1699module_init(ip6gre_init);
1700module_exit(ip6gre_fini);
1701MODULE_LICENSE("GPL");
1702MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1703MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1704MODULE_ALIAS_RTNL_LINK("ip6gre");
1705MODULE_ALIAS_NETDEV("ip6gre0");
Linux 6.x block layer and io_uring tree(s)