]> Pileus Git - ~andy/linux/blob - net/ipv6/ip6_gre.c
tunnels: harmonize cleanup done on skb on rx path
[~andy/linux] / net / ipv6 / ip6_gre.c
1 /*
2  *      GRE over IPv6 protocol decoder.
3  *
4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
5  *
6  *      This program is free software; you can redistribute it and/or
7  *      modify it under the terms of the GNU General Public License
8  *      as published by the Free Software Foundation; either version
9  *      2 of the License, or (at your option) any later version.
10  *
11  */
12
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/mroute.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
35 #include <linux/hash.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38
39 #include <net/sock.h>
40 #include <net/ip.h>
41 #include <net/ip_tunnels.h>
42 #include <net/icmp.h>
43 #include <net/protocol.h>
44 #include <net/addrconf.h>
45 #include <net/arp.h>
46 #include <net/checksum.h>
47 #include <net/dsfield.h>
48 #include <net/inet_ecn.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/rtnetlink.h>
53
54 #include <net/ipv6.h>
55 #include <net/ip6_fib.h>
56 #include <net/ip6_route.h>
57 #include <net/ip6_tunnel.h>
58
59
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63
64 #define 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
70 static int ip6gre_net_id __read_mostly;
71 struct ip6gre_net {
72         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
73
74         struct net_device *fb_tunnel_dev;
75 };
76
77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
78 static int ip6gre_tunnel_init(struct net_device *dev);
79 static void ip6gre_tunnel_setup(struct net_device *dev);
80 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
81 static 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))
102 static 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]
113
114 /* Given src, dst and key, find appropriate for input tunnel. */
115
116 static 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
130         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
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
155         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
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
179         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
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
205         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
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
238 static 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
256 static 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
262 static 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
270 static 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
285 static 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
310 static 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;
338         nt->net = dev_net(dev);
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
352 failed_free:
353         free_netdev(dev);
354         return NULL;
355 }
356
357 static 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
367 static 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;
371         __be16 *p = (__be16 *)(skb->data + offset);
372         int grehlen = offset + 4;
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 */
388         if (!pskb_may_pull(skb, grehlen))
389                 return;
390         ipv6h = (const struct ipv6hdr *)skb->data;
391         p = (__be16 *)(skb->data + offset);
392
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)
398                 return;
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;
443 }
444
445 static 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;
456         int err;
457
458         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
459                 goto drop;
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))
470                         goto drop;
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
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                 skb->protocol = gre_proto;
513                 /* WCCP version 1 and 2 protocol decoding.
514                  * - Change protocol to IP
515                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
516                  */
517                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
518                         skb->protocol = htons(ETH_P_IP);
519                         if ((*(h + offset) & 0xF0) != 0x40)
520                                 offset += 4;
521                 }
522
523                 skb->mac_header = skb->network_header;
524                 __pskb_pull(skb, offset);
525                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
526
527                 if (((flags&GRE_CSUM) && csum) ||
528                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
529                         tunnel->dev->stats.rx_crc_errors++;
530                         tunnel->dev->stats.rx_errors++;
531                         goto drop;
532                 }
533                 if (tunnel->parms.i_flags&GRE_SEQ) {
534                         if (!(flags&GRE_SEQ) ||
535                             (tunnel->i_seqno &&
536                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
537                                 tunnel->dev->stats.rx_fifo_errors++;
538                                 tunnel->dev->stats.rx_errors++;
539                                 goto drop;
540                         }
541                         tunnel->i_seqno = seqno + 1;
542                 }
543
544                 /* Warning: All skb pointers will be invalidated! */
545                 if (tunnel->dev->type == ARPHRD_ETHER) {
546                         if (!pskb_may_pull(skb, ETH_HLEN)) {
547                                 tunnel->dev->stats.rx_length_errors++;
548                                 tunnel->dev->stats.rx_errors++;
549                                 goto drop;
550                         }
551
552                         ipv6h = ipv6_hdr(skb);
553                         skb->protocol = eth_type_trans(skb, tunnel->dev);
554                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
555                 }
556
557                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
