Merge tag 'wireless-drivers-next-for-davem-2016-05-02' of git://git.kernel.org/pub...
[cascardo/linux.git] / 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/init.h>
28 #include <linux/in6.h>
29 #include <linux/inetdevice.h>
30 #include <linux/igmp.h>
31 #include <linux/netfilter_ipv4.h>
32 #include <linux/etherdevice.h>
33 #include <linux/if_ether.h>
34 #include <linux/hash.h>
35 #include <linux/if_tunnel.h>
36 #include <linux/ip6_tunnel.h>
37
38 #include <net/sock.h>
39 #include <net/ip.h>
40 #include <net/ip_tunnels.h>
41 #include <net/icmp.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
44 #include <net/arp.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
48 #include <net/xfrm.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
52
53 #include <net/ipv6.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.h>
57 #include <net/gre.h>
58
59
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63
64 #define HASH_SIZE_SHIFT  5
65 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
66
67 static int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
70
71         struct net_device *fb_tunnel_dev;
72 };
73
74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80
81 /* Tunnel hash table */
82
83 /*
84    4 hash tables:
85
86    3: (remote,local)
87    2: (remote,*)
88    1: (*,local)
89    0: (*,*)
90
91    We require exact key match i.e. if a key is present in packet
92    it will match only tunnel with the same key; if it is not present,
93    it will match only keyless tunnel.
94
95    All keysless packets, if not matched configured keyless tunnels
96    will match fallback tunnel.
97  */
98
99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
100 static u32 HASH_ADDR(const struct in6_addr *addr)
101 {
102         u32 hash = ipv6_addr_hash(addr);
103
104         return hash_32(hash, HASH_SIZE_SHIFT);
105 }
106
107 #define tunnels_r_l     tunnels[3]
108 #define tunnels_r       tunnels[2]
109 #define tunnels_l       tunnels[1]
110 #define tunnels_wc      tunnels[0]
111
112 /* Given src, dst and key, find appropriate for input tunnel. */
113
114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
115                 const struct in6_addr *remote, const struct in6_addr *local,
116                 __be32 key, __be16 gre_proto)
117 {
118         struct net *net = dev_net(dev);
119         int link = dev->ifindex;
120         unsigned int h0 = HASH_ADDR(remote);
121         unsigned int h1 = HASH_KEY(key);
122         struct ip6_tnl *t, *cand = NULL;
123         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
124         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
125                        ARPHRD_ETHER : ARPHRD_IP6GRE;
126         int score, cand_score = 4;
127
128         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
129                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
130                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
131                     key != t->parms.i_key ||
132                     !(t->dev->flags & IFF_UP))
133                         continue;
134
135                 if (t->dev->type != ARPHRD_IP6GRE &&
136                     t->dev->type != dev_type)
137                         continue;
138
139                 score = 0;
140                 if (t->parms.link != link)
141                         score |= 1;
142                 if (t->dev->type != dev_type)
143                         score |= 2;
144                 if (score == 0)
145                         return t;
146
147                 if (score < cand_score) {
148                         cand = t;
149                         cand_score = score;
150                 }
151         }
152
153         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
154                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
155                     key != t->parms.i_key ||
156                     !(t->dev->flags & IFF_UP))
157                         continue;
158
159                 if (t->dev->type != ARPHRD_IP6GRE &&
160                     t->dev->type != dev_type)
161                         continue;
162
163                 score = 0;
164                 if (t->parms.link != link)
165                         score |= 1;
166                 if (t->dev->type != dev_type)
167                         score |= 2;
168                 if (score == 0)
169                         return t;
170
171                 if (score < cand_score) {
172                         cand = t;
173                         cand_score = score;
174                 }
175         }
176
177         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
178                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
179                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
180                                  !ipv6_addr_is_multicast(local))) ||
181                     key != t->parms.i_key ||
182                     !(t->dev->flags & IFF_UP))
183                         continue;
184
185                 if (t->dev->type != ARPHRD_IP6GRE &&
186                     t->dev->type != dev_type)
187                         continue;
188
189                 score = 0;
190                 if (t->parms.link != link)
191                         score |= 1;
192                 if (t->dev->type != dev_type)
193                         score |= 2;
194                 if (score == 0)
195                         return t;
196
197                 if (score < cand_score) {
198                         cand = t;
199                         cand_score = score;
200                 }
201         }
202
203         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
204                 if (t->parms.i_key != key ||
205                     !(t->dev->flags & IFF_UP))
206                         continue;
207
208                 if (t->dev->type != ARPHRD_IP6GRE &&
209                     t->dev->type != dev_type)
210                         continue;
211
212                 score = 0;
213                 if (t->parms.link != link)