558
559                 skb_reset_network_header(skb);
560
561                 err = IP6_ECN_decapsulate(ipv6h, skb);
562                 if (unlikely(err)) {
563                         if (log_ecn_error)
564                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
565                                                      &ipv6h->saddr,
566                                                      ipv6_get_dsfield(ipv6h));
567                         if (err > 1) {
568                                 ++tunnel->dev->stats.rx_frame_errors;
569                                 ++tunnel->dev->stats.rx_errors;
570                                 goto drop;
571                         }
572                 }
573
574                 tstats = this_cpu_ptr(tunnel->dev->tstats);
575                 u64_stats_update_begin(&tstats->syncp);
576                 tstats->rx_packets++;
577                 tstats->rx_bytes += skb->len;
578                 u64_stats_update_end(&tstats->syncp);
579
580                 netif_rx(skb);
581
582                 return 0;
583         }
584         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
585
586 drop:
587         kfree_skb(skb);
588         return 0;
589 }
590
591 struct ipv6_tel_txoption {
592         struct ipv6_txoptions ops;
593         __u8 dst_opt[8];
594 };
595
596 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
597 {
598         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
599
600         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
601         opt->dst_opt[3] = 1;
602         opt->dst_opt[4] = encap_limit;
603         opt->dst_opt[5] = IPV6_TLV_PADN;
604         opt->dst_opt[6] = 1;
605
606         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
607         opt->ops.opt_nflen = 8;
608 }
609
610 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
611                          struct net_device *dev,
612                          __u8 dsfield,
613                          struct flowi6 *fl6,
614                          int encap_limit,
615                          __u32 *pmtu)
616 {
617         struct net *net = dev_net(dev);
618         struct ip6_tnl *tunnel = netdev_priv(dev);
619         struct net_device *tdev;    /* Device to other host */
620         struct ipv6hdr  *ipv6h;     /* Our new IP header */
621         unsigned int max_headroom;  /* The extra header space needed */
622         int    gre_hlen;
623         struct ipv6_tel_txoption opt;
624         int    mtu;
625         struct dst_entry *dst = NULL, *ndst = NULL;
626         struct net_device_stats *stats = &tunnel->dev->stats;
627         int err = -1;
628         u8 proto;
629         struct sk_buff *new_skb;
630
631         if (dev->type == ARPHRD_ETHER)
632                 IPCB(skb)->flags = 0;
633
634         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
635                 gre_hlen = 0;
636                 ipv6h = (struct ipv6hdr *)skb->data;
637                 fl6->daddr = ipv6h->daddr;
638         } else {
639                 gre_hlen = tunnel->hlen;
640                 fl6->daddr = tunnel->parms.raddr;
641         }
642
643         if (!fl6->flowi6_mark)
644                 dst = ip6_tnl_dst_check(tunnel);
645
646         if (!dst) {
647                 ndst = ip6_route_output(net, NULL, fl6);
648
649                 if (ndst->error)
650                         goto tx_err_link_failure;
651                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
652                 if (IS_ERR(ndst)) {
653                         err = PTR_ERR(ndst);
654                         ndst = NULL;
655                         goto tx_err_link_failure;
656                 }
657                 dst = ndst;
658         }
659
660         tdev = dst->dev;
661
662         if (tdev == dev) {
663                 stats->collisions++;
664                 net_warn_ratelimited("%s: Local routing loop detected!\n",
665                                      tunnel->parms.name);
666                 goto tx_err_dst_release;
667         }
668
669         mtu = dst_mtu(dst) - sizeof(*ipv6h);
670         if (encap_limit >= 0) {
671                 max_headroom += 8;
672                 mtu -= 8;
673         }
674         if (mtu < IPV6_MIN_MTU)
675                 mtu = IPV6_MIN_MTU;
676         if (skb_dst(skb))
677                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
678         if (skb->len > mtu) {
679                 *pmtu = mtu;
680                 err = -EMSGSIZE;
681                 goto tx_err_dst_release;
682         }
683
684         if (tunnel->err_count > 0) {
685                 if (time_before(jiffies,
686                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
687                         tunnel->err_count--;
688
689                         dst_link_failure(skb);
690                 } else
691                         tunnel->err_count = 0;
692         }
693
694         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
695
696         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
697
698         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
699             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
700                 new_skb = skb_realloc_headroom(skb, max_headroom);
701                 if (max_headroom > dev->needed_headroom)
702                         dev->needed_headroom = max_headroom;
703                 if (!new_skb)
704                         goto tx_err_dst_release;
705
706                 if (skb->sk)
707                         skb_set_owner_w(new_skb, skb->sk);
708                 consume_skb(skb);
709                 skb = new_skb;
710         }
711
712         if (fl6->flowi6_mark) {
713                 skb_dst_set(skb, dst);
714                 ndst = NULL;
715         } else {
716                 skb_dst_set_noref(skb, dst);
717         }
718
719         proto = NEXTHDR_GRE;
720         if (encap_limit >= 0) {
721                 init_tel_txopt(&opt, encap_limit);
722                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
723         }
724
725         skb_push(skb, gre_hlen);
726         skb_reset_network_header(skb);
727         skb_set_transport_header(skb, sizeof(*ipv6h));
728
729         /*
730          *      Push down and install the IP header.
731          */
732         ipv6h = ipv6_hdr(skb);
733         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
734         ipv6h->hop_limit = tunnel->parms.hop_limit;
735         ipv6h->nexthdr = proto;
736         ipv6h->saddr = fl6->saddr;
737         ipv6h->daddr = fl6->daddr;
738
739         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
740         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
741                                    htons(ETH_P_TEB) : skb->protocol;
742
743         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
744                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
745
746                 if (tunnel->parms.o_flags&GRE_SEQ) {
747                         ++tunnel->o_seqno;
748                         *ptr = htonl(tunnel->o_seqno);
749                         ptr--;
750                 }
751                 if (tunnel->parms.o_flags&GRE_KEY) {
752                         *ptr = tunnel->parms.o_key;
753                         ptr--;
754                 }
755                 if (tunnel->parms.o_flags&GRE_CSUM) {
756                         *ptr = 0;
757                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
758                                 skb->len - sizeof(struct ipv6hdr));
759                 }
760         }
761
762         ip6tunnel_xmit(skb, dev);
763         if (ndst)
764                 ip6_tnl_dst_store(tunnel, ndst);
765         return 0;
766 tx_err_link_failure:
767         stats->tx_carrier_errors++;
768         dst_link_failure(skb);
769 tx_err_dst_release:
770         dst_release(ndst);
771         return err;
772 }
773
774 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
775 {
776         struct ip6_tnl *t = netdev_priv(dev);
777         const struct iphdr  *iph = ip_hdr(skb);
778         int encap_limit = -1;
779         struct flowi6 fl6;
780         __u8 dsfield;
781         __u32 mtu;
782         int err;
783
784         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
785                 encap_limit = t->parms.encap_limit;
786
787         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
788         fl6.flowi6_proto = IPPROTO_IPIP;
789
790         dsfield = ipv4_get_dsfield(iph);
791
792         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
793                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
794                                           & IPV6_TCLASS_MASK;
795         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
796                 fl6.flowi6_mark = skb->mark;
797
798         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
799         if (err != 0) {
800                 /* XXX: send ICMP error even if DF is not set. */
801                 if (err == -EMSGSIZE)
802                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
803                                   htonl(mtu));
804                 return -1;
805         }
806
807         return 0;
808 }
809
810 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
811 {
812         struct ip6_tnl *t = netdev_priv(dev);
813         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
814         int encap_limit = -1;
815         __u16 offset;
816         struct flowi6 fl6;
817         __u8 dsfield;
818         __u32 mtu;
819         int err;
820
821         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
822                 return -1;
823
824         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
825         if (offset > 0) {
826                 struct ipv6_tlv_tnl_enc_lim *tel;
827                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
828                 if (tel->encap_limit == 0) {
829                         icmpv6_send(skb, ICMPV6_PARAMPROB,
830                                     ICMPV6_HDR_FIELD, offset + 2);
831                         return -1;
832                 }
833                 encap_limit = tel->encap_limit - 1;
834         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
835                 encap_limit = t->parms.encap_limit;
836
837         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
838         fl6.flowi6_proto = IPPROTO_IPV6;
839
840         dsfield = ipv6_get_dsfield(ipv6h);
841         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
842                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
843         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
844                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
845         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
846                 fl6.flowi6_mark = skb->mark;
847
848         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
849         if (err != 0) {
850                 if (err == -EMSGSIZE)
851                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
852                 return -1;
853         }
854
855         return 0;
856 }
857
858 /**
859  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
860  *   @t: the outgoing tunnel device
861  *   @hdr: IPv6 header from the incoming packet
862  *
863  * Description:
864  *   Avoid trivial tunneling loop by checking that tunnel exit-point
865  *   doesn't match source of incoming packet.