214                         score |= 1;
215                 if (t->dev->type != dev_type)
216                         score |= 2;
217                 if (score == 0)
218                         return t;
219
220                 if (score < cand_score) {
221                         cand = t;
222                         cand_score = score;
223                 }
224         }
225
226         if (cand)
227                 return cand;
228
229         dev = ign->fb_tunnel_dev;
230         if (dev->flags & IFF_UP)
231                 return netdev_priv(dev);
232
233         return NULL;
234 }
235
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237                 const struct __ip6_tnl_parm *p)
238 {
239         const struct in6_addr *remote = &p->raddr;
240         const struct in6_addr *local = &p->laddr;
241         unsigned int h = HASH_KEY(p->i_key);
242         int prio = 0;
243
244         if (!ipv6_addr_any(local))
245                 prio |= 1;
246         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
247                 prio |= 2;
248                 h ^= HASH_ADDR(remote);
249         }
250
251         return &ign->tunnels[prio][h];
252 }
253
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255                 const struct ip6_tnl *t)
256 {
257         return __ip6gre_bucket(ign, &t->parms);
258 }
259
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
261 {
262         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
263
264         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265         rcu_assign_pointer(*tp, t);
266 }
267
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
269 {
270         struct ip6_tnl __rcu **tp;
271         struct ip6_tnl *iter;
272
273         for (tp = ip6gre_bucket(ign, t);
274              (iter = rtnl_dereference(*tp)) != NULL;
275              tp = &iter->next) {
276                 if (t == iter) {
277                         rcu_assign_pointer(*tp, t->next);
278                         break;
279                 }
280         }
281 }
282
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284                                            const struct __ip6_tnl_parm *parms,
285                                            int type)
286 {
287         const struct in6_addr *remote = &parms->raddr;
288         const struct in6_addr *local = &parms->laddr;
289         __be32 key = parms->i_key;
290         int link = parms->link;
291         struct ip6_tnl *t;
292         struct ip6_tnl __rcu **tp;
293         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
294
295         for (tp = __ip6gre_bucket(ign, parms);
296              (t = rtnl_dereference(*tp)) != NULL;
297              tp = &t->next)
298                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
299                     ipv6_addr_equal(remote, &t->parms.raddr) &&
300                     key == t->parms.i_key &&
301                     link == t->parms.link &&
302                     type == t->dev->type)
303                         break;
304
305         return t;
306 }
307
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309                 const struct __ip6_tnl_parm *parms, int create)
310 {
311         struct ip6_tnl *t, *nt;
312         struct net_device *dev;
313         char name[IFNAMSIZ];
314         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315
316         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317         if (t && create)
318                 return NULL;
319         if (t || !create)
320                 return t;
321
322         if (parms->name[0])
323                 strlcpy(name, parms->name, IFNAMSIZ);
324         else
325                 strcpy(name, "ip6gre%d");
326
327         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328                            ip6gre_tunnel_setup);
329         if (!dev)
330                 return NULL;
331
332         dev_net_set(dev, net);
333
334         nt = netdev_priv(dev);
335         nt->parms = *parms;
336         dev->rtnl_link_ops = &ip6gre_link_ops;
337
338         nt->dev = dev;
339         nt->net = dev_net(dev);
340         ip6gre_tnl_link_config(nt, 1);
341
342         if (register_netdevice(dev) < 0)
343                 goto failed_free;
344
345         /* Can use a lockless transmit, unless we generate output sequences */
346         if (!(nt->parms.o_flags & GRE_SEQ))
347                 dev->features |= NETIF_F_LLTX;
348
349         dev_hold(dev);
350         ip6gre_tunnel_link(ign, nt);
351         return nt;
352
353 failed_free:
354         free_netdev(dev);
355         return NULL;
356 }
357
358 static void ip6gre_tunnel_uninit(struct net_device *dev)
359 {
360         struct ip6_tnl *t = netdev_priv(dev);
361         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
362
363         ip6gre_tunnel_unlink(ign, t);
364         dst_cache_reset(&t->dst_cache);
365         dev_put(dev);
366 }
367
368
369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
370                 u8 type, u8 code, int offset, __be32 info)
371 {
372         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
373         __be16 *p = (__be16 *)(skb->data + offset);
374         int grehlen = offset + 4;
375         struct ip6_tnl *t;
376         __be16 flags;
377
378         flags = p[0];
379         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
380                 if (flags&(GRE_VERSION|GRE_ROUTING))
381                         return;
382                 if (flags&GRE_KEY) {
383                         grehlen += 4;
384                         if (flags&GRE_CSUM)
385                                 grehlen += 4;
386                 }
387         }
388
389         /* If only 8 bytes returned, keyed message will be dropped here */
390         if (!pskb_may_pull(skb, grehlen))
391                 return;
392         ipv6h = (const struct ipv6hdr *)skb->data;
393         p = (__be16 *)(skb->data + offset);
394
395         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
396                                 flags & GRE_KEY ?