866  *
867  * Return:
868  *   1 if conflict,
869  *   0 else
870  **/
871
872 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
873         const struct ipv6hdr *hdr)
874 {
875         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
876 }
877
878 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
879 {
880         struct ip6_tnl *t = netdev_priv(dev);
881         int encap_limit = -1;
882         struct flowi6 fl6;
883         __u32 mtu;
884         int err;
885
886         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
887                 encap_limit = t->parms.encap_limit;
888
889         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
890         fl6.flowi6_proto = skb->protocol;
891
892         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
893
894         return err;
895 }
896
897 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
898         struct net_device *dev)
899 {
900         struct ip6_tnl *t = netdev_priv(dev);
901         struct net_device_stats *stats = &t->dev->stats;
902         int ret;
903
904         if (!ip6_tnl_xmit_ctl(t))
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
924 tx_err:
925         stats->tx_errors++;
926         stats->tx_dropped++;
927         kfree_skb(skb);
928         return NETDEV_TX_OK;
929 }
930
931 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
932 {
933         struct net_device *dev = t->dev;
934         struct __ip6_tnl_parm *p = &t->parms;
935         struct flowi6 *fl6 = &t->fl.u.ip6;
936         int addend = sizeof(struct ipv6hdr) + 4;
937
938         if (dev->type != ARPHRD_ETHER) {
939                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
940                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
941         }
942
943         /* Set up flowi template */
944         fl6->saddr = p->laddr;
945         fl6->daddr = p->raddr;
946         fl6->flowi6_oif = p->link;
947         fl6->flowlabel = 0;
948
949         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
950                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
951         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
952                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
953
954         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
955         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
956
957         if (p->flags&IP6_TNL_F_CAP_XMIT &&
958                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
959                 dev->flags |= IFF_POINTOPOINT;
960         else
961                 dev->flags &= ~IFF_POINTOPOINT;
962
963         dev->iflink = p->link;
964
965         /* Precalculate GRE options length */
966         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
967                 if (t->parms.o_flags&GRE_CSUM)
968                         addend += 4;
969                 if (t->parms.o_flags&GRE_KEY)
970                         addend += 4;
971                 if (t->parms.o_flags&GRE_SEQ)
972                         addend += 4;
973         }
974
975         if (p->flags & IP6_TNL_F_CAP_XMIT) {
976                 int strict = (ipv6_addr_type(&p->raddr) &
977                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
978
979                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
980                                                  &p->raddr, &p->laddr,
981                                                  p->link, strict);
982
983                 if (rt == NULL)
984                         return;
985
986                 if (rt->dst.dev) {
987                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
988
989                         if (set_mtu) {
990                                 dev->mtu = rt->dst.dev->mtu - addend;
991                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
992                                         dev->mtu -= 8;
993
994                                 if (dev->mtu < IPV6_MIN_MTU)
995                                         dev->mtu = IPV6_MIN_MTU;
996                         }
997                 }
998                 ip6_rt_put(rt);
999         }
1000
1001         t->hlen = addend;
1002 }
1003
1004 static int ip6gre_tnl_change(struct ip6_tnl *t,
1005         const struct __ip6_tnl_parm *p, int set_mtu)
1006 {
1007         t->parms.laddr = p->laddr;
1008         t->parms.raddr = p->raddr;
1009         t->parms.flags = p->flags;
1010         t->parms.hop_limit = p->hop_limit;
1011         t->parms.encap_limit = p->encap_limit;
1012         t->parms.flowinfo = p->flowinfo;
1013         t->parms.link = p->link;
1014         t->parms.proto = p->proto;
1015         t->parms.i_key = p->i_key;
1016         t->parms.o_key = p->o_key;
1017         t->parms.i_flags = p->i_flags;
1018         t->parms.o_flags = p->o_flags;
1019         ip6_tnl_dst_reset(t);
1020         ip6gre_tnl_link_config(t, set_mtu);
1021         return 0;
1022 }
1023
1024 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1025         const struct ip6_tnl_parm2 *u)
1026 {
1027         p->laddr = u->laddr;
1028         p->raddr = u->raddr;
1029         p->flags = u->flags;
1030         p->hop_limit = u->hop_limit;
1031         p->encap_limit = u->encap_limit;
1032         p->flowinfo = u->flowinfo;
1033         p->link = u->link;
1034         p->i_key = u->i_key;
1035         p->o_key = u->o_key;
1036         p->i_flags = u->i_flags;
1037         p->o_flags = u->o_flags;
1038         memcpy(p->name, u->name, sizeof(u->name));
1039 }
1040
1041 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1042         const struct __ip6_tnl_parm *p)
1043 {
1044         u->proto = IPPROTO_GRE;
1045         u->laddr = p->laddr;
1046         u->raddr = p->raddr;
1047         u->flags = p->flags;
1048         u->hop_limit = p->hop_limit;
1049         u->encap_limit = p->encap_limit;
1050         u->flowinfo = p->flowinfo;
1051         u->link = p->link;
1052         u->i_key = p->i_key;
1053         u->o_key = p->o_key;
1054         u->i_flags = p->i_flags;
1055         u->o_flags = p->o_flags;
1056         memcpy(u->name, p->name, sizeof(u->name));
1057 }
1058
1059 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1060         struct ifreq *ifr, int cmd)
1061 {
1062         int err = 0;
1063         struct ip6_tnl_parm2 p;
1064         struct __ip6_tnl_parm p1;
1065         struct ip6_tnl *t;
1066         struct net *net = dev_net(dev);
1067         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1068
1069         switch (cmd) {
1070         case SIOCGETTUNNEL:
1071                 t = NULL;
1072                 if (dev == ign->fb_tunnel_dev) {
1073                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1074                                 err = -EFAULT;
1075                                 break;
1076                         }
1077                         ip6gre_tnl_parm_from_user(&p1, &p);
1078                         t = ip6gre_tunnel_locate(net, &p1, 0);
1079                 }
1080                 if (t == NULL)
1081                         t = netdev_priv(dev);
1082                 memset(&p, 0, sizeof(p));
1083                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1084                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1085                         err = -EFAULT;
1086                 break;
1087
1088         case SIOCADDTUNNEL:
1089         case SIOCCHGTUNNEL:
1090                 err = -EPERM;
1091                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1092                         goto done;
1093
1094                 err = -EFAULT;
1095                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1096                         goto done;
1097
1098                 err = -EINVAL;
1099                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1100                         goto done;
1101
1102                 if (!(p.i_flags&GRE_KEY))
1103                         p.i_key = 0;
1104                 if (!(p.o_flags&GRE_KEY))
1105                         p.o_key = 0;
1106
1107                 ip6gre_tnl_parm_from_user(&p1, &p);
1108                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1109
1110                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1111                         if (t != NULL) {
1112                                 if (t->dev != dev) {
1113                                         err = -EEXIST;
1114                                         break;
1115                                 }
1116                         } else {
1117                                 t = netdev_priv(dev);
1118
1119                                 ip6gre_tunnel_unlink(ign, t);
1120                                 synchronize_net();
1121                                 ip6gre_tnl_change(t, &p1, 1);
1122                                 ip6gre_tunnel_link(ign, t);
1123                                 netdev_state_change(dev);
1124                         }
1125                 }
1126
1127                 if (t) {
1128                         err = 0;
1129
1130                         memset(&p, 0, sizeof(p));
1131                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1132                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1133                                 err = -EFAULT;
1134                 } else
1135                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1136                 break;
1137
1138         case SIOCDELTUNNEL:
1139                 err = -EPERM;
1140                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1141                         goto done;
1142
1143                 if (dev == ign->fb_tunnel_dev) {
1144                         err = -EFAULT;
1145                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1146                                 goto done;
1147                         err = -ENOENT;
1148                         ip6gre_tnl_parm_from_user(&p1, &p);
1149                         t = ip6gre_tunnel_locate(net, &p1, 0);
1150                         if (t == NULL)
1151                                 goto done;
1152                         err = -EPERM;
1153                         if (t == netdev_priv(ign->fb_tunnel_dev))
1154                                 goto done;
1155                         dev = t->dev;
1156                 }
1157                 unregister_netdevice(dev);
1158                 err = 0;
1159                 break;
1160
1161         default:
1162                 err = -EINVAL;
1163         }
1164
1165 done:
1166         return err;
1167 }
1168
1169 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1170 {
1171         struct ip6_tnl *tunnel = netdev_priv(dev);
1172         if (new_mtu < 68 ||
1173             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1174                 return -EINVAL;
1175         dev->mtu = new_mtu;
1176         return 0;
1177 }
1178
1179 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1180                         unsigned short type,
1181                         const void *daddr, const void *saddr, unsigned int len)
1182 {
1183         struct ip6_tnl *t = netdev_priv(dev);
1184         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1185         __be16 *p = (__be16 *)(ipv6h+1);
1186
1187         ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1188         ipv6h->hop_limit = t->parms.hop_limit;
1189         ipv6h->nexthdr = NEXTHDR_GRE;
1190         ipv6h->saddr = t->parms.laddr;
1191         ipv6h->daddr = t->parms.raddr;
1192
1193         p[0]            = t->parms.o_flags;
1194         p[1]            = htons(type);
1195
1196         /*
1197          *      Set the source hardware address.