397                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
398                                 p[1]);
399         if (!t)
400                 return;
401
402         switch (type) {
403                 __u32 teli;
404                 struct ipv6_tlv_tnl_enc_lim *tel;
405                 __u32 mtu;
406         case ICMPV6_DEST_UNREACH:
407                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
408                                     t->parms.name);
409                 break;
410         case ICMPV6_TIME_EXCEED:
411                 if (code == ICMPV6_EXC_HOPLIMIT) {
412                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
413                                             t->parms.name);
414                 }
415                 break;
416         case ICMPV6_PARAMPROB:
417                 teli = 0;
418                 if (code == ICMPV6_HDR_FIELD)
419                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
420
421                 if (teli && teli == be32_to_cpu(info) - 2) {
422                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
423                         if (tel->encap_limit == 0) {
424                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
425                                                     t->parms.name);
426                         }
427                 } else {
428                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
429                                             t->parms.name);
430                 }
431                 break;
432         case ICMPV6_PKT_TOOBIG:
433                 mtu = be32_to_cpu(info) - offset;
434                 if (mtu < IPV6_MIN_MTU)
435                         mtu = IPV6_MIN_MTU;
436                 t->dev->mtu = mtu;
437                 break;
438         }
439
440         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
441                 t->err_count++;
442         else
443                 t->err_count = 1;
444         t->err_time = jiffies;
445 }
446
447 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
448 {
449         const struct ipv6hdr *ipv6h;
450         struct ip6_tnl *tunnel;
451
452         ipv6h = ipv6_hdr(skb);
453         tunnel = ip6gre_tunnel_lookup(skb->dev,
454                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
455                                       tpi->proto);
456         if (tunnel) {
457                 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
458
459                 return PACKET_RCVD;
460         }
461
462         return PACKET_REJECT;
463 }
464
465 static int gre_rcv(struct sk_buff *skb)
466 {
467         struct tnl_ptk_info tpi;
468         bool csum_err = false;
469         int hdr_len;
470
471         if (gre_parse_header(skb, &tpi, &csum_err, &hdr_len) < 0)
472                 goto drop;
473
474         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
475                 goto drop;
476
477         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
478                 return 0;
479
480         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
481 drop:
482         kfree_skb(skb);
483         return 0;
484 }
485
486 struct ipv6_tel_txoption {
487         struct ipv6_txoptions ops;
488         __u8 dst_opt[8];
489 };
490
491 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
492 {
493         return iptunnel_handle_offloads(skb,
494                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
495 }
496
497 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
498                                struct net_device *dev, __u8 dsfield,
499                                struct flowi6 *fl6, int encap_limit,
500                                __u32 *pmtu, __be16 proto)
501 {
502         struct ip6_tnl *tunnel = netdev_priv(dev);
503         __be16 protocol = (dev->type == ARPHRD_ETHER) ?
504                           htons(ETH_P_TEB) : proto;
505
506         if (dev->type == ARPHRD_ETHER)
507                 IPCB(skb)->flags = 0;
508
509         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
510                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
511         else
512                 fl6->daddr = tunnel->parms.raddr;
513
514         if (tunnel->parms.o_flags & TUNNEL_SEQ)
515                 tunnel->o_seqno++;
516
517         /* Push GRE header. */
518         gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
519                          protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
520
521         skb_set_inner_protocol(skb, proto);
522
523         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
524                             NEXTHDR_GRE);
525 }
526
527 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
528 {
529         struct ip6_tnl *t = netdev_priv(dev);
530         const struct iphdr  *iph = ip_hdr(skb);
531         int encap_limit = -1;
532         struct flowi6 fl6;
533         __u8 dsfield;
534         __u32 mtu;
535         int err;
536
537         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
538
539         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
540                 encap_limit = t->parms.encap_limit;
541
542         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
543
544         dsfield = ipv4_get_dsfield(iph);
545
546         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
547                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
548                                           & IPV6_TCLASS_MASK;
549         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
550                 fl6.flowi6_mark = skb->mark;
551
552         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
553         if (err)
554                 return -1;
555
556         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
557                           skb->protocol);
558         if (err != 0) {
559                 /* XXX: send ICMP error even if DF is not set. */
560                 if (err == -EMSGSIZE)
561                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
562                                   htonl(mtu));
563                 return -1;
564         }
565
566         return 0;
567 }
568
569 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
570 {
571         struct ip6_tnl *t = netdev_priv(dev);
572         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
573         int encap_limit = -1;
574         __u16 offset;
575         struct flowi6 fl6;
576         __u8 dsfield;
577         __u32 mtu;
578         int err;
579
580         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
581                 return -1;
582
583         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
584         if (offset > 0) {
585                 struct ipv6_tlv_tnl_enc_lim *tel;
586                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
587                 if (tel->encap_limit == 0) {
588                         icmpv6_send(skb, ICMPV6_PARAMPROB,
589                                     ICMPV6_HDR_FIELD, offset + 2);
590                         return -1;
591                 }
592                 encap_limit = tel->encap_limit - 1;
593         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
594                 encap_limit = t->parms.encap_limit;
595
596         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
597
598         dsfield = ipv6_get_dsfield(ipv6h);
599         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
600                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
601         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
602                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
603         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
604                 fl6.flowi6_mark = skb->mark;
605
606         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
607                 return -1;
608
609         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
610                           &mtu, skb->protocol);
611         if (err != 0) {
612                 if (err == -EMSGSIZE)
613                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
614                 return -1;
615         }
616
617         return 0;
618 }
619
620 /**
621  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
622  *   @t: the outgoing tunnel device
623  *   @hdr: IPv6 header from the incoming packet
624  *
625  * Description:
626  *   Avoid trivial tunneling loop by checking that tunnel exit-point
627  *   doesn't match source of incoming packet.