1198          */
1199
1200         if (saddr)
1201                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1202         if (daddr)
1203                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1204         if (!ipv6_addr_any(&ipv6h->daddr))
1205                 return t->hlen;
1206
1207         return -t->hlen;
1208 }
1209
1210 static const struct header_ops ip6gre_header_ops = {
1211         .create = ip6gre_header,
1212 };
1213
1214 static const struct net_device_ops ip6gre_netdev_ops = {
1215         .ndo_init               = ip6gre_tunnel_init,
1216         .ndo_uninit             = ip6gre_tunnel_uninit,
1217         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1218         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1219         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1220         .ndo_get_stats64        = ip_tunnel_get_stats64,
1221 };
1222
1223 static void ip6gre_dev_free(struct net_device *dev)
1224 {
1225         free_percpu(dev->tstats);
1226         free_netdev(dev);
1227 }
1228
1229 static void ip6gre_tunnel_setup(struct net_device *dev)
1230 {
1231         struct ip6_tnl *t;
1232
1233         dev->netdev_ops = &ip6gre_netdev_ops;
1234         dev->destructor = ip6gre_dev_free;
1235
1236         dev->type = ARPHRD_IP6GRE;
1237         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1238         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1239         t = netdev_priv(dev);
1240         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1241                 dev->mtu -= 8;
1242         dev->flags |= IFF_NOARP;
1243         dev->iflink = 0;
1244         dev->addr_len = sizeof(struct in6_addr);
1245         dev->features |= NETIF_F_NETNS_LOCAL;
1246         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1247 }
1248
1249 static int ip6gre_tunnel_init(struct net_device *dev)
1250 {
1251         struct ip6_tnl *tunnel;
1252
1253         tunnel = netdev_priv(dev);
1254
1255         tunnel->dev = dev;
1256         tunnel->net = dev_net(dev);
1257         strcpy(tunnel->parms.name, dev->name);
1258
1259         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1260         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1261
1262         if (ipv6_addr_any(&tunnel->parms.raddr))
1263                 dev->header_ops = &ip6gre_header_ops;
1264
1265         dev->tstats = alloc_percpu(struct pcpu_tstats);
1266         if (!dev->tstats)
1267                 return -ENOMEM;
1268
1269         return 0;
1270 }
1271
1272 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1273 {
1274         struct ip6_tnl *tunnel = netdev_priv(dev);
1275
1276         tunnel->dev = dev;
1277         tunnel->net = dev_net(dev);
1278         strcpy(tunnel->parms.name, dev->name);
1279
1280         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1281
1282         dev_hold(dev);
1283 }
1284
1285
1286 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1287         .handler     = ip6gre_rcv,
1288         .err_handler = ip6gre_err,
1289         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1290 };
1291
1292 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1293         struct list_head *head)
1294 {
1295         int prio;
1296
1297         for (prio = 0; prio < 4; prio++) {
1298                 int h;
1299                 for (h = 0; h < HASH_SIZE; h++) {
1300                         struct ip6_tnl *t;
1301
1302                         t = rtnl_dereference(ign->tunnels[prio][h]);
1303
1304                         while (t != NULL) {
1305                                 unregister_netdevice_queue(t->dev, head);
1306                                 t = rtnl_dereference(t->next);
1307                         }
1308                 }
1309         }
1310 }
1311
1312 static int __net_init ip6gre_init_net(struct net *net)
1313 {
1314         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1315         int err;
1316
1317         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1318                                            ip6gre_tunnel_setup);
1319         if (!ign->fb_tunnel_dev) {
1320                 err = -ENOMEM;
1321                 goto err_alloc_dev;
1322         }
1323         dev_net_set(ign->fb_tunnel_dev, net);
1324
1325         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1326         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1327
1328         err = register_netdev(ign->fb_tunnel_dev);
1329         if (err)
1330                 goto err_reg_dev;
1331
1332         rcu_assign_pointer(ign->tunnels_wc[0],
1333                            netdev_priv(ign->fb_tunnel_dev));
1334         return 0;
1335
1336 err_reg_dev:
1337         ip6gre_dev_free(ign->fb_tunnel_dev);
1338 err_alloc_dev:
1339         return err;
1340 }
1341
1342 static void __net_exit ip6gre_exit_net(struct net *net)
1343 {
1344         struct ip6gre_net *ign;
1345         LIST_HEAD(list);
1346
1347         ign = net_generic(net, ip6gre_net_id);
1348         rtnl_lock();
1349         ip6gre_destroy_tunnels(ign, &list);
1350         unregister_netdevice_many(&list);
1351         rtnl_unlock();
1352 }
1353
1354 static struct pernet_operations ip6gre_net_ops = {
1355         .init = ip6gre_init_net,
1356         .exit = ip6gre_exit_net,
1357         .id   = &ip6gre_net_id,
1358         .size = sizeof(struct ip6gre_net),
1359 };
1360
1361 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1362 {
1363         __be16 flags;
1364
1365         if (!data)
1366                 return 0;
1367
1368         flags = 0;
1369         if (data[IFLA_GRE_IFLAGS])
1370                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1371         if (data[IFLA_GRE_OFLAGS])
1372                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1373         if (flags & (GRE_VERSION|GRE_ROUTING))