628  *
629  * Return:
630  *   1 if conflict,
631  *   0 else
632  **/
633
634 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
635         const struct ipv6hdr *hdr)
636 {
637         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
638 }
639
640 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
641 {
642         struct ip6_tnl *t = netdev_priv(dev);
643         int encap_limit = -1;
644         struct flowi6 fl6;
645         __u32 mtu;
646         int err;
647
648         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
649                 encap_limit = t->parms.encap_limit;
650
651         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
652         fl6.flowi6_proto = skb->protocol;
653
654         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
655         if (err)
656                 return err;
657
658         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
659
660         return err;
661 }
662
663 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
664         struct net_device *dev)
665 {
666         struct ip6_tnl *t = netdev_priv(dev);
667         struct net_device_stats *stats = &t->dev->stats;
668         int ret;
669
670         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
671                 goto tx_err;
672
673         switch (skb->protocol) {
674         case htons(ETH_P_IP):
675                 ret = ip6gre_xmit_ipv4(skb, dev);
676                 break;
677         case htons(ETH_P_IPV6):
678                 ret = ip6gre_xmit_ipv6(skb, dev);
679                 break;
680         default:
681                 ret = ip6gre_xmit_other(skb, dev);
682                 break;
683         }
684
685         if (ret < 0)
686                 goto tx_err;
687
688         return NETDEV_TX_OK;
689
690 tx_err:
691         stats->tx_errors++;
692         stats->tx_dropped++;
693         kfree_skb(skb);
694         return NETDEV_TX_OK;
695 }
696
697 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
698 {
699         struct net_device *dev = t->dev;
700         struct __ip6_tnl_parm *p = &t->parms;
701         struct flowi6 *fl6 = &t->fl.u.ip6;
702         int addend = sizeof(struct ipv6hdr) + 4;
703
704         if (dev->type != ARPHRD_ETHER) {
705                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
706                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
707         }
708
709         /* Set up flowi template */
710         fl6->saddr = p->laddr;
711         fl6->daddr = p->raddr;
712         fl6->flowi6_oif = p->link;
713         fl6->flowlabel = 0;
714
715         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
716                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
717         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
718                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
719
720         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
721         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
722
723         if (p->flags&IP6_TNL_F_CAP_XMIT &&
724                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
725                 dev->flags |= IFF_POINTOPOINT;
726         else
727                 dev->flags &= ~IFF_POINTOPOINT;
728
729         /* Precalculate GRE options length */
730         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
731                 if (t->parms.o_flags&GRE_CSUM)
732                         addend += 4;
733                 if (t->parms.o_flags&GRE_KEY)
734                         addend += 4;
735                 if (t->parms.o_flags&GRE_SEQ)
736                         addend += 4;
737         }
738         t->hlen = addend;
739
740         if (p->flags & IP6_TNL_F_CAP_XMIT) {
741                 int strict = (ipv6_addr_type(&p->raddr) &
742                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
743
744                 struct rt6_info *rt = rt6_lookup(t->net,
745                                                  &p->raddr, &p->laddr,
746                                                  p->link, strict);
747
748                 if (!rt)
749                         return;
750
751                 if (rt->dst.dev) {
752                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
753
754                         if (set_mtu) {
755                                 dev->mtu = rt->dst.dev->mtu - addend;
756                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
757                                         dev->mtu -= 8;
758                                 if (dev->type == ARPHRD_ETHER)
759                                         dev->mtu -= ETH_HLEN;
760
761                                 if (dev->mtu < IPV6_MIN_MTU)
762                                         dev->mtu = IPV6_MIN_MTU;
763                         }
764                 }
765                 ip6_rt_put(rt);
766         }
767 }
768
769 static int ip6gre_tnl_change(struct ip6_tnl *t,
770         const struct __ip6_tnl_parm *p, int set_mtu)
771 {
772         t->parms.laddr = p->laddr;
773         t->parms.raddr = p->raddr;
774         t->parms.flags = p->flags;
775         t->parms.hop_limit = p->hop_limit;
776         t->parms.encap_limit = p->encap_limit;
777         t->parms.flowinfo = p->flowinfo;
778         t->parms.link = p->link;
779         t->parms.proto = p->proto;
780         t->parms.i_key = p->i_key;
781         t->parms.o_key = p->o_key;
782         t->parms.i_flags = p->i_flags;
783         t->parms.o_flags = p->o_flags;
784         dst_cache_reset(&t->dst_cache);
785         ip6gre_tnl_link_config(t, set_mtu);
786         return 0;
787 }
788
789 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
790         const struct ip6_tnl_parm2 *u)
791 {
792         p->laddr = u->laddr;
793         p->raddr = u->raddr;
794         p->flags = u->flags;
795         p->hop_limit = u->hop_limit;
796         p->encap_limit = u->encap_limit;
797         p->flowinfo = u->flowinfo;
798         p->link = u->link;
799         p->i_key = u->i_key;
800         p->o_key = u->o_key;
801         p->i_flags = u->i_flags;
802         p->o_flags = u->o_flags;
803         memcpy(p->name, u->name, sizeof(u->name));
804 }
805
806 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
807         const struct __ip6_tnl_parm *p)
808 {
809         u->proto = IPPROTO_GRE;
810         u->laddr = p->laddr;
811         u->raddr = p->raddr;
812         u->flags = p->flags;
813         u->hop_limit = p->hop_limit;
814         u->encap_limit = p->encap_limit;
815         u->flowinfo = p->flowinfo;
816         u->link = p->link;
817         u->i_key = p->i_key;
818         u->o_key = p->o_key;
819         u->i_flags = p->i_flags;
820         u->o_flags = p->o_flags;
821         memcpy(u->name, p->name, sizeof(u->name));
822 }
823
824 static int ip6gre_tunnel_ioctl(struct net_device *dev,
825         struct ifreq *ifr, int cmd)
826 {
827         int err = 0;
828         struct ip6_tnl_parm2 p;
829         struct __ip6_tnl_parm p1;
830         struct ip6_tnl *t = netdev_priv(dev);
831         struct net *net = t->net;
832         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
833
834         memset(&p1, 0, sizeof(p1));
835
836         switch (cmd) {
837         case SIOCGETTUNNEL:
838                 if (dev == ign->fb_tunnel_dev) {
839                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
840                                 err = -EFAULT;
841                                 break;
842                         }
843                         ip6gre_tnl_parm_from_user(&p1, &p);
844                         t = ip6gre_tunnel_locate(net, &p1, 0);
845                         if (!t)
846                                 t = netdev_priv(dev);
847                 }
848                 memset(&p, 0, sizeof(p));
849                 ip6gre_tnl_parm_to_user(&p, &t->parms);
850                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
851                         err = -EFAULT;
852                 break;
853
854         case SIOCADDTUNNEL:
855         case SIOCCHGTUNNEL:
856                 err = -EPERM;
857                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
858                         goto done;
859
860                 err = -EFAULT;
861                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
862                         goto done;
863
864                 err = -EINVAL;
865                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
866                         goto done;
867
868                 if (!(p.i_flags&GRE_KEY))
869                         p.i_key = 0;
870                 if (!(p.o_flags&GRE_KEY))
871                         p.o_key = 0;
872
873                 ip6gre_tnl_parm_from_user(&p1, &p);
874                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
875
876                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
877                         if (t) {
878                                 if (t->dev != dev) {
879                                         err = -EEXIST;
880                                         break;
881                                 }
882                         } else {
883                                 t = netdev_priv(dev);
884
885                                 ip6gre_tunnel_unlink(ign, t);
886                                 synchronize_net();
887                                 ip6gre_tnl_change(t, &p1, 1);
888                                 ip6gre_tunnel_link(ign, t);
889                                 netdev_state_change(dev);
890                         }
891                 }
892
893                 if (t) {
894                         err = 0;
895
896                         memset(&p, 0, sizeof(p));
897                         ip6gre_tnl_parm_to_user(&p, &t->parms);
898                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
899                                 err = -EFAULT;
900                 } else
901                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
902                 break;
903
904         case SIOCDELTUNNEL:
905                 err = -EPERM;
906                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
907                         goto done;
908
909                 if (dev == ign->fb_tunnel_dev) {
910                         err = -EFAULT;
911                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
912                                 goto done;
913                         err = -ENOENT;
914                         ip6gre_tnl_parm_from_user(&p1, &p);
915                         t = ip6gre_tunnel_locate(net, &p1, 0);
916                         if (!t)
917                                 goto done;
918                         err = -EPERM;
919                         if (t == netdev_priv(ign->fb_tunnel_dev))
920                                 goto done;
921                         dev = t->dev;
922                 }
923                 unregister_netdevice(dev);
924                 err = 0;
925                 break;
926
927         default:
928                 err = -EINVAL;
929         }
930
931 done:
932         return err;
933 }
934
935 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
936                         unsigned short type,
937                         const void *daddr, const void *saddr, unsigned int len)
938 {
939         struct ip6_tnl *t = netdev_priv(dev);
940         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
941         __be16 *p = (__be16 *)(ipv6h+1);
942
943         ip6_flow_hdr(ipv6h, 0,
944                      ip6_make_flowlabel(dev_net(dev), skb,
945                                         t->fl.u.ip6.flowlabel, true,
946                                         &t->fl.u.ip6));
947         ipv6h->hop_limit = t->parms.hop_limit;
948         ipv6h->nexthdr = NEXTHDR_GRE;
949         ipv6h->saddr = t->parms.laddr;
950         ipv6h->daddr = t->parms.raddr;
951
952         p[0]            = t->parms.o_flags;
953         p[1]            = htons(type);
954
955         /*
956          *      Set the source hardware address.