1374                 return -EINVAL;
1375
1376         return 0;
1377 }
1378
1379 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1380 {
1381         struct in6_addr daddr;
1382
1383         if (tb[IFLA_ADDRESS]) {
1384                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1385                         return -EINVAL;
1386                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1387                         return -EADDRNOTAVAIL;
1388         }
1389
1390         if (!data)
1391                 goto out;
1392
1393         if (data[IFLA_GRE_REMOTE]) {
1394                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1395                 if (ipv6_addr_any(&daddr))
1396                         return -EINVAL;
1397         }
1398
1399 out:
1400         return ip6gre_tunnel_validate(tb, data);
1401 }
1402
1403
1404 static void ip6gre_netlink_parms(struct nlattr *data[],
1405                                 struct __ip6_tnl_parm *parms)
1406 {
1407         memset(parms, 0, sizeof(*parms));
1408
1409         if (!data)
1410                 return;
1411
1412         if (data[IFLA_GRE_LINK])
1413                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1414
1415         if (data[IFLA_GRE_IFLAGS])
1416                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1417
1418         if (data[IFLA_GRE_OFLAGS])
1419                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1420
1421         if (data[IFLA_GRE_IKEY])
1422                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1423
1424         if (data[IFLA_GRE_OKEY])
1425                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1426
1427         if (data[IFLA_GRE_LOCAL])
1428                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1429
1430         if (data[IFLA_GRE_REMOTE])
1431                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1432
1433         if (data[IFLA_GRE_TTL])
1434                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1435
1436         if (data[IFLA_GRE_ENCAP_LIMIT])
1437                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1438
1439         if (data[IFLA_GRE_FLOWINFO])
1440                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1441
1442         if (data[IFLA_GRE_FLAGS])
1443                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1444 }
1445
1446 static int ip6gre_tap_init(struct net_device *dev)
1447 {
1448         struct ip6_tnl *tunnel;
1449
1450         tunnel = netdev_priv(dev);
1451
1452         tunnel->dev = dev;
1453         tunnel->net = dev_net(dev);
1454         strcpy(tunnel->parms.name, dev->name);
1455
1456         ip6gre_tnl_link_config(tunnel, 1);
1457
1458         dev->tstats = alloc_percpu(struct pcpu_tstats);
1459         if (!dev->tstats)
1460                 return -ENOMEM;
1461
1462         return 0;
1463 }
1464
1465 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1466         .ndo_init = ip6gre_tap_init,
1467         .ndo_uninit = ip6gre_tunnel_uninit,
1468         .ndo_start_xmit = ip6gre_tunnel_xmit,
1469         .ndo_set_mac_address = eth_mac_addr,
1470         .ndo_validate_addr = eth_validate_addr,
1471         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1472         .ndo_get_stats64 = ip_tunnel_get_stats64,
1473 };
1474
1475 static void ip6gre_tap_setup(struct net_device *dev)
1476 {
1477
1478         ether_setup(dev);
1479
1480         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1481         dev->destructor = ip6gre_dev_free;
1482
1483         dev->iflink = 0;
1484         dev->features |= NETIF_F_NETNS_LOCAL;
1485 }
1486
1487 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1488         struct nlattr *tb[], struct nlattr *data[])
1489 {
1490         struct ip6_tnl *nt;
1491         struct net *net = dev_net(dev);
1492         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1493         int err;
1494
1495         nt = netdev_priv(dev);
1496         ip6gre_netlink_parms(data, &nt->parms);
1497
1498         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1499                 return -EEXIST;
1500
1501         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1502                 eth_hw_addr_random(dev);
1503
1504         nt->dev = dev;
1505         nt->net = dev_net(dev);
1506         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1507
1508         /* Can use a lockless transmit, unless we generate output sequences */
1509         if (!(nt->parms.o_flags & GRE_SEQ))
1510                 dev->features |= NETIF_F_LLTX;
1511
1512         err = register_netdevice(dev);
1513         if (err)
1514                 goto out;
1515
1516         dev_hold(dev);
1517         ip6gre_tunnel_link(ign, nt);
1518
1519 out:
1520         return err;
1521 }
1522
1523 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1524                             struct nlattr *data[])
1525 {
1526         struct ip6_tnl *t, *nt;
1527         struct net *net = dev_net(dev);
1528         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1529         struct __ip6_tnl_parm p;
1530
1531         if (dev == ign->fb_tunnel_dev)
1532                 return -EINVAL;
1533
1534         nt = netdev_priv(dev);
1535         ip6gre_netlink_parms(data, &p);
1536
1537         t = ip6gre_tunnel_locate(net, &p, 0);
1538
1539         if (t) {
1540                 if (t->dev != dev)
1541                         return -EEXIST;
1542         } else {
1543                 t = nt;
1544
1545                 ip6gre_tunnel_unlink(ign, t);
1546                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1547                 ip6gre_tunnel_link(ign, t);
1548                 netdev_state_change(dev);
1549         }
1550
1551         return 0;
1552 }
1553
1554 static size_t ip6gre_get_size(const struct net_device *dev)
1555 {
1556         return