957          */
958
959         if (saddr)
960                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
961         if (daddr)
962                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
963         if (!ipv6_addr_any(&ipv6h->daddr))
964                 return t->hlen;
965
966         return -t->hlen;
967 }
968
969 static const struct header_ops ip6gre_header_ops = {
970         .create = ip6gre_header,
971 };
972
973 static const struct net_device_ops ip6gre_netdev_ops = {
974         .ndo_init               = ip6gre_tunnel_init,
975         .ndo_uninit             = ip6gre_tunnel_uninit,
976         .ndo_start_xmit         = ip6gre_tunnel_xmit,
977         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
978         .ndo_change_mtu         = ip6_tnl_change_mtu,
979         .ndo_get_stats64        = ip_tunnel_get_stats64,
980         .ndo_get_iflink         = ip6_tnl_get_iflink,
981 };
982
983 static void ip6gre_dev_free(struct net_device *dev)
984 {
985         struct ip6_tnl *t = netdev_priv(dev);
986
987         dst_cache_destroy(&t->dst_cache);
988         free_percpu(dev->tstats);
989         free_netdev(dev);
990 }
991
992 static void ip6gre_tunnel_setup(struct net_device *dev)
993 {
994         dev->netdev_ops = &ip6gre_netdev_ops;
995         dev->destructor = ip6gre_dev_free;
996
997         dev->type = ARPHRD_IP6GRE;
998
999         dev->flags |= IFF_NOARP;
1000         dev->addr_len = sizeof(struct in6_addr);
1001         netif_keep_dst(dev);
1002 }
1003
1004 static int ip6gre_tunnel_init_common(struct net_device *dev)
1005 {
1006         struct ip6_tnl *tunnel;
1007         int ret;
1008         int t_hlen;
1009
1010         tunnel = netdev_priv(dev);
1011
1012         tunnel->dev = dev;
1013         tunnel->net = dev_net(dev);
1014         strcpy(tunnel->parms.name, dev->name);
1015
1016         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1017         if (!dev->tstats)
1018                 return -ENOMEM;
1019
1020         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1021         if (ret) {
1022                 free_percpu(dev->tstats);
1023                 dev->tstats = NULL;
1024                 return ret;
1025         }
1026
1027         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1028
1029         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1030
1031         dev->needed_headroom    = LL_MAX_HEADER + t_hlen + 4;
1032         dev->mtu                = ETH_DATA_LEN - t_hlen - 4;
1033
1034         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1035                 dev->mtu -= 8;
1036
1037         return 0;
1038 }
1039
1040 static int ip6gre_tunnel_init(struct net_device *dev)
1041 {
1042         struct ip6_tnl *tunnel;
1043         int ret;
1044
1045         ret = ip6gre_tunnel_init_common(dev);
1046         if (ret)
1047                 return ret;
1048
1049         tunnel = netdev_priv(dev);
1050
1051         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1052         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1053
1054         if (ipv6_addr_any(&tunnel->parms.raddr))
1055                 dev->header_ops = &ip6gre_header_ops;
1056
1057         return 0;
1058 }
1059
1060 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1061 {
1062         struct ip6_tnl *tunnel = netdev_priv(dev);
1063
1064         tunnel->dev = dev;
1065         tunnel->net = dev_net(dev);
1066         strcpy(tunnel->parms.name, dev->name);
1067
1068         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1069
1070         dev_hold(dev);
1071 }
1072
1073
1074 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1075         .handler     = gre_rcv,
1076         .err_handler = ip6gre_err,
1077         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1078 };
1079
1080 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1081 {
1082         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1083         struct net_device *dev, *aux;
1084         int prio;
1085
1086         for_each_netdev_safe(net, dev, aux)
1087                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1088                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1089                         unregister_netdevice_queue(dev, head);
1090
1091         for (prio = 0; prio < 4; prio++) {
1092                 int h;
1093                 for (h = 0; h < HASH_SIZE; h++) {
1094                         struct ip6_tnl *t;
1095
1096                         t = rtnl_dereference(ign->tunnels[prio][h]);
1097
1098                         while (t) {
1099                                 /* If dev is in the same netns, it has already
1100                                  * been added to the list by the previous loop.
1101                                  */
1102                                 if (!net_eq(dev_net(t->dev), net))
1103                                         unregister_netdevice_queue(t->dev,
1104                                                                    head);
1105                                 t = rtnl_dereference(t->next);
1106                         }
1107                 }
1108         }
1109 }
1110
1111 static int __net_init ip6gre_init_net(struct net *net)
1112 {
1113         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1114         int err;
1115
1116         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1117                                           NET_NAME_UNKNOWN,
1118                                           ip6gre_tunnel_setup);
1119         if (!ign->fb_tunnel_dev) {
1120                 err = -ENOMEM;
1121                 goto err_alloc_dev;
1122         }
1123         dev_net_set(ign->fb_tunnel_dev, net);
1124         /* FB netdevice is special: we have one, and only one per netns.
1125          * Allowing to move it to another netns is clearly unsafe.