1557                 /* IFLA_GRE_LINK */
1558                 nla_total_size(4) +
1559                 /* IFLA_GRE_IFLAGS */
1560                 nla_total_size(2) +
1561                 /* IFLA_GRE_OFLAGS */
1562                 nla_total_size(2) +
1563                 /* IFLA_GRE_IKEY */
1564                 nla_total_size(4) +
1565                 /* IFLA_GRE_OKEY */
1566                 nla_total_size(4) +
1567                 /* IFLA_GRE_LOCAL */
1568                 nla_total_size(sizeof(struct in6_addr)) +
1569                 /* IFLA_GRE_REMOTE */
1570                 nla_total_size(sizeof(struct in6_addr)) +
1571                 /* IFLA_GRE_TTL */
1572                 nla_total_size(1) +
1573                 /* IFLA_GRE_TOS */
1574                 nla_total_size(1) +
1575                 /* IFLA_GRE_ENCAP_LIMIT */
1576                 nla_total_size(1) +
1577                 /* IFLA_GRE_FLOWINFO */
1578                 nla_total_size(4) +
1579                 /* IFLA_GRE_FLAGS */
1580                 nla_total_size(4) +
1581                 0;
1582 }
1583
1584 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1585 {
1586         struct ip6_tnl *t = netdev_priv(dev);
1587         struct __ip6_tnl_parm *p = &t->parms;
1588
1589         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1590             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1591             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1592             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1593             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1594             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1595             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1596             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1597             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1598             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1599             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1600             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1601                 goto nla_put_failure;
1602         return 0;
1603
1604 nla_put_failure:
1605         return -EMSGSIZE;
1606 }
1607
1608 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1609         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1610         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1611         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1612         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1613         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1614         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1615         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1616         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1617         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1618         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1619         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1620 };
1621
1622 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1623         .kind           = "ip6gre",
1624         .maxtype        = IFLA_GRE_MAX,
1625         .policy         = ip6gre_policy,
1626         .priv_size      = sizeof(struct ip6_tnl),
1627         .setup          = ip6gre_tunnel_setup,
1628         .validate       = ip6gre_tunnel_validate,
1629         .newlink        = ip6gre_newlink,
1630         .changelink     = ip6gre_changelink,
1631         .get_size       = ip6gre_get_size,
1632         .fill_info      = ip6gre_fill_info,
1633 };
1634
1635 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1636         .kind           = "ip6gretap",
1637         .maxtype        = IFLA_GRE_MAX,
1638         .policy         = ip6gre_policy,
1639         .priv_size      = sizeof(struct ip6_tnl),
1640         .setup          = ip6gre_tap_setup,
1641         .validate       = ip6gre_tap_validate,
1642         .newlink        = ip6gre_newlink,
1643         .changelink     = ip6gre_changelink,
1644         .get_size       = ip6gre_get_size,
1645         .fill_info      = ip6gre_fill_info,
1646 };
1647
1648 /*
1649  *      And now the modules code and kernel interface.
1650  */
1651
1652 static int __init ip6gre_init(void)
1653 {
1654         int err;
1655
1656         pr_info("GRE over IPv6 tunneling driver\n");
1657
1658         err = register_pernet_device(&ip6gre_net_ops);
1659         if (err < 0)
1660                 return err;
1661
1662         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1663         if (err < 0) {
1664                 pr_info("%s: can't add protocol\n", __func__);
1665                 goto add_proto_failed;
1666         }
1667
1668         err = rtnl_link_register(&ip6gre_link_ops);
1669         if (err < 0)
1670                 goto rtnl_link_failed;
1671
1672         err = rtnl_link_register(&ip6gre_tap_ops);
1673         if (err < 0)
1674                 goto tap_ops_failed;
1675
1676 out:
1677         return err;
1678
1679 tap_ops_failed:
1680         rtnl_link_unregister(&ip6gre_link_ops);
1681 rtnl_link_failed:
1682         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1683 add_proto_failed:
1684         unregister_pernet_device(&ip6gre_net_ops);
1685         goto out;
1686 }
1687
1688 static void __exit ip6gre_fini(void)
1689 {
1690         rtnl_link_unregister(&ip6gre_tap_ops);
1691         rtnl_link_unregister(&ip6gre_link_ops);
1692         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1693         unregister_pernet_device(&ip6gre_net_ops);
1694 }
1695
1696 module_init(ip6gre_init);
1697 module_exit(ip6gre_fini);
1698 MODULE_LICENSE("GPL");
1699 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1700 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1701 MODULE_ALIAS_RTNL_LINK("ip6gre");
1702 MODULE_ALIAS_NETDEV("ip6gre0");