1126          */
1127         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1128
1129
1130         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1131         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1132
1133         err = register_netdev(ign->fb_tunnel_dev);
1134         if (err)
1135                 goto err_reg_dev;
1136
1137         rcu_assign_pointer(ign->tunnels_wc[0],
1138                            netdev_priv(ign->fb_tunnel_dev));
1139         return 0;
1140
1141 err_reg_dev:
1142         ip6gre_dev_free(ign->fb_tunnel_dev);
1143 err_alloc_dev:
1144         return err;
1145 }
1146
1147 static void __net_exit ip6gre_exit_net(struct net *net)
1148 {
1149         LIST_HEAD(list);
1150
1151         rtnl_lock();
1152         ip6gre_destroy_tunnels(net, &list);
1153         unregister_netdevice_many(&list);
1154         rtnl_unlock();
1155 }
1156
1157 static struct pernet_operations ip6gre_net_ops = {
1158         .init = ip6gre_init_net,
1159         .exit = ip6gre_exit_net,
1160         .id   = &ip6gre_net_id,
1161         .size = sizeof(struct ip6gre_net),
1162 };
1163
1164 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1165 {
1166         __be16 flags;
1167
1168         if (!data)
1169                 return 0;
1170
1171         flags = 0;
1172         if (data[IFLA_GRE_IFLAGS])
1173                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1174         if (data[IFLA_GRE_OFLAGS])
1175                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1176         if (flags & (GRE_VERSION|GRE_ROUTING))
1177                 return -EINVAL;
1178
1179         return 0;
1180 }
1181
1182 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1183 {
1184         struct in6_addr daddr;
1185
1186         if (tb[IFLA_ADDRESS]) {
1187                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1188                         return -EINVAL;
1189                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1190                         return -EADDRNOTAVAIL;
1191         }
1192
1193         if (!data)
1194                 goto out;
1195
1196         if (data[IFLA_GRE_REMOTE]) {
1197                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1198                 if (ipv6_addr_any(&daddr))
1199                         return -EINVAL;
1200         }
1201
1202 out:
1203         return ip6gre_tunnel_validate(tb, data);
1204 }
1205
1206
1207 static void ip6gre_netlink_parms(struct nlattr *data[],
1208                                 struct __ip6_tnl_parm *parms)
1209 {
1210         memset(parms, 0, sizeof(*parms));
1211
1212         if (!data)
1213                 return;
1214
1215         if (data[IFLA_GRE_LINK])
1216                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1217
1218         if (data[IFLA_GRE_IFLAGS])
1219                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1220
1221         if (data[IFLA_GRE_OFLAGS])
1222                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1223
1224         if (data[IFLA_GRE_IKEY])
1225                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1226
1227         if (data[IFLA_GRE_OKEY])
1228                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1229
1230         if (data[IFLA_GRE_LOCAL])
1231                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1232
1233         if (data[IFLA_GRE_REMOTE])
1234                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1235
1236         if (data[IFLA_GRE_TTL])
1237                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1238
1239         if (data[IFLA_GRE_ENCAP_LIMIT])
1240                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1241
1242         if (data[IFLA_GRE_FLOWINFO])
1243                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1244
1245         if (data[IFLA_GRE_FLAGS])
1246                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1247 }
1248
1249 static int ip6gre_tap_init(struct net_device *dev)
1250 {
1251         struct ip6_tnl *tunnel;
1252         int ret;
1253
1254         ret = ip6gre_tunnel_init_common(dev);
1255         if (ret)
1256                 return ret;
1257
1258         tunnel = netdev_priv(dev);
1259
1260         ip6gre_tnl_link_config(tunnel, 1);
1261
1262         return 0;
1263 }
1264
1265 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1266         .ndo_init = ip6gre_tap_init,
1267         .ndo_uninit = ip6gre_tunnel_uninit,
1268         .ndo_start_xmit = ip6gre_tunnel_xmit,
1269         .ndo_set_mac_address = eth_mac_addr,
1270         .ndo_validate_addr = eth_validate_addr,
1271         .ndo_change_mtu = ip6_tnl_change_mtu,
1272         .ndo_get_stats64 = ip_tunnel_get_stats64,
1273         .ndo_get_iflink = ip6_tnl_get_iflink,
1274 };
1275
1276 #define GRE6_FEATURES (NETIF_F_SG |             \
1277                        NETIF_F_FRAGLIST |       \
1278                        NETIF_F_HIGHDMA |                \
1279                        NETIF_F_HW_CSUM)
1280
1281 static void ip6gre_tap_setup(struct net_device *dev)
1282 {
1283
1284         ether_setup(dev);
1285
1286         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1287         dev->destructor = ip6gre_dev_free;
1288
1289         dev->features |= NETIF_F_NETNS_LOCAL;
1290         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1291 }
1292
1293 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1294         struct nlattr *tb[], struct nlattr *data[])
1295 {
1296         struct ip6_tnl *nt;
1297         struct net *net = dev_net(dev);
1298         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1299         int err;
1300
1301         nt = netdev_priv(dev);
1302         ip6gre_netlink_parms(data, &nt->parms);
1303
1304         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1305                 return -EEXIST;
1306
1307         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1308                 eth_hw_addr_random(dev);
1309
1310         nt->dev = dev;
1311         nt->net = dev_net(dev);
1312         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1313
1314         dev->features           |= GRE6_FEATURES;
1315         dev->hw_features        |= GRE6_FEATURES;
1316
1317         if (!(nt->parms.o_flags & GRE_SEQ)) {
1318                 /* TCP segmentation offload is not supported when we
1319                  * generate output sequences.
1320                  */
1321                 dev->features    |= NETIF_F_GSO_SOFTWARE;
1322                 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1323
1324                 /* Can use a lockless transmit, unless we generate
1325                  * output sequences
1326                  */
1327                 dev->features |= NETIF_F_LLTX;
1328         }
1329
1330         err = register_netdevice(dev);
1331         if (err)
1332                 goto out;
1333
1334         dev_hold(dev);
1335         ip6gre_tunnel_link(ign, nt);
1336
1337 out:
1338         return err;
1339 }
1340
1341 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1342                             struct nlattr *data[])
1343 {
1344         struct ip6_tnl *t, *nt = netdev_priv(dev);
1345         struct net *net = nt->net;
1346         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1347         struct __ip6_tnl_parm p;
1348
1349         if (dev == ign->fb_tunnel_dev)
1350                 return -EINVAL;
1351
1352         ip6gre_netlink_parms(data, &p);
1353
1354         t = ip6gre_tunnel_locate(net, &p, 0);
1355
1356         if (t) {
1357                 if (t->dev != dev)
1358                         return -EEXIST;
1359         } else {
1360                 t = nt;
1361         }
1362
1363         ip6gre_tunnel_unlink(ign, t);
1364         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1365         ip6gre_tunnel_link(ign, t);
1366         return 0;
1367 }
1368
1369 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1370 {
1371         struct net *net = dev_net(dev);
1372         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1373
1374         if (dev != ign->fb_tunnel_dev)
1375                 unregister_netdevice_queue(dev, head);
1376 }
1377
1378 static size_t ip6gre_get_size(const struct net_device *dev)
1379 {
1380         return
1381                 /* IFLA_GRE_LINK */
1382                 nla_total_size(4) +
1383                 /* IFLA_GRE_IFLAGS */
1384                 nla_total_size(2) +
1385                 /* IFLA_GRE_OFLAGS */
1386                 nla_total_size(2) +
1387                 /* IFLA_GRE_IKEY */
1388                 nla_total_size(4) +
1389                 /* IFLA_GRE_OKEY */
1390                 nla_total_size(4) +
1391                 /* IFLA_GRE_LOCAL */
1392                 nla_total_size(sizeof(struct in6_addr)) +
1393                 /* IFLA_GRE_REMOTE */
1394                 nla_total_size(sizeof(struct in6_addr)) +
1395                 /* IFLA_GRE_TTL */
1396                 nla_total_size(1) +
1397                 /* IFLA_GRE_TOS */
1398                 nla_total_size(1) +
1399                 /* IFLA_GRE_ENCAP_LIMIT */
1400                 nla_total_size(1) +
1401                 /* IFLA_GRE_FLOWINFO */
1402                 nla_total_size(4) +
1403                 /* IFLA_GRE_FLAGS */
1404                 nla_total_size(4) +
1405                 0;
1406 }
1407
1408 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1409 {
1410         struct ip6_tnl *t = netdev_priv(dev);
1411         struct __ip6_tnl_parm *p = &t->parms;
1412
1413         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1414             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1415             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1416             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1417             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1418             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1419             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1420             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1421             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1422             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1423             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1424             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1425                 goto nla_put_failure;
1426         return 0;
1427
1428 nla_put_failure:
1429         return -EMSGSIZE;
1430 }
1431
1432 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1433         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1434         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1435         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1436         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1437         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1438         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1439         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1440         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1441         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1442         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1443         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1444 };
1445
1446 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1447         .kind           = "ip6gre",
1448         .maxtype        = IFLA_GRE_MAX,
1449         .policy         = ip6gre_policy,
1450         .priv_size      = sizeof(struct ip6_tnl),
1451         .setup          = ip6gre_tunnel_setup,
1452         .validate       = ip6gre_tunnel_validate,
1453         .newlink        = ip6gre_newlink,
1454         .changelink     = ip6gre_changelink,
1455         .dellink        = ip6gre_dellink,
1456         .get_size       = ip6gre_get_size,
1457         .fill_info      = ip6gre_fill_info,
1458         .get_link_net   = ip6_tnl_get_link_net,
1459 };
1460
1461 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1462         .kind           = "ip6gretap",
1463         .maxtype        = IFLA_GRE_MAX,
1464         .policy         = ip6gre_policy,
1465         .priv_size      = sizeof(struct ip6_tnl),
1466         .setup          = ip6gre_tap_setup,
1467         .validate       = ip6gre_tap_validate,
1468         .newlink        = ip6gre_newlink,
1469         .changelink     = ip6gre_changelink,
1470         .get_size       = ip6gre_get_size,
1471         .fill_info      = ip6gre_fill_info,
1472         .get_link_net   = ip6_tnl_get_link_net,
1473 };
1474
1475 /*
1476  *      And now the modules code and kernel interface.
1477  */
1478
1479 static int __init ip6gre_init(void)
1480 {
1481         int err;
1482
1483         pr_info("GRE over IPv6 tunneling driver\n");
1484
1485         err = register_pernet_device(&ip6gre_net_ops);
1486         if (err < 0)
1487                 return err;
1488
1489         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1490         if (err < 0) {
1491                 pr_info("%s: can't add protocol\n", __func__);
1492                 goto add_proto_failed;
1493         }
1494
1495         err = rtnl_link_register(&ip6gre_link_ops);
1496         if (err < 0)
1497                 goto rtnl_link_failed;
1498
1499         err = rtnl_link_register(&ip6gre_tap_ops);
1500         if (err < 0)
1501                 goto tap_ops_failed;
1502
1503 out:
1504         return err;
1505
1506 tap_ops_failed:
1507         rtnl_link_unregister(&ip6gre_link_ops);
1508 rtnl_link_failed:
1509         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1510 add_proto_failed:
1511         unregister_pernet_device(&ip6gre_net_ops);
1512         goto out;
1513 }
1514
1515 static void __exit ip6gre_fini(void)
1516 {
1517         rtnl_link_unregister(&ip6gre_tap_ops);
1518         rtnl_link_unregister(&ip6gre_link_ops);
1519         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1520         unregister_pernet_device(&ip6gre_net_ops);
1521 }
1522
1523 module_init(ip6gre_init);
1524 module_exit(ip6gre_fini);
1525 MODULE_LICENSE("GPL");
1526 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1527 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1528 MODULE_ALIAS_RTNL_LINK("ip6gre");
1529 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1530 MODULE_ALIAS_NETDEV("ip6gre0